research on improving methods for estimating crop...

127
SYNTHESIS OF LITERATURE AND FRAMEWORK Research on Improving Methods for Estimating Crop Area, Yield and Production under Mixed, Repeated and Continuous Cropping January 2016 Working Paper No. 5

Upload: voque

Post on 24-Mar-2018

218 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

SYNTHESIS OF LITERATURE AND

FRAMEWORK

Research on Improving Methods

for Estimating Crop Area, Yield

and Production under Mixed,

Repeated and Continuous

Cropping

January 2016

Working Paper No. 5

Page 2: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

Global Strategy Working Papers

Global Strategy Working Papers present intermediary research outputs (e.g.

literature reviews, gap analyses etc.) that contribute to the development of

Technical Reports.

Technical Reports may contain high-level technical content and consolidate

intermediary research products. They are reviewed by the Scientific Advisory

Committee (SAC) and by peers prior to publication.

As the review process of Technical Reports may take several months, Working

Papers are intended to share research results that are in high demand and should

be made available at an earlier date and stage. They are reviewed by the Global

Office and may undergo additional peer review before or during dedicated

expert meetings.

The opinions expressed and the arguments employed herein do not necessarily

reflect the official views of Global Strategy, but represent the author’s view at

this intermediate stage. The publication of this document has been authorized

by the Global Office. Comments are welcome and may be sent to

[email protected].

Page 3: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

Synthesis of Literature and Framework

Under the project

Research on Improving

Methods for Estimating Crop Area, Yield and Production under Mixed, Repeated and

Continuous Cropping

Page 4: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

Drafted By

U.C. Sud

Tauqueer Ahmad

V.K. Gupta

Hukum Chandra

Prachi Misra Sahoo

Kaustav Aditya

Man Singh

Ankur Biswas

and

ICAR-Indian Agricultural Statistics Research Institute

New Delhi, India

2015

Page 5: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

Table of contents Abstract……………………………………………………………………………... i

Acronyms and Abbreviations………………………………………………………. ii

1. Introduction……………………………………………………………………... 1

2. Concepts and definitions……………………………………………….............. 3

3. Methods based on mapping…………………………………………………….. 12

3.1. Methods based on mapping……………………………………………. 13

3.2. Land surveying methods………………………………………………. 22

3.3. Farmer assessment of crop area……………………………………….. 30

3.4. Farmer assessment of crop area……………………………………….. 31

4. Methods of crop yield estimation………………………………………………. 37

4.1. Whole plot harvest……………………………………………………... 37

4.2. Crop cut and farmers’ estimate methods……………………………… 38

4.3. Other methods of crop yield estimation………………………………. 48

5. Crop acreage and yield estimation under mixed cropping…………………... 56

6. Crop acreage and yield estimation under continuous cropping……………... 64

7. Small area estimation………………………………………………………….... 66

7.1. Synthetic estimation…………………………………………………... 67

7.2. Mixed models in small area estimation……………………………….. 68

7.3. Extension of mixed models in small area estimation…………………. 72

7.4. Some applications of small area estimation in agriculture……………. 73

8. Country experiences on crop acreage and yield estimation…………………. 75

8.1. Bulgaria………………………………………………………………... 75

8.2. Brazil…………………………………………………………………... 76

8.3. Canada………………………………………………………………… 76

8.4. Egypt…………………………………………………………………... 78

8.5. Ethiopia………………………………………………………………... 79

8.6. Nigeria…………………………………………………………………. 80

8.7. France………………………………………………………………….. 81

8.8. United States of America……………………………………………… 82

8.9. India…………………………………………………………………… 83

8.10. South Africa…………………………………………………………... 84

8.11. Sudan…………………………………………………………………. 85

8.12. Morocco……………………………………………………………… 86

8.13. Rwanda………………………………………………………………. 86

Page 6: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

8.14. Indonesia……………………………………………………………… 87

8.15. Jamaica……………………………………………………………….. 88

9. Conclusions……………………………………………………………………… 89

Annex I……………………………………………………………………... 91

Annex II…………………………………………………………………….. 95

References…………………………………………………………………………... 105

Page 7: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

i

Abstract This technical report focuses on methodologies used for estimation of a crop

area and crop yield under mixed and continuous cropping. It provides a review

of the literature on methodologies used for estimation of crop area and crop

yield in developed and developing countries. The report begins with

descriptions of relevant concepts and definitions and follows with descriptions

of various methods used for estimation of crop area and crop yield along with

their positive and negative attributes. Also in the report, gold standard methods

of crop area and crop yield estimation are identified and various methods are

compared against those gold standards and country experiences with regard to

crop area and crop yield estimation are discussed. The potential of a small area

estimation technique to provide reliable estimates of crop area and crop yield in

cases of small sample size is also described in detail along with some

applications. Based on a critical review of literature, some relevant issues are

highlighted and recommendations are made.

The authors would like to thank Naman Keita, Christophe Duhamel and

Michael Austin Rahija of the Food and Agriculture Organization of the United

Nations (FAO) for their helpful and constructive comments that contributed to

improving the final version of the technical report. The authors are also grateful

to Consuelo Señoret of FAO for her continuous administrative support. The

authors thank Dr. S.D. Sharma for improving the quality of the technical report.

The authors gratefully acknowledge the suggestions made by peer reviewers,

which led to further improvement in the technical report.

Page 8: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

ii

Acronyms and Abbreviations

CONAB National Food Supply Company, Brazil

CSA Central Statistical Agency

EBLUE Empirical Best Linear Unbiased Estimator

EBLUP Empirical Best Linear Unbiased Predictor Estimator

ESA Egyptian Survey Authority

FAO Food and Agriculture Organization of the United Nations

FASAL Forecasting Agricultural output using Space, Agrometeorology

and Land based observations

FCC False Color Composite

GCES General Crop Estimation Surveys

GDP Gross Domestic Product

GHS General Household Survey

GIS Geographic Information System

GPS Global Positioning System

IBGE Brazilian Institute of Geography and Statistics

ICAR India Council of Agriculture Research

IFPRI International Food Policy Research Institute of Niger

INRAN National Agricultural Research Institute of Niger

LACIE Large Area Crop Inventory Experiment

LISS Linear Imaging Self Scanner

LUCAS Land Use/Cover Area frame Survey

MAAIF Ministry of Agriculture Animal Industry and Fishery, Uganda

MAPA Ministry of Agriculture, Livestock, and Supply, Brazil

MSS Multispectral Scanner System

MXL Maximum Likelihood

NBS National Bureau of Statistics

NDVI Normalized Difference Vegetation Index

NSSO National Sample Survey Office

PC Principal Components

PSU Primary Sampling Unit

SSU Secondary Sampling Unit

UBOS Uganda Bureau of Statistics

USDA United States Department of Agriculture

Page 9: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

1

1 Introduction Agriculture, with its related sectors, such as horticulture, animal husbandry,

fishery and forestry, is the largest livelihood provider and contributes

significantly to the national gross domestic product (GDP) in most of

developing/under developed countries. Lack of quality agricultural statistics

may lead to misallocation of scarce resources and policy formulations that fail

to resolve critical development problems (Kelly et al. 1995). As such, the

generation of timely, reliable and quality agricultural statistics is critical for

policy planning and administrative decision-making. Reviewing and upgrading

a mechanism for the continuous generation of timely and reliable agricultural

statistics, therefore, is of paramount importance.

Two major approaches for development of appropriate methodologies for the

generation of agricultural statistics are (a) complete enumeration and (b) sample

survey. Sample survey is generally adopted because it provides an output that is

cost effective, timely, precise and of high quality. The choice of an appropriate

sampling design and the estimation procedure are therefore critical in this

context.

The availability and quality of agricultural statistics have been declining in the

developing and underdeveloped countries. Some of these countries even lack

the capacity to produce a minimum set of data as evidenced by the poor

response rates to Food and Agriculture Organization of the United Nations

(FAO) questionnaires (World Bank 2010). The global strategy to improve

agricultural and rural statistics is a groundbreaking effort to strengthen

agricultural statistics. At its forty-first session, in February 2100, the United

Nations Statistical Commission endorsed the technical content and strategic

directions of the global strategy and urged the rapid development of an action

plan for implementation (hereafter, Global Action Plan). One of the issues

identified in the global strategy under the component data collection methods is

estimating area, yield and production of mixed, repeated and continuous

cropping. The purpose of this document is to provide a synthesis of literature

Page 10: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

2

and framework for estimating area, yield and production, in general, and in

mixed, repeated and continuous cropping, in particular. To begin with, in

section 2, the related concepts and definitions of various terms used in the

document are given. The different methods used for estimation of crop area and

crop yield are discussed in sections 3 and 4, respectively. Sections 5and 6

depict the methods of estimating crop area and yield under mixed and

continuous cropping systems, respectively. Section 7 deals with small area

estimation technique. A glimpse of various methods of crop area and yield

estimation procedures employed in different countries is given in section 8.The

concluding remarks on "synthesis of literature and framework" are given in

section 9.

Page 11: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

3

2 Concepts and definitions

The purpose of this section is to describe concepts and give definitions of

various terms used in the report. Before giving definitions, it should be noted

that crop production estimates are usually derived from two key components,

namely total area harvested/planted and yield per unit area. The area under the

crop and the yield per unit area are estimated separately and the product of

those two estimates provides the estimate of the total output or crop production.

However, there is another approach of estimation of crop yield wherein crop

yield is estimated by dividing crop production by crop area. Under this

approach, crop production estimates are obtained using farmers recall/diary

method. The former approach is widely practiced in most of the countries.

Crop area

Estimated crop area is one of the two major components of estimated crop

production. To estimate crop production properly, the crop area must be

estimated precisely, accurately and correctly. In the following paragraph,

concepts and definitions of the crop area are given.

Crop area can be defined as “the horizontal projection of a particular extent of

earth’s surface “which corresponds to the area shown on cadastral maps (FAO

1982). This definition takes care of crop areas in the plains and in hilly regions.

It also ensures that the total area is equal to the sum of the component area,

which is not the case when an area is measured on slopes. Some of the

commonly used methods for estimating crop area are land surveying methods,

farmers’ appraisal and remote sensing. Because of natural calamities or

economic considerations, certain areas planted or sown with a given crop are

not harvested or are harvested before the crop reaches maturity. Reasons cited

for this are poor germination, pest or disease damage, animal grazing, floods,

lack of labor or lack of market. Thus, the area planted may not be equal to the

area harvested. In addition, some crops, such as cassava, may be grown as an

insurance measure and are only fully harvested during a drought or a food

shortage. In any of the above circumstances, the definition of crop area that is

Page 12: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

4

used has a large influence on area, yield and production estimates. The concept

of crop area, therefore, needs to be subdivided into sown or planted area and

harvested area.

Sown or planted area: The area that corresponds to the total sown area for

producing a specific crop during a given year is termed as the sown or planted

area. Regulation (EC) No. 543/20091 on crops statistics defines this as cropped

area. Sown or planted area figures are required to estimate quantities used for

seeding purposes. Data on sown area, also, helps to come up with a rough

estimate of the production. The concept of planted area is used in such countries

as India and Bangladesh.

Harvested area: FAO (Martinez et al. 2015) and Regulation (EC) No.

543/2009 on crops statistics defines harvested areas the part of the sown or

planted area that is harvested. The harvested area may, therefore, be equal to or

less than the planted area. It serves as an important basis for obtaining a reliable

and accurate yield and production estimates. Both planted area and harvested

area concepts are used in Sri Lanka.

For practical reasons, most agricultural surveys and censuses record crop area

as the planted area instead of a harvested area. Casley &Kumar (1988),

however, argue that the harvested area is always the most relevant area

measurement for recording crop area and estimating crop yield at the plot level.

Yield estimates by daily recording, sampling harvest units, farmer recall, and

crop cards are based on planted area, whereas yield estimates by crop cut and

whole plot harvesting are based on the harvested area. On the other hand, for

the Uganda National Census of Agriculture and Livestock (1990-1991), crop

yields were determined on the basis of crop cuts using the planted area as the

crop area (MAAIF 1992).

Some additional concepts of area that may be useful in considering area

statistics (FAO 1982) are described below:

Area intended for planting (or sowing) refers to the area that the holders plan

or intend to sow under various crops. Area intended for planting may be equal

to the planted area, but it can be smaller or larger than the planted area. These

data are usually collected before planting starts.

Area tilled describes the part of arable land on which work has been done to

make the land fit for raising crops at a given point of time. The work involved

may comprise different practices, such as ploughing, harrowing and manuring.

1Regulation (EC) No 543/2009 of the European Parliament and of the Council. 2009.Crop

statistics and repealing Council Regulations (EEC) No 837/90 and (EEC) No 959/93, Official

Journal of the European Union, Special edition in Croatian Chapter 03, 27, 318 – 328

Page 13: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

5

Area damaged gives an account of loss due to the effect of unfavorable factors,

such as floods, rain, winds, snowfall and insect attack.

Area abandoned is the part of the area intended or area planted that has been

abandoned for raising crop or harvesting crop for different reasons, such as

difficult weather conditions for raising the crop. Crop areas are sometimes

abandoned from harvesting if a poor harvest is expected.

Regulation (EC) No 543/2009 on crops statistics provides some more important

concepts and definitions:

Harvest year: The calendar year in which the harvest begins.

Utilized agricultural area: Total area taken up by arable land, permanent

grassland, permanent crops and kitchen gardens used by the holdings,

regardless of the type of tenure or its use as common land.

Area under cultivation corresponds to the total area sown or planted, but after

the harvest, it excludes the ruined area resulting from, for example, natural

disasters and calamities. Area under cultivation may be the same or smaller

than the sown or planted area.

Production area: In connection with permanent crops, production area means

the area that can potentially be harvested in the reference harvest year. It

excludes all non-producing areas, such as new plantations that have not yet

started to produce crops.

Main area of a given parcel: The area where the parcel has been used only

once during a given crop year is unequivocally defined as the main area of a

given parcel.

Crop Yield

The concept of crop yield is generally used to represent the average amount of

produce obtained per unit of the crop area, while the concept of production

covers the total amount produced (FAO 1982). Regulation (EC) No 543/2009

on crops statistics defines crop yield as the harvested production per unit area

under cultivation. In cases of tree crops the concept of yield covers the average

amount of produce per tree and the production is calculated as the product of

the average yield per tree and the number of producing trees. Some of the

commonly used methods for measuring crop yield are crop cut and farmers’

appraisal.

The three main concepts of crop yield being used by many countries are

described below: (Fermont & Benson 2011):

Page 14: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

6

Biological yield or gross yield is the yield obtained before any loss

occurs during and after harvest;

Harvested yield is the biological yield minus harvest losses;

Economic yield is the quantity that the farmer can use after post-

harvest losses that may occur during cleaning, threshing, winnowing

and drying (Casley &Kumar 1988; Keita 2003).

Some of the important concepts and definitions relating to crops are discussed

below.

Primary crops: Primary crops are crops that come directly from the land

without having undergone any real processing apart from cleaning (FAO

Statistics 2011). They can be further divided into temporary and permanent

crops:

Temporary crops: Crops that are sown and harvested during the same

agricultural year, sometimes more than once.

Permanent crops: Crops that are sown or planted once and need not be

replanted after each annual harvest.

Sole crop: A crop grown in pure stand.

Mono-cropping: The practice of growing only one crop on a piece of land year

after year is termed as mono-cropping, such as growing only winter season

crops in a dryland area. This may be due to climatic or socioeconomic

conditions or the result of specialization of a farmer in growing a particular

crop. For instance, groundnut, cotton and sorghum are grown year after year

due to scarce rainfall in different countries. This term can be better described as

continuous cropping or continuous mono-cropping.

Mixed cropping: Mixed crops refer to two or more different temporary and

permanent crops grown simultaneously in the same field or plot (FAO 1982).

Each of these crops is referred to as associated crops. As per Regulation (EC)

No. 543/2009 on crop statistics, a combination of crops that are cultivated on a

parcel of agricultural land at the same time is termed as “combined cropping”.

For example, in Brazil, cocoa is planted with clove and rubber and to some

extent with coconut (Alvim &Nair 1986).Uganda National Census of

Agriculture and Livestock (1990-1991) indicates that80 to 90 percent of their

planted area are in mixed stands (MAAIF 1992).Under this cropping scenario,

it is recommended that the estimated area for each one of the associated crops

be the area that the particular crop would have covered if it had been grown

Page 15: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

7

alone (FAO 1982).The area under cultivation is distributed between the crops in

proportion to the area of the land they are cultivated on.

To elaborate on mixed cropping, the following terms need to be defined:

Intercropping: The practice of intercropping refers to growing more than one

crop in the same land area in rows of definite proportion and pattern. When a

particular crop is planted between rows of another crop it is usually referred to

as an interplanted crop. A good example of this is planting sorghum and

groundnut between rows of cotton. In Asian and African countries,

intercropping is a widespread and traditional cropping system because of scarce

land and small field size. Thus, intercropping has the potential to increase

natural resources in space and time. Intercropping is commonly used because it

gives higher returns per unit area than sole cropping. It acts as an insurance

against crop failure in an abnormal year. It also helps the soil fertility as the

nutrient uptake is made from both layers.

Intercropping can be further divided into the following subcategories

(Vandermeer 1992):

Mixed Intercropping: Growing two or more crops simultaneously on

the same piece of land with no distinct row arrangement is called mixed

intercropping. This practice is more commonly applied in traditional

and subsistence farming in many developing countries. For example, in

some parts of India, large numbers of crops are sown in mixed intercrop

arrangement.

Page 16: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

8

Figure 1: Pearl millet, Maize, Sunflower, Ragi, 2 types of Sorghum, Groundnut, 2

Green leafy Vegetables in India

Source: http://agropedia.iitk.ac.in/content/mixed-cropping-pearl-milllet.

Row Intercropping: Growing two or more crops simultaneously where

one or more crops are planted in rows is called row intercropping. This

pattern is usually found in areas involved in intensive agriculture where

the plough has replaced the machete and fire is used for land

preparation. As an example, in parts of India, pearl millet-groundnut

intercropping and sorghum-pigeon pea combinations are sown as row

intercropping (figure 2). Figure 3 shows wheat-pea intercrop fields in

the United States of America.

Page 17: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

9

Figure 2: Sorghum-pigeon pea row intercropping in India

Source: http://agropedia.iitk.ac.in/content/intercropping-pearl-millet

Figure 3: Wheat-pea row intercrop in the United

Source: Machado 2009

Strip Intercropping: Growing two or more crops simultaneously in

different strips wide enough to carry out independent cultivation, but

narrow enough for the crops to interact agronomically is called strip

intercropping. This form of intercropping is more common in highly

modernized systems, especially in systems that require intensive use of

machinery. For example, maize, soybean and other cereals are planted

as strip intercropping in the United States (figure 4).

Page 18: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

10

Figure 4. Maize, soybean and other cereals in the United States as strip intercropping

Source: http://orgprints.org/18950/

Relay intercropping: A technique in which different crops are planted

during different time periods in the same field and both (or all) crops

are being gown simultaneously at least part of the time is called relay

intercropping. This form of intercropping may actually include the other

three as subsets, as its primary categorization variable is time. In

Northern China, farmers harvest wheat and maize within one growing

season under a relay intercropping system (Knörzeret al. 2009).

Continuous cropping

Continuous crops or successive crops or sequential crops or catch crops are

crops that are sown and harvested from the same piece of land previously

occupied by another crop, or even by the same crop, during the same

agricultural year. The area of crops growing under this condition is accounted

for in the total crop area and if necessary, ad hoc surveys for this purpose

should be conducted (FAO 1982). Continuous cropping can take different

forms:

Continuing planting/harvesting: Repeated planting and harvesting of

crops at particular intervals of time in an agricultural year is termed as

continuing planting/harvesting. The practice of continuous planting is

very common in many African countries. For example, farmers in

Uganda plant and harvest crops throughout the year.

Page 19: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

11

Successive cropping: Planting and harvesting either the same crop or

different crops more than once in the same field during the agricultural

year (one crop is planted after the other crop is harvested) is termed as

successive cropping.

Some other forms of continuous cropping are found by:

Replanting the same crop on the same land after it has been damaged

(totally or partially) through natural or other causes.

Enlarging gradually (at interval of times) the area of land planted to

one or several crops.

Page 20: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

12

3 Methods of estimating crop

area The information on crop area statistics forms the backbone of an agricultural

statistical system. The crop area should cover the entire area devoted to each

crop, including, when necessary, estimates for smaller areas not covered in the

current annual area surveys.

The area under crop in each season and yield rate is required to estimate the

production of a particular crop. As farmers grow crops under different cropping

systems, such as mono-cropping, mixed cropping, continuous cropping or

repeated cropping, a sound methodology is required to obtain the area under a

particular crop. This can be obtained with relative ease when a single crop is

grown in a field in a particular season, but it becomes difficult under mixed

cropping or intercropping. It becomes even more difficult in areas where the

land records and crop registers are not maintained properly.

Crop area often has a strong inter annual variability. Crop yield remains

relatively stable, barring arid countries during normal climatic conditions, but in

cases in which there is variability in weather conditions with droughts or floods,

crop yield may be variable. Area statistics is generally generated using a sample

survey approach. However, with more emphasis being placed on disaggregate

level planning, there is an increased need for estimating crop area with respect

to, among other factors, different varieties, irrigation availability and soil type,

etc. Advancements in computer and space technology during the past few

decades has led to the availability of compact and high performance computing

systems that are well suited to the demands of remote sensing satellite data

processing. Remote sensing satellite data are advantageous because of the vast

area coverage, synoptic view and online information. A number of

methodologies are suitable for providing acreage estimates for major crops

using remote sensing satellite data. Remote sensing integrated with geographic

information system (GIS) technologies are also being used for crop acreage

estimation. The Global Positioning System (GPS) is being used to identify the

Page 21: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

13

selected plots for estimation of the corresponding plot area. The various

methods for area estimation are broadly grouped into two categories, as

discussed below.

3.1. METHODS BASED ON MAPPING

Mapping is essential for crop area estimation. Crop areas can be easily

estimated when large-scale accurate maps depicting the actual positions of

parcels and fields are available. In many countries such maps are compiled and

kept up-to-date by the cadastral services to generate revenue. Cadastral maps

show the boundaries and area of each parcel of land together with an

identification number. The name of the owner and other characteristics of the

parcel are also kept in a separate register. In many developing countries such

cadastral maps do not exist or do not cover the whole country. In the European

Union, cadastral maps for taxing purposes seldom correspond to the real

cropped parcels and cannot be used for crop statistics. This situation is worse in

most developing countries. Yearly administrative registers are usable in a few

countries, such as Scandinavian countries, which have few large farms and only

a marginal amount of small plots for hobby, which are often excluded from the

registers. In many countries, survey maps are prepared on the basis of aerial

photographs. This process can be problematic because coverage based on aerial

photographs may not be complete. Additionally, the scale may vary from

region to region and the maps and photographs may not be up-to-date. Some

developing countries, such as Algeria, have undertaken plot mapping activities,

in order to define a sampling frame, but not for direct area estimation. This may

be useful, but is costly. Direct area estimation from classified satellite images

usually has a large bias that becomes enormous if the agricultural landscape is

complex. Various methods adopted for estimating crop area using maps were

described in detail by FAO (1982). Craig & Atkinson (2013) provide a vide

literature review on crop area estimation methodologies. Various method of

crop area estimation are elaborated below.

3.1.1. AREA FRAME OF SEGMENTS WITH PHYSICAL

BOUNDARIES

For effective implementation of area sampling, a complete set of large-scale,

accurate and detailed maps and/or aerial photos/imagery of a study area is

required. Updating regularly the maps, so that they are error-free (there is no

omission or duplication), is a practical necessity. In addition, the maps should

cover the entire study area. Subdividing maps/photos/imagery into segments is

a prerequisite. To the extent possible, the segments should be "natural

geographic areas" established by natural borders, such as ridges, rivers and

Page 22: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

14

roads. When these segments are numbered and perhaps characterized through

the recording of ancillary information, the maps or photos/imagery are

transformed into frames of area sampling units from which samples of

segments to be observed are selected.

When the cadastral maps are available, the fields are numbered and those

falling within the sample segment can be identified and listed. The lists of

selected fields are sent to the enumerator who is assigned a given area. The

enumerator records the name of the crop for each of the fields falling in his/her

assigned area. A precondition for proper implementation of this technique is

that the boundaries of the listed fields should be clear and well known and

clearly designated in local records. In situations in which more than one crop is

grown in the field, the area occupied by each crop can be determined either

subjectively or by measurement. Subjective determination of a crop area is

usually a rapid and cheap process, but it also may not be very accurate. On the

other hand, determination of a crop area through measurements may be very

accurate, but it can be very expensive.

3.1.2. AREA FRAMES OF SEGMENTS WITH A REGULAR SHAPE

This method is implementable when large-scale, accurate and detailed maps or

photographs/imagery are available. It is also known as grid sampling. Under

this method, small administrative or agro economic regions are covered by a

large number of non-overlapping photographs/imagery or maps, which can be

used as primary sampling units. On each sample map or photograph (when

maps or photos/imagery are considered as primary sampling units), a grid is

superimposed in which each square has a fixed known area, such as 10

hectares. The ultimate sampling units are squares on the grid, which are

identified based on coordinates. The crops in the different fields within the

selected sample square are then identified on the ground. Thus, in this method,

the total area of each of the sampling units is fixed. It is expected that the data

in the grid method of sampling shall have smaller variability because of the

equal size of the ultimate sampling unit. Furthermore, development of estimates

at a higher level is also easy, but identification of the boundaries of a square

may be difficult. The boundaries of the square do not need to be identified on

the ground, but problems can arise if the actual field boundaries do not coincide

with the boundaries visible on images (this may also happen for segments with

physical boundaries).

3.1.3. POINT SAMPLING

As for point sampling, the second stage sampling units within the primary

sampling units are points and the primary sampling units are the area covered

Page 23: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

15

by a map or a photograph/imagery. The points are selected either randomly or

systematically, such as by the points of intersection of the lines in a grid. The

enumerator is provided with an enlarged photograph/imagery on which the

sample points are drawn as the intersection of two branches of a cross. The

enumerator goes to the precise place and takes note of the type of soil, the land

use category and the crop. Notably, point sampling methods are not necessarily

a two-stage sampling.

Merits: A positive aspect of the point sampling technique is that the estimation

of crop areas appears to be simpler than the actual measurement of the areas of

the fields. Also, an enumerator can visit from 50 to 100 points on the ground

identified on an aerial photograph/imagery in a day's work, a far greater number

that is possible when using other techniques.

Drawbacks: The main problems faced during the implementation of this

method are: difficulty in identifying the exact location of the sample point on

the ground; identifying the type of cultivation; and potential inaccuracy of the

results when the area under the crop is a small fraction of the total area covered

by the photo. Many countries have needed to estimate crop areas using this

method. However, the results have not been satisfactory. The introduction of

GPS/Tablet/Smart phones for the field work has changed substantially the

problem of locating the point. An issue that remains problematic, however, is

clearly stating the priority between GPS and imagery if they are inconsistent

with each other. Priority should be given to the image.

The LUCAS (Land Use/Cover Area-frame Survey) was launched by the

European Union in 2001 based on a non-stratified, systematic, two-stage

sampling scheme, with primary sampling units (PSUs) defined as a rectangle of

1500 × 600 m following a grid of 18 km (Delincé 2001; Bettio et al. 2002). In

each PSU, 10 points (SSUs) were selected arranged on two rows of 5 points

with a step of 300 m. The “point” (SSU) was defined as a circle with a radius of

3 m to be consistent with ground survey specifications. For LUCAS 2006

(Gallego 2007), the sampling plan changed and became a two-phase sampling

plan of unclustered points (always defined as squares of 3 m). From 2002, the

Italian AGRIT survey (Martino 2003; Carfagna 2007) also adopts a sample

design based on unclustered points. It is a single stage two-phase sampling: the

first phase gives a systematic sample of unclustered points that are

photo/imagery-interpreted and subsampled (second phase) with higher rates in

agricultural strata. A similar approach was tested in Greece in 2004.

Jinguji (2014) introduced a new survey method known as the dot sampling

method by combining a traditional attribute survey method with two current

information technologies, namely Excel and Google Earth, to estimate rice

Page 24: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

16

planted area in Japan. This method has been effective in developing countries,

such as Sri Lanka and Thailand, and thus shows that it is possible to execute a

survey that is simple, reliable, and cost-effective in comparison to existing

methods. The survey results obtained through dot sampling method were closer

to official estimates. However, it must be noted that it may be difficult to adopt

this method in countries where critical Internet infrastructure required for using

Google Earth is inadequate. Another complication of using Google Earth may

be use of multi-date images with varying resolution resulting into inconsistency

in delineation of various crops. However, this application can be used even if

Internet access is not good with screen captures (geometrically corrected) or

even printouts on paper, but rules need to be established when the images are

updated and not geometrically consistent.

3.1.4. REMOTE SENSING AND GEOGRAPHIC INFORMATION

SYSTEM

Remote sensing is an important tool for generating agricultural statistics.

Remote sensing and geographic information system (GIS) technology has been

widely adopted to estimate crop area statistics.Classified satellite images and

land cover maps produced by photo-interpretation are useful tools for this

purpose. Direct use of satellite images in terms of pixel counting for area

measurement or simple area measurements of polygons of a land cover map is

not recommended.

Initially, three broad approaches for use of remote sensing to generate crop

statistics were recognized:

(1) Remote sensing forms a base for estimating parameters of spatial

variability through area frame sample design. It provides an efficient

and low cost stratification based on crop proportion derived from visual

interpretation or digital classification of remote sensing data;

(2) Direct and independent estimation that uses remote sensing data and a

recognition technique to estimate the crop area in the study region.

(3) The use of remote sensing data as an auxiliary variable helps make the

estimates based on ground surveys more precise and reduces the amount

of the field data to be collected, if the precision to be reached is fixed.

On the contrary, if the sample size is fixed, this approach provides

higher precision of the estimate.

Page 25: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

17

Crop acreage estimation using remote sensing data

The steps involved in the crop acreage estimation using remote sensing data

are:

(a) Study area extraction

As satellite data are available in the form of scenes covering fixed areas, the

acreage estimates are obtained by overlaying the administrative boundary on

the scenes and masking out pixels outside the boundary using a point-in-

polygon algorithm. Raw scenes can be used for overlaying boundaries by using

control points, but the scene rectification is carried out for mosaicing map

outputs or analysis in the GIS system. To sample an area and conduct a rapid

crop inventory, a rigorous sampling approach is required in which sample

segments in the form of square or rectangular sub scenes are extracted for

analysis. This sampling approach requires that the following parameters be

defined in a study: segment size; spatial stratification; sample allocation; and

sampling fraction. For LACIE 5 nmi ×6 nmi (9km×11km) segments and a2.5

percent sampling fraction was adopted (MacDonald and Hall 1980). Other

schemes have been adopted for other studies. Hallum and Perry (1984) have

proposed an objective methodology for defining the optimum sampling unit

size, which takes into consideration non-sampling errors in remote sensing, as

well as a model of sampling error variance as a function of segment size.

(b) Crop discrimination/identification from satellite data

Crop discrimination/mapping using space data is carried out either by visual or

digital interpretation techniques. Visual techniques are generally based on

standard false color composite (FCC) generated using green, red and near-

infrared (IR) bands assigned blue, green and red colors. Haack and Jampoler

(1995) demonstrated that a color composite formed by the best three bands (TM

bands 3, 4 and 5) gave better discrimination in comparison to the standard FCC

over a study site in the Imperial Valley in California. The digital techniques are

applied to each pixel, use a full dynamic range of observations and are preferred

for crop discrimination. A multi-temporal approach is used when single-date

data do not permit accurate crop discrimination.

In this case, the procedure entails the following three stages:

(i) Pre-processing: The pre-processing includes multi-date registration and

removal/ minimization of atmospheric effects through radiometric

normalization or atmospheric correction. The bands from the multiple

acquisition dataset are used in the multi-date classification procedures (Bizzell

et al. 1975; Bauer et al. 1979; Hixson et al. 1980).

Page 26: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

18

(ii) Data compression: Data compression techniques are commonly adopted.

For more than two dates, the number of bands becomes very large. The first

four principal components (PC) from three date data from TM (15 bands,

Band1 to Band5) in Germany (Mauser 1989) gave higher accuracies than

various band combinations.

(iii) Image classification: This is the major step involved in area measure as

the estimate depends on classification and accuracy of classification only. The

main techniques applied for classification of remote sensing satellite imageries

are (Vibhuteet al. 2013):

Supervised classification: In this approach, it is assumed that prior

knowledge for the classification of land, namely land cover types in

specific sites, is available. The most useful classifiers of this type are

the maximum likelihood classifier, the minimum distance classifier, the

parallelepiped classifier and the Mahalanobis classifier.

Unsupervised classification: In this category of methods, there is no

prior knowledge of the area to be classified. The two most common

classifiers of this group are K-means clustering and Iterative Self-

Organizing Data Analysis.

Hybrid classifier: These methods are a combination of supervised and

unsupervised classifications.

Fuzzy classifier: These classification methods are based on fuzzy logic.

Supervised/unsupervised and hybrid approaches are commonly used to classify

images. Techniques involving graphical shape, such as profile modeling

techniques, angular measures and Delta classifiers have also been used

(Badhwar 1984). Campbell et al. (1987) evaluated the direct use of temporal

spectral data for wheat acreage estimation in Australia. Using a discriminant

function on total data set improved separation, as compared to a profile-based

approach, with loss in separability occurring at both the data reduction steps,

namely spectral to the vegetation index (VI) and multi-date VI to spectral

profile step. Belward and de Hoyos (1987) compared accuracy for crop

classification between supervised maximum likelihood (MXL) and binary tree

classification approaches. Although similar accuracies were obtained by the two

classification procedures, the binary tree approach seems to be a viable

alternative to MXL in terms of ease of application and training of datasets.

Knowledge-based crop classification was suggested by Janssen & Middelkoop

(1992) for situations in which the crop rotation information about the area was

formalized using Markov chains and a transition matrix. These conditional

probabilities and remote sensing data were input for a Bayesian image

Page 27: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

19

classification. Many of these approaches are still at the research level and are

not being applied for practical purposes.

(c) Estimation of area under a crop in the study area

The estimation of a crop area from the classified image generated by the crop

discrimination procedure is the final step for obtaining crop area estimates.

Some of the estimators used are:

Direct estimator

This is the simplest estimator, where in a study region of known total area D,

the crop area (Zc) is given as:

Zc = D × (Xc/N)

where, Xc and N are number of pixels in crop C and total number of pixels,

respectively.

Depending upon the classification accuracy, the estimate could be biased. The

information on the classification accuracy is used in the next set of estimators

for obtaining improved crop area estimates (Maselli et al. 1990).

Global estimate using confusion matrices

Let A={Aij}be the confusion matrix on a test set, Ai the number of pixels

classified into crop Ci(ground truth), Aj the number of pixels classified into land

use Cj. Then, the elements of Pr= {Prij} and Pc= {Pcij} matrices are defined as:

Prji = Aij/Ai and Pcij = Aij/Aj

If Pr and Pc are unbiased estimators of the corresponding matrices for the whole

population, the next area estimators are unbiased too (Hay 1988; 1989; Jupp

1989):

Zdir = D ×(Pc×X/N) and Zinv = D ×(Pr-1×X/N)

This estimator exploits a larger part of the information contained in the

confusion matrix and gives good results, but has not been widely adopted.

Regression estimator

Regression estimator uses both ground data and classified remote sensing data

for acreage estimation. The regression estimators are described in standard

statistical texts. The formulation of separate form is:

(Re )

1

ˆ L

R h h g

h

Y N y

Page 28: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

20

where, (Re )ˆ ( - ) h g h h h hy y b X x

hy = average ground-reported crop area per sample segment of stratum h, i.e.,

1

1 hn

h hj

jh

y yn

ˆhb = regression coefficient of ground-reported area on remote sensing-derived

area based on nh segments for stratum h.

hX = average remote sensing-based area for all frame units of stratum h (Thus.

the entire area must be classified to obtain this mean of the population, namely,

1

1 hN

h hj

jh

X XN

hx = average remote sensing-based crop area per sample segment of stratum h,

i.e., 1

1 hn

h hj

jh

x xn

)

This approach has been extensively studied in the United States where in the

June Enumeration Survey, the crop data are collected on a sample basis through

a questionnaire method. Other examples of application of regression approach

are potato and canola-rapeseed in Canada using Landsat MSS (Ryerson et al.

1985) and wheat in Brazil using Landsat MSS (Moriera et al. 1986). Gonzalez-

Alonso & Cuevas (1993) show improved performance by using regression

estimator with confusion matrix information over a test site in Spain. Gonzalez-

Alonso et al. (1994) have compared results from the regression approach using

two different sampling approaches, namely square sample segments and

irregular segments over a test site (900 km2), in Spain. The weighted relative

efficiency in the case of the square segment approach was higher than for the

irregular segment approach.

In recent times, much work has been done in the field of remote sensing for

area estimation. Gallego (2004) gave an overview of different ways to use

satellite images for land cover area estimation. Carfagna & Gallego (2005)

discussed remote sensing as a valuable tool for agricultural statistics when area

frames or multiple frames were used. Sahoo et al. (2005) developed an

integrated approach based on remote sensing, GIS along with survey data for

crop area estimation under paddy crop in the North-Eastern hilly regions of

India. In those regions, particularly Meghalaya, different crops are grown at

different elevations. For example, paddy is grown in valleys, pineapple is

grown on hill slopes and potato and ginger are grown on relatively flat surfaces

Page 29: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

21

on the hill top. Thus, elevation plays a significant role in growing crops in those

regions. Furthermore, the total cropped area in the state is also much less (only

10 percent) and is mostly scattered throughout the state. These conditions make

it difficult for the usual stratification criterion based on administrative

boundaries to provide accurate/efficient strata. Based on the fact that elevation

and extent of cultivation plays a major role in influencing the acreage under a

crop, Sahoo et al. (2010) suggested using spatial stratification based on

elevation and the extent of cultivation for crop acreage estimation of multiple

crops in North-Eastern hilly regions in India.

Sahoo et al. (2012) extended the integrated methodology in Jaintia Hills district

of Meghalaya for estimation of acreage under paddy crop in India. Extensive

cloud coverage makes it difficult for researchers to estimate crop area using

remote sensing satellite data for kharif crops. Sahoo et al. (2013) developed

methodology for generating cloud-free images, which can be used to provide

reliable estimates of crop acreage and a methodology for estimating crop area

directly from satellite images having cloud cover and shadows. Goswami et al.

(2012) highlighted the application of remote sensing and GIS technologies for

the wheat acreage estimation for Indore district, Madhya Pradesh, India.

Wu and Li (2004) described a new method of crop area estimation using remote

sensing based on a stratified two-stage sampling method. To develop the strata,

physical factors, such as temperature, precipitation, soil type, sun eradiation and

proportions of main crop types, were considered. Crop areas were estimated

using cluster sampling assisted by remotely sensed images where the cluster-

sampling frame was built using 1:100000 scaled map-sheet. It had been difficult

to extract crop area from remotely sensed data, such as Landsat-TM. However,

this hurdle was overcome by first estimating crop proportion (Wu and Li

2004).Remote sensing technology is being used by some countries to generate

agricultural statistics, as this technology enables the actual ground situation to

be presented efficiently in which the agricultural features are evident and

clearly visible on the images. This method, however, is difficult to use in

countries where agricultural practices are dominated by small plots, diverse

planting dates, dispersed trees and intercropping systems. Even in the United

States where agriculture is dominated by mono-cropping on large fields, the use

of remote sensing techniques is still limited to supplementing area estimates

obtained using other methods in a few selected states. There are other factors

that limit the intensive application of remote sensing for collection or

generation of agricultural statistics, particularly crop area and crop yield.

Among them is crop discrimination, which depends on classification techniques

adopted for satellite data analysis. Classification of satellite data is a subjective

Page 30: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

22

method, as it depends on a number of factors, including, among them, ground

truthing, number of ground control points (GCPs) considered and method of

classification adopted. To test the classification, an accuracy assessment is done

which is also a subjective method, as it depends on the number of GCPs

considered for the accuracy assessment. An important aspect to be considered

when using remote sensing satellite data is the question of choosing satellite

data with appropriate spatial resolution. The satellite images are of varying

spatial resolutions. The most appropriate resolution for crop area or yield

estimation is debatable as satellite data of different pixel size/pixel purity are

available and should be carefully analyzed before finalizing the satellite data to

be used. The complexity of the landscape is also an important element that must

be investigated as crops also grow in hilly regions with undulating topography

and variable topographic geometry. This also plays an important role in satellite

data classification and affect the accuracy of classification. Another important

aspect is the cost involved in setting up the infrastructure, which includes cost

of software and satellite data and training of manpower for using satellite data,

which tends to on the high side.

Merits: This method provides quick crop area estimates covering a vast

geographical area. It is also useful for obtaining estimates of areas in hilly

terrains and in areas that are inaccessible.

Drawbacks: This method is expensive. There may be problems in getting

estimates for areas under cloud cover. The area estimates may not be accurate

for small plots.

3.2. METHODS BASED ON LAND SURVEYING

Land surveying determines the form and extent of a portion of the earth’s

surface by measurements. It may be linear or angular. Choice of a particular

method depends on the availability of resources, including, for example,

manpower or instruments used for the survey. Another important factor in these

surveys is the requirement of the level of accuracy of the results. In general, the

methods for measuring crop area in various countries are costly and time-

consuming. Various methods of measurement of crop area are described below:

3.2.1. POLYGON METHOD

This method, also known as the traverse measurement, traversing, chain and

compass, or the Topofil method is one of the most prevalent traditional ones

used to measure crop area (MAC 1965; Schøning et al. 2005). It can be

considered the gold standard for crop area estimation in view of its potential to

provide accurate area. In cases in which the plots are of a regular shape, the

Page 31: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

23

method involves measuring the length of each side and the angle of each corner

using a measuring tape and a compass. The surface area of the plot can then be

calculated using trigonometry (FAO 1982). In cases involving an irregular

shaped plot, an approximate polygon is obtained with straight sides by

demarcating its vertices on the ground. Due care is taken to ensure balancing

the protruding pieces left out in the process by including other small pieces that

are not part of the plot. During the give-and-take process and during the

measurement process, errors are introduced. According to Casley & Kumar

(1988), in situations in which the polygon does not close and the closing error is

larger than 3 percent of the perimeter of the polygon, the measurement

procedure should be repeated. The polygon method is commonly considered the

most objective method to accurately estimate crop areas. Diskin (1997)

observes that it may even be worth spending extra time, training and cost to

pursue this method. Nigeria widely deploys this method for estimating crop

area.

In this method, first the boundaries of a field to be measured are identified by

use of sight poles, and taking compass hearings and measuring the length of

each side of the obtained polygon. The traditional procedure of evaluating the

area of a field on the basis of measurements entails plotting the field in the

office by using a ruler and a protractor and then measuring the area of the

sketch by using a planimeter or grid paper. These methods were first

implemented in the 1974 Census of Agriculture in the Côte d'Ivoire. The

Statistics Division of FAO (FAO 1982) has developed several methods for

calculating areas with programmable calculators. A description of the

calculation of the area using the polygon method is given in annex1.

Merits: This method often provides accurate area measurements and can be

used directly in the field when measurements are made. The closure error can

be evaluated directly on the spot, and when the error of the measurement is

considered to be too large, the process can be repeated.

Drawbacks: Obtaining area measurement through this method is time-

consuming.

3.2.2. RECTANGULATION METHOD

Some land parcels and crop fields have irregular boundary lines that need not

be the straight lines. To accomplish this, the rectangulation method is

frequently used to measure the crop areas. The method consists of determining

the length of the parcel or field by measurement somewhere more or less across

the middle and then the area is obtained by determining the average width

through eye estimation. The method was slightly improved by having three

Page 32: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

24

measures of the width: two near the two ends of the field and one in the middle.

It was further improved by measuring the width at a large number of equidistant

positions, thus dividing the total area into a number of rectangles, more

precisely a number of trapeziums. A graphical illustration of rectangulation

(Figure 5), as given in FAO (1982), is shown below:

Figure 5: Rectangulation method

Feasibility: The method is feasible for land parcels having almost regular

shapes.

Merits: When the shape of the field is not too complicated, the method of

rectangulation for estimating crop areas can be useful and reliable.

Drawbacks: Application of this method on the ground can be difficult,

resulting in substantial errors in the measurement of area. Additionally, in order

to measure the length and the different widths, the enumerator may have to

enter in the field and in the process step on the crop, which would be

disagreeable to the farmer.

3.2.3. TRIANGULATION METHOD

To carry out this method, the enumerators split the field into a number of

imaginary triangles and measure the height and base of each triangle using a

measuring tape (FAO 1982). The total area of the field is obtained by adding

the area falling under each of the individual rectangles and triangles. A

graphical illustration of the triangulation method (figure 6) as given in FAO

(1982) is highlighted below:

Page 33: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

25

Figure 6: Triangulation method

Feasibility:

When the boundaries of the field are rectilinear and the field is a plane

polygon with well-defined vertices;

When all the vertices of the polygon representing the field can be seen

from one particular point;

When all the distances between the fixed point and the vertices can be

easily measured. This condition implies that either the farmer allows the

enumerator to measure the distances across the field (with the risk of

trampling the crop) or that the distance measuring instrument eliminates

the need for the surveyor enters the field.

Merits: The advantage of triangulation over other area measuring techniques is

that it requires that only the distances be measured as the triangle is uniquely

determined when its three sides are known.

Drawbacks: Triangulation does not permit the direct discovery of errors of

measurement.

The triangulation method is being used in Nigeria for crop area estimation. A

pilot study was conducted in 1963 for the 1965 agricultural census in Uganda,

in which the rectangulation and triangulation method were compared with the

polygon method (MAC 1965). It was found that though the polygon method

gave accurate area figures, it took a longer time. In addition, the polygon

method required two enumerators instead of just one needed for the

triangulation method. The study results also revealed that the rectangulation and

triangulation method underestimated the total cultivated area and area per

Page 34: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

26

holding by approximately 5 percent whereas some individual crop areas were

underestimated by 12-15 percent enumerator error, tedious and time-consuming

measurements are some other problems associated with this method. However,

Muwanga-Zake (1985) pointed out that the enumerator bias may be substantial

in districts that are covered by only a few enumerators.

3.2.4. P2/A METHOD (PERIMETER SQUARED OVER AREA)

If the perimeter is known, the area of a field can be quickly estimated, which is

used to check the gross errors of calculation, such as misplacing decimal point.

This is done by dividing the perimeter by a number between four and five and

then squaring the result (FAO 1982). The choice of the divisor is subjective and

depends on the degree of complexity of the boundary, such as the number of

sides or number of concavities. Thus, if the field is very complicated, the

divisor should be nearer to five and if the field is close to being a square or

rectangle then the divisor is nearer to four. When the field is highly variable, the

divisor might exceed even five.

The P2/A method, where P stands for perimeter and A for area, is a subjective

method based on a relative stable relation between the perimeter squared of a

field and its area (Mpyisi 2002; Fermont & Benson 2011). By using this

method, it is possible to get an estimate of a field area quickly, making it a good

option in situations in which there is an inadequate number of enumerators. The

method is relatively inexpensive and less time- consuming. The value of ratio

between P2 and A depends on the complexity of the shape of the plot.

A study in Rwanda showed that the simple correlation between perimeter

squared and area was found to be 0.95. Thus, a field's perimeter could be used

as a rough substitute of its area by utilizing the fixed ratio. A comparison

between P2/A method and the polygon method showed that P

2/A method

provided accurate estimates of area with a net error as low as 2 percent (Mpyisi

2002). This methodology has been used in the Food Security Research Project

under the Agricultural Statistics Division of the Ministry of Agriculture,

Rwanda to conduct surveys on a national sample with a limited number of

enumerators at a reasonable cost. However, based on experiences and analysis

conducted on thousands of plots measured during the Agricultural Census of

the Ivory Coast, the P2/A method failed produce any acceptable results and at

the same time, no mathematical relation could be established between the area

and perimeter of a complex polygon.

Merits: The main advantage of this method is that it is possible to get estimates

of a field area quickly with limited manpower. Another advantage is that it is

relatively inexpensive.

Page 35: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

27

Drawbacks: This method is subjective and therefore may not give accurate

estimates.

Measuring distances

The above-mentioned methods require measurement of distances whether it is

length or breadth of a rectangle or a base and the height of a triangle or the

perimeter. Consequently, the methods for measuring distance form an important

component of all these methods. In developing countries, most of the farmers

are unaware of the magnitude of the areas under the crops grown in their fields.

Accordingly, field staffs are employed to collect area statistics.

i) Pacing

Pacing means walking at a normal gait and counting the number of steps to

cover a certain distance. The steps are then converted to standard units. To

begin with, an average length of steps (usually 0.83m) is used by the

enumerators. Since there may be individual differences in steps, the step of each

enumerator needs to be calibrated by pacing a well-known distance and use that

as the conversion factor.

Merits and drawbacks: It has been observed that the length of the pace of

even the same enumerator changes with the change in the type of surface, such

as sandy soil, clay and uneven ground. The length of the pace also varies with

the enumerator's state of health and level of fatigue. Therefore, it has become

necessary to calibrate the step several times a day, which has taken away most

of the advantages of the pacing method.

There is also a risk of miscounting the number of paces, especially when this

number is large. In order to eliminate this risk, a simple instrument, the

pedometer, was proposed to be used. A pedometer, which consists of a digital

reader and a dial, measures the movements of the body; each step taken is

registered and shown on the dial. Each pedometer needs to be tested before uses

in case some of them may be out of order.

For the above reasons, the pacing method for measuring the length of sides or

diagonals of a field is not recommended and, in fact, has been discontinued in

almost all countries. However, it can be used, even without calibration, when

random points are to be selected within a field for the purpose of laying crop-

cutting plots.

ii) Measuring Distances with Instruments

A commonly used method for measuring distances involves standardized cord

and a wooden pole. A cord of fifty meters has been used for the allotment of

Page 36: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

28

parcels 50m × 50m of communal land to village members in many African

countries. To estimate crop areas, the cord needs to be of a non-extensible

material and care should be taken to avoid getting it wet, otherwise, its length

would be altered.

For example a wooden pole, the kassaba, of 3.55m, has been used to measure

the sides of fields in Egypt. A common source of error encountered in large

fields is the miscounting of the number of kassabas.

a) Surveyor's chain

The classical method for measuring distances is with a surveyor's chain.

The metallic chain is composed of straight links with circular ends connected

by rings with a handle. Each link is 0.20m long measured from the center of

one connecting ring to the center of the next. The usual chain length is 20m

(100 links) but there are 10mand 50m chains. Similar chains graduated in yards

and feet are also available (FAO 1982).

Two men are required to measure a distance, say AB, with a chain: one man

holds one end of the chain at the point A while the other stretches the chain on

the ground along the direction AB and marks the point A1, corresponding to the

end of the chain. Then, the first man moves to point A1 and the operation is

repeated as many times as necessary. The distance AB is calculated as so many

complete chains plus a number of links.

Merits: The advantage of the chain is that it is a cheap and secure instrument.

Drawbacks: It is heavy and not easy to handle. If not handled carefully, the

links often tend to bend, which reduces its length and results in overestimating

the distance over a long distance. There is also the risk of forgetting to count a

chain length.

When the ground is uneven and the chain is not placed on the ground but held a

few centimeters above, a slight error may be introduced due to catenary effect.

b) Tapes

A low cost instrument for measuring distance is metallic or plasticized tape,

which is a substitute for surveyor's chain. Tapes are wound on a special reel and

are graduated in meters, decimeters and centimeters or in yards, feet and inches.

They are available in different lengths, say 20m, 30m, and 50 m or 50 feet and

100 feet. The distances are determined in the same way as with chains.

Page 37: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

29

Merits: The tape is not liable to bend and the catenary effect is almost

inexistent. It is easier to handle and generally more accurate.

Drawbacks: The tape is likely to break easily. It may get rusted if not cleaned

after use and the plasticized tapes may lose their markings of the graduations.

Although the above measuring instruments (chains and tapes) are not very

costly, the operating expenses are high as two persons are needed to do the

measurements. The remuneration of two persons, even if one of them need not

be a professional enumerator but simply a laborer, in the long run, is more

expensive than the use of more costly instruments that can be managed by a

single person. Such instruments are the topofil, the Trumeter or Smith Wheel

and the optical range finder.

c) Topofil

The topofil is a distance measuring device fitted with a non-recoverable, light

but strong string and a counter that registers distances in decimeters, meters and

hectometers. The string runs out of the instrument as the enumerator walks the

distance to be measured. It can be easily carried and used by the enumerator.

The process is the following: the enumerator fastens the end of the string to a

fixed point and sets the counter at zero; as the enumerator walks the distance,

the string unrolls and the counter registers the length of the string unrolled; at

the terminal point, the enumerator reads on the counter the length of string

unrolled, which is then cut and discarded.

Merits: Any distance not longer than the length of the string on the reels can be

measured in one single operation (maximum length 5480m) using topofil. The

speed of measurement matches the gait of the enumerator; the enumerator can

read the counter at any intermediate point and set back the counter to zero, and

as the distance is recorded mechanically, there is no danger of miscalculating

long distances.

Drawbacks: The apparatus is costly, with a high running cost as a reel can be

used to measure at the most 20 fields of small dimensions (about 5 ha.).The

topofil case is a bit heavy and therefore inconvenient for carrying over long

distances, and the string sags slightly and may even rest on the ground or on the

plants.

d) Graduated wheel/ Trumeter / Smith wheel

The main elements of a Trumeter or a Smith wheel are a graduated wheel, a

handle to push it and a counter, which registers the number of revolutions of the

wheel. The circumference of the wheel is either one meter or one yard. At the

starting point, the enumerator sets the counter at zero and pushes the wheel

Page 38: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

30

along the line until the length of the distance to be measured. The reading on

the counter plus the length corresponding to an incomplete revolution gives the

length of the distance under consideration.

Merits: The cost of a graduated wheel is not very high and there are no running

expenses. The instrument is easy to use and, therefore, does not require skilled

enumerators. Accuracy of the instrument is high on smooth dry land, and the

mechanical recording of the number of revolutions eliminates the risk of

mistakes in counting.

Drawbacks: The instrument is not ideal for use in number of areas, including

those with rough ground, ploughed land and irrigated and humid land. It cannot

be used for direct measurement of horizontal distances when the land is sloping,

and when the land surface is undulating, the wheel measures the wavy curve

and not the straight line.

3.3. FARMER ASSESSMENT OF CROP AREA

In this method, the farmers are enquired to estimate the area of their fields. The

enumerator and the farmer may visit all fields of the farmer and estimate the

surface area by visual inspection (David 1978). In Uganda, the agricultural

module of the population census of 2002 used farmer estimates to obtain area

estimates (Menyha 2008). David (1978) concluded from two studies conducted

in Philippines that farmers overestimated their area by just 6 to 8 percent, while

in a third study, it was found that farmers slightly underestimated their area.

Ajayi &Waibel (2000) observed that in the Côte d'Ivoire, coast farmers were

able to confidently estimate crop areas when their plot size was larger than the

local area unit(± 0.25 ha), but when plot size was smaller than the local area

unit, farmers greatly overestimated the plot size (125 percent error). Several

authors (David 1978; Ajayi &Waibel 2000; de Groote &Traoré 2005) observed

that the accuracy of farmer estimates reduces with increasing plot size, resulting

in underestimations.

De Groote and Traoré (2005) mentioned several problems with the use of

farmer area assessments. Farmers may not trust the enumerators out of concern

that they may be subjected to more taxes. The problem is compounded if a

farmer is illiterate and thus may not be able to provide accurate information.

David (1978) found that farmers were likely to round off figures to the nearest

quarter or third of an acre. Due to the multitude of possible problems, FAO

(1982) considers the accuracy of farmer surface area estimations to be

insufficient. To increase the accuracy of farmer area estimates, one or more

subsamples of the sample can be defined from which enumerators obtain both

farmer estimates and direct area measurements. A correction factor may be

Page 39: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

31

defined on the basis of their correlation (David 1978). De Groote &Traoré

(2005) suggested that the enumerator should discuss the area estimates with the

farmer in the field, which would improve the accuracy of the estimates. This

method is adopted for acreage estimation under crops in such countries as

Egypt and Sudan.

Merits: This method is relatively less time-consuming and inexpensive.

Demerits: This method is highly subjective as it depends on farmers’

knowledge and, experience. The farmers are likely to misreport crop area over

fear of being burdened with increased taxes.

3.4. GLOBAL POSITIONING SYSTEM

GPS is a space-based satellite navigation system that provides location and time

information anywhere on earth. GPS hardware determines coordinates for x, y

and z axis, with x and y being the geographic coordinates that determine the

location and z being the coordinate that determines the elevation. Initially, GPS

was used to determine the location of a particular point, but with the

advancement in the technology, it is now capable of determining the elevation

and even the area covered. As a result, GPS has become a very important tool

for measuring the area under a crop with an added advantage of reduced time

and labor. There are, however, some reservations regarding the accuracy of the

instrument. The working of GPS, steps for using GPS for crop area

measurement, conditions in which GPS could be used for crop area

measurement, issues related to the use of GPS for area measurement and some

studies done in past for crop area estimation using GPS are discussed in this

section.

3.4.1. STEPS FOR USING THE GLOBAL POSITIONING SYSTEM

FOR CROP AREA MEASUREMENT

The following steps must be carefully observed when acquiring data using GPS

(Keita et al.2009):

The GPS device should be thoroughly tested before use;

The acquisition strategy should be designed in advance;

The material required to check data in the field, such as maps, should be

prepared or procured accurately and in advance;

The available data or maps should be uploaded into the receiver;

Identification of the fields to be measured;

Mark the borders of the identified fields accurately;

Page 40: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

32

Proper care should be taken regarding data storage and reporting. The

data storage can be done either on the paper (each measurement is

annotated on the paper) or on the GPS directly;

The shape of the plot on the screen of the devise during the test and in

the operational data acquisition phase should be observed to avoid

errors;

A post-acquisition integrity check is advisable for acquisition of

redundant control points;

Post-acquisition control, which includes checking the authenticity of the

data, computing the perimeter/area ratio, visualizing the plot borders,

should be conducted and inserted into the virtual globe.

The enumerator holds the GPS device in his/her hand and walks the whole

perimeter of the field from a specific starting point at one corner of the field.

The points are tracked chronologically in the memory of the device to make

lines. Lines are considered to define an area if the first point is connected to the

last point. If the loop is not closed by the enumerator, then the program will

compute the closed area obtained by connecting the last point in the log to the

first point. This geometry is used by GPS to calculate the area of the polygon. If

the user walks in an overlapping path, an impossible surface will be defined,

resulting in a grossly erroneous calculation. The calculation uses signed

numbers so if the track is crossed over at some point creating an “8” like

picture, then the program will compute the difference in area between the two

circles. The data are stored in a track-log on the device, which can be used to

calculate the area of the field. Most GPS models allow for direct area

calculation.

During the time needed for measuring the area of a plot logging along the

perimeter, the GPS constellation can be considered as being relatively stable,

thus the positioning errors do affect the measurement. Area measurement error

is linked to the operator speed and to the acquisition rate of the GPS device.

Field area measurement errors can be limited if an appropriate combination of

operator speed and GPS acquisition rate is selected, for parcels up to 4 ha, the

“optimum” range of speeds for operators on foot is between 0.5 m/sand 2 m/s

(1.8-7.2 km/h) (Bogaert et al. 2005).

Page 41: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

33

3.4.2. ISSUES RELATED TO THE USE OF THE GLOBAL

POSITIONING SYSTEM FOR AREA MEASUREMENT

Area measurements with GPS receivers are more rapid, efficient in terms of

time, feasible, digital and thus traceable and easy to incorporate into a database.

However, issues with the accuracy and precision of the results remain. Some of

them are mentioned below:

Accuracy of measurements with GPS: If the area is measured with a compass

and a meter is considered as the true area then the most GPS measurements are

very near to the true area, although, in general, GPS receivers tend to

underestimate the area of plots. According to some experiences, about 80

percent of GPS measurements have an absolute value of the relative error less

than or equal to 0.025.

(a) Factors affecting the accuracy of GPS: The accuracy of GPS

measurements is influenced by (i) the tree canopy cover (accuracy is

high with no tree canopy cover and lower with partial or dense tree

canopy cover), (ii) the weather conditions (accuracy is higher under

sunny conditions than under cloudy conditions), (iii) the plot size (the

larger the size of the plots, more accurate are the results).

(b) Battery storage problem: Securing ample power supply is one of the

major problems faced while using a GPS device for measurement. Some

GPS devices may be plugged into the car’s 12-volt power port. Some

GPS units use large capacity battery while others use an external box of

dry-cell batteries. Small GPS units typically save weight and space by

using two AAA cells, but they may last for only four or five hours on

one set of batteries. Larger handhelds use up to four AA cells (which

mean more weight), but run for as long as 12 hours per set.

(c) Speed: The time required for taking measurements with GPS is much

less as compared to other methods that involve making measurements,

such as a compass and tape method.

(d) Improvement in accuracy by repeating the measurements: The

accuracy of the measurements with GPS units does not improve from

the first to the third measurement. The GPS measurement is, in general,

3.3 times more rapid than other measurements and if two measurements

have to be performed, the GPS becomes 1.65 times more rapid.

Therefore, when using GPS, careful cost-benefit analysis has to be done,

Page 42: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

34

considering also the cost of the GPS receiver, before deciding to replace

the compass and tape method with GPS.

3.4.3. RECOMMENDED CONDITIONS FOR USE OF GPS FOR CROP

AREA MEASUREMENT BASED ON STUDIES

The empirical field experience conducted by FAO and several other institutions

in various regions and under different conditions provides a scientific basis for

recommending the use of GPS for crop area measurement under specific

conditions. These studies also identify the conditions in which the use of GPS

may not lead to accurate estimation of crop areas. The main conclusions of

these studies were validated during an international expert meeting, which was

held in Addis Ababa in November 2008. These studies have collected

information on 207 plots. FAO and some other institutions conducted field

studies in order to investigate the use of GPS for crop area estimation under

different conditions and compared the results with traditional estimates of the

area (Carfagna & Keita, 2009). On the same plots, the area was measured with

the traditional method first and then with the various kinds of GPS receivers,

such as Garmin 12 xl (G12), Garmin 72 (G72), Garmin 60 (G60) Garmin Etrex

Ventura (GE) and Magellan Explorist 400 (M400), often repeating the

measurement three times. The study concluded that most GPS measurements

were found to be very near to the true area when the area measured with a

compass and a meter was considered as the true area. About 80 percent of the

measurements were found to have a relative error of less than or equal to 2.5

percent. A linear regression of the measures using a compass and a tape and the

measures made by GPS shows that the linear model explains a very high

percentage of the variability of the compass and tape measures (R2 = 0.9633),

with parameters significantly different from zero and the slope very near 1

(0.9600301).

The accuracy of GPS for crop area measurements was found to be higher for

larger plots, which suggested that for large and very large plots (from 10,200

square meters), the GPS measures are very similar to the compass and meter

ones; for medium-size plots (form 5,300 to 9,999 square meters), they are less

similar. Finally, for small and very small plots (less than 0.5 hectares) the two

distributions are quite different. This lower limit of a half of a hectare as the

measuring area using current GPS devices with acceptable accuracy was also

confirmed by the experiments.

The time needed for measurement with a compass and tape and GPS receivers

shows a high variability. For some small and medium plots, the time needed

when using a compass and tape is 17 times longer than when using GPS. The

Page 43: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

35

distributions of the time needed for the first, the second and the third

measurements are not significantly different. The accuracy of the measurements

does not improve from the first to the third measurement. As the repetition

involves cost, these data do not suggest repeating the measurement. The studies

also revealed that some types of GPS perform better than others in different

conditions. Therefore, the selection of GPS should be done with great care.

Also, regarding operational use, battery life, robustness under difficult field

conditions and simplicity of use should be taken into account when selecting a

GPS type.

The Uganda Bureau of Statistics has been testing the accuracy of GPS estimates

for crop area (Schøning et al. 2005). During the 2003 Pilot Census of

Agriculture in Uganda, area estimates obtained using GPS were compared with

the area estimates obtained using the polygon method. Lesser time consumption

was found to be the main advantage of using GPS. GPS resulted in an overall

time saving of more than 300 percent. On average, GPS area estimates were 6-

12 percent lower than area estimates obtained from the polygon method.

Analyzing the results by plot size showed that GPS estimates were strongly

correlated (R2 = 0.90) with the polygon estimates for larger plots (greater than

0.5 hectare. For smaller plots (less than 0.5 hectare), the correlation was found

to be very poor (R2 = 0.12). However, recently under the WB/LSMS project, a

researcher in Burundi observed that even for plots as small as 0.25 acres (1000

m2), the accuracy of the GPS method is still reasonable as compared to the rope

and compass method.

The accuracy of any GPS receiver is around ± 15 m. Therefore, for small plots,

there may be large errors in the estimation of area. A GPS receiver might record

a plot measuring 30m × 33.3m (0.1 hectare) as a plot of 60m × 63.3m,

overestimating the area by 385 percent. The accuracy of a GPS receiver may be

enhanced by installing a second GPS receiver in a location with known

coordinates – a Differential Global Positioning System (D-GPS) (Fermont &

Benson 2011). The precision of GPS can also be improved by repeating the

measurements using D-GPS, or using the Satellite Based Augmentation System

(SBAS) differential corrections, such as the Wide Area Augmentation System

(WAAS), the European Geostationary Overlay Service (EGNOS) and the

Japanese Multifunctional Transport Satellite Augmentation System (MSAS). A

GPS receiver capable of SBAS differential correction can give position

accuracy on an average of three meters. The precision can also be improved by

Assisted Global Positioning System (A-GPS) which provides a reliable position

under poor signal conditions, such as under trees or indoors.

Page 44: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

36

There may be erroneous readings in GPS measurements resulting from

interference of trees and projection problems in hilly areas (Schøning et al.

2005; Sempungu 2010). It should be noted, however, that the latter is a problem

for any method measuring crop area on steep slopes. This is related to the fact

that the measured crop area should not be the physical area measured on the

ground, but instead it is projection onto a horizontal plane (Muwanga-Zake

1985). The projection problem may become acute on slopes greater than 10

degrees. As the introduced error is 0.4 and 1.5 percent for slopes of 5 and 10

degrees, respectively, the projection problem may be ignored on slopes of less

than 10 degrees (Fermont &Benson 2011). Therefore, use of clinometers with a

compass and rope in hilly regions with steep slopes is recommended.

The Living Standard Measurement Study by World Bank, which was carried

out in 2013 and 2014 in Ethiopia, Nigeria and the Untied Republic of Tanzania

compared the crop area measured through compass and rope, GPS (Garmin E

Trex 30 was used in the United Republic of Tanzania and Ethiopia, while, in

Nigeria, the Garmin GPS Maps 62 and self-reporting by farmers (the compass

and rope method of measurement was assumed to be the standard) was used.

This study revealed that, on average, GPS measures were very accurate

estimates of plot size, including for very small plots and for reasonably small

samples. Thus, the GPS method of area measurement is recommended over

compass and rope and self-reporting by the farmer. In large-scale surveys, GPS

measures were often plagued by missing observations. The study, therefore,

recommended that GPS measurement (where feasible) should be complemented

by a farmer self-reported estimated area (for all plots) as the ladder involves

negligible fieldwork costs, and, more importantly, it can serve as a baseline for

imputation where objective measurements may be missing. Another important

finding of the study was that the farmers were able to report land area correctly

in non-standard/local units. In addition, the compass and rope method may be

repeated whenever the closing error is more than 3 percent.

Page 45: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

37

4 Methods of crop yield

estimation Crop area, crop production and crop yield are recognized as three key variables

by World Bank (2010) in the Global Strategy to Improve Agricultural and Rural

Statistics that should be part of the minimum core data set that all countries

should be able to provide. Crop productivity or crop yield is one of the essential

indicators for agricultural development. Crop yield is normally expressed in

kilogram (kg) or metric ton (MT) per hectare (ha). The estimation of crop yield

involves both the estimation of the crop area and estimation of the quantity of

produce obtained from that area. In many circumstances, it may not be easy to

estimate crop yield as both are prone to error and bias and the measuring

process may be time-consuming.

There are several methods available for crop yield estimation with known

merits and drawbacks. The use of these methods varies across countries.

Among these, the whole plot harvest method may be treated as the gold

standard for crop yield estimation. The two paramount methods for estimating

crop yield are crop cut and farmers' estimate. In addition to them, there are

some other methods available for crop yield estimation. These methods along

with their critical analysis are described discussed below:

4.1. WHOLE PLOT HARVEST

The whole plot harvest method is employed during detailed farm surveys and in

demonstration plots (Norman et al.1995). Harvesting of the whole plot is done

in cases of on-farm trials. Before harvesting, plot boundaries need to be clearly

marked and then the harvest area should be calculated. The crops having a

defined maturity date, such as cereals or legumes, with a determinate growth

that can be harvested in a single operation. Legumes with an indeterminate

growth habit, such as common bean, cowpea and mung bean, or crops with

harvests spread over different seasons or year, such as cotton, banana or

cassava, require multiple harvests per plot. Each harvest is dried and weighed

Page 46: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

38

separately. Individual harvest weights are summed up in order to obtain total

production and to calculate crop yield.

Whole plot harvest measures the harvested yield. This method is regarded as

the absolute standard for crop yield estimation, especially if done together with

the farmer (Casley & Kumar 1988).

Merits: The main advantage is that it is almost bias-free, as all sources of

upward bias reported for crop cuts can be eliminated when the whole field is

harvested. This method is suitable for small-scale investigations of a case-study

nature (Poate 1988). The complete harvesting generates more accurate data than

crop cuts because the bias from within-field variability, which commonly is 40–

60 percent of total yield variability, is eliminated. Wherever the whole field

average size is less than 0.5 hectare, complete harvesting takes a similar amount

of time as the crop cuts in two or three subplots per field (Casley &Kumar

1988).

Drawbacks: Murphy et al. (1991) pointed out that the area measurement for

the whole plot introduces a limited source of downward bias. This is especially

the case with irregular shaped plots when enumerators have to approximate

curved lines with straight lines for calculating the surface area. It has been

observed that enumerators tend to minimize the exclusion of planted areas and

forget to include non-planted areas. This may introduce an upward bias of up to

5 percent in the area estimation, which results into a limited underestimation of

the harvested yield. In addition, whole plot harvesting requires the enumerator

to be present at the time of harvest, which does not always happen. The main

drawback of the method is that it involves a large volume of work, making it

unsuitable for moderate-to-large sample sizes or multiple crop studies.

4.2. CROP CUT AND FARMERS’ESTIMATE METHODS

The crop cuts and farmers eye estimates are the two most commonly used

methods to estimate crop production. This section provides methodological

details and their merits and drawbacks and then presents the results of

comparative studies.

4.2.1. CROP CUT METHOD

The crop cut method was developed in the late 1940s in India for estimating

crop yield based on sampling of small subplots within cultivated fields by

pioneers in sampling and survey design, especially based on the work of P.C.

Mahalanobis of the Indian Statistical Institute and P.V. Sukhatme of the Indian

Council of Agricultural Research (ICAR). Within a decade, the crop-cut

method were adopted as the standard method recommended by FAO to estimate

Page 47: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

39

crop production (FAO 1982; Murphy et al. 1991). Since then, this method has

been commonly regarded as the reliable and objective method for estimating

crop yield. It involves randomly locating, prior to the harvest, a small subplot

within each field. The shape of the subplot is usually considered as a

square/rectangle/triangle/circle. The subplot is harvested by the survey

enumerator or farmer himself. Then the crop is weighed after proper drying and

processing (figure 7). Crop yield is calculated as total production divided by

total harvested area of the crop cut plot or subplots. Detailed description of

steps involved in conducting crop cutting experiments is given in annex II.

Figure 7. Steps involved in Crop Cut Method

Page 48: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

40

Spencer (1989) described the yield plot method in cases in which the

enumerator would take out a randomly chosen quadrant in a farm field before

harvest. When the farmer harvests his or her plot, he or she leaves the quadrant

unharvested, which is cut and measured by the enumerator soon afterward.

However, one randomly sampled subplot may not represent the variability in

crop performance within a field. For that reason, the concept of sampling of

multiple small subplots came under consideration. Fielding & Riley (1997)

suggested using two large quadrants (in the order of 50-75 m2 each). In a study

based on Botswana, Norman et al. (1995) suggested using a systematic

sampling scheme with multiple small quadrants. It was also suggested to use a

measuring stick instead of a quadrant to facilitate the logistics and for ease of

operation. In a detailed study in five African countries, Verma et al. (1988)

found that quadrant sampling gave more accurate and reliable results than the

row sampling, as subplot boundaries were clearer in the former method. Rozelle

(1991) suggested another multiple quadrant crop cut method known as the

“five-point” method, in which the enumerator harvests five “one square meter”

quadrants located in the corners and the center of a plot.

National statistical institutes in for example, Benin, India, Niger and Zimbabwe,

prefer to use the crop-cut method (MoS & PI 2008; Murphy et al. 1991). The

United States Department of Agriculture (USDA) uses crop cuts for yield

estimation of specific major crops in specific states. Uganda used crop cuts in

the annual crop yield surveys of 1967 and 1968 and in the agricultural census of

1990-1991.

The equipment/materials required for conducting the crop cutting experiments

are:

a) Measuring tape as per requirement (30m or 50m)

b) Weighing balance as per requirement (beam or spring balance)

c) Set of weights (1g, 2g, 5g, 10g, 20g, 50g, 100g, 200g, 500g, 1kg, 2kg

and 5kg)

d) String or rope (30m)

e) Four pegs

f) Hessian cloth: a coarsely woven fabric usually made from vegetable

fibers and jute. Known for its plain weaving and durable quality, it is

ecofriendly and used for the packaging of a variety of goods including

grains, sugar and pulses.

g) Cloth bags for keeping the produce for drying

h) Two strong waterproof bags (one for keeping crop cutting equipment

and the other for keeping, such things as schedules and papers)

Page 49: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

41

i) Blank schedules, instruction manual, random number tables and other

related documents.

Page 50: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

42

Figure 8. Equipment required for conducting the crop cutting experiments

Measuring tape Compass Rope

Beam balance with weights Beam balance pan with weights Pegs

Spring balance Set of Weights Hessian cloth

Page 51: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

43

Merits: Since being endorsed by FAO in the 1950s, the crop cut method has

been commonly regarded as the most reliable and objective method for

estimating crop yield. A sufficient number of cuts in a sufficient number of

fields provides a valid estimate of average yield (Murphy et al. 1991). An

advantage of the crop-cut method is that the area of the cut is known and thus

an error cannot be introduced into the final yield computation

(Poate 1988).

Drawbacks: The crop cut method measures the biological yield, which does

not take into account post-harvest losses and therefore does not reflect the

economic yield that is of use to the farmer. Obtaining yield estimation through

crop cuts is both time-consuming and labor-intensive. To facilitate field work

and reduce costs and the time involved, a clustered sampling procedure is

usually applied when crop cuts are used for larger scale surveys. As crop

cutting experiments involve intensive field work requiring many field

investigators, the objective method for data collection is unaffordable, which

has prompted many countries and international agencies to consider the

farmer’s interview method to obtain relatively cheaper and quicker estimates of

average yield.

4.2.2. FARMERS’ESTIMATE METHOD

The most common alternative for the crop cut method is to use farmers'

estimates, which gives a measure of total produce and the economic yield.

Estimating crop production through farmer interviews involves asking farmers

what quantity they did harvest or what quantity they expect to harvest in order

come up with an estimate for an individual plot, field or farm. The first one is

commonly known as farmer recall, whereas the second one is referred to as

farmer prediction. As harvest quantities are farmer estimates, they are generally

expressed in local harvest units instead of kilograms or tons. To convert harvest

quantities to standard units, conversion factors are required.

4.2.2.1. FARMER RECALL

This method is the post-harvest estimation commonly made at the farmer’s

house or at the site where the harvest is stored for the enumerator to cross-check

the estimates with the available storage capacity (Casley &Kumar 1988).

Depending on rainfall distribution, recall periods may range from six months or

one season to three years or three-to-six seasons (Howard et al. 1995; Lekasi et

al. 2001; Erenstein et al. 2007). Smale et al. (2010) gave examples of longer-

term subjective recall, namely two-year average production data, instead of

asking farmers to estimate harvested quantities for individual growing seasons.

Developed countries, such as Sweden, are obtaining farmer recall data through

Page 52: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

44

web-based surveys or telephone interviews (personal communication with G.

Ländell, Sweden Statistics, 2010).

To estimate the crop yield, the production data obtained from farmer recall

requires dividing the plot area from which the crop was harvested. This,

however, introduces an additional source of error. Fermont et al. (2009)

obtained a direct estimate of average crop yield to remove this error by asking

farmers to estimate the number of local harvest units they would have obtained

from a well-known unit of land, often the farm compound, if it had been planted

with a specific crop.

4.2.2.2. FARMER PREDICTION

This method is the pre-harvest estimation commonly obtained on a plot-by-plot

basis, in which the enumerator and the farmer are in visual contact with the

growing crop. The enumerator needs to be able to judge the validity of the

farmer’s response. The basis of the farmer’s predictions of expected yield is

their previous experiences, by comparing the current crop performance to

previous crop performances (David 1978). Singh (2003) suggested that yield

estimation should be made at the time of maximum crop growth. In the United

States, to obtain production forecasts, monthly telephonic interviews are

conducted with farmers (USDA 2009).

Merits: The use of farmer estimation does not require laborious measurements

and allows for a more efficient, random sampling design (Murphy et al. 1991;

Casley & Kumar 1988). In comparison to the crop cut method, the use of

farmer estimation is less costly and quicker to carry out. Consequently, farmers'

estimation with the same resources allow for a larger number of yield estimates

to be collected than crop cuts.

Drawbacks: Even though farmer recall and predictions were found to be

effective and inexpensive in several studies, they too had their own

shortcomings. These included (i) ignorance of in-kind payments, (ii) non-

standard harvest units, (iii) intentional over/underreporting, (iv) low accuracy

with longer recall periods, (v) historical average production factors, (vi) poor

quality responses in lengthy interviews, (vii) insufficient supervision, (viii)

illiteracy, especially in African countries (David 1978; Casley &Kumar 1988;

Poate 1988; Rozelle 1991;Howard et al. 1995; Kelly et al. 1995; Diskin 1997;

UBOS 2002; Ali et al. 2009; Fermont &Benson 2011). Several studies indicate

that the use of farmers’ estimates is affected by the bias in estimation.

Page 53: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

45

Comparison of crop cut and farmers’ estimate with the gold standard

method

As the crop-cut method measures biological yield, the whole-plot-harvest

method measures harvested yield and the farmer recall method measures

economic yield, each method takes into account different amounts of harvest

losses. As a result, theoretically, the three yield estimates obtained for the same

plot can never have the same value. The estimated yield levels for all estimates

are completely free of sampling and non-sampling errors should be in the

following order: crop cuts > whole plot > farmer recall.

For a small and controlled survey in Nigeria to compare crop-cut estimates

obtained from 60 m2 subplots with farmer predictions and whole plot harvests

of millet and sorghum, Casley & Kumar (1988) reported an average bias of 14

percent for both crop cuts and farmer predictions. They also quoted a study on

rice in Bangladesh that showed that crop cuts had an average 20 percent upward

bias, compared with whole plot harvests, while a small study in Bangladesh

showed 15 percent bias in farmer recall data (Poate 1988). A small study in the

United States using precise procedures on soybeans showed that the bias in

yield estimated using crop cuts might be as low as 5 percent. In that study, the

crop cuts even underestimated yield in comparison to whole plot harvest in

some cases (Rogers and Murfield 1965).

For years, it was assumed that farmer estimates were too subjective and

unreliable to obtain reliable data on crop yield (Verma et al. 1988), whereas

crop cuts were assumed to be unbiased (Murphy et al. 1991). Thus, when

farmer estimates differed from crop-cutting measurements, it was automatically

assumed that the differences reflected “farmer error”. The idea that crop cut

measurements were not seriously affected by bias, such as consistent over- or

under- estimation, was based on early evidence from crop cut work in India.

However, in the late 1980s, evidence started to emerge that biases associated

with crop cuts were often substantial if not carried out properly or the

enumerators were not experienced. Especially in the case of small, irregularly

shaped fields with uneven plant density, biases were found to be large, which

was the situation of many smallholder farmers in Africa (Poate 1988; Murphy

et al. 1991).

In this context, the most significant study done so far is by Verma et al. (1988)

on maize crop in five African countries, namely, Benin, Central African

Republic, Kenya, Niger and Zimbabwe, in which crop cut and farmers’

interview methods were compared with the complete harvest from the selected

fields. The study revealed that (i) the farmers generally overestimated the

planted area, (ii) farmers' post-harvest estimates were close to actual production

Page 54: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

46

and superior to the objective method of crop cutting, (iii) farmers' pre-harvest

estimates were also good for predicting production levels but had high variance

and (iv) the crop-cut method overestimated average yield by about 30 percent.

The study, however, had several weaknesses, namely (i) the enumerators who

interviewed the farmers also carried out crop-cut work and measured the total

harvest of the sample fields, (ii) the sample size in each country was small,

resulting in high within-country variability and (iii) the selected farmers were

fully aware and willing participants at each stage of the study and thus

independence of the method was lost.

In contrast to the above study, Rozelle (1991) reported that farmers in Malawi

had great difficulties estimating crop production after harvest. Diskin (1997)

pointed out that Verma et al. (1988) evaluated production estimates, not crop

yield estimates. Thus, their study only provided evidence of the merits of

farmers' estimation over crop cuts while estimating production not yield. To

convert production estimates to yield estimates, the production estimates are

divided by the area estimates. Diskin (1997) argued that the results of Verma et

al. (1988) only supported the use of farmers' interviews over crop cuts to

estimate crop yield in cases in which farmers’ estimates of area had a minimum

source of error.

Casley &Kumar (1988) compared the crop cuts with farmer estimates of yield

on smallholder maize fields for two consecutive years across six regions using

data from the Central Statistical Office in Zimbabwe. Crop-cut estimates were

found to be on average 86 percent (with a range of 32 to 100 percent among

regions) higher than the farmers' estimates. Supervision of the crop cuts was

much tighter in the second year. As a result of this, crop-cut estimates were, on

average, 37 percent (with a range of 27 to 78 percent between regions) higher

than farmers' estimates. Though, this shows that crop cuts likely overestimate

crop yield, it does not rule out a substantial margin of bias in farmers' estimates.

Thus, crop-cut estimates, if carried out by an experienced enumerator along

with proper training and supervision, may provide reasonably accurate

estimates of crop yield.

Minot (2008) conducted a study in Ethiopia, which was in line with the

Zimbabwe study. He reported that as per agricultural sample survey of the

Ethiopian Central Statistical Agency in 2008, an average cereal yield estimates

using crop cuts were 31 to 46 percent higher than the farmer yield estimates for

the same season as observed in a large household survey carried out by the

International Food Policy Research Institute.

Page 55: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

47

A five-year study of Statistics Sweden showed first, that farmer recall did not

systematically underestimate cereal yield and, second, that farmer estimates did

not strongly deviate from the cereal yield observed with crop cuts (‒4.9 to +9.5

percent) at the country level (Hagblad 1998). The results of this study were in

contrast to those of African studies. However, it is possible that Swedish

farmers are better able to recall production because of the higher levels than

African farmers because of mechanization, commercialization, and

recordkeeping within Swedish farming systems.

The farmer predictions (as opposed to farmer recall) were compared with crop

cuts during other studies. In two studies in Asia, crop cuts were found to be

strongly correlated (R2 = 0.86) to farmer predictions of rice yield, but 25 and 37

percent higher (David 1978). This was in line with results from India, where

farmer predictions of wheat yield were also strongly correlated (R2 = 0.87) to

crop-cut yield data (Singh 2003).

In India, the Indian Agricultural Statistics Research Institute, New Delhi

conducted a study on wheat crop in Lucknow and Aligarh districts during 1988-

1989 and 1989-1990. The unpublished results of this study reveals close

agreement between farmers' post-harvest estimate of average yield, yield

obtained through crop cutting and the whole field harvest. Kathuria (1995)

conducted a study for FAO on maize crop in Zambia and concluded that the

farmers interview method needed to be further experimented as some farmers

tended to under/overestimate the average yield suited to their interests. Mathur

et al. (2003) compared the farmers’ eye estimation method with the objective

crop-cut method in a study on wheat production of Karnal district of Haryana

state in India. They showed that the farmers’ estimate was very close to actual

production values. Ahmad et al. (2004) did a comparative study of farmers eye

estimate and crop cut estimate for general crop estimation surveys and

concluded that there was no significant difference between farmers’ eye

estimate and the crop cut estimate provided that the farmers’ eye estimate was

taken one week prior to the harvest. Thus, to obtain a more efficient estimate,

one may combine crop cut estimates taken one week prior to the harvest with

farmers eye estimate as an auxiliary variable.

The crop-cut method is regarded as a reliable and objective method for

estimating crop yield. This method, however, may be accompanied with an

inherent upward bias because of measurement errors and the increased cost and

time involved. But these deficiencies can be largely overcome by appropriate

training and supervision. The use of optimum sample sizes and auxiliary data

available in the system may provide reliable estimates.

Page 56: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

48

The choice of these methods is spread over various countries. For example,

national statistical institutes in Kenya, Rwanda and Sweden prefer to use farmer

recall data to obtain production estimates, while those in Benin, India, Niger

and Zimbabwe favor the crop-cut method. The United States Department of

Agriculture uses a combination of farmer recall for their agricultural census and

crop cuts for yield estimation of specific major crops in a few States. Several

European countries prefer to use more expensive crop cuts for potatoes, while

cheaper methodologies such as farmer recall, or expert assessment for other

crops.

4.3. OTHER METHODS OF CROP YIELD ESTIMATION

In this section, other methods used to estimate crop yield are described. These

include daily recording, sampling of harvest units, expert assessment, crop

cards, crop modeling, purchase records, allometric models and remote sensing.

4.3.1. DAILY RECORDING

This is an intensive method for estimating crop production at the smallholder

farmer’s level. In this method, first the enumerators visit each plot of a farm

household and record its surface area. Over a certain time period, such as a

cropping season or a year, the enumerators visit the farmers regularly (ideally

daily) to record the weight and state of harvest of any crop that has been

harvested since the previous visit. In order to determine factors for converting

to a standard state of harvest for each crop, the enumerator may take

subsamples of the harvested crops from time to time. This method was used

during Ugandan agricultural census of 1965.

This method estimates the economic yield of a crop. As it is frequently

recorded, this method is able to capture multiple harvesting of the same plot, a

common practice for crops with an extended harvest period, such as cassava,

banana or coffee, and for crops in which the ripening period is spread over a

period of time, such as green maize and indeterminate legumes. Under this

method, unrecorded “losses” attributed to eating, selling or other activities

causing post-harvest losses are reduced. Furthermore, as detailed area

measurement of each plot is taken at the start of the exercise, crop yield can be

calculated without an additional source of error.

When this method was used for the Uganda agricultural census, in some regions

of the country, the published crop yield estimates per region were based on a

very limited number of plots (a maximum of 235 plots, very often less than 100

plots, and in a few cases only 2 plots per crop per region. In the census report, it

was argued that the sampling error might be expected to be large, especially in

Page 57: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

49

those estimates that were obtained on the basis of a very limited number of

plots (MAC 1967). In addition, non-sampling errors could also have been high

because of possible errors in weighing or failure to weigh the entire harvested

crop. The other reason for high non-sampling errors might be attributed to the

use of average national conversion factors to convert the recorded weights into

standard harvest conditions. Though the conversion factors that were used were

calculated from data collected during the survey, it was found that in order to

estimate conversion factors with a high degree of accuracy, more work was

required.

Merits: This method may generate very high quality data when enumerators do

their job properly and farmers do not harvest a specific crop from more than

one field per day.

Drawbacks: This is a very labor-intensive method and thus requires cluster

sampling (Muwanga-Zake 1985), which has a negative impact on overall

sampling error. The likelihood of measurement and recording errors increases

because of the daily weighing and recording operations. Other disadvantages

observed with this method was that enumerators often lack motivation to visit

each farmer every day and farmers sometimes mix harvests from various plots

in cases in which they harvest the same crop from several plots in one day

(Muwanga-Zake 1985).

4.3.2. SAMPLING OF HARVEST UNITS

In this method, instead of harvesting and weighing the whole field, the

enumerator may wait for the farmer to harvest his or her field. An attempt is

made to estimate the number of units, such as sacks, baskets and bundles that

has been harvested by the farmer. The enumerator then randomly selects a

number of harvest units and weighs them to obtain an average unit weight.

Sampling of harvest units is generally done just before storage, which includes

a measurement of the moisture content of the harvested product (Casley

&Kumar 1988).

The method of sampling of harvest units measures either harvested yield or

economic yield, depending on the time between harvesting and sampling (that

is, the amount of post-harvest losses). To estimate crop production for a specific

plot using the sampling of the harvest units method, Casley & Kumar (1988)

suggested that the harvest be collected in identifiable and complete units and

reviewed before being stored in a granary or otherwise disposed. The units

should not be too variable so that average unit weight could be estimated

without too much error. Crops should be harvested at once and the enumerator

should make the estimation shortly after harvesting. In addition, the harvested

Page 58: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

50

units should originate from one specific plot. This is especially a concern if a

household has multiple plots with the same crop or one field with the crop of

interest partially intercropped with a second crop. Poate & Casley (1985) found

this method more appropriate for estimating crop production at the farm level

than to estimate crop yield at the individual plot level as the above conditions

are usually not met. Rozelle (1991) pointed out that, in cases in which the

enumerator was unable to visit the household directly after harvesting, this

problem could be overcome by including questions to estimate the amounts of

the harvest that have already been used.

Merits: The technique is straightforward and can be used on larger samples as

compared to the crop-cut and whole-plot harvesting methods. Unlike farmer

estimates, it does not matter if the harvest units are particular to each individual

farmer, as the enumerator either weighs the complete harvest or weighs a

random, unbiased, selection of harvest units of each farmer (Poate 1988).

Drawbacks: When the harvest is stored in one or several large granaries or

stores, the enumerator needs to use analytical skills to accurately estimate total

production (Rozelle 1991). This method is considered unsuitable for crops with

an extended harvest period and multiple pickings, such as root crops, banana,

cotton and similar crops.

4.3.3. EXPERT ASSESSMENT

Experts that have extensive experience with crops, such as extension staff, field

technicians or subject matter experts, can estimate crop yield by either visually

assessing the field or by estimating yield on the basis of a combination of tools.

This technique gives an estimation of biological yield. Experts are often able to

estimate crop production or yield by visually assessing the condition (for

example, color, plant vigor and plant density) of the crop in the field. This is

known as eye assessment. In the 1990s, several European countries, including

Belgium, Ireland, Germany and the Netherlands, used eye assessment to

estimate crop yield for their annual agricultural statistics (Bradbury 1994). In

Australia and the United States, eye assessment has been upgraded through a

combination of visual assessment, field measurements and empirical formulas

to a so-called expert assessment method.

For cereals and grain legumes, the yield in t/ha is estimated by multiplying the

average number of grains per head by the average number of heads per 5m row

and dividing this by a constant K that depends on the row spacing and grain

weight are carried out in at least 10 representative sites within a field (DPI

2010). For cotton, extension staff counts the number of cotton bolls that are

open or expected to open by harvest in 10 representative one row-feet sections

Page 59: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

51

in the field. To determine average boll weight, the bolls on three plants are

picked and weighed in each section. The lint yield is estimated assuming a

certain picker efficiency and gin turnout and knowing the row spacing

(Goodman and Monks 2003). The expert assessment may become so detailed

that the difference between this method and that of crop cuts on the basis of row

segments may become blurred, though expert assessments will never involve

harvesting the whole row segment.

Few studies have compared expert assessments with other yield or production

estimation methods and their results are contradictory. A poor correlation of

rice yield that were eye estimated by experts and actual crop yield was observed

by David (1978);he concluded that eye estimation of yield should not be used.

However, In Zimbabwe, Casley and Kumar (1988) observed that expert

assessment was closely related (<10 percent difference) to farmer estimates.

Merits: Two advantages of the expert assessment method are that it can be

applied on a relatively large scale as compared to the crop-cut and farmer

estimation methods and it does not require area estimation and eliminates a

source of potential bias. Another important advantage is that one team of

experts can be used throughout a study, which results in a similar bias for all

yield estimation (Rozelle 1991).

Drawbacks: Eye estimations of crop yield require not only practical but also

technical familiarity with the yield potential of different varieties of a crop and

their relative performance in different environments (David 1978). The

accuracy of the yield assessment, therefore, strongly depends on the level of

expertise of the expert. When assessments are made by extension officers, yield

estimation may be upward biased, especially if the assessments are made in

their own work area and the information collected thus pertains to the quality of

their own work (Casley and Kumar 1988). In contrast, Bradbury (1996b)

reported that yield estimates by expert judgment in Europe were generally

considered to be biased downward. Considering that a national survey or an

agricultural census requires yield estimates of a large range of crops, it is

difficult to find experts that have the required practical and technical expertise

to provide accurate estimation across all crops.

4.3.4. CROP CARDS

The crop card method is a refined version of the farmer recall method. It also

entails estimating the economic yield. The method has evolved to obtain more

reliable yield estimates of crops with an extended period of harvest, such as

cassava, banana and sweet potato, because farmers may have problems in

remembering the amounts they harvested over time for one or several plots.

Page 60: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

52

Under this method, each farmer participating in the survey is given a set of crop

cards by a crop card monitor and receives training on how to use the cards. The

cards are used to record the quantity the farmer harvested in local harvesting

units after each harvest operation. The card crop monitor is expected to visit

each farmer on a regular basis to monitor the farmers' recordings and to make

corrections if the farmer has any problem. After a specified period, the crop

card monitor collects all the cards for processing.

This method was tested in Uganda during the Uganda National Household

Survey of 2005-2006 and was compared with farmer recall estimates.

Using the data collected for the Survey, Carletto et al. (2010) showed that crop

card production estimates were 40 to 60 percent lower than the farmer recall

production estimates for both crops with an extended harvest time (cassava and

banana) and for other crops (maize and beans).This was in line with the

findings of Sempungu (2010), who, using the same data set, found that cassava

and sweet potato yield estimates from the crop card method were 30 and 46

percent, respectively, lower than those obtained from farmer recall. The above

studies suggested, first, that farmers were either seriously overestimating crop

production during the recall exercise or underestimating crop production with

the crop card method and, second, the upward or downward bias that resulted

did not seem to depend on the type of crop. This is contradictory to the

assumption that farmers have difficulties in accurately recalling multiple

harvests of crops over an extended harvest period.

Merits: This method provides more reliable yield estimates of crops with an

extended period of harvest than the farmer recall method, as farmers may have

problems in remembering the amounts they harvested over time for one or

several plots.

Drawbacks: There are several problems with this method. Among them are

irregular monitoring by enumerators, illiterate farmers who were not able to fill

in the crop cards and, some recordings often include crop purchases and a very

large range of observed harvest units (Ssekiboobo 2007).

4.3.5. CROP MODELING

This method is widely used to estimate average biological yield of smallholder

farmers. Crop models vary widely in terms of complexity. The simplest sets of

models are of an empirical-statistical nature, whereas the most complex models

are based on crop physiology. The former is intended to find the best

correlation between crop yield and environmental factors, such as weather

parameters (temperature, humidity, rainfall) from long-term data sets. Using the

established relations, the model attempts to predict crop yield at a regional or

Page 61: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

53

national level on the basis of actual environmental observations, whereas crop

growth models estimate crop yield as a function of physiological processes and

environmental conditions. They range from relatively simple models taking into

account only basic crop physiology processes, such as the Penman-Monteith

models, which are based on estimation of actual evapo-transpiration, to

extremely complex models that estimate daily gains in biomass production by

taking into account all known interactions between the environment and

physiological processes (Sawasawa 2003). The crop modeling approach is used

in India for multiple season crop forecasting using weather parameters as well

as parameters, such as crop area and price in the previous years under the

Forecasting Agricultural Output using Space, Agro-meteorology and Land-

based Observations (FASAL) project (Parihar & Oza 2006).

Merits: Crop models can be used to predict crop yield in specific conditions or

a range of conditions and are an extremely useful tool in research studies

exploring the impact of specific factors on average crop yield.

Drawbacks: Crop models cannot be used to predict crop yield for individual

farmer fields, as this requires far too much input data.

4.3.6. OTHER METHODS

In addition to those mentioned above, other methods for crop yield estimation

are described in this section. They include:

Purchase records from agro-industries

Generally in the case of cash crops, such as coffee, cotton, tea, cocoa and

sugarcane, purchase records for individual farmers can be obtained from agro-

industries and linked to farmers involved in the agricultural survey. Purchase

records can be a valuable source of production estimates (economic yield) at

regional or national levels. If the produce is sold to the

agro-industries and the records can be linked to the individual farmers in the

survey, production estimates at the farm level can be obtained. These may be

converted to crop yield if the total crop area is known. This works well in

developed countries. For example, Sweden and Norway obtain records on sugar

beet production from the agro-industry and the national grain administrator,

respectively (Bradbury 1996a). In Uganda, this may work for cotton, though

linking records from the cotton ginneries to individual farmers in the

agricultural survey may not be straightforward. Data on total cotton production

from each ginnery and aggregated data on cotton area in the same region can be

used to obtain a proxy estimation of cotton yield in a region. This may be more

difficult for coffee, as this crop is sold in several batches to one or more buyers,

and some of it is consumed locally.

Page 62: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

54

Allometric Models

Allometric models are mathematical models based on plant morphological

attributes and crop yield. When these relationships are sufficiently accurate

(R2>0.75), then measurements of several morphological characteristics of a

selected number of plants can be used to predict biological yield in a field.

These models need to be based on variables that can be quantified easily using

rapid, inexpensive and nondestructive methods of data collection. Tittonell et

al. (2005) used plant height and ear length to predict maize yield in Western

Kenya. Wairegi et al. (2009) found that with a multivariate model using girth of

the pseudostem at the base and at 1 m, the number of hands and the number of

fingers gave a robust prediction of bunch weight for bananas in Uganda. Both

models were valid for a range of cultivars and soil fertility levels, whereas the

banana model was also valid for a range of agro-ecologies and not specific to

the development stage. This indicates that such models can be used in a wide

range of conditions. Labor demands for data collection for use in allometric

models are likely to be somewhat lower than for crop cuts, but enumerators

need additional training and an adapted datasheet for data collection. These

models are mainly applicable for fruit crops.

Remote sensing

Crop yield is the result of environmental factors, such as soil, weather, pest and

disease outbreaks, and farmer management. The total effect of these factors

translates into the production of green biomass and finally yield. Green plants

have a unique spectral reflectance or spectral signature. The proportion of

radiation reflected in different parts of the spectrum depends on the state,

structure, and composition of the plant. This information is captured in satellite

images as spectral data (that is, spectral reflection in various bands), which can

be used to construct several vegetation indices such as the Normalized

Difference Vegetation Index (NDVI) and the ratio vegetation index. High

correlations are found between NDVI and green biomass in studies done at the

field level (Groten 1993). Ground truthing in the form of field visits to

determine crop types and actual yield estimation in selected fields (pixels) that

cover the full range of observed NDVI values is required in order to correlate

NDVI values to crop types and crop yield.

In many countries, including China, India and the United States, using remote

sensing to estimate biological crop yield is being explored and is likely to

become the keystone of agricultural statistics in the future (Zhao et al. 2007).

However, considerable research is still needed before remote sensing can be

widely applied to estimate crop yield. In India, for example, vegetation indices

from satellite images showed only a moderate correlation (R2 between 0.45 and

Page 63: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

55

0.54) with crop cut data (Singh 2003). One important shortcoming for the use

of satellite images to estimate crop yield of smallholder farmers is the spatial

resolution of the sensor, which is not sufficient to capture the variability of

crops and crop performance in smallholder fields that often are less than 0.1

hectare and may be intercropped as well. A detailed field level study by

Sawasawa (2003) on rice in India highlighted this problem by showing that,

even with high resolution images, only 52 percent of observed yield variability

was captured. In developing countries, the problems of cloud coverage and the

need for expensive ground truthing, specialist knowledge and expensive image

processing software, limits the current usefulness of remote sensing (Reynolds

et al. 2000).

Das and Singh (2013) used the multiple frame approach to estimate wheat yield

for the state of Haryana in India. For this purpose, information was collected

from Wide Field Sensor (WiFS) and Linear Imaging Self Scanner-III (LISS-III)

data from the Indian Remote Sensing-1D (IRS-1D) satellite and crop-cutting

experiment data collected by probability sampling design from a list frame of

villages. Multiple-frame estimators were found to be more precise in

comparison to single-frame estimators.

Page 64: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

56

5 Crop acreage and yield

estimation under mixed

cropping Accurate estimation of crop yield is a difficult task, which becomes even more

challenging with the involvement of diverse farming systems, including a wide

range of crops and cropping patterns, such as mixed cropping, intercropping,

continuous cropping and staggered harvesting, for crops with an extended

harvest period, such as banana, cassava and coffee, and heterogeneous crop

performance in the field. An excellent review on estimation of crop area and

yield under mixed cropping or intercropping is available in Fermont & Benson

(2011).Recently, farmers have diversified into new farming systems that are

poorly enumerated in most of the national surveys. They tend to employ

intercropping or relay cropping in their fields in order to (i) evenly spread the

weather/calamities associated risks, and (ii) increase the production/total output

of individual fields as compared to pure stands. In the 1970s, researchers

identified eight different crops grown in mixtures in the middle belt of Nigeria

(Norman et al. 1982). Intercropping is the dominant agricultural practice in

many countries. As an example, intercropped fields in the IFPRI/INRAN

(National Agricultural Research Institute of Niger) data set of Niger have six

crops per field. Figures from the Uganda National Census of Agriculture and

Livestock (1990-1991) indicated that 80 to 90 percent of the area was under

mixed stands. A more recent survey by Henrietet al. (1997) showed that mixed

cropping such as millet/cowpea, sorghum/cowpea, sorghum/groundnut

intercropping was the predominant system in the Sudan Savannah of Nigeria.

The Tanzania Agriculture Sample Census, 2007/08 (NSCASHA 2012)

indicated that about 23% of national area comes under temporary mixed crops,

permanent mixed crops and permanent/annual mixed crops. Intercropping

became more essential for such crops as maize, beans, millet groundnuts and

cassava. Much of the early work focused on the motivations for planting mixed

Page 65: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

57

rather than single-cropped fields, owing to risk aversion, labor constraints or

higher profits (Norman 1973a; Abalu & D'Silva 1979; Just 1981; Just &

Candler 1985). The final impact of intercropping on crop yield is the result of

complex interactions of many factors, including, among others, relative time of

planting, plant density, rainfall, soil fertility and crop management. The overall

effect of intercropping on the production and yield of individual crops is mostly

negative, though, in some cases intercropping may actually increase crop yield.

Hopkins & Berry (1994) discussed potential underestimation in national

production and yield by not fully enumerating mixed cropping fields which

could be very high. They reported that cereal (millet or maize or sorghum) was

the principal crop for 64 percent and pulse (peanuts or bambara nuts) for 36

percent of the intercropped fields. Hence, if only the principal crop was

enumerated on an intercropped field, the value of output captured represented

only 74 percent of the total value produced per hectare and only 72 percent of

the total value produced per labor day. Kelly et al. (1995) recommended that

only the two principal crops in an intercropped field be accounted for. Thus,

would be useful to include a minimum percentage (say, 20 percent of plot area

occupied by a crop) before a crop is counted as an intercrop.

The problem of correct recording of area under mixed crops becomes difficult.

This is because cultivators plant a large number of crops with considerable

variation in the proportion of seed of the component crops. The crops in

mixture are sown either row-wise, separately or mixed altogether. It has been

observed that even the row-ratio may vary. Kathuria (1998) highlighted the

problems of area and yield estimation under mixed and continuous cropping.

He pointed out that mixed cropping was a subsistence need and a widely used

practice by small farmers in many African countries. In the case of a mixed

cropping situation, several crops are planted at the same time in the same field

with the seeds either mixed or planted in rows in some fixed ratios. Harvesting

of these mixed crops may not always be at the same time and hence

enumerators are required to make repeated visits to record yield data.

Cassava is a staple food crop grown in many African countries. It is not

completely harvested at any particular time of the year, but it may be taken out

from the ground as and when required or used as a "reserve crop" in

emergencies. In some countries, it is not feasible to identify the area planted

with cassava at any moment as some may have just been harvested in a

continuous cropping sequence.

Kathuria (1998) concluded that preparation of a sampling frame, sample

selection and measurement of area are problems that need to be addressed in

these situations. Each of these situations requires an approach for conducting an

Page 66: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

58

average yield estimation. When and how often the data on yield should be

collected in cases of continuous cropping are important points for

consideration.

Four strategies have been identified for estimating crop area, production and

yield in farming systems that have a significant proportion of crops produced

under intercropping (Kelly et al. 1995). Fermont &Benson (2011) provided an

excellent review on these four strategies with their advantages and

disadvantages.

Strategy 1 - Ignore intercropping

For the first strategy, intercropped plots are completely ignored. Crop areas are

recorded for sole cropped plots only, resulting in the estimated total crop area

being an underrepresentation of the actual area and the average crop yield being

an overestimation of actual crop yield. The estimated total production for crop

A is given as

Total production of crop A = (ΣArea crop Apure)× Avg. yield crop Apure.

Strategy 2 - Only record main crop

Under second strategy, though the intercropped plots are not ignored, but only

the main or predominant crop is recorded and the minor crop is ignored.

Estimates of crop area and crop yield are presented as if they are obtained in

sole cropped plots, though in reality they are obtained from a mixture of pure

and mixed stands. Total area for crop A is estimated as the sum of the total pure

area of crop A and the total area in which crop A is the main crop. As areas

with crop A as a minor crop are ignored, the estimated crop area is still an

underrepresentation of the actual area, though the underrepresentation is

significantly less than as in case of strategy 1. Average yield is determined from

a random selection of fields that have crop A either as a pure stand or as the

main intercrop. Total production for crop A is estimated as

Total production of crop A = (ΣArea crop Apure / main crop) × Avg. yield crop Apure

/ main crop.

Strategy 3 - Use whole plot as a denominator for each crop in the mixture

The third strategy uses the entire plot size as a denominator for each crop in a

mixture during both area and yield estimations. The crop mixture is indicated

for each area and yield estimation. The total area for crop A consists of the total

area for crop A as a pure stand plus the total area for crop A in all its recorded

mixtures. If crop A is a minor crop in prevalent mixtures, the total area for crop

A will be overestimated. The average yield for crop A is determined separately

for crop A as a pure stand and for each of its recorded mixtures.

Page 67: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

59

Capturing total production for crop A involves inclusion of the area and average

yield of crop A for all recorded mixtures. According to Kelly et al. (1995), the

included mixtures should be limited to the two most important mixtures for

crop A within a region in order to capture most of the production value,

simplifying data recording and reducing possible recording errors.

Alternatively, a threshold (for example, area of mixture × 10 per cent of total

area of crop A) may be used to decide whether to include a certain mixture in

the estimation. Total production for crop A is then estimated as

Total production of crop A = (ΣArea crop Apure) × Avg. yield crop Apure +

(ΣArea crop Amix_1) × Avg. yield crop Amix_1+

(ΣArea crop Amix_2) × Avg. yield crop Amix_2

where mix_1 and mix_2 represent the most common crop mixture for crop A.

These may be mixtures in which crop A is the main crop or the minor crop. In

4situations in which the excluded mixtures represent relatively important areas,

their exclusion will result in the underreporting of the total production of crop

A.

Strategy 4 - Allocate part of plot size to each crop in the mixture

For the fourth strategy, the plot size is proportionally divided between the crops

planted in the mixture during both area and yield estimations in order to adjust

the observed area and yield estimations to pure stand estimation. The division

of the area between various crops can be done in three different ways:

Strategy 4a: visual estimation of the proportion occupied by each crop;

Strategy 4b: examining the seeding rates or measurements of crop density;

Strategy 4c: using fixed area ratios for each intercrop combination.

Total area for crop A is estimated as the sum of the adjusted crop areas,

whereas average adjusted yield is determined from a random selection of plots

that have crop A either as a pure stand or as a major or minor intercrop. Total

production for crop A is then estimated as

Total production crop A = (ΣArea crop Aadjusted to pure stand) ×Avg. yield crop

Aadjusted to pure stand

According to Kelly et al. (1995), strategies 3 and 4 are the most commonly used

around the world. Mortensson et al. (2004) emphasizes that, FAO and some

European countries use strategy 3 whereas all European countries reporting to

Eurostat use strategy 4a. Kelly et al. (1995) reported that the strategy 4b is used

in Rwanda. Strategy 4c is used in many states in India (MoS&PI 2008). For

strategy 3, the entire plot size is used as the denominator, which allows

Page 68: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

60

comparison to be made between yield of sole crop versus intercrop. The main

disadvantage of this strategy is area reported under intercropped fields is double

counted. Thus, it does not allow for aggregating crop area at a higher level.

Strategy 4 proportionally allocates area to each intercrop as it adjusts all area

and yield figures to pure stand data. This facilitates a comparison across regions

or countries and removes the risk of double counting areas. Under strategy 4a,

estimating area proportions may be a questionable and difficult exercise,

especially if the crops are planted at random or more than two crops are present

in the plot. In strategy 4c, use of a fixed area ratio simplifies data collection by

the enumerators. Although it does not result in correct estimations of crop yield

at the individual plot level, it may generate satisfactory results at higher

aggregate levels (MoS&PI 2008). Fermont & Benson (2011) has suggested that

in future agricultural censuses or surveys, complex intercropping scenarios

should be taken into account.

Strategy 5 – Estimation of crop area based on imputed area

The estimation of the crop area through the two approaches yields completely

different statistical data pertaining to two different concepts of area. The first

denoted by "allocated area" is that fraction of the physical field area in which

the particular crop is cultivated. The sum of the allocated areas of the different

crops in the mixture should be equal to the total physical area of the field. The

second denoted by "imputed area" is the area which would have been occupied

by the crop had it been cultivated in pure stand. In general, the sum of the

imputed areas is not equal to the physical area of the field. The ratio between

the imputed area and the physical area can be considered as an indicator of the

intensity of cultivation of the land.

Under this strategy, as discussed in FAO (1982), the concept of imputed area is

utilized. The imputed area is described as the area occupied by the crop when it

is in pure stand. The sum of the imputed area is not equal to the physical area of

the field whereas the previously mentioned methods are based on the concepts

of allocated area. Allocated area is the fraction of physical field or area where

the particular crop is cultivated. The sum of allocated areas of different crops in

the mixture is equal to the physical area of the field. To calculate the imputed

area under the condition of mixed cropping, the following formulas are used.

Page 69: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

61

Let, A be the physical area of the field

i is the subscript denoting the crop in the mixture.

ci is the numerical value of the characteristic for crop i under the conditions of

mixed cropping.

Ci is the value of the same characteristic of crop i under the conditions of pure

stand.

Then the imputed area Ai of crop i is given as

ii

i

cA A

C

and the allocated area of crop i is given as

i

i ii

i i

i

c

C AA A A .

c A

C

Imputation of crop areas in mixed cropping can be based on different criteria.

However, these criteria should depend on those characteristics of the crops are

highly correlated with either the area or production. The main characteristics

that can be used for the imputation of areas are:

Amount of seeds;

Density of the plants, such as mound or hills;

Volume of production; commercial value of the produce.

The choice of the proper characteristic depends on its relevancy to the

objectives of the survey and also on the availability of the data.

The different characteristics to be used in the imputation or allocation of areas

to different crops are not always readily available. The holder usually knows

the quantity of seeds he has utilized (in some local unit of measurement) but is

not always aware of the amount to be sown in the case of pure stand cultivation.

The average amount of seeds in the case of pure stand cultivation could be

collected from the farmers, theoretically determined or decided upon a

posteriori. Information on crop density in mixed cropping can be obtained

through the use of density plots or by counting the number of plants within the

crop-cutting plot.

Page 70: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

62

The theoretical average density for the crop in pure stand can be derived from

averages

calculated for regions, districts and provinces, among others, depending on

available information. Production under conditions of mixed cropping and also

in pure stand can only be obtained after the crop has been harvested and the

production is measured. The commercial value of the produce requires a study

of crop prices to supplement the information on the volume of the production.

When the imputation of crop areas is based on the criteria of the amount of the

seeds sown or the volume of the production obtained during the totality of the

time reference period (agricultural year), in cases in which some of the crops in

the mixture are harvested before other crops are planted and thus added to the

mixture (a combination of mixed and successive cropping), is automatically

covered. In such a system of estimation of crop areas, the problem of successive

cropping reduces to a special case of mixed cropping.

In the presentation and/or tabulation of the results on crop areas under

conditions of mixed cropping, it would be very useful to present separately the

following four types of areas for each particular crop:

(i) Total area of the crop in pure stand;

(ii) Total area of the crop mixed with others;

(iii) Total imputed area of the crop;

(iv) Total allocated area of the crop.

This would permit different types of aggregation, as follows:

(i) + (ii) is the total physical area on which the crop is cultivated;

(i) + (iii) is the total area which could be used for the calculation of the crop

production (multiplying it by the average yield in pure stand)

(i) + (iv) is the total land-use area of the crop.

Method of recording area under mixed crops in India

The practice of sowing intercrops in the same field is quite common in almost

every part of India. This practice of mixed cropping provides protection to

cultivators against weather uncertainties. Meanwhile, the method of sowing

Page 71: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

63

intercrop is not uniform across the country. The crop mixtures are sown either

row-wise, separately or mixed altogether. The proportionate net area of each

component crop from all crop mixtures involving it are obtained and added to

the area sown singly (pure) with it to give its net area which is then published.

The allocation of gross area of a crop-mixture to its different component crops

is done either at the source, for example at the field level, by the primary

worker during the crop inspection (Girdawari or partal) and the net area of

each component crop is recorded separately in the crop statement (Jinswar), or

the primary worker records the whole area of crop-mixture, treating it as a

single crop and the total area of the mixture is separated to the component crops

at the district level.

The assignment of net area to different component crops at the field level is

made in proportion to the number of their rows, if they are sown in separate

lines. In cases in which the crops in the mixture are sown after thoroughly

mixing the seeds, this allocation is done in proportion to the actual amount of

seeds sown or seed-rates adjusted for mixed sowing or by eye estimation of the

relative stands of component crops. The components occupying a negligible

area or area below certain specified minimum are ignored and their area is

allocated to the chief component along or proportionately to all component

crops of a mixture.

The apportionment of net areas of component crops of a mixture at the district-

level is done on the basis of a fixed ratio, which is supposed to represent the

average conditions with regard to one or more of the above-mentioned factors

for the fields of the mixture in the district.

The procedure followed in the allocation of net areas of component crops of

mixture allocations may be grouped under the following three categories:

1. Allocation is done entirely at the field level;

2. Certain major crop mixtures are recognized as single crops and

allocation of net area of their components is done not at the field level

but at the district level, while in the case of unrecognized mixtures, the

allocation is done at the field level;

3. Allocation is done entirely at the district level on the basis of fixed

ratios.

Page 72: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

64

6 Crop acreage and yield

estimation under continuous

cropping As per recommendations of FAO (1982), the estimation of crop areas under

continuous cropping should be carried out in multi round surveys depending on

the number of different configurations of the crops in the fields. In a system of

regular periodic reporting (monthly, bimonthly or quarterly), the number of

rounds for estimation of crop areas under a system of continuous cropping

could be the same.

In Rwanda, continuous or sequential cropping (one crop following another

during the same year on a given parcel of land) and intercropping (several crops

planted at the same time on the parcel of land) are common. Although the

potential for underestimating output is high because mixed cropping and

sequential cropping are common, surveys conducted to collect national

production data enumerate all crops and use measures of relative crop densities

to determine how much land is occupied by each crop. Norman et al. (1995)

states “although progress has been made in developing methods for assessing

crop densities and yields for both systematic and random crop mixtures,

researchers and statistical services need to consider the research question at

hand, as well as the cost and feasibility of getting accurate density estimates,

before adopting these procedures”.

Due to even distribution of rainfall, farmers may plant and harvest crops

throughout the year. This may extremely complicate the estimation of crop area

and production. In order to capture the continuous cropping patterns during the

1965 agricultural census in Uganda, the crop area of each farmer was recorded

three times during the census year (MAC 1965). However, it was noted that the

holders changed the constituency of crops within a plot and plot boundaries so

frequently that it was difficult to link the records of the various visits.

Page 73: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

65

Subsequent Ugandan surveys, therefore, used a single visit to measure crop

areas.

Crops that have an extended harvest period and multiple pickings, such as

cassava, sweet potato, banana, cotton, and coffee, pose a problem in crop yield

estimation studies. Crop-cutting and whole plot harvesting methods do not take

into account the extended harvest period of the crop under study. They are

carried out at one given point of time when the crop is assumed to have

matured. If harvesting is always done at the same time after planting, the

resulting crop yield can be compared across regions and years. When crop

yields are obtained through the whole plot harvesting method, it may be

regarded as the most objective yield measurement that can be obtained for this

type of crop. Notably, crop cuts and whole plot harvesting cannot be used to

estimate banana yield due to its uneven ripening throughout the year. Wairegi et

al. (2009) developed a method to estimate banana bunch weight on individual

plants in East Africa using non-destructive field observations.

Page 74: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

66

7 Small area estimation

techniques Surveys are generally planned to produce estimates for population,

subpopulation or larger domains, such as national or subnational/province/state

levels. Sample sizes are fixed in such a way that direct survey estimator

(defined using domain-specific survey data only) provides reliable estimates

with a predetermined level of precision for planned domains in those surveys.

Policy planners, researchers, government and public agencies often require

estimates for many unplanned domains. Such unplanned domains can be a

small geographic area or a demographic group or a cross classification of both.

The sample sizes for such unplanned domains in the existing survey data may

be very small or at times even zero. In the survey literature, a domain is

regarded as small if the domain-specific sample is not large enough to support a

direct survey estimator of adequate precision. These small domains are also

called small areas, so called because the sample size in the area or domain from

the survey is small. If the sample size is small, domain-specific direct

estimators can provide an unacceptably large coefficient of variation. Therefore,

it becomes necessary to employ indirect small area estimators that make use of

sample data from related areas or domains through linking models, and thus

increase the effective sample size in the small areas. Such estimators can give a

significantly smaller coefficient of variation than direct estimators provided the

linking models are valid. These estimators are often referred to as the indirect

estimators as they use values of survey variables (and auxiliary variables) from

other small areas or times, and possibly from both. An underlying theory that

resolves the problem of small sample sizes is often referred to as small area

estimation techniques in the survey literatures, see Rao (2003).For example, the

traditional domain estimation approach provides reliable estimates of crop area

and crop yield in a mixed cropping scenario. However, for some of the crop

mixtures, the sample sizes may turn out to be extremely small, which may

affect the reliability of the estimates. This small sample size problem can be

Page 75: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

67

addressed through the small area estimation approach by combining the survey

data and the already available secondary data. The small area estimation

techniques are generally based on model-based methods (Pfeffermann 2002;

Rao 2003). The idea is to use statistical models to link the variable of interest

with auxiliary information, namely a census and administrative data, for the

small areas to define model-based estimators for these areas. The traditional

indirect estimation techniques based on implicit linking models are a synthetic

estimation.

7.1. SYNTHETIC ESTIMATION

When producing the synthetic estimates for small areas, availability of direct

estimates for a set of larger domains of the population is assumed. Appropriate

weights or proportions are then applied to these large population domain

estimates to obtain the desired small area estimates. This class of estimators

implicitly assumes that small areas that are being considered are similar, in

some sense, to some larger areas which contain them and for which the reliable

direct estimate is available. Gonzalez (1973) described synthetic estimator as

one in which an unbiased estimator of a large area was used to derive estimates

for subareas under the assumption that the small areas had the same

characteristics as the larger areas. The term “synthetic” refers to the fact that an

estimator computed from a large domain is used for each of the separate areas

comprising that domain, assuming that the areas are “homogeneous” with

respect to the quantity being estimated. Thus, synthetic estimators already

borrow information from other “similar areas”. (National Health Interview

Survey (1968) first used synthetic estimates to calculate state estimates of long

and short-term physical disabilities from the National Health Interview Survey

data. As the sample size in a small area increases, the direct estimator becomes

more desirable than a synthetic estimator. This is true whether or not the sample

was designed to produce estimates for small areas. Also, this motivates the use

of a weighted sum of direct estimator and synthetic estimator as a desirable

alternative than choosing one over the other. This weighted estimator is termed

as the composite estimator. These estimators are of interest because they permit

a trade-off among the advantages and disadvantages of direct and synthetic

estimators through their weighted combination (Rao 2003).

The synthetic methods have the advantage of being simple to implement. These

estimation techniques provide a more efficient estimate than the corresponding

design-based direct estimator for each small area through the use of implicit

models which “borrow strength” across the small areas. These models assume

that all the areas of interest behave similarly with respect to the variable of

interest and do not take into account the area-specific variability. However, it

Page 76: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

68

can sometimes lead to severe bias if the assumption of homogeneity within the

larger domain is violated. That is the area specific variability typically remains

even after accounting for the auxiliary information. This limitation is handled

by an alternative estimation technique based on an explicit linking model,

which provides a better approach to small area estimation by incorporating

random area-specific effects that account for the between area variation beyond

what is explained by auxiliary variables included in the model, referred to as the

mixed effect model. It should be noted that the random area effects in the mixed

model capture the dissimilarities between the areas. In general, estimation

methods based on an explicit model are more efficient than traditional methods

based on an implicit model.

7.2. MIXED MODELS IN SMALL AREA ESTIMATION

The explicit models used in small area estimation are a special case of the linear

mixed model and are very flexible in formulating and handling complex

problems in small area estimation. Based on the level of auxiliary information

available, small area estimation methods are classified into two broad types:

i) The area level mixed effect models (or area level models), which are

used when auxiliary information is available only at the area level. They

relate small area direct estimates to area-specific covariates (Fay &

Herriot 1979);

ii) Unit level mixed effect models (or unit level models), proposed

originally by Battese et al. (1988). These models relate the unit values

of a study variable to unit-specific covariates.

These are special cases of the linear mixed model, usually referred to as area

level and unit level small area models.

7.2.1. AREA LEVEL MODELS

Fay and Herriot (1979) proposed an area-level small area model that relates

small area direct survey estimates to area-level covariates. The area-level model

is widely used by statistical organizations because of its flexibility in combining

different sources of information with different error structures, and can be

described as follows. Let i index the m small areas (small domains) of interest

and let y

i be an unbiased direct survey estimator of an unobservable population

parameter (for example, the population mean) Y

i of a variable of interest y for

Page 77: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

69

small area i. Let x

i be a p-vector of known auxiliary variables for area i that are

related to the population mean Y

i.

i i iy Y eand

Ti i iY u x

, (7.1)

Where the first (measurement) equation accounts for the sampling variability of

the observable survey estimator y

iof the true area i population mean

Y

i, while

the second (process) equation is a linear regression model for the unobservable

Y

iin terms of the vector

x

i. Combining these two equations leads to an area

level linear mixed model of form

; i =1,...,m. (7.2)

Here is a p-vector of unknown fixed effect parameters, the regression errors

u

i are often assumed to be independent and identically distributed Gaussian

errors, with E(u

i) = 0 and

2var( ) ,i uu and the sampling errors e

i are

similarly assumed to be independently distributed Gaussian errors with

E(e

i|Y

i) = 0 and

2var( | )i i ie Y D . The regression and sampling errors are

assumed to be independent of each other within and across areas. An important

additional assumption is that the constants 2

iD are known. The parameter s

u

2 is

typically referred to as the variance component of (7.2). Under (7.2), replacing

unknown model parameters by estimates values, the empirical best linear

unbiased predictor (EBLUP) estimate of Y

i is (Henderson 1975; Fay & Herriot

1979)

ˆˆ ˆEBLUP T

i i iY u x , (7.3)

where ˆˆˆ y T

i i i iu x is the EBLUP of area-specific effects iu ,

1

2 2 2ˆ ˆ ˆi u u iD

is the shrinkage effect for small area i, is the empirical

best linear unbiased estimator of . See Rao (2003, chapter 5) for further

details. For non-sampled areas, the approach for estimating small areas is

synthetic estimation (Rao 2003) based on a suitable model fitted to the data

Page 78: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

70

from the sampled areas. This is equivalent to setting the area effect for such an

area to zero. Under model (7.2), the synthetic EBLUP predictor for Y

i is

. (7.4)

Sudet al. (2012) considered an application of small area estimation techniques

to derive model-based estimates of average yield for paddy crop at districts

levels in the State of Uttar Pradesh in India by linking data generated under the

Improvement on Crop Statistics scheme by National Sample Survey Office

(NSSO) (data collected with much reduced sample size, however, the quality of

data is very high) and the Population Census 2001. They adopted the area level

model (7.2) as covariates were available only at the area level. In particular,

they illustrated how the small area estimation technique can be satisfactorily

applied to produce reliable district-level estimates of crop yield using a crop

cutting experiment supervised under the Improvement of Crop Statistics (ICS)

scheme. Further small area estimation techniques provided estimates for those

districts also where there was no sample information under ICS and so direct

estimates could not be computed. They recommended that wherever it is not

possible to conduct an adequate number of crop cutting experiments due to

constraints of cost or infrastructure or both, the small area estimation techniques

can be gainfully used to generate reliable estimates of crop yield based on a

smaller sample. Sisodia and Chandra (2012) used small area estimation

techniques for crop yield estimation at the developmental block level.

7.2.2. UNIT LEVEL MODELS

Battese et al. (1988) first employed the “nested error unit level regression

model”, which has since become one of the simplest models commonly used in

small area estimation. This model is also known as Random Intercept model.

They used small area estimation under a unit level model to estimate county

crop areas using sample survey data in conjunction with satellite information. In

particular, they were interested in estimating the area under corn and soybeans

for each of the 12 counties in North-Central Iowa using farm-interview data as

dependent and LANDSAT satellite data as independent variables. Each county

was divided into area segments and the areas under corn and soybeans were

ascertained for a sample of segments by interviewing farm operators. Auxiliary

data in the form of numbers of pixels (a term used for “picture elements” of

about 0.45 hectares) classified as corn and soybeans were also obtained for all

the area segments, including the sampled segments, in each county using the

LANDSAT satellite readings. In the model, it is assumed that the values of

auxiliary variables are known for every unit in the sample and that the true area

Page 79: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

71

means of these variables are also known. Denoting by ijx the auxiliary values

for unit j in small area i , the model has the form,

T

ij ij i ijy u e x (7.5)

where ijy denotes the value of variable of interest for sampled unit

( 1,...., ) ij j n in area ( 1,...., )i i m , ijx is a 1p vector of unit level auxiliary

variables, is a 1p vector of the unknown fixed effects, in is the number of

sample units in area i, iu is the area specific random effect associated with

area i with mean zero and variance 2

uσ , and ije is individual level random error

with mean zero and variance 2

eσ . The two error terms are mutually

independent. The random error iu represents the joint effect of small areas that

are not accounted for by the auxiliary variables, also known as the model error

for area i. The normality of iu and ije is often assumed. Let population mean of

Y in small area i beT

i ii i iY u e X , where 1

1

iN

i i jjN

X x , is assumed to be

known. For sufficiently large iN , 1

10

iN

i i jje N e

, and then mean of Y in

small area i is approximated by i i iu X . Under (7.5), the EBLUP of the

mean of Y for small area i, is

ˆ ˆˆˆ ( )T T

i ii i i iy X x , (7.6)

where 1

2 2 1 2ˆ ˆ ˆ ˆi u u i en

and in sample size for small area i. Here, iy and

ix are the sample mean of y and x, respectively. Similar to (7.4), for non-

sampled areas, we can define synthetic estimator under (7.5). Note that the

EBLUE of used in (7.6) is defined under unit level model (7.5). In contrast,

the EBLUE in (7.3) and (7.4) is based on area level model (7.2).

Page 80: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

72

7.3. EXTENSION OF MIXED MODELS IN SMALL AREA

ESTIMATION

It is noteworthy that small area estimation methods are based on a linear mixed

model, with area-specific random effects to account for between areas variation

beyond that explained by auxiliary variables included in the fixed part of the

model. In commonly used small area estimation methods, these random area

effects are assumed to be independent. That is, different small areas are

considered as independent to each other. However, in practice most small area

boundaries are arbitrary and there appears to be no good reason why units on

only one side of such a boundary should not generally be correlated with units

just on the other side. For example, in agricultural data, neighboring areas

exhibit strong spatial dependency and therefore independence assumption of

random area effects seems questionable. The EBLUP method can be improved

by including spatial structure in the random area effects. Singh et al. (2005)

defined the spatial-EBLUP approach in small area estimation under area level

model. Chandra et al. (2007) compared the EBLUP and MBDE approaches for

the spatially correlated populations under the unit level model. Chandra (2013)

described an improved method of small area estimation using spatial

information in estimating the crop yield at district level combining

Improvement of Crop Statistics data and known population level auxiliary

information. He exploited spatial association between the districts through a

spatial model, in particular, a Simultaneous Autoregressive error process in

random area effects under an area level model. He further explored various

ways to define this spatial weight matrix to exploit spatial information to

produce reliable estimates for small areas and suggested that spatial association

effects (or spatial dependence) should also be used to improve small area level

estimates.

As noted above, in order to increase the overall sample size in small area

estimation, information from other data sets must be used. This information can

be borrowed from “similar” areas or from a previous occasion. In the time

series modelling approach, we exploit information in data over time, namely

repeated surveys in order to obtain further improvement in the efficiency of

estimators. In general, empirical studies show that small area estimates that

draw upon information across time are more efficient than those that draw

information across an area, as the time series data usually represent the same

information about the target variable from the past (Pfeffermann et al. 1998;

Datta et al. 1998). Sometimes cross sectional and times series data are

combined to obtain further improvement in efficiency of the small area

Page 81: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

73

estimators. In general, empirical studies show that for repeated surveys, a

considerable gain in efficiency can be achieved by borrowing strength across

both small areas and time (Rao and Yu 1994). Singh et al. (2005) used spatial-

temporal models in small area estimation. They used spatial models to exploit

spatial auto-correlation among the small area units and a spatial temporal model

fitted through Kalman filtering for the time series data. Chambers and Tzavidis

(2006) introduced the M-quantile approach to small area estimation. Standard

approaches of small area estimation assume that the underlying relationship

between the variable of interest y and the set of covariates x are linear.

However, in practice, in a lot of survey data (for example, agricultural), this

linear relationship is not valid. Chandra and Chambers (2011) proposed a small

area estimation method for the variable, which follows the linear model under

log transformation. Furthermore, the existing approaches of small area

estimation assume that the relationship between the variable of interest y and

covariates are stationary over the study space (the same for all areas). However,

this assumption may not be correct for a lot of survey data, such as agricultural

and environmental data. Chandra et al. (2012) described the small area

estimation for such data using a geographical weighted regression approach to

capture the spatial non-stationarity in the data.

7.4. SOME APPLICATIONS OF SMALL AREA

ESTIMATION IN AGRICULTURE

Rao (2003) discussed some approaches of small area estimation with

applications to agriculture data. Dorota (2006) applied small area estimation

methods in agricultural sample surveys in Poland, using the latest census of

agriculture as an auxiliary source of data. They used empirical and hierarchical

Bayes estimators, and some auxiliary information from the last census of

agriculture to obtain more precise estimates of agricultural characteristics from

agricultural sample surveys by county. Two different regression models were

considered: an area-level regression model and a unit-level one. The unit-level

approach required matching particular farms in the agricultural sample surveys

to the census of agriculture. The precision of the model-based estimates is

significantly increased compared to direct estimates.

In India, early development in small area estimation, for crop yield estimation

can be dated back to 1966 and 1968 when Panse et al. (1966) attempted to

estimate the crop yields at the developmental block (small area) level using a

double sampling approach. An estimate of crop yield from large number of

plots prior to harvest were obtained from farmers by enquiry while data on crop

yield through crop-cutting experiments was collected on a subsample from the

Page 82: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

74

larger sample. The data from the two sources were suitably combined in the

form of a double sampling regression estimator to obtain precise estimators of

crop yield at the small area level. The approach was applied in two districts of

India. In one of the districts, the results were not encouraging because of the

poor correlation between the farmer estimate and the crop cut estimate. The

reason for the poor correlation was that the farmer appraisal data was obtained

well before the harvest time of the crop. Srivastava et al. (1999) used a

synthetic method for crop estimation at the block level. The population was

classified into two dimensions with small area on one side and post-strata

(homogeneous groups) on the other side. For crop yield, the cell weights were

estimated by ranking ratio methods using the data collected in the crop-cutting

approach. In fact, auxiliary information collected during crop-cutting

experiments was used in conjunction with small area level data for crop area for

estimating the cell-weights. This approach was applied to estimate crop yield at

the block level for wheat and paddy crops on the basis of data obtained from

crop estimation surveys in the Haryana state of India during the period 1987-

1988. The results were consistent and satisfactory. However, the results were

based on certain assumptions. One assumption was that different blocks are

homogeneous with respect to the target variable under study. This assumption is

widely referred to as the synthetic assumption in small area literature. It is a

very strong assumption, which may often not stand up. When this assumption

fails, estimates can be seriously biased. The synthetic approach of estimation

was also applied by Singh and Goyal (2000) to estimate crop yield for wheat

crop at the tehsil (sub-district) level, using remote sensing data. Post-strata were

formed using the vegetation index derived from remote sensing satellite data.

Wheat crop data from the General Crop Estimation Survey (GCES) during

1995-96 in Rohtak district of Haryana State in India while the spectral data of

IRS-IBLISS-II for February 17, 1996 was taken for the vegetation index. The

method improved the efficiency of the estimators, to some extent, in terms of

standard error. However, neglecting the bias remains a serious limitation.

Within a framework of sampling design conforming to the GCES approach,

Sud et al. (2001) developed crop yield estimates at blocks level using farmers’

estimates. The technique was similar to one used by Panse et al. (1966). In this

case, the farmer appraisal data was obtained within 15 days of harvest of the

crop. This resulted in a reasonably high correlation between the farmer

appraisal and the crop-cut data, which led to precise estimates of crop yield at

the small area level. These estimates were, in fact, direct estimates and were

based on usual sample survey techniques for improvement of estimators. One of

the merits of this approach is that it does not involve any explicit model that is

difficult to standardize in the context of finite populations. Sharma et al. (2004)

Page 83: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

75

proposed two different types of gram panchayat level estimators of crop yield

using small area crop estimation methodology. Ahmad &Kathuria (2010)

conducted a study on estimation of crop yield at the block level using double

sampling approach.

Page 84: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

76

8 Country experiences on crop

acreage and yield estimation This section contains an overview of the methodologies for estimation of crop

acreage and yield being followed in different countries.

8.1. BULGARIA

Agriculture covers half the national Bulgarian territory, forests cover one third.

The methods involved in estimation of crop parameters in Bulgaria are as

follows:

Crop area estimation

Data of Bulgaria on land cover and land use are obtained by a survey called

BANCIK, similar to the survey on land cover and land use LUCAS, carried out

in the European Union. The area frame is constructed by a systematic area

frame sampling in two stages where PSUs that are cells in a regular grid with

size 6 km × 6 km (3,123 segments) and SSUs that are 36 points, arranged in a

6×6 point grid of 234 m each. The survey is carried out during May to July

throughout the country. The surveyors observe the same points in the same

segments each year. This shows the land cover variations and the structural

changes in the land use. The used nomenclature is fully harmonized with the

LUCAS nomenclature. (Bulgaria 2001)

Crop yield estimation

In Bulgaria forecast data on wheat and barley production and yields are

obtained through in-situ observation (expert estimation) of points in the field

during the conducting of the land use and land cover survey (BANCIK). The

data on harvested area and production of main crops result from the survey on

main crops. The survey is carried out in November through interviews with

farmers. The holdings are defined through a random stratified sample based on

the census holdings list. Data on vegetables are obtained from a separate sample

Page 85: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

77

survey on vegetable production. A list sample is used, based on the agricultural

census list of holdings (Bulgaria 2013).

8.2. BRAZIL

The Ministry of Agriculture, Livestock, and Supply, through the National Food

Supply Company (Conab), systematically carries out assessments of

agricultural crops to quantify and to follow Brazilian production. The Brazilian

Institute of Geography and Statistics is developing a new System of Integrated

Household Surveys based on a master sample. This process consists of many

steps, such as developing and updating a master frame, designing a sample

survey that can cover the demands of most household surveys and modifying

current household surveys in order to adapt them to the new master sample. The

methods involved in estimation of crop parameters in Brazil are as follows.

Crop area estimation

To obtain information about the estimated area of major crops, Conab uses

satellite imagery, aerial photography and geo-referenced information for

mapping cultivated areas in the main producing states. This activity began with

the launch of the GeoSafras Project, the purpose of which is improve the

methodology of the crop forecasting in Brazil through the development of

technologies related to remote sensing, satellite positioning, geographic

information systems and statistical, spectral and agro-meteorological models to

be applied in the estimates of area and yield(Brazil2013).

Crop yield estimation

By providing accurate information about the location of crop fields, mappings

are used in the spectral and agro-meteorological monitoring of crops, through

the monitoring of meteorological conditions, such as rainfall, soil moisture and

temperature, and of vegetation indices calculated from satellite images, which

reflect the condition of the vegetation and provide an indication of the level of

productivity. Historical data of agrometeorological and spectral parameters in

the mapped areas in each crop year are being used in the development,

calibration and application of models/systems to estimate productivity.

8.3. CANADA

The Agriculture Division of Statistics Canada conducts an extensive statistical

program with several highly integrated components comprised of the Census of

Agriculture, crop and livestock surveys, farm-economic statistics, agri-

environmental statistics, tax and other administrative data, research and analysis

and remote sensing data. Data are collected through computer-assisted

Page 86: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

78

telephone interviews from the regional offices and transferred electronically to

headquarters in Ottawa. The methods involved in estimation of crop parameters

are discussed as follows (Statistics Canada2015):

Crop area estimation

For estimation of crop area in Canada, the Agriculture Division of Statistics

Canada uses a sample survey approach with a cross-sectional design. During

surveys, two types of sampling frames: list and area, are used. In the farm

surveys, only the list frame is used for the sample selection. This list frame is

stratified into homogenous groups on the basis of census characteristics, such as

farm size and crop area, and sub provincial geographic boundaries. Sample

sizes, namely the number of farms corresponding to each survey, varies. The

March Farm Survey estimates farmers' seeding intentions. In June, preliminary

seeded acreage results are published and in November the estimates are revised

with information from the surveys conducted in the fall. The survey data

collected are weighted in order to produce unbiased level indicators that are

representative of the population. Presently, the agricultural area sample survey

has been redesigned and is referred to as the area farm survey. It is used to

complement the list samples of other agricultural surveys with sometimes

different list frames. As the Census of Agriculture in Canada is conducted at the

same time as the Census of Population and uses the same enumerators, the

enumeration areas created for the Population Census serve as useful units for

the area frame. To estimate crop acreage of potato, Agriculture and Agri-Food

Canada conducts sample survey on remote sensing data, using a regression

estimator to obtain the crop acreage in major agricultural states.

Crop yield estimation

For estimation of crop yield and production, Statistics Canada uses a sample

survey approach. Since March 2014, for response burden matters, the

operations with, total arable land that is lower than a provincial threshold are

excluded from farm survey samples (null sampling). This threshold, which

varies from one province to the other, is such that all farms having an area

greater than the threshold itself represent in total 95percent of the total arable

land of a given province. The estimate share of the non-surveyed farms is then

modelled and added to the estimation derived from surveyed farms, in order to

draw a complete picture of field crop productions. The July, September and

November farm surveys provide data on the harvested area, expected yield and

production of crops on farms. The survey data are weighted to estimate

production at the provincial and crop district levels.

Page 87: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

79

8.4. EGYPT

The structure of farming in Egypt has totally changed over the past 50 years. It

has gone from being a small number of very large holdings, managed by

landowners, with little government control to consisting of a very large number

of small holdings with total government control. The methods involved in

estimation of crop parameters in Egypt are as follows (Fawzy et al. 1998):

Crop area estimation

The method formerly used by the Egyptian Survey Authority to obtain crop

area estimates entailed complete enumeration, which was very expensive. To

overcome this, a sampling technique to reduce costs and effort was developed,

with the area under major food grain crops estimated by using sample survey

methods. Later, the Ministry of Agriculture and Land Reclamation proposed a

more accurate and less expensive technique based on a check-sample of the

area determined by subjective methods of the agricultural local staff to remove

its bias. The country also adopts new technologies, such as GPS and remote

sensing, to improve the quality of crop area estimates. Currently, crop area is

estimated in Egypt by using following four methods:

(a) Direct measurements by the monitoring, verification, and evaluation

unit team, using a modern optical instrument;

(b) Direct measurement by the sampling staff using a tape on the

ground;

(c) Inquiry from the local extension staff in the village;

(d) Farmers’ estimate for his or her crop area.

Crop yield estimation

Both subjective and objective methods are used for estimating crop production

in Egypt. The best yield data come from the sampling offices, which conduct

crop-cutting surveys at harvest time. The sampling frames used for the survey

vary by governorates, a region headed by a governor. The national headquarters

determines the number of crop cutting samples for each governorate and crop.

These sample sizes are based on analysis of the previous year’s data. A

stratified multistage sampling procedure is followed to select samples. The land

areas are classified into strata based on type of irrigation and age of tile

drainage. Groupings of similar land areas are formed into clusters. The cluster

sizes vary depending on the governorate. Sampling units are formed within the

selected clusters consisting of about three feddans (1 feddan=4,200.833 square

Page 88: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

80

meters). A random sample of two sampling units is selected from each cluster

for the crop cutting survey.

8.5. ETHIOPIA

Agriculture is the primary activity in Ethiopia, with about 84 percent of the

country’s population engaged in various agricultural activities that generate

income for household consumption. To produce crop statistics, a sample survey

is used to produce estimates of crop area and the volume of production by farm

and crop type by the Central Statistical Agency (CSA, 2011).The methods of

collection of agricultural statistics are discussed below.

Crop area estimation

The sampling methodology of crop acreage estimation consists of construction

of sampling frames, which is a list of commercial farms from all parts of the

country that includes their cropland area size and livestock number is collected

from all part of the country through the Central Statistical Agency Branch

Statistical Offices. The collected farm list is compiled at the head office and the

functional and non-functional farms at the time of updating are identified.

Before the sample selection is completed, the cutoff point for the farms is

decided. Separate cutoff points for farms involved in crop production and those

involved in livestock is set. Farms with a total area that exceeds the cutoff point

are selected with certainty whereas farms with a total area that falls below the

cutoff point is sampled using probability proportional to size, with size being

the total area of the farms. Modern technologies, such as GPS and remote

sensing, are also being used for crop area estimation. The areas of commercial

farms are directly measured by GPS while those that are state owned are

measured based on an interview.

Crop yield estimation

For crop yield estimation, a stratified two-stage cluster sample design is

implemented, where, enumeration areas are taken to be the primary sampling

units and the agricultural households are the secondary sampling units. The

sample size is determined by fixing the precision and the amount of resources

allocated for the survey. In the first stage sample, the selection of enumeration

areas is done by probability proportional to size. The second stage sample

(household) selection is done by systematic random sampling. Through its 25

branch offices, the Central Statistical Agency has put in place a comprehensive

field organization to follow up on and monitor the survey field work.

Page 89: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

81

8.6. NIGERIA

Agriculture in Nigeria is characterized by considerable regional and crop

diversity. In most of the surveys and censuses conducted by the National

Bureau of Statistics (NBS, 2007), which is the major producer of agricultural

statistics in Nigeria, crops and livestock are considered together because of the

tendency for most of the farmers to practice crops and livestock husbandry

simultaneously. The three major agricultural statistics in Nigeria are crops and

livestock statistics, forestry and wildlife statistics, and fisheries statistics. The

most important sources of survey-census-based official statistics on crops are

the National Bureau of Statistics and the Livestock Department of the Federal

Ministry of Agriculture and Rural Development. The agricultural surveys and

censuses conducted on an ad-hoc basis by the National Bureau of Statistic were

integrated into one operational program called the National Integrated Survey

of Households. The design of that survey is based on guidelines set by the

National Households Survey Capability Program, which is sponsored by the

United Nations. Notably, the National Bureau of Statistics launched the revised

General Household Survey in 2010-2011 referred to as the GHS-Panel, to

collect panel data on households, their characteristics, welfare and their

agricultural activities. The GHS-Panel is a cross-sectional survey of 22,000

households is carried out annually throughout the country. The panel

component, which is applied to 5,000 households that participate in the survey,

entails collecting additional data on multiple agricultural activities and

household consumption. In this survey, the use of the computer-assisted

personal interview was proposed for the paperless collection of the GHS-Panel.

Crop area estimation

The National Agricultural Sample Census of Nigeria covers all land holdings

(except kitchen Gardens), which are traditionally operated. A two-stage

stratified sampling design is used with enumeration areas as first stage units and

households as second stage units. In the first stage, 300 enumerated areas are

selected with equal probability from each state using a systematic selection

procedure. These are grouped into 150 strata. In the second stage, two

enumerated areas are selected from each stratum. Finally, 20 households (10

households that engage in arable crops and/or livestock rearing and 10 that keep

livestock and poultry only) are selected from each stratum. The enumerated

areas are drawn from those demarcated by the National Population Commission

for the 2006 Housing and Population Census. The parcels owned by households

are measured with the help of a tape and compass, chain or by the triangulation

method. In addition, a GPS device and an area frame survey is also used for

estimation of crop area in Nigeria.

Page 90: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

82

Crop yield estimation

For crop yield estimation in Nigeria, a stratified two-stage sampling design is

adopted for the selection of household samples. The objective crop-cutting

experiments along with farmer’s prediction is used for crop yield estimation.

8.7. FRANCE

France has been one of the most dominant agricultural centers of Europe for

centuries. With 516,100 farms (Agricultural Census 2010), approximately

3percent of the workforce is employed in agriculture or similar sectors, such as

fishing or forestry. An annual survey is conducted by the surveyor as a part of a

land survey using aerial photography during the months of June and July, on a

date that falls between the sowing of seeds and harvesting for estimating areas

under cereals and fruit crops. Furthermore, cereal surveys are conducted to

measure wheat, barley and maize production from area and yield data.

Crop area estimation

To determine the land use and land cover statistics France relies on the

TERUTI system, which is similar to the Land cover/use (LUCAS) system, to

provide statistics on land use and land cover across the European Union (FAO,

2015). This is a two-stage system. For the first stage, square segments of 36

points, each 300m apart, are observed within each square. For basic

observations, the observation of the point in the field covers a circumference

having a diameter of 3m; in extended observations, the circumference has a

diameter of 40m.

Crop yield estimation

In France, yield figures are estimated from samples selected from sample plots

of known areas at random in several stages. Both subjective and objective

methods of crop yield estimation are used for crop yield estimation. Two types

of methods, namely selection of holdings and land use survey are used for

estimating yield. Regarding the selection of holdings technique, the sampling is

done at four stages, namely communes, holdings, fields, and sample plots. The

survey is conducted by interviewing farmers. Measurements of sample ears of

wheat or barley from plots of approximately one square meter and samples of

ears of maize from two rows five meters in length are taken for an objective

crop-cutting experiment.

Page 91: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

83

8.8. UNITED STATES OF AMERICA

The National Agricultural Statistics Service is tasked with generating

agricultural production statistics in the United States. The data collection work

is assigned to state statistical offices. The March/June survey of crop acreage is

used for framing estimates of crop acreage. The major emphasis in the survey is

on planted areas. Each sampled segment is divided into tracts, which are

delineated on photographs. A tract is a parcel of land within a segment under

one management. An area frame for a land area, such as a state or country,

consists of a collection or listing of all parcels of land for the area of interest

from which to sample from. These land parcels can be defined based on

ownership or simply on easily identifiable boundaries as is done by the National

Agricultural Statistics Service. The major area frame survey conducted by the

Service is the June Agricultural Survey. This mid-year survey provides area

frame estimates primarily for crop acreages and livestock inventories. Two

questionnaires are developed, one for the people living in the segment and the

other for those who live outside the segment. See Davies (2009).

Crop area estimation

For crop area estimation, area sample is used in addition to the sample selected

from a list of large farm operators. The area sample is the outcome of a single

stage stratified sampling. The basic stratification employed by National

Agricultural Statistics Service involves: (1) dividing the land into land-use

strata, such as intensively cultivated land, urban areas and range land, and (2)

further dividing each land-use stratum into substrata by grouping areas that are

agriculturally similar. Within each stratum, the land can be divided into the

sampling units or segments and then a sample of segments is selected for a

survey. This is a very time-consuming endeavor. The time spent developing and

sampling a frame can be greatly reduced by: (1) dividing the land into larger

sampling units called first-step or PSUs, (2) selecting a sample of then

delineating the segments only for these units , and (3) selecting a sample of

segments from the selected PSUs. Boundaries of sample segments are

delineated on an aerial photograph. The area of segment is determined by a

planimeter. The open segment concept is useful when farms are used as

reporting units and is applied in novel way. The June survey is conducted

annually and each year only 20 percent of the earlier selected sample is

replaced. The enumerators and supervisors are trained before the survey.

Page 92: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

84

Crop yield estimation

Objective yield surveys are executed for reporting yield statistics where

specifically cropped fields are randomly selected based on probability

proportional to size. These yield surveys involve making counts and

measurements of selected crops and weighing them. The yield estimation is

made on the basis of measurement of plant characteristics. Generally, two units

are measured in a selected field. Each unit consists of a specified number of

rows of predetermined length or rectangular units if crop rows are

indistinguishable. The acreage reported in the March/June agriculture survey

provides the sampling frame for yield surveys. Probability samples are selected

for developing estimates. Sample allocations are made in such a manner that the

sampling errors of the estimates are minimized. The March agriculture survey

uses a multiple frame, namely a list and area frame. The enumerators are

provided with aerial photographs along with area segments for data collection.

8.9. INDIA

India is primarily an agriculture-based country; its economy is largely

dependent on agriculture. The Directorate of Economics and Statistics, Ministry

of Agriculture, is the pivotal agency for the coordination and compilation of

agricultural statistics at all administrative levels. Other principal agencies that

collect data and conduct methodological studies on agricultural statistics are the

National Sample Survey Office, the Indian Agricultural Statistics Research

Institute and the State Directorate of Economics and Statistics. For more details

see MoS & PI (2008).

Crop area estimation

The crop area statistics are generated in large parts (84 percent) of the country

through complete enumeration. Cadastral maps of these areas exist. In areas

that have not been cadastrally surveyed (15 per cent), a sample survey approach

is used for data collection. A stratified unistage sampling design is adopted for

data collections in areas where in blocks are the strata, and a 20 per cent sample

of villages is selected. In successive years, a new group of non-overlapping 20

percent sample of villages is selected. Thus, over a five-year period, the entire

area is enumerated. In the remaining portions (1 percent), crop area statistics are

generated by the village headman. The data quality in those areas is not very

high, as the village headmen are usually not trained to in the field of statistics.

Crop yield estimation

Yield of crops in the country is estimated on the basis of crop-cutting

experiments. A stratified multistage random sampling design where tehsils are

Page 93: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

85

strata, villages are the first stage unit of sampling, fields growing a particular

crop are the second stage unit of sampling and a plot of a specified size is the

ultimate unit of sampling. Generally, square plots are used for the crop-cutting

experiments. For crops, such as cotton, rectangular plots of a larger size are

used. In some parts of the country, triangular and circular plots are also used.

8.10. SOUTH AFRICA

The South African Department of Agriculture Crop Estimates Committee is

tasked with producing crop estimates for the country on a monthly basis. To

perform this task, the Committee receives data from various input suppliers.

The National Department of Agriculture is the custodian of the Crop Estimates

Committee. The grain crop production estimates are published monthly. The

Committee meets monthly to review and debate the current status of cropped

areas and conditions to determine crop production estimates for grain crops.

Grain crops in South Africa consist of two groups: the first group is comprised

of crops, such as maize, sunflower, Soya beans, and sorghum, that are

cultivated during summer and second one is made up of crops grown during the

winter, such as wheat, barley, oats and canola. Information about crops was

previously only supplied by provincial and industry representatives and through

qualitative reports on weather patterns and crop conditions that resulted in

subjective calculations. See Ferreira et al. 2006.

Crop area estimation

The Producer Independent Crop Estimate System was developed in 2005.

Implemented after a successful pilot study conducted in the Gauteng province,

the System uses crop field boundaries digitized from satellite imagery with a

point frame sampling system to objectively estimate the area planted with grain

crops. It involves several steps, starting with the procurement of satellite

imagery. Then, digitization of crop field boundaries from satellite imagery is

completed, followed by designing the point frame and selection of random

sample point. The next step entails using aerial survey sample points to capture

crop data and as the final step, statistical analysis is performed.

Crop yield estimation

The Crop Estimates Committee is responsible for the official crop forecasts and

estimates of summer and winter field crops for the country. The summer crops

for which estimates obtained are maize, sorghum, groundnuts, sunflower seed,

soya beans and dry beans. For the purposes of the Committee, white maize and

yellow maize are treated as two separate crops and then added together to

obtain total maize. For yield estimation, two kinds of surveys are done, namely

an area and farmer expected (subjective) yield survey and an objective yield

Page 94: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

86

survey. The subjective survey is conducted by randomly selecting a number of

points over the relevant provinces. The points where maize is located are then

used to select subsamples for the objective yield surveys. For the winter crops

survey, data are collected for wheat, malting barley, canola and sweet lupines.

The points where wheat is located are then used for the objective yield survey

subsample. The subsample is selected using a probability proportional to size

sampling design. The fields of the subsample are visited and two plots are

randomly chosen in the fields and the plots are laid out. Measurements are then

taken on the plots, the number of plants in a selected area is counted, the

number of ears and the number of seeds per ear is counted and the mass is

calculated. The counted plants and measurements are then analyzed for

estimation of crop yield.

8.11. SUDAN

Agriculture plays a very important role in the economy of the country as more

than 70 percent of the population is engaged directly or indirectly in this

activity. The General Directorate of Planning under the Ministry of Agriculture

in Sudan conducts the crop estimation surveys (Elsaied & Ahmed 2013).

Crop area estimation

For crop area estimation, a stratified two-stage sampling design with an uniform

sampling fraction (proportionate allocation) is adopted where the sheikh ships

(villages in Sudan) are the first stage unit of sampling and holdings growing

crops form the second stage unit of sampling. For the wheat crop, area statistics

are compiled through complete enumeration. Farmers’ eye estimation is also

used for estimation of the areas for some crops.

Crop yield estimation

Crop yield estimation is done using crop-cutting experiments. A stratified

multistage sampling design is adopted for estimating the yield rate of wheat and

sorghum. For -cutting experiments, plots of size 7m×6m are randomly

demarcated. The statistics division of the Agricultural Planning Administration

carries out an annual survey to estimate yield of the main crops, namely

sorghum, wheat, groundnuts and cotton. The blocks are considered as strata; the

first stage unit of sampling are tenancies (i.e. farm operator) and the ultimate

stage units are the plots of size 2m×2m. Then the crops are cut dried, threshed

and weighed, and the weighted average is calculated to estimate the yield of

crop.

Page 95: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

87

8.12. MOROCCO

The country’s agricultural survey program is based on area sampling methods

conducted by the Division of Statistics and Computer Science of the Directorate of

Planning and Economic Affairs, Ministry of Agriculture and Agricultural

Development. Some of the main crops are cereals, legumes, such as beans, peas and

lentils, olives, vegetables, sunflower and citrus.

Crop area estimation

In the early 1980’s, a stratified area and list sampling design was used for crop

area estimation (Bouzaffour & Hanuschak 1998).Stratification of the area frame

was carried out on a variety of cartographic materials depending on what was

available for an area when the work was being done. The stratification was also

carried out by using aerial photography and maps of various ages and, in some

instances, the frames were stratified using only maps. The stratification is now

based on recent photography and/or recent TM and SPOT images. The survey is

conducted in two phases. The first round takes place from 15 February to 15

April. It entails gathering data on autumn crops planting and on planting

intentions for the spring. The second round, which takes place in May and June,

involves gathering final planting data for the spring crops. The sample includes

about 70,000 farmers and provides estimates at the provincial level and at the level of

"special action zones". The area frame sample covers 90percent of the area surveyed,

the remaining 10 percent is covered by a village sample.

Crop yield estimation

The objective crop-cutting experiment and farmers eye estimation method is

used to obtain the crop yield estimates at different provincial levels by the

Director of Programming and Economic Affairs. The crop yield surveys are

carried out from May to September, with the exact date depending on the time of

maturity of each crop. Samples of the mature crops are harvested and sent to a

laboratory for threshing and assessing moisture content, among other things, to make

an objective estimate of yield. In addition, forecasting of crop yield is done using

agro-meteorological models based on remote sensing data, meteorological

inputs from ground stations and meteorological satellites, soil types and texture,

plant growth processes and measurements, and remotely sensed vegetative

index data.

8.13. RWANDA

The National Institute of Statistics of Rwanda is coordinating the National

Statistics System and is responsible for providing timely, accurate, and useful

statistics in all sectors of the country, including agricultural statistics. It has

Page 96: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

88

embarked on designing and implementing a new and improved system of

agricultural statistics.

Crop area estimation

The multiple frame survey design combines a probability sample of segments

selected from an area sampling frame, with a complementary list of large scale

farms to be completely enumerated. A stratified two-stage random sampling

design is adopted wherein PSUs are selected by a probability proportional to

size with replacement sampling design and within selected PSU segments of 10

hectares are randomly selected and completely enumerated. The area statistics

is collected through GPS. Personal digital assistant devices are also used for

data collection.

Crop yield estimation

For the purpose of estimation of crop yield, 25 percent of the already-selected

segments are used. The crop production estimates are obtained on the basis of

farmers eye estimate. The production estimates are divided by the crop area

harvested in order to obtain crop yield.

Both the crop area and yield estimates are obtained at the national level using

the appropriate sample sizes.

8.14. INDONESIA

Indonesia has a centralized national statistical system. BPS-Statistics Indonesia,

the main agency for generating agricultural statistics, is responsible for

conducting major censuses and surveys in the country and the dissemination of

all official statistics.

Crop area estimation

Crop area estimates are obtained using complete enumeration of all the sub

districts in the country. The data on planted area, harvested area, damage and

standing crop area in wet land and dry land are collected using eye estimates,

seed utilization, the irrigation block system and reports from farmers.

Crop yield estimation

A stratified multistage random sampling design is adopted wherein census

blocks are the first stage units, households in the selected census blocks are the

second stage units, fields are the third stage units and plots of specific size and

shape are the ultimate stage units. The crop- cutting experiment approach is

used for yield estimation.

Page 97: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

89

8.15. JAMAICA

The agricultural sector of Jamaica is one of the most important contributors to

the country’s GDP. The sector, with its diverse areas of specialization, has been

a major player in the country’s food supply, which ranges from domestic crop

production to high quality meat and milk supply to a wide cross-section of the

society.

Crop area estimation

Areas under cultivation are measured using measuring tapes or wheels on the

plains and eye estimation on sloped area.

Crop yield estimation

Crop yield is measured or estimated based on plant population by ascertaining

the planting distance for each crop.

Page 98: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

90

9 Conclusions The purpose of this report has been to define relevant concepts related to crop

area and crop yield estimation, in general, and in the context of mixed, repeated

and continuous cropping, in particular. In addition to presenting a critical

review of literature on this subject, country-experiences have also been

reported. While reviewing the literature, it has been found that there are several

issues and problems with regard to crop area and crop yield estimation.

A number of crop area estimation methods are available in practice, such as the

polygon method, triangulation/rectangulation and P2/A method. The polygon

method is accurate, but is costly and time-consuming. Thus, there is growing

demand for a cost-effective method for crop area estimation, especially in

developing and underdeveloped countries. Technologies, such as geographic

information systems, GPS and remote sensing are gaining importance for crop

area estimation. Area measurements using these technologies are, by and large,

more rapid, time efficient, digital and easy to incorporate into a database.

However, there are some concerns with regard to their accuracy and precision.

In the future, remote sensing may become an important alternative for crop area

estimation, but this method remains difficult to use in underdeveloped and

developing countries, where agriculture is primarily dominated by small plots,

different planting dates, scattered trees and intercropping systems.

The whole plot harvest method of crop yield estimation is almost bias-free, as

all sources of upward bias reported for crop cuts can be eliminated when the

entire field is harvested. However, it involves a large volume of work, making it

impractical for moderate-to-large sample sizes or multiple crop studies. Two

widely used methods in various countries are farmers' estimation and the crop-

cut method. Both methods have their own inherent pros and cons. The farmers’

estimation method is quick and cheap, but may result in poor quality data

because of intentional over/underreporting of crop production. The problem is

further aggravated by illiteracy among farmers and inappropriate timing of the

interview before/after the crop harvest.

Page 99: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

91

The crop-cut method, on the other hand, has been regarded as a reliable and

objective method for estimating crop yield. This method, however, may be

accompanied with an inherent upward resulting from measurement errors and

increased cost and time. Nevertheless, these deficiencies can be largely

overcome by appropriate training and supervision and by using optimum

sample sizes and auxiliary data available in the system. A strong advantage of

the crop-cut method is that the area of the cut is known and therefore errors are

not introduced into the final yield computation.

The choice of these two methods, however, is divided among countries. For

example, national statistical institutes in Kenya, Rwanda and Sweden prefer to

use farmer recall data to obtain production estimates, while Benin, India, Niger

and Zimbabwe opt for the crop-cut method. The United States Department of

Agriculture uses a combination of farmer recall for its agricultural census and

crop cuts for yield estimation of specific major crops in specific states. Several

European countries favor more expensive crop cuts for potatoes, but use

cheaper methodologies, such as farmer recall, expert assessment or purchase

records, for other crops.

Therefore, to estimate crop areas and yields, several region-specific issues, in

general, and the crop-specific issues, in particular, in the context of mixed,

repeated and continuous cropping need to be addressed. For crop area and yield

estimation, there is a need to conduct more studies in different countries to

establish supremacy of one method over another. For crop area and yield

estimation in the context of mixed, repeated and continuous cropping, relevant

issues are (i) non-availability of an updated sampling frame, (ii) determination

of optimum sample size relevant to the sampling design and estimation

procedure employed, (iii) choice of an appropriate sampling interval in

continuous cropping scenario, (iv) procedures for measurement of area and

yield under a particular crop, (v) procedure for apportioning of crop area when

temporary and permanent crops are grown together, (vi) suitable procedures for

capturing crop produce with extended harvest, (vii) appropriate schedules for

capturing such data, (viii) cost effective procedures for efficient collection and

processing of such data, (ix) procedures for reducing non-sampling errors for

good quality data and (x) developing user-friendly software for processing

survey data and analysis, including the sampling weights and computation of

standard errors of estimators so as to enable timely release.

Page 100: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

92

Annex I Calculation of the area of a polygon

The procedure as given in FAO 1982 is provided here. Let a polygon with n

sides be defined as

ai, αi i = 1, 2,…,n

where ai is the length of the side i and αi is the angle this side formed with

North measured in clockwise direction. Again let, a vector iar

represents the

side i in a two dimensional space XOY in which Y-axis coincides with the

North. Thus, the horizontal and vertical projections of the vector iar

(figure

A1.1) are ai sin αi and ai cos αi, respectively.

Figure A1.1. Horizontal and vertical projections of a vector

Define vectors

i

i jj 1

R a ,

r

i = 1, 2,…,n. (A1.1)

Their horizontal and vertical projections will be, respectively

i

i j jj 1

X a Sin a

(A1.2)

Page 101: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

93

i

i j jj 1

Y a Cosa .

(A1.3)

If the polygon is closed, then nR 0.r

The area of a triangle formed by two vectors, which start from the same point,

can be calculated as a function of their horizontal and vertical projections. Thus,

the area of the triangle between vectors 1Rr

and 2Rr

(figure A1.2) is given by

1 2 1 1 21

A X Y X Y2

Figure A1.2

It should be noted that this area will have a positive value if the vector 1Rr

precedes the vector 2Rr

while looking clockwise, otherwise it will have a

negative value. The area of the whole polygon, calculated as a sum of areas of

triangles, each formed by the two consecutive vectors iRr

, will be

n 2

i 1 i i i 2i 1

1A X Y X Y

2

(A1.4)

where Xi and Yi are given by Equation (A1.2) and (A1.3).

Page 102: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

94

Closure error and corrected area of a polygon

In practice, the polygon defined by the data which are collected in the field will

never close. In this case

nR 0.r

The-length of the vector nRr

2 2n n nR X Y

which can be used as a measure of error. However, a common practice is to

express the closure error as a percent of the perimeter of the polygon:

nn

ii 1

RC 100

a

(A1.5)

The errors are considered acceptable if the closing error is less than 2 percent.

There are different methods of closing the polygon (FAO 1982), including:

A. Closure by connecting the last but one point with the starting point;

B. Closure from the mid-point;

C. Closure by shifting all vertices on an equal basis;

D. Closure by shifting all vertices on a proportionate basis.

Page 103: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

95

Figure A1.3. Methods A, B and C

An advantage of the first three methods is that there is no need to keep in the

memory all input data until the end of the calculation. In these three methods,

each pair of input data can be elaborated when they are entered, and required

sums can be aggregated. The moment the last pair of data is entered, corrected

area and closure error can be evaluated. On the other hand, in the fourth

method, there is a need to keep in the memory all input data for one polygon

until the calculations are completed. It is important to note that the four

methods give an unbiased estimate provided measurement errors are absent.

Page 104: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

96

Annex II Steps involved in conducting crop-cutting experiments

The steps involved in conducting crop=cutting experiments are as follows:

a) Selection of field where the crop-cutting experiment is to be carried out;

b) Locating and marking the experimental plot of a given size and shape in

the selected field;

c) Harvesting the crop of the experimental plot;

d) Threshing of crop harvested from the experimental plot;

e) Winnowing of the threshed crop;

f) Weighing the produce obtained from the threshed crop;

g) Drying the produce, in case of excess moisture;

h) Weighing the dry produce.

Size and shape of the crop cutting experiment plot

During earlier attempts in crop surveys, greater attention was paid in deciding

the size of the crop-cutting experiment plot in a selected field. Various plot

sizes were tried varying from 1/160 of an acre for paddy in Orissa State, India

to 1/10 of an acre for cotton in Madhya Pradesh State, India. The plot size

adopted in the earlier attempts by Hubback (1946) and Mahalanobis (1939) was

very small, being of the order of 1/2000 of an acre. Attempts have been made

since 1944 to study the relative efficiency of various plot sizes for yield rates.

On the basis of detailed investigations, the size and shape of the crop cutting

experiment plot for various crops, in respect of different states, are specified.

The shapes of the cuts for various crops vary to some extent in different states.

In most of the states and for many crops, the plots are either square of size 5m ×

5m, 10m × 10m or rectangle of size 10m × 5m. In the Uttar Pradesh State,

India, the experimental plot is an equilateral triangle of side 10 m for most of

the crops and in West Bengal State, India, it is a circle with a radius of

approximately 1.7145m. For some crops, especially fruits, it consists of specific

number of trees.

Page 105: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

97

The plot size adopted for different food and non-food crops is listed in the table

below.

Name of the crop Shape Length (m) Breadth (m) Diagonal (m)

Paddy, wheat, sorghum, pearl

millet, maize, groundnut, tobacco,

sugarcane, green gram, chilly,

horse gram, black gram, chickpea,

sunflower

Square 5 5 7.07

Redgram, sesamum, caster, cotton Square 10 10 14.14

Selection of field

Field is a distinct piece of land where the crop is grown. It is clearly demarcated

on all its sides either by bunds or by patches of other crops or left uncultivated.

As per the existing methodology of estimation of yield rates of crops, two fields

of the crop are selected in each selected village and one experimental plot of the

crop is selected in each selected field (Sukhatme & Panse 1951). When

selecting two fields in each selected village, two random numbers are assigned

to the primary worker. The complete land of the selected village is divided into

fields. Each field has its own identification number called the survey number or

khasra number. The highest survey number in the selected village may be

greater than, equal to or less than the random number assigned for selection of

the field. If the assigned random number is equal to or less than the highest

survey number, the survey number corresponding to the random number is

selected and if it is greater than the highest survey number, the assigned random

number is divided by the highest survey number and the survey number

corresponding to the remainder is selected. In cases in which the remainder is 0,

the highest survey number is selected. If the crop is not grown in the selected

survey number, the next survey number is selected.

Page 106: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

98

If the selected survey number is further divided into subdivisions, only one

subdivision is selected randomly. In cases in which the selected

survey/subdivision number contains more than one field where the crop is

grown, the field nearest to the south-west corner of the survey/subdivision

number is selected. The selected field must satisfy the following conditions:

a) The area of the selected field should be more than the area of crop

cutting experiment plot, so that the crop cutting experiment plot of the

recommended size fits in the selected field.

b) If the selected field is sown with mixed crops, the experimental crop

must constitute at least 10 percent of its crop area.

c) The experimental crop in the field is not meant for prize competition or

seed production or demonstration.

d) The experimental crop is not grown for fodder purpose.

The field must be considered for conducting a crop cutting experiment and the

yield obtained from the cross cutting experiment plot must be recorded, if the

a) Experimental crop has not germinated or has failed but its area is

recorded by the village accountant;

b) The field where the experimental crop is being grown is being grazed

by cattle or damaged partially or completely by wild animals;

c) The experimental crop is affected by pests/diseases or heavy

rainfall/inadequate rainfall;

d) The yield must be recorded as zero in case the experimental crop is

completely damaged.

The field need not be considered for selection for conducting a crop cutting

experiment, if the experimental crop has

a) Not germinated or has failed and its area is not recorded by the village

accountant.

b) Withered or dried up and another crop has been raised in its place in the

same season and the area of the new crop has been recorded by the

village accountant.

c) Substitution of fields is not allowed on concerns over poor growth or

prior harvest by cultivators without intimation to the primary worker or

due to a late visit by the primary worker. Furthermore, if a part or all of

the selected field has been already harvested, the experiment should not

be conducted in that field, and it has to be treated as lost.

Page 107: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

99

Identification of south-west corner of the selected field

After the field is selected, the south-west corner of the field needs to be

identified. If one stands at the south-west corner facing north of the selected

field, the selected field will be in the front and at the right hand side of the

person. Fixing the south-west corner of the selected field has been made

mandatory to ensure similarity. It also helps the supervisor to identify the crop-

cutting experiment plot in the absence of an enumerator. In cases in which the

selected field is not exactly in north-south and east-west direction, the corner

that is approximately south-west may be taken as the south-west corner of the

selected field. The south-west corner of the -cutting experiment plot is

randomly located with reference to the south-west corner of the selected field.

Measurement of the length and breadth of the selected field

The method for of measuring the selected field differs for regular shaped and

irregular-shaped fields.

Regular shaped field: When the selected field is of a regular shape then the

longest side is measured as the length and the measurement of the shorter side

is the width in the steps from the south-west corner of the field (figure A2.1).

Figure A2.1. Field in regular shape

Irregular shaped field: In cases in which the selected field is irregular in

shape, the selected field needs to be enclosed in a regular shape by the outer

least possible dimensions for the purpose of locating the south-west corner of

the experimental plot in the selected field. The longest side should be measured

as the length and the shorter side measured as the breadth of the outer regular

shape of the irregular selected field in steps. The south-west corner of the

experimental plot should be fixed with reference to the south-west corner of the

outer regular shape of the irregular selected field (figure A2.2).

Page 108: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

100

Figure A2.2. Field in irregular shape

Determination of the random number pair

To ensure that the whole experimental plot fits in the selected field, seven steps

have to be deducted from the length and the breadth of the selected field (7

steps are equal to approximately 5 meter).

Example:

Length of the selected field measured in steps = 120 Steps

Number of steps to be deducted from length = 7 Steps

Number of steps in length after deducting 7 steps = 113 Steps

Breadth of the selected field measured in steps = 70 Steps

Number of steps to be deducted from breadth = 7 Steps

Number of steps in breadth after deducting 7 steps = 63 Steps

Two random numbers, one for length and the other for the breadth are selected.

These selected random numbers should be less than or equal to the number of

steps obtained after deducting 7 steps from the length and 7 steps from breadth

of the selected field. The random number is selected from the column of the

random number table assigned to the enumerator/primary worker for

determination of the south-west corner of the experimental plot.

Column number 1 of random number table should be assigned to the primary

worker. First, a random number for the length needs to be selected and then the

same thing needs to done for the breadth. In the above example, 113 steps are

obtained after deducting 7 steps from the length of a selected field. This

comprises three digits. Therefore, by referring to the three digitized random

Page 109: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

101

number table, a random number that is equal to or less than 113 is selected. By

referring to column 1 of the three digit random number table, the first random

number is 058. Therefore, random number 058 is selected for length. The

second random number is selected for breadth. After deducting 7 steps from the

breadth of the selected field, the remainder is 63 steps. Since, 63 comprises two

digits, by referring to the column number, one of the two digit random number

table, a random number which is equal to or less than 63 is selected. By

referring to column 1 of the two digit random number table, the first random

number is 51. Accordingly, random number 51 is selected for the breadth. The

pair (58, 51) is the pair of random numbers selected for locating the south-west

corner of the experimental plot in the selected field. If the assigned column of

the random number table is exhausted during the process of selecting the

random numbers, the next column on the right hand side is the reference

column. If all of or part of the experimental plot goes beyond the boundary of

the field, owing to the irregular shape of the field, the pair of random numbers

is rejected and a new pair of random numbers is selected until all of the

experimental plot is accommodated within the field.

Marking of the experimental plot

The selected random number for the length is 58. Therefore, it is necessary to

move 58 steps along the length of the field from the south-west corner of the

field and from the point reached by measuring 58 steps, and then moving 51

steps perpendicular to the length and parallel to breadth of the field. The point

reached is the south-west corner of the experimental plot shown as point “A” in

figure A2.3. The point “A” is also referred to as the key point of the

experimental plot. A peg at the key point of the experimental plot should be set.

Figure A2.3: Marking of experimental plot (Step-1)

Page 110: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

102

Five meters along the length of the field from corner “A” needs to measure to

reach the next corner, which is the second corner of the experimental plot say

corner “B”. At corner “B”, a peg should be fixed (figure A2.4). The line joining

the point “A” and “B” is the base of the experimental plot.

Figure A2.4: Marking of experimental plot (step-2)

To mark the third and fourth corner of the experimental plot, the right angle

triangle method should be applied. To mark the third corner, the first person

should stand at corner “A” while holding the measuring tape at the zero meter

mark on the measuring tape and second person should stand at corner “B” while

holding the same measuring tape at 12.07, namely the 7.07+5.0 meter mark.

The third person holding the measuring tape at 7.07 [square root of (52 + 52)]

meter mark should stretch the measuring tape in the inner side in the direction

of the breadth of the field. The point reached shall be the third corner of the

experimental plot say corner “C”. A peg should be fixed at corner “C” (figure

A2.5). Corner “C” is 5.0 meter away from corner “B” and 7.07 meter (diagonal)

from corner “A”.

Figure A2.5. Marking of experimental plot (step-3)

To locate the fourth corner of the experimental plot, the third person standing at

corner “C” should hold the measuring tape at 5.0 meter mark and the stretch it

Page 111: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

103

towards inner side in the direction of the breadth of the field, the point reached

is the fourth corner of the experimental plot referred to as corner “D”. A peg

should then be fixed at corner “D” (figure A2.6). The corner “D” is5.0mfrom

corner “A” and 7.07 m from corner “B”.

Figure A2.6. Marking of experimental plot (step-4)

A, B, C and D are the four corners of the experimental plot. The distance

between A and B, B and C, C and D, A and D should be checked. The distance

between A and B, B and C, C and D, A and D should be equal to 5.0 m. The

distance between both the diagonals AC and BD should also be checked. The

distance of each diagonal should be equal to 7.07 m (figure A2.7). It is

important that the pegs should be tall, straight and firmly fixed on the ground.

Figure A2.7. Marking of experimental plot (step-5)

Page 112: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

104

Harvesting of experimental plot

A well-stretched string should be tied around the pegs. The string should be

lowered gradually to the ground level. The position of the string on the ground

demarcates the boundary of the experimental plot. The decision about whether

the plants lie within the experimental plot is based on the position of their roots.

The plants on the boundary of the experimental plot should be harvested only if

the roots are more than half inside the experimental plot. Care should be taken

to collect all the harvested plants and no ear heads should be left in the

experimental plot. Farmers should be requested not to harvest the field until the

whole experimental plot is harvested and harvested crop is gathered and

brought to the threshing floor. The bundle of the harvested crop should be

marked/tagged to prevent it from be mixed with other crops.

Threshing and winnowing of the harvested experimental crop

A piece of Hessian cloth should be used for drying and threshing the

experimental crop. The harvested experimental crop should be spread on the

cloth at the threshing floor for drying and threshed carefully as per the usual

method. All the grains of the threshed experimental crop should be separated by

winnowing. The clean grains should be weighed at the nearest possible

weighing unit. After being weighed, the produce should be returned to the

farmer. If the produce has more moisture, a sample of the recommended

quantity of the produce has to be taken and kept in a cloth bag until the

moisture is dried up.

Driage

It is necessary to carry out drying experiments to obtain final estimate of yield

in terms of dry produce, the experiments for different crops should be

conducted at the district level by the district statistical officer. Crop-cutting

experiments supervised by the district statistical supervisor must be selected for

drying. The drying experiments are conducted in respect of 15 percent of the

experiments planned for the specific crops or subject to a minimum of four

experiments per crop. Generally, a one kilogram sample of harvested produce

should be taken at random for drying by the district statistical supervisor. If, the

produce obtained from the experimental plot is less than one kilogram, the

entire produce is to be taken. With regard to sugarcane, the final produce should

be expressed in terms of cane only. In the case of cotton, the final produce

should be expressed in terms of lint. The cotton is converted into lint by using

ginning percentage (cotton to lint), which is obtained from the ginning

factories. A required sample of the produce is taken in a small bag and kept for

drying by the usual method for a specified period. The dry weight should be

Page 113: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

105

taken at the nearest possible weighing unit after the moisture of the produce has

been completely eliminated.

Page 114: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

106

References

Abalu, G. &D'Silva, B. 1979.Socioeconomic aspects of existing farming

systems and practices in northern Nigeria. Paper presented at the International

Workshop onSocio-economic Constraints to Development of Semi-Arid

Tropical Agriculture, Hyderabad, India, International Crop Research Institute of

Semi-Arid Tropical Agriculture ( ICRISAT).

Aditya, K., Sud, U.C. and Chandra, H. 2014. Estimation of domain mean

using two stage sampling with sub-sampling of non-respondents. Journal of the

Indian Society of Agricultural Statistics, 68(1: 39-54.

Ahmad, T. & Kathuria, O.P. 2010.Estimation of crop yield at block level.

Advances in Applied Research, 2(2):164-172.

Ahmad, T., Kathuria, O.P. &Rai, A.2004.Comparative study of farmer’s eye

estimate and crop cut estimate for general crop estimation surveys. Annals of

Agricultural Research, 25(3):394-397.

Ajayi, O.C. &Waibel, H. 2000. How accurate are farm size estimates obtained

from smallholder farmers in West Africa”? Lessons from Côte d’Ivoire. In

Farmers and Scientists in a Changing Environment: Assessing Research in

West Africa. Margraf Verlag Publishers:, Weikersheim, Germany, 543-550.

Ali, N., Kahn, K., Patel, P. &Gorelick, J. 2009. Moving the learning forward:

From incremental to transformational impact on empowering smallholder

farmers and women. Impact Assessment of the Agricultural Marketing

Initiative, West Nile, Uganda. Capstone Report George Washington University,

Washington, DC.

Alvim, R. &Nair, P.K.R.1986. Combination of cacao with other plantation

crops: An agroforestry system in Southeast Bahia, Brazil. Agroforestry Systems,

4(1): 3-15.

Badhwar, G.D.1984. Automatic corn-soybean classification using Landsat

MSS data. I. Near-harvest crop proportion estimation. Remote Sensing of

Environment,14:15-29.

Battese, G.E., Harter, R.M& Fuller, W.A. 1988.An error component model

for prediction of county crop areas using survey and satellite data. Journal of

the American Statistical Association, 83: 28-36.

Page 115: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

107

Bauer, M.E., J.E. Cipra, P. E. Anuta &J.A. Etheridge. 1979. Identification

and area estimation of agricultural crops by computer classification of Landsat

MSS data. Remote Sensing of Environment,8: 75-92.

Belward, A.S. &de Hoyos, A.1987.A comparison of supervised maximum-

likelihood and decision tree classification for crop cover estimation from multi-

temporal Landsat MSS data. International Journal of Remote Sensing,8: 229-

235.

Bettio, M., Delincé, J., Bruyas, P., Croi, W. & Eiden. G. 2002. Area frame

surveys: aim, principals and operational surveys. In Building Agro

Environmental Indicators: focusing on the European area frame survey

LUCAS, Bettiopp.12-28. European Communities: Rome.

Bizzell, R.M., Hall, F.G., Feiveson, A.H., Bauer, M.E., Davis, B.J., Manila,

W.A. & Rice, P.C. 1975. Results from the crop identification technology

assessment for remote sensing (CITARS) projects. In proceedings of the Tenth

International Symposium on Remote Sensing of Environment,6-10 October,

195, Environmental, Ann Arbor, Michigan, United States.

Bogaert P., Delincé J. &Kay S. 2005.Assessing the error of polygonal area

measurements: a general formulation with applications to agriculture.

Measurement Science and Technology, 16: 1170–1178.

Bosecker, R.R. 1988. Sampling methods in agriculture. National Agricultural

Statistical Service, U.S. Department of Agriculture.

Bouzaffour, S. &Hanuschak, G. 1998. Crop monitoring in Morocco using

remote sensing and geographic information systems: From a management

perspective. International Conference on Agricultural Statistics, Agricultural

Statistics 2000, Washington D.C., 18-20 March 1998.

Bradbury, D. 1994.Cereals in Europe-Statistical systems for measuring area,

production, and yield. European Commission, Luxembourg: Eurostat. 1996a.

Cereals in Europe: A supplement - Statistical systems for measuring area,

production, and yield in Austria, Finland, Sweden and Norway. Luxembourg:

Eurostat.

1996b. Crops in Europe-Statistical systems for measuring area, production, and

yield of non-cereal crops. Luxembourg: Eurostat.

Brazil, Ministry of Agricultural, Livestock, and Supply,2013. Brazilian crop

Assessment: Grain. Companhia Nacional de Abastecimento –Conab.

Page 116: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

108

Breidt, F. &Fuller, W. 1999. Design of supplemented panel surveys with

application to the National Resources Inventory. Journal of Agricultural,

Biological and Environmental Statistics, 4(4): 391-403.

Bulgaria, Ministry of Agriculture and Forestry, 2001.“Employment and use

the territory of BULGARIA in 2001results. BANSIK 2001.

Campbell, N.A., De Boer, E.S. &Hick, P.T.1987. Some observations on crop

profile modeling. International Journal of Remote Sensing,8: 193-201.

2007. A comparison of area frame sample designs for agricultural statistics,. In

Bulletin of the International Statistical Institute, the 56thsession proceedings,

Lisbon, 22-29 August 2007.

Carfagna, E. & Gallego, F.J. 2005.Using remote sensing for agricultural

statistics”. International Statistical Review, 73(3), 389-404.

Carfagna, E. & Keita, N. 2009.Use of modern geo-positioning devices in

agricultural censuses and surveys. Bulletin of the International Statistical

Institute, the 57th

Session, Proceedings, Special Topics Contributed Paper

Meetings (STCPM22) organized by Naman Keita (FAO), Durban, August 16-

22, 2009.

Carletto, C., Deininger, K., Muwonge, J. & Savastano, S. 2010.Using diaries

to improve production statistics: Evidence from Uganda. Paper presented at the

fifth International Conference on Agricultural Statistics, Kampala, 13-15 2010,

Kampala.

Casley, D.J. &Kumar, K. 1988. The collection, analysis and use of monitoring

and evaluation data. Baltimore, Maryland, United States: Johns Hopkins

University Press for the World Bank.

Chambers, R. &Tzavidis, N. 2006. M-quantile models for small area

estimation. Biometrika, 93(2): 255-268.

Chandra, H. 2013. Exploring spatial dependence in area level random effect

model for disaggregate level crop yield estimation. Journal of Applied

Statistics,40(4): 823-842.

Chandra, H. &Chambers, R. 2011. Small area estimation under

transformation to linearity. Survey Methodology, 37(1): 39-51.

Chandra H., Salvati N. & Chambers R. 2007. Small area estimation for

spatially correlated populations. A comparison of direct and indirect model-

based methods. Statistics in Transition, 8: 887-906.

Page 117: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

109

Chandra, H., Salvati, N., Chambers, R. &Tzavidis, N. 2012. Small area

estimation under spatial non stationarity. Computational Statistics and Data

Analysis, 56(10): 2875-2888.

Craig, M. &Atkinson, D. 2013. A literature review of crop area estimation.

Technical report for FAO.

Das S.K. &Singh R. 2013. A multiple-frame approach to crop yield estimation

from satellite- remotely sensed data. International Journal of Remote Sensing,

34(11), 3803-3819.

Datta, B., Day, B.&Maiti, T. 1998. Multivariate Bayesian small area

estimation: an application to survey and satellite data”. Sankhya: The Indian

Journal of Statistics, Series A, 69(3) 1-19.

David, I.P. 1978.Non-sampling errors in agricultural surveys: Review, current

findings, and suggestions for future research. Presented at the Philippine

Statistical Association Annual Conference, 18 June 1978, Manila.

Davies, C. 2009.Area frame design for agricultural surveys. RDD Research

Report, Research and Development Division, United States Department of

Agriculture, - Fairfax, Virginia, United States.

De Groote, H. &Traoré, O. 2005.The cost of accuracy in crop area estimation.

Agricultural Systems, 84(1): 21–38.

Delincé, J. 2001. A European approach to area frame survey. Proceedings of

the Conference on Agricultural and Environmental Statistical Applications in

Rome (CAESAR)2: 463–472.

Diskin, P. 1997. Agricultural productivity indicators measurement guide .Food

and nutrition technical assistance project, Washington, DC: US Agency for

International Development.

Dorota, B. 2006. Attempts at applying small area estimation methods in

agricultural sample surveys in Poland. Statistics in Transition, 7(6): 1203-1218.

Department of Primary Industries (DPI). 2010. Estimating crop yields and

crop losses. Melbourne, Australia: DPI.

Elsaied, M. &Ahmed, N.M.E. 2013. Status of agricultural statistics in Sudan.

Proceedings of the International Conference on Agricultural Statistics.

Ethiopia, Central Statistical Agency (CSA), 2011. Statistical report on area

and production of crops, and farm management practices. Volume VIII.

Erenstein, O., Malik, R.K. &Singh, S. 2007. Adoption and impact of zero-

tillage in the rice-wheat zone of irrigated Haryana, India. New Delhi:

Page 118: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

110

International Centre for Maize and Wheat Improvement (CIMMYT) and the

Rice-Wheat Consortium for the Indo-Gangetic Plains.

FAO Statistics. 2011. Crops statistics: concepts, definitions and classifications.

Available at: www.fao.org/economic/the-statistics-division-

ess/methodology/methodology-systems/crops-statistics-concepts-definitions-

and-classifications/en. Accessed on 10 September, 2015.

Fay, R.E. &Herriot, R.A. 1979. Estimation of income from small places: An

application of James-Stein procedures to census data. Journal of the American

Statistical Association, 74(366): 269-277.

Fawzy, M.A., Pope, L., &Amer, Y.M.E.G. 1998. Availability and quality of

agricultural data in Egypt. Impact Assessment No. 4, Monitoring, verification

and evaluation unit Agricultural policy reform program, Ministry of Agriculture

and Land Reclamation, Egypt.

Ferreira, S. L., Newby,T. &Preez, E. 2006. Use of remote sensing in support

of crop area estimates in South Africa. Compilation of ISPRS WG VIII/10

workshop proceedings, Remote Sensing Support to Crop Yield Forecast and

Area Estimates, Stresa, Italy, 30 Novermber-1 December 2006.

Fermont, A.M., Asten, P.J.A., Tittonell, P., Wijk, M.T. &Giller, K.E. 2009.

Closing the cassava yield gap: an analysis from small-holder farms in East

Africa. Field Crops Research, 112(1): 24–36.

Fermont, A.M. &Benson, T. 2011.Estimating yield of food crops grown by

smallholder farmers-A Review in the Uganda context. Development Strategy

and Governance Division, International Food Policy Research Institute (IFPRI)

Discussion Strategy. Washington D.C.: IFPR.

Fielding, W.J. &Riley. J. 1997. How big should on-farm trials be and how

many plots should be measured”? PLA Notes 29:19–22. London: International

Institute of Economic Development.

Food and Agriculture Organization of the United Nations (FAO). 1982.

Estimation of crop areas and yields in agricultural statistics. FAO economic and

social development paper No. 22. Rome: FAO.

2004. Remote sensing and land cover area estimation. International Journal of

Remote Sensing, 25(15): 3019-3047.

2006. Review of the main remote sensing methods for crop area estimates”.

ISPRS Archives XXXVI workshop proceedings: Remote Sensing Support to

Crop Yield Forecast and Area Estimates.

Page 119: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

111

2007. Sampling efficiency of the EU point survey LUCAS 2006,Proceeding of

the 56th

session of the ISI, 22-29 August 2007, Lisbon.

Gonzalez-Alonso, F. & Cuevas, J.M. 1993. Remote sensing and agricultural

statistics: crop area estimation through regression estimators and confusion

matrices. International Journal of Remote Sensing,14(6): 1215-1219.

Gonzalez-Alonso, F., Soria, S.L. & Cuevas-Gozalo, J.M. 1994. Comparing

two methodologies for crop area estimation in Spain using Landsat TM images

and ground gathered data. Remote Sensing of Environment, 35(1): 29-35.

Gonzalez, M.E. 1973.Use and evaluation of synthetic estimators. Proceedings

of the Social Statistics Section, American Statistical Association.

Goodman, B. & Monks, C.D. 2003. “A farm demonstration method for

estimating cotton yield in the field for use by extension agents and specialists”.

Journal of Extension, 41(6).

Goswami, S.B., Saxena, A. & Bairagi, G.D. 2012. Remote sensing and GIS

based wheat crop acreage estimation in Indore district, M.P. International

Journal of Emerging Technology and Advanced Engineering,2(3), 200-203.

Groten, S.M.E. 1993. NDVI—Crop monitoring and early yield assessment of

Burkina Faso. International Journal of Remote Sensing. 14(8): 1495-1515.

Haack, B. & Jampoler, S. 1995. Color composite comparisons for agricultural

assessments. International Journal of Remote Sensing.,16(9) 1589-1598.

Hagblad, L. 1998. Crop cutting versus farmer reports–review of Swedish

findings. Statistik Rapport 1998:2, Örebro: Statistics Sweden.

Hallum, C.R. &Perry Jr., C.R. 1984. Estimating optimal sampling unit sizes

for satellite surveys. Remote Sensing of Environment, 14(1-3): 183-196.

Hay, A.M. 1988. The derivation of global estimates from a confusion matrix.

International Journal of Remote Sensing, 9(8):1395-1398.

Hay, A.M. 1989. Global estimates from confusion matrices: a reply to Jupp.

International Journal of Remote Sensing, 10(9): 1571-1573.

Henderson, C. R. 1975. “Best linear unbiased estimation and prediction under

a selection model. Biometric, 31 (2): 423-447.

Henriet, J.,van Ek, G.A., Blade, S.F &Singh, B.B. (1997). “Quantitative

Savannah of Northern Nigeria: Rapid survey of prevalent Cropping System”.

Samarau Journal of Agriculture Research, 14: 37-45.

Page 120: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

112

Hixson, M.M., Bauer, ME & D.K. Scholz. 1980. An assessment of Landsat

data acquisition history on identification and aerial estimation of corn and

soybeans.

Hopkins, J., &Berry, P. 1994. Determinants of land and labor productivity in

crop production in Niger. Report to the United States Agency for International

Development, Niamey, Niger (November), Washington, D.C.: International

Food Policy Research Institute.

Howard, J., Said, A., Molla, D., Diskin, P. & Bogale, S. 1995. Toward

increased domestic cereals production in Ethiopia: using a commodity systems

approach to evaluate strategic constraints and opportunities”. East Lansing,

Michigan, U.S.: Michigan State University.

Hubback, J.A. 1946. Sampling for rice yields in Bihar and Orissa. Sankhya:

The Indian Journal of Statistics,7(3): 281–294.

India, Ministry of Statistics and Program Implementation (MoS & PI).

2008. Manual for area and crop production statistics.

Janssen, L.L.F. & Middelkoop, H. 1992. Knowledge-based crop classification

of a Landsat Thematic Mapper image. International Journal of Remote

Sensing,13(15): 2827-2837.

Jinguji, I.2014. Dot sampling method for area sampling.

Jupp, D.L.B. 1989. The stability of global estimates from confusion matrices.

International Journal of Remote Sensing, 10(9): 1563-1569.

Just, R. 1981. The productivity of mixed cropping practice: A study of survey

data from Northern Nigeria. Parts I and II: West Africa projects. Prepared for

World Bank, West Africa Projects. Washington, D.C.: World Bank.

Just, R. &Candler, W. 1985. Production functions and rationality of mixed

cropping. European Review of Agricultural Economics, 12(2): 207-331.

Kathuria, O.P. 1995. A report of area and yield estimation surveys in Zambia.

Kathuria, O.P. 1998.Africa-A fertile ground for agricultural statistics research.

Journal of the Indian Society of Agricultural Statistics, 51(2&3): 303-314.

Keita, N. 2003.Root and tuber crops: concepts and methods recommended by

FAO and operational issues. Paper presented at the Expert Consultation on Root

Crop Statistics, 3-6 December, Harare.

Keita, N., Carfagna, E. & Mu’Ammar, G. 2009. Issues and guidelines for the

emerging use of GPS and PDAs in agricultural statistics in developing

Page 121: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

113

countries. The Fifth International Conference on Agricultural Statistics (ICAS

V), Kampala,12-15 October 2010.

Kelly, V., Hopkins, J., Reardon, T. & Crawford, E. 1995. Improving the

measurement and analysis of African agricultural productivity: Promoting

complementarities between micro and macro data. International Development

Paper No. 16, East Lansing, Michigan, USA: Michigan State University.

Knörzer, H, Graeff-Hönninger, S., Claupein, W. 2009. Developing an

improved model for simulating a relay intercropping system of wheat and

maize. Paper prepared for 52. Jahrestagung der fur Pflanzebauwissenschaften

Halle, Germany, 1-3 September.

Lekasi, J.K., Tanner, J.C., Kimani, S.K. &Harris, J.P.C. 2001.Managing

manure to sustain smallholder livelihoods in the East African highlands for high

potential production systems of the National Resources Systems Programme

Renewable Natural Resources Knowledge Strategy. Department of

International Development, HDRA publications.

Li Q.Z. & Wu B.F. 2004. Crop proportion monitoring precise assessment

Chinese Journal of Remote Sensing,8: 581-587.

MacDonald, R.B. & Hall, F.G. 1980. Global crop forecasting. Science, 208:

670-679.

Machado, S. 2009. Does intercropping have a role in modern agriculture?

Journal of Soil and Water Conservation, 64(2): 55A.

Mahalanobis, P.C. 1939.A sample survey of the acreage under jute in Bengal.

Sankhya: The Journal of Statistics,4 (4): 511-531.

Mahalanobis, P.C. 1944. On large-scale sample surveys. Philosophical

Transactions of the Royal Society.31(584): 329-451.

Martinez, L.I., Flores, A,L., Rodriquez, R.G., &de La Vega Panizo,

R.2015.Technical report on improving the use of GPS, GIS and remote sensing

in setting up master sampling frames. Technical Report Series GO-06-2015.

Martino, L. 2003. The Agrit system for short-term estimates in agriculture: a

project for 2004. In DRAGON seminar, Krakov, Poland, July, 2003, pp.9-11.

Maselli, F., Conese, C., Zipoli, G. & Pittau, M.A. 1990. Use of error

probabilities to improve area estimates based on maximum likelihood

classifications. Remote Sensing of Environment,31(2): 155-160.

Mathur, D.C.,Sud, U.C., Bathla, H.V.L. &Sharma, D.P. 2006. Comparison

of crop cut and farmers' estimates. Indian Journal of Agriculture Sciences,

73(2):628-29.

Page 122: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

114

Mauser, W. 1989.Agricultural land-use classification in the upper Rhine valley

using multi-temporal TM data. Proceedings on Earthnet Pilot Project on

Landsat TM Application (ESA-SP - 1102). Frascati, Italy, p.191-198.

Ministry of Agriculture and Co-operatives (MAC). 1965. “Report on

Uganda census of agriculture, Vol. I”. Entebbe, Uganda.

Ministry of Agriculture and Food, Bulgaria. 2013. “CROP - crop

production”.

Available at:

http://www.mzh.government.bg/MZH/bg/ShortLinks/SelskaPolitika/Agrostatist

ics /Crop/Methodology_copy3.aspx

Ministry of Agriculture, Animal Industries, and Fisheries, Uganda

(MAAIF). 1992. “Report on Uganda National Census of Agriculture and

Livestock (1990-1991), Vol. III”.Crop Areas, Yields and Production, Entebbe,

Uganda.

Ministry of Statistics and Program Implementation, India (MoS&PI).

2008. “Manual for crop area and production statistics”. Ministry of Statistics

and Program Implementation, New Delhi, India.

Minot, N. 2008. Notes on Ethiopian cereal yield estimates. International Food

Policy Research Institute, Addis Ababa, Ethiopia.

Moriera, M.A., Chen, S.C. and Batista, G.T. 1986. Wheat-area estimation

using digital Landsat MSS data and aerial photographs. International Journal of

Remote Sensing, 7(9): 1109-1120.

Mortensson, B., Ländell, G. & Wahlstedt, G. 2004. Crop production

statistics in Albania, Bosnia-Herzegovina, Croatia, Cyprus, Macedonia and

Malta. Orebro: Statistics Sweden.

Mpyisi, E. 2002. Estimation of area and production of root and tuber crops in

Rwanda”. Paper presented at FAO Expert Consultation on Root Crop Statistics,

Harare, 2-6 December 2002.

Murphy, J., Casley, D.J. & Curry, J.J. 1991. Farmers’ estimations as a

source of production data. World Bank Technical Paper 132. Washington, DC:

World Bank.

Muwanga-Zake, E.S.K. 1985. Sources of possible errors and biases in

agricultural statistics in Uganda: A review. Kampala: Institute of Statistics and

Applied Economics, Makerere University.

Mwebaze, S.M.N. 1999. Country pasture/forage resources — Uganda. FAO:

Rome.

Page 123: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

115

National Health Interview Survey (1968). Synthetic State Estimates of

Disability. PHS Publication No. 1759. Public Health Service, Washington: U.S.

Government Printing Office.

National Sample Census of Agriculture Small Holder Agriculture

(NSCASHA). 2012. Volume II: Crop Sector - National Report. The National

Bureau of Statistics and the Office of the Chief Government Statistician,

Zanzibar.

Nigeria, National Bureau of Statistics (NBS). 2007. 2006 Annual

collaborative survey of socio-economic activities in Nigeria”. Volume I,

Federal Republic of Nigeria.

Norman, D. 1973a. Crop mixtures under indigenous conditions in the northern

part of Nigeria. In Factors Affecting Growth in Western Africa, Ofevi, I.M. (ed)

Accra: University of Ghana.

Norman, D.W., Simmons, E.B. & Hays, H.M. 1982. Farming Systems in the

Nigerian Savana: Research and Strategies for Development. Westview Press:

Boulder, Colorado, United States.

Norman, D.W., Worman, F.D., Siebert, J.D. & Modiakgotla, E. 1995.The

Farming Systems Approach to Development and Appropriate Technology

Generation”. Rome: FAO.

Panse, V.G.1947.Plot size in yield surveys.Nature, 159: 820. 1954. Estimation

of Crop Yields. Rome: FAO.

Panse, V.G., Rajagopalan, M. &Pillai, S. 1966. “Estimation of crop yields for

small areas”. Biometrics, 22(2):374-388.

Parihar J.S. &Oza M.P. 2006. FASAL: an integrated approach for crop

assessment and production forecasting”. Agriculture and Hydrology

Applications of Remote Sensing. Asia-Pacific Remote Sensing Symposium.

International Society for Optics and Photonics.

Pfeffermann, D. 2002. Small area estimation: new developments and

directions”. International Statistical Review, 70(1):125-143.

Pfeffermann, D., Feder, M., and Signorelli, D. (1998). Estimation of

autocorrelations of survey errors with application to trend estimation in small

areas. Journal of Business and Economic Statistics, 16, 339-348.

Poate, C.D. 1988. A review of methods for measuring crop production from

smallholder producers. Experimental Agriculture, 24(1):1–14.

Page 124: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

116

Poate, C.D. &Casley, D.J. 1985. Estimating crop production in development

projects: methods and their limitations”. Washington, D.C.: World Bank.

Rao, J.N.K. 2003. Small Area Estimation. New York: John Wiley& Sons.

Rao, J.N.K., &Yu, M. 1994. Small area estimation by combining time series

and cross-sectional data. Canadian Journal of Statistics, 22(4): 51 1-528.

Reynolds, C.A., Yitayew, M., Slack, D.C., Hutchison, C.F., Huele, A.

&Petersen, M.S. 2000. Estimating crop yields and production by integrating

FAO crop specific water balance model with real-time satellite data and

ground-based auxiliary data. International Journal of Remote Sensing, 21(18)

3487–3508.

Rogers, C.E. &Murfield, D.E. 1965. Validation of objective method of

estimating soybean yield. Agricultural Economics Research, 17 (3): 90–91.

Rozelle, S. 1991. Rural household data collection in developing countries:

designing instruments and methods for collecting farm production data.

Working Paper in Agricultural Economics 91-17. Ithaca, NY: Cornell

University.

Ryerson, R.A., Dobbins, R.N. &Thibault, C. 1985. Timely crop area

estimates from Landsat. Photogrammetric Engineering and Remote Sensing,

51(11): 1735-1743.

Sahoo, P.M., Ahmad, T., Singh, K.N. & Gupta, A.K. 2013.Study to develop

methodology for crop acreage estimation under cloud cover in the satellite

imageries. Project report, Indian Agricultural Statistics Research Institute, New

Delhi Publication.

Sahoo, P.M., Rai, A., Ahmad, T., Singh, R. &Handique B.K. 2012.

Estimation of acreage under paddy crop in Jaintia Hills district of Meghalaya

using remote sensing and GIS. International Journal of Agricultural and

Statistical Sciences, 8(1): 193-202.

Sahoo, P.M., Rai, A., Bathla, H.V.L., Ahmad, T. &Farooqui, S. 2010.

Developing remote sensing based methodology for collection of agricultural

statistics in North East hilly region. Project report, Indian Agricultural Statistics

Research Institute, New Delhi Publication.

Sahoo, P.M., Rai, A., Singh, R., Handique, B.K. and Rao, C.S.

2005Integrated approach based on remote sensing and GIS for estimation of

area under paddy crop in North-Eastern hilly region. Journal of the Indian

Society of Agricultural Statistics, 59(2): 151-160.

Page 125: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

117

Sawasawa, H.L.A. 2003. Crop yield estimation: Integrating RS, GIS and

management factors: A case study of Birkoor and Kortgirimandals, Nizamabad

District, India., ITC–Faculty of Geo-Information Science and Earth

Observation of the University of Twente: Enschede, the Netherlands (M.Sc.

Thesis).

Schøning, P., Apuuli, J.B.M., Menyha, E. &Muwanga-Zake, E.S.K. 2005.

Handheld GPS equipment for agricultural statistic surveys: experiments on area

measurements done during fieldwork for the Uganda Pilot Census of

Agriculture, 2003. Statistics Norway.

Sempungu, A. 2010. Comparison of the crop card and farmer recall method for

cassava and sweet potato yields. Makerere University, Kampala. (M.Sc.

Thesis).

Sharma, S.D., Srivastava, A.K. & Sud, U.C. 2004. Small area crop estimation

methodology for crop yield estimates at gram panchayat level. Journal of the

Indian Society of Agricultural Statistics,5 7(special volume): 26-37

Singh, R. 2003.Use of satellite data and farmers eye estimate for crop yield

modeling. Journal of Indian Society Agricultural Statistics, 56(2):166-176.

Singh, B.B., Shukla, G.K. & Kundu, D. 2005. Spatial-temporal models in

small area estimation. Survey Methodology, 31(2): 183-195.

Sisodia, B.V.S. & Chandra, H. 2012. Estimation of crop production at smaller

geographical level in India. Journal of the Indian Society of Agricultural

Statistics, 66(2): 313-319.

Smale, M., Diatkite, L., Sidibe, A., Grum, M., Jones, H., Traore, I.

&Guindo, H. 2010.The impact of participation in diversity field fora on farmer

management of millet and sorghum varieties in Mali. African Journal of

Agricultural Resource Economics, 4(1): 23-47.

Spencer, D.S.C. 1989.Micro-level farm management and production

economics research among traditional African farmers: lessons from Sierra

Leone. African Rural Employment Paper No. 3, East Lansing, Michigan,

United States: Department of Agricultural Economics, Michigan State

University.

Srivastava, A.K., Ahuja, D.L., Bhatia, D.K. &Mathur, D.C. 1999. Small

area estimation techniques in agriculture. In Souvenir in Advances in

Agricultural Statistics and Computer Application (pp. 32-45), Indian

Agricultural Statistics Research Institute, New Delhi,.

Page 126: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

118

Ssekiboobo, A.M. 2007. Practical problems in the estimation of performance

indicators for the agricultural sector in Uganda. Paper presented at the Fourth

International Conference on Agricultural Statistics, 22-24 October2007,

Beijing.

Statistics Canada, 2015. Field Crop Reporting Series, Available at:

www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3401&lang=

en&db=imdb&adm=8&dis=2. Accessed 10 September 2015.

Sud, U.C., Chandra, H. &Srivastava, A.K. 2012.Crop yield estimation at

district level using Improvement of Crop Statistics Scheme data — an

application of small area estimation technique. Journal of the Indian Society of

Agricultural Statistics, 66(2): 321-326.

Sud, U.C., Mathur, D.C., Srivastava, A.K., Bathla, H.V.L. & Jha, G.K.

2001. Crop yield estimation at small area level using farmers' estimates. Project

report, Indian Agricultural Statistics Research Institute, New Delhi.

Sukhatme, P.V. 1946a. Bias in the use of small size plots in sample surveys for

yield. Current Science, 15(5): 119-120.

1946b. Bias in the use of small size plots in sample surveys for yield. Nature,

157(1946): 630.

Sukhatme, P.V. &Panse V.G. 1951. Crop surveys in India II. Journal of the

Indian Society of Agricultural Statistics, 3(1-2): 97-168.

Tittonell, P. Vanlauwe, B., Leffelaar, P.A. &Giller, K.E.,2005.Estimating

yields of tropical maize genotypes from non-destructive, on-farm plant

morphological measurements. Agriculture, Ecosystem, and Environment,

105(1-2): 213–220.

Uganda Bureau of Statistics (UBOS). 2002.Uganda national household

survey 1999/2000: Report on the Crop Survey Module. Uganda Bureau of

Statistics, Kampala.

Uganda, Ministry of Agriculture, Animal Industries, and Fisheries,

(MAAIF). 1992. Report on Uganda National Census of Agriculture and

Livestock (1990-1991), Vol. III. Entebbe, Republic of Uganda, Ministry of

Agriculture, Animal Industry and Fisheries.

Uganda, Ministry of Agriculture and Co-operatives (MAC). 1965. Report

on Uganda census of agriculture, Vol. I.

Vandermeer, J.H. 1989. The Ecology of Intercropping”, New York:

Cambridge University press, 6-7.

Page 127: Research on Improving Methods for Estimating Crop …gsars.org/.../WP_Synthesis-of...for-Estimation-of-Crop-Area-190116.pdffor Estimating Crop Area, Yield and Production under Mixed,

119

Verma, V., Marchant, T. &Scott, C. 1988.Evaluation of Crop-cut Methods

and Farmer Reports for Estimating Crop Production: Results of a

Methodological Study in Five African Countries. London: Longacre

Agricultural Development Centre Ltd.

Vibhute, A.D., Nagne, A.D., Gawali, B.W., &Mehrotra, S.C. 2013.

Comparative analysis of different supervised classification techniques for

spatial land use/land cover pattern mapping using RS and GIS. International

Journal of Scientific and Engineering Research, 4(7).

Wairegi, L.W.I., van Asten, P.J.A., Tenywa, M.& Bekunda, M. 2009.

Quantifying bunch weights of the East African highland bananas (Musa spp.

AAA-EA) using non-destructive field observations”. Scientia Horticulturae,

121(1): 63–72.

World Bank. 2010. “Global strategy to improve agricultural and rural

statistics”. Report No. 56719-GLB. Washington D.C.: Bank for Reconstruction

and Development/World Bank.

Wu, B. and Li, Q. 2004. Crop area estimation using remote sensing on two-

stage stratified sampling. Proceedings of the XXth

Congress of International

Society of Photogrammetry and Remote Sensing (ISPRS), Technical

Commission, Istanbul, Turkey, 22-23 July 12-23 2004 2004.

Zhao, J., Shi, K. &Wei, F. 2007. “Research and application of remote sensing

techniques in Chinese agricultural statistics.” Paper presented at the Fourth

International Conference on Agricultural Statistics, 22-24 October 2007,

Beijing.