study on the impacts of climate change on china’s ... · climate impacts on production and land...

STUDY ON THE IMPACTS OF CLIMATE CHANGEON CHINA’S AGRICULTURE �

HUI LIU 1, XIUBIN LI 1, GUENTHER FISCHER 2 and LAIXIANG SUN 2

1Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing, 100101, ChinaE-mail: [email protected]

2International Institute for Applied System Analysis, A-2361 Laxenburg, Austria

Abstract. This paper measures the economic impacts of climate change on China’s agriculturebased on the Ricardian model. By using county-level cross-sectional data on agricultural net revenue,climate, and other economic and geographical data for 1275 agriculture dominated counties, we findthat under most climate change scenarios both higher temperature and more precipitation wouldhave an overall positive impact on China’s agriculture. However, the impacts vary seasonally andregionally. Autumn effect is the most positive, but spring effect is the most negative. Applying themodel to five climate scenarios in the year 2050 shows that the East, the Central part, the South, thenorthern part of the Northeast, and the Plateau would benefit from climate change, but the Southwest,the Northwest and the southern part of the Northeast may be negatively affected. In the North, mostscenarios show that they may benefit from climate change. In summary, all of China would benefitfrom climate change in most scenarios.

1. Introduction

Since recognition of potential climate change, efforts have been underway to esti-mate the economic impacts of projected changes in climate on important sectors,such as agriculture, forestry and ecosystem, coastal zones and fisheries, water re-sources, and energy development. Although several sectors have been studied, nonehave received more attention than agriculture. Only a few, however, have lookedat the agricultural impacts of climate change in developing countries like Chinawhere agriculture is a large component of GDP.

Geographically, China touches the tropical belt in the south and extends intothe cold temperate zone in the north. China is also a large agricultural countrywhere agriculture constituted 18.7% of GDP in 1997. China’s agriculture has tofeed more than one-fifth of the world’s population, and, historically, China hasbeen famine prone. As recently as the late 1950s and early 1960s a great famineclaimed about thirty million lives (Ashton et al., 1984, Cambridge History of China1987). Since economic reform, there has been an unprecedented conversion ofarable land into non-agricultural uses following rapid economic development andindustrialization. This loss of agricultural land, together with the trend towards

� Foundation item: Young Scientist Summer Program at the International Institute for AppliedSystem Analysis, YSSP 1999, Austria.

Climatic Change 65: 125–148, 2004.© 2004 Kluwer Academic Publishers. Printed in the Netherlands.

126 HUI LIU ET AL.

a much higher demand for agricultural products for the growing and wealthierpopulation, has resulted in a debate about the country’s long-term capacity to feeditself. How China will avoid national chronic food insecurity in the future is anissue, which is inevitably of significant global implication. Whether China can feeditself in the future not only depends on agricultural land resources but also dependson the impacts of climate change on its agriculture in the future.

Some studies on the impacts of climate change on China’s agriculture havebeen done. However, these studies were either in a limited area or for the yield ofa single crop, such as rice, wheat or maize (Ying, 1995; Gao, 1993; Deng, 1993).Recently, some scientists have used the AEZ model developed by the InternationalInstitute of Applied System Analysis (IIASA) and FAO to assess the impact ofclimate change on China’s agricultural land productivity (Tang et al., 2000). Otherscientist use FARM (Future Agriculture Resource Model) promulgated by Darwinet al. to provide an estimate of economic impact of climate change on agriculturefor all of China plus South Korea (Darwin, 1999a). Although some scientists havestudied the economic impact of climate change on agriculture in China (Darwin,1999a), it only estimated the total average impact that includes not only China butalso Korea. A study on the impacts of climate change on China’s entire agriculturaloutput in value and their regional diversity has not been done yet.

This paper provides the first regionally detailed estimates of the economic im-pact of climate change on agriculture in China. The analysis computed the impactsof changes in temperature and precipitation on agricultural net revenue per Mu(1 Hectare = 15 Mu) and the aggregate effects across Mu’s for different scenarios.We utilized county-level agricultural, climate, social economic and edaphic datafor 1275 agriculture-dominated counties, for the period of 1985–1991, to examinefarmer-adapted response to climate variations across the country. Although we ap-plied methodologies developed in the United States, careful attention was paid toadapting these methods to China’s conditions. For example, the studies paid carefulattention to irrigation, farm labor, management, and technology development.

2. Methodology

We use the Ricardian approach to estimate the impacts of climate change onChina’s agriculture. The Ricardian approach examines how climate in differentplaces affects the net rent or value of farmland (Mendelsohn et al., 1994). Thisapproach is a cross-sectional empirical analysis designed to capture the effect of‘natural experiments’ practiced by farmers across different climate zones or loca-tions. In other words, the farming activities across a large country with sufficientlyvarying climate can be used as samples for comparing farmers’ response to changeclimate. The method uses the typical economic measure of farm performance: netrevenue or net farm income. By examining the economic performance of farmsacross different climate and regressing farm performance on long-term climate, one

STUDY ON THE IMPACTS OF CLIMATE CHANGE ON CHINA’S AGRICULTURE 127

can empirically estimate long-term climate sensitivity. The Ricardian approach hasbeen used to estimate the impacts of climate change on agriculture in the UnitedStates (Mendelsohn et al., 1994, 1996, 1999; Schlenker et al., 2002), India (Dinaret al., 1998) and Brazil.

The most important advantage of the Ricardian approach is its ability to capturethe adaptation that farmers make in response to local environmental conditions.It captures the actual response rather than the controlled ones. In addition, it iscapable of capturing the farmers’ choices over crop mix instead of yield.

A valid criticism of the Ricardian approach is that it has historically assumedthe price to be equilibrium, and in case of significant climate change the crop pricecould change for a prolonged period. Under such circumstances, the Ricardianestimate would be either over- or underestimating the climate change impacts,depending on how the prices change. The bias was calculated to be small in mostrelevant examples of climate change (Mendelsohn et al., 1996). However, Darwin(1999a) showed that in a global generated equilibrium analysis, changes in Ricar-dian rents systematically overestimate both benefits and losses and on average areupwardly biased because inflated benefits are larger than exaggerated costs. Thebias can be quite large. Another valid criticism is that it is difficult to incorporatecarbon dioxide fertilization into the regression. Moreover, when using the Ricar-dian method, the selection of weights and failure to control for irrigation couldinfluence the estimate of climate change on agriculture (Schlenker et al., 2002).

2.1. THE RICARDIAN MODEL

This section summarizes the theoretical understandings of the Ricardian model byRoy Mendelsohn et al. (Mendelsohn et al., 1996; Dinar et al., 1998).

If use i is the best use for the land Li given the environment E and factor pricesR, the observed market rent of the land will be equal to the annual net profit fromthe production of crop i, therefore, land rent per hectare is equal to net revenue perhectare (Dinar et al., 1998), i.e.:

pL = [PiQi − Ci(Qi, R,E)]/Li , (1)

where pL is land rent per hectare, Pi and Qi are respectively the price andquantity of crop i, Ci( ) is the function of all purchase inputs other than land.R = [R1. . . . . .Rj ] is the vector of factor prices, E is an exogenous environmentalinput into the production of goods, e.g., temperature, precipitation, and soils, whichwould be the same for different goods’ production.

The present value of the stream of current and future revenues gives land value:

Vl =∫ ∞

0pLe−rtdt =

∫ ∞

0[PiQi − Ci(Qi, R,E)]e−rt/Lidt . (2)

128 HUI LIU ET AL.

We assume that the change of environment from EA to EB will leave marketprices of inputs (fertilizer, labor) and outputs unchanged. The change in annualwelfare from this environment change is given by:

W(EB − EA) = PQB − �Ci(Qi, R,EB) − [PQA − �Ci(Qi, R,EA)] . (3)

Substituting Equation (1) into Equation (3) gives:

W(EB − EA) = �(pLBLB − pLALA) , (4)

where pLA and LA are respectively net revenue per hectare and planted land area atEA and pLB and LB are at EB . The present value of this welfare change is thus:

∫ ∞

0W(EB − EA)e−rtdt = �(VLBLB − VLALA) , (5)

where VLB and VLA are respectively land value at EB and EA.The Ricardian model takes the form of either (1) or (5) depending on whether

the dependent variable is annual net revenues or farm value. The value of thechange in the environmental variables is captured exactly by the change in landvalues across different conditions. Cross-sectional observation, where normal cli-mate and edaphic factors vary, can hence be utilized to estimate farmer-adaptedclimate impacts on production and land value.

2.2. MODIFIED RICARDIAN MODEL IN CHINA

Due to imperfect land markets and lack of documentation of agricultural farmvalues in China, in this paper, we estimate (1), using annual net revenues as theindependent variable.

According to the Ricardian model, theoretically, the net revenue should be:

Net revenue = [�PiQi − �Ci(Qi, R,E)]/�Li , (6)

where Pi and Qi are respectively the price and quantity of crop i, Ci( ) is thefunction of all purchase inputs other than land, Li is the area planted for crop i.

However, in practice, due to the limitations of available data in China, it is betterto use net income as net revenue.

Net income = [�(PiQi) − �(PnMn) − (Wg ∗ La ∗ days)]/�Li , (7)

where Pi and Qi are respectively the price and quantity of crop i, Pn and Mn

are respectively the price and quantity of the input material n, such as, fertilizers,pesticides, seeds, electricity, etc., Wg, La, and days are respectively the labor cost,numbers of agricultural labor, and the average number of days worked in the statesby farm workers.

Actually, there is no price for each crop and farm labor in the statistical data.There is only agricultural output in China’s statistical yearbook. In China, the pure


agricultural output is defined as the total output of agriculture minus the wholeexpenditure of material input except labor forces. Therefore, the net income can bedefined as:

Net income = (PAO − Wg ∗ La ∗ days)/�Li , (8)

where PAO is the pure agriculture output.A key characteristic of Chinese agricultural sectors is the large share of non-

hired household labor in agriculture and the absence of the labor price. Ignoringnon-hired labor underestimates input costs, and hence overestimates net revenue.For these reasons, the cost of labor forces is removed from the function of netincome and will be considered as an independent variable in the Ricardian climateregression.

From the above analysis, the net revenue in China is modified as pure agricul-tural output per Mu of agricultural land. For county K in year y, the net revenueis:

Net revenueKy = PAOK

y /LKy , (9)

where L is the total agricultural area including the cultivated land area, forest area,grassland area, and water area of a county because the pure agricultural output inChina includes the output of agriculture, forestry, stock-raising and fishery.

3. Data Processing and Empirical Specification

Using the Ricardian technique, we estimate the value of climate in China’s agri-culture. Agriculture is the most appealing application of the Ricardian techniqueboth because of the significant impact of climate on agricultural productivity andbecause of the extensive county-level data on farm inputs and outputs.

3.1. DATA SOURCES AND DATA PROCESSING

For the most part, the data are actual county averages, from the IIASA database ofLUC project for the period of 1985–1991. Those data are the source for much ofthe agricultural data used here, including pure agricultural output, area irrigated,land use data, agricultural labor, etc. All of these are at county level in all of Chinaexcept Taiwan and Hong Kong.

3.1.1. The Dependent Variable: Net Revenue per Agricultural MuAs discussed in the second section, the pure agricultural output per agricultural Muis used as the net revenue for each county, which can be calculated from formula(9).

However, as inflation was very high in the late 1980s in China, the pure agricul-tural output of each county in different years needs to be modified by price index

130 HUI LIU ET AL.

Table I

The agricultural net revenue in China (Yuan/Mu)

Year Original net Index of GOVA a Modified net

revenue revenue

1985 88 1 88

1987 113 0.9360–1.2947 97

1988 138 1.2006–1.6456 96

1989 149 1.2487–1.7113 97

1990 172 1.5588–1.9671 103

1991 188 1.5594–1.9614 109

a Sources: 1985–1990, IIASA database; 1991, calculatedfrom China Statistic Year Book 1992, State Statistical Bu-reau of the People’s Republic of China.

of gross output value of agriculture (GOVA). In addition, because the price indexof GOVA varied much from province to province, it needs to use a different indexin different provinces. The index in different years was calculated based on theassumption that the index of 1985, the first year for analysis, is 1. Table I shows thedifference between the original agricultural net revenue and modified agriculturalnet revenue in different years.

3.1.2. Explanatory VariablesTable II shows the explanatory variables and their processing.

3.2. CLIMATE SCENARIOS

Three general circulation models (GCMs), i.e., HadCM2, CGCM1, and ECHAM4,were used to simulate China’s climate change under five climate change scenariosfor the periods of 2020s,1 2050s,1 and 2080s� (Tang et al., 2000). Table III showssome information of the GCMs, climate scenarios and simulated climatic factorsthat would be used to simulate the impacts of climate change on China’s agricul-ture, which will be discussed in the fourth part of this article. Table IV presents theaverage changes of temperature and precipitation in China for the various climatescenarios. HadCM2 is based on 0.5% increase of atmospheric carbon dioxide levelper year and the other models are based on 1%. That is why the average increaseof temperature based on HadCM2 is lower than that based on the other models.Some of the scenarios, such as HadCM2-gs and CGCM1-gs, include the coolingeffects of sulfate emissions (‘gs’) and others, such as HadCM2-gx, CGCM1-ggand ECHAM4-gg do not (‘gg’, ‘gx’). Air temperature is expected to increase in

� 2020s, 2050s and 2080s represent the year of 2010 ∼ 2039, 2040 ∼ 2069 and the year of2070 ∼ 2099 respectively.


Table II

Explanatory variables

Variables Sources Processing

Soil Soil map of China, Selecting the major type and

IIASA database characteristic of each county

Slope IIASA database Weighted average

Elevation IIASA database Weighted average

Climate data 310 meteorological stations Interpolating county climate

in China data from station data

Distance to market IIASA database Distance to the provincial

capital city of each county

Social and IIASA database Directly extracted from

economic data IIASA database

China under all of the five scenarios. Most of the scenarios (except CGCM1-gs,and CGCM1-gg in 2080s) show that the total precipitation in China is expected toincrease in the future.

3.3. UNITS OF ANALYSIS

The units of observations for this analysis are 1275 counties in thirty provincesand autonomous regions of China, which are not the whole counties of China.Some counties are removed from our analysis because of changes in administrativedivisions and variation in county codes from year to year. So, first of all, it needs tomatch the county code to the standard county code. During the match work, somecounties are removed from the analysis, which are: (1) the counties for which thereare no county code in the standard code system, (2) the counties missing crucialdata, such as, pure agricultural output, land use data, climate date or soil data etc.

In order to reduce the impacts of none-cropping factors, such as forestry, fishery,and livestock raising, on the agricultural net revenue and emphasize the influenceof climate on the grains, the counties in which the arable land area is less than 25%of the total macro-agriculture area (total area of arable land, forest land, grass land,and water area) are also removed from our analysis.

Finally, 1275 agriculture-dominated counties are chosen as the basic observa-tions for this analysis.

132 HUI LIU ET AL.

Table III

Three GCMs and climate scenarios

Name of HadCM2 CGCM1 ECHAM4

GCMs

Source of the Britain Canada Germany

models

Climate HadCM2-gs, CGCM1-gg, ECHAM4-gg

scenarios HadCM2-gx CGCM1-gs

Simulating 2020s, 2050s, 2080s 2020s, 2050s, 2080s 2020s, 2050s, 2080s

periods

Simulating Temperature and Temperature and Temperature and

factors precipitation in winter, precipitation in winter, precipitation in winter,

spring, summer, and spring, summer, and spring, summer, and

autumn for each county autumn for each county autumn for each county

Table IV

Average changes in temperature and precipitation in China for the variousScenarios

Climate scenarios Temperature (◦C) Precipitation (%)

2020s 2050s 2080s 2020s 2050s 2080s

HadCM2-gs 0.8 1.3 2.2 1.0 0.2 2.9

HadCM2-gx 1.6 2.5 3.8 5.8 10.4 18.6

CGCM1-gg 2.5 4.7 7.9 1.2 4.7 –2.5

CGCM1-gs 1.7 3.2 5.8 –2.8 –6.2 –8.1

ECHAM4-gg 1.8 3.1 4.4 8.2 10.4 13.6

3.4. RICARDIAN REGRESSION

The data is pooled and county level net agricultural revenue per Mu are regressedon climate, soil, and other controlled variables, such as agricultural labor, distanceto market, etc., to estimate the best-use value function across different counties.There are 1275 pooled cross-sectional observations.

The independent variables include temperature and precipitation terms forspring, summer, autumn, and winter, pH and OM% (organic materials) of soil,slope, elevation, agricultural labor per 100 Mu agricultural land, and distance to


market. For each variable, linear and quadratic terms are included to capture itsnonlinear effects on agricultural net revenue.

On the other hand, we should separate the irrigated agriculture and rain-fedagriculture for the cultivated land and horticultural land because parameters onclimate variables in counties that rely heavily on irrigation differ from parameterson climate variables in counties where there is no irrigation (Darwin, 1999b). Toillustrate these problems, the agricultural net revenue should be described as:

NR = SCrf NR(C+H)rf + SCirNR(C+H)ir + SGFNRGF and

NRi = f (T , P,X) ,(10)

where NR is net revenue per Mu, SCrf is the share of rain-fed cultivated andhorticultural land, NR(C+H)rf is net revenue per Mu on rain-fed cultivated andhorticultural land, SCir is the share of irrigated cultivated and horticultural land,NR(C+H)ir is net revenue per Mu on irrigated cultivated and horticultural land,SGF is the share of grassland and forest land, NRGF is net revenue per Mu ongrassland and forest land, NRi represents any of the three net revenue categories,T is temperature, P is precipitation, and X represents other variables. Parameterson the climate variables are expected to differ for each net revenue category.

Table V shows the original regression model. Where DIS.M, LAB._P, ELEVand SLR are respectively distance to market, number of labor for agriculture per100 Mu, elevation, and slope; spr.Pir(C + H), win.Trf(C + H), aut.P(GF) arerespectively spring precipitation for the share of irrigated cultivated and horticul-tural land, winter temperature for the share of rain-fed cultivated and horticulturalland, and autumn precipitation for the share of grassland and forestland. Throughstepwise regression, some variables, such as organic materials, pH, and slope,which are insignificant in the analysis, are removed. Table VI shows the trimmedregression model from which it can be seen that the squared terms for most ofthe climate variables are significant, implying that the observed relationship isnonlinear. However, some of the squared terms are positive, implying that there is aminimal production level of those terms and that either more or less of these termswill increase net agricultural revenue according to the observation’s current con-dition. The negative quadratic coefficient implies that there is an optimal level ofthese climate variables from which the value function decreases in both directions.For the grassland and forest, precipitation is more significant than temperature inwinter and spring, but in summer temperature is more significant. For the cultivatedland and horticulture land, both temperature and precipitation are significant in allseasons.

The remaining control variables behave largely as expected. Social economicvariables, such as labor and distance to market, play a role in determining the valueof a farm. Agricultural labor has a positive impact on net revenue as expected.Although both the relationship between net revenue and distance to market and therelationship between net revenue and elevation are U-shaped, the minimal produc-tion level of distance to market and elevation is 509 Km and 1742 M respectively

134 HUI LIU ET AL.

Table V

Result of Ricardian Regression (enter method)

B Std. error t Sig.

(Constant) 92.955 113.833 0.817 0.414

DIS.M –7.542E-02 0.029 –2.604 0.009

DIS.M2 6.584E-05 0.000 1.748 0.081

OM (%) –1.290 2.779 –0.464 0.643

OM2 0.110 0.274 0.401 0.689

PH –15.076 15.643 –0.964 0.335

PH2 1.059 1.131 0.936 0.349

LAB._P 4.888 0.984 4.968 0.000

LAB2 –9.093E-03 0.026 –0.344 0.731

SLR –1.978E-02 0.016 –1.216 0.224

SLR2 8.483E-06 0.000 1.072 0.284

ELEV –3.549E-02 0.020 –1.771 0.077

ELEV2 8.860E-06 0.000 1.620 0.106

win Pir(C + H) 0.182 0.688 0.264 0.792

spr. Pir(C + H) –0.871 0.353 –2.467 0.014

sum.Pir(C + H) –2.334E-02 0.109 –0.214 0.831

aut.Pir(C + H) 1.136 0.359 3.160 0.002

win Prf(C + H) 1.436 0.848 1.693 0.091

spr. Prf(C + H) –0.103 0.438 –0.235 0.814

sum.Prf(C + H) –0.288 0.165 –1.748 0.081

aut.Prf(C + H) 0.712 0.406 1.755 0.080

win P2ir(C + H) 3.009E-04 0.003 0.094 0.925

spr. P2ir(C + H) 8.859E-04 0.001 1.286 0.199

sum.P2ir(C + H) –1.605E-04 0.000 –1.437 0.151

aut.P2ir(C + H) –5.532E-03 0.002 –2.854 0.004

win P2rf(C + H) –1.501E-02 0.009 –1.710 0.088

spr. P2rf(C + H) –1.745E-04 0.002 –0.107 0.915

sum.P2rf(C + H) 4.709E-04 0.000 1.474 0.141

aut.P2rf(C + H) –5.845E-03 0.003 –2.117 0.035

win Tir(C + H) 0.143 0.796 0.180 0.857

spr. Tir(C + H) –2.443 2.305 –1.060 0.289

sum.Tir(C + H) 0.196 1.020 0.192 0.848

aut.Tir(C + H) 2.235 1.909 1.171 0.242

win Trf(C + H) –0.163 0.782 –0.208 0.835

spr. Trf(C + H) 0.206 2.127 0.097 0.923

sum.Trf(C + H) –0.945 0.891 –1.061 0.289

aut.Trf(C + H) 1.018 1.874 0.543 0.587


Table V

(Continued)

B Std. error t Sig.

win T2ir(C + H) 1.643E-02 0.004 3.858 0.000

spr. T2ir(C + H) –2.439E-02 0.013 –1.915 0.056

sum.T2ir(C + H) 1.763E-03 0.002 0.992 0.321

aut.T2ir(C + H) 2.203E-02 0.010 2.164 0.031

win T2rf(C + H) 1.843E-03 0.002 1.062 0.288

spr. T2rf(C + H) –1.115E-02 0.014 –0.817 0.414

sum.T2rf(C + H) 3.508E-03 0.002 1.903 0.057

aut.T2rf(C + H) 2.735E-03 0.013 0.208 0.835

win.T(GF) 0.165 0.637 0.259 0.796

spr.T(GF) 1.127 1.747 0.645 0.519

sum.T(GF) 1.033 0.770 1.342 0.180

aut.T(GF) –1.155 1.499 –0.771 0.441

win.P(GF) –0.958 0.486 –1.972 0.049

spr.P(GF) 0.230 0.239 0.960 0.337

sum.P(GF) 1.425E-02 0.124 0.115 0.908

aut.P(GF) –0.376 0.269 –1.394 0.164

win.T2(GF) 7.140E-04 0.002 0.327 0.744

spr.T2(GF) –9.999E-03 0.008 –1.231 0.219

sum.T2(GF) –5.173E-03 0.002 –2.636 0.008

aut.T2(GF) 8.482E-03 0.008 1.062 0.289

win.P2(GF) 2.505E-03 0.002 1.262 0.207

spr.P2(GF) 1.068E-04 0.000 0.352 0.725

sum.P2(GF) 7.814E-05 0.000 0.544 0.586

aut.P2(GF) 1.129E-03 0.001 1.238 0.216

a Dependent variable: net revenue.b Weighted by cropland.c Number of observation: 1275.R2 = 0.645.

according to the trimmed model. For almost all agriculture dominated counties inChina, the distance to market is less than 509 Km and the elevation is lower than1742 M, that is to say, farther to the market or higher elevation decrease the value.The characteristics of soil and slope are not significant to the value of farm inChina.

136 HUI LIU ET AL.

Table VI

Result of Ricardian regression (stepwise method)

B Std. error t Sig.

(Constant) 5.702 29.293 0.195 0.846

DIS.M –8.050E-02 0.025 –3.181 0.002

DIS.M2 7.908E-05 0.000 2.504 0.012

LAB._P 4.548 0.358 12.711 0.000

ELEV –4.720E-02 0.009 –4.982 0.000

ELEV2 1.355E-05 0.000 3.280 0.001

spr. Pir(C + H) –0.834 0.187 –4.462 0.000

aut.Pir(C + H) 1.035 0.227 4.564 0.000

win Prf(C + H) 1.023 0.452 2.265 0.024

sum.Prf(C + H) –0.160 0.078 –2.045 0.041

spr. P2ir(C + H) 8.109E-04 0.000 1.887 0.059

sum.P2ir(C + H) –1.662E-04 0.000 –5.322 0.000

aut.P2ir(C + H) –5.468E-03 0.001 –3.762 0.000

win P2rf(C + H) –1.690E-02 0.005 –3.190 0.001

sum.P2rf(C + H) 4.609E-04 0.000 2.521 0.012

aut.P2rf(C + H) –2.270E-03 0.001 –2.277 0.023

aut.Tir(C + H) 0.831 0.299 2.780 0.006

win T2ir(C + H) 1.864E-02 0.003 6.446 0.000

spr. T2ir(C + H) –3.822E-02 0.004 –9.658 0.000

sum.T2ir(C + H) 1.911E-03 0.001 2.363 0.018

aut.T2ir(C + H ) 3.103E-02 0.004 8.535 0.000

spr. T2rf(C + H ) –5.568E-03 0.002 –3.633 0.000

sum.T2rf(C + H) 1.802E-03 0.001 2.362 0.018

sum.T(GF) 0.676 0.236 2.859 0.004

win.P(GF) –0.258 0.154 –1.679 0.093

spr.P(GF) 0.232 0.073 3.193 0.001

sum.T2(GF) –3.520E-03 0.001 –4.294 0.000

a Dependent variable: net revenue.b Weighted by cropland.c Number of observation: 1275.R2 = 0.639.

4. Simulating the Impacts and Interpretation of Climate Coefficient

4.1. SIMULATING OF NET AGRICULTURAL REVENUE PER MU

Climate change data on the county level are used in simulating their impacton China’s agriculture. First, the temperature and precipitation for each grid(0.5◦×0.5◦) are calculated relying on GCMS. Then the grid data are translated into


average temperature and precipitation for each county in order to link the climatedata with agricultural data that are organized by county level.

The trimmed regression model is used to simulate the changes of net revenuefor each of 1275 counties for the analysis year. County-level changes in net revenueper Mu are then aggregated to get a measure of the net impact for all of China as awhole. The change of net revenue per Mu of the given scenario in year y is givenby:

CNy =1275∑d=1

[Netrevny

d(Td + �TCS, Pd + �PCS) − Netrevny

d(Td, Pd)]/1275 , (11)

where y is the year of 1985, 1987, 1988, 1989, 1990, 1991; (Td, Pd) describes theclimate for county d; (Td + �TCS, Pd + �PCS) describes the new climate under asimulated climate scenario; Netrevny

d (Td, Pd) is predicted value of the net revenueper Mu for county d in year y; Netrevny

d (Td + �TCS, Pd + �PCS) is the forecastedvalue of net revenues under a climate scenario for county d in year y .

Yearly changes in the net revenue per Mu are correspondingly averaged overthe period (1985–1991) to yield an average net impact.

The change in net revenue per Mu is calculated for the new climate for each ofthe four seasons. Table VII presents these impacts by season.

Overall, an increase in precipitation level is beneficial for increasing net revenueper Mu of China’s agricultural land, whereas rise in temperature is whether harmfulor beneficial depending on different GCMs.

As shown in Table VII, there is significant seasonal variation in both temper-ature and precipitation effects. Temperature rise in all seasons except spring ispositive. Summer and autumn temperature effects are positive because warmingtemperature during these two harvest seasons may facilitate the ripening processand ensure optimal crop production. Positive winter temperature effect could bethe result of increasing the growing period by a warmer winter. However, a warmerspring not only makes the winter crop grow too quickly, which leads to an increasein the incidence of pests and insects that reduce the production, but also causeshigh evaporation of soil water which would increase the damage to crops by springdrought that happens frequently in China.

Increased precipitation in all of the four seasons is beneficial under most scenar-ios. More precipitation in winter can increase soil moisture for the winter crop. Thepositive impact of spring precipitation under most scenarios is due to the increaseof rainfall, which in the spring can increase soil moisture and reduce spring droughtthat is good for winter crop to turn green again and spring sowing. Autumn precip-itation is good for planting winter crop. The negative effect of increased rainfall insummer under some scenarios is expected for a summer-dominated rainfall climateregime. It may lead to more flood disasters in China.

138 HUI LIU ET AL.

Table VII

Changes in net revenue per Mu (1990 Rs) by seasonal temperature and precipitation(Yuan/Mu)

Climate scenarios E4-gg H2-gs H2-gx CG-gs CG-gg

Temperature Winter 2.961 0.917 1.574 8.255 16.287

effects Spring –55.835 –17.309 –45.982 –155.128 –166.999

Summer 5.375 2.398 4.792 4.267 6.651

Autumn 53.958 29.936 60.663 41.128 65.794

Total effects of temperature 6.459 15.942 21.048 –101.479 –78.261

Precipitation Winter 4.366 3.466 2.358 3.465 3.623

effects Spring 1.754 6.666 –1.026 4.419 –2.514

Summer –0.100 0.855 1.311 0.996 0.434

Autumn 1.118 3.147 2.849 3.802 2.631

Total effects of precipitation 7.137 14.114 5.491 12.618 4.174

Effects of Winter 7.326 4.363 3.932 11.720 19.909

temperature Spring –54.081 –10.643 –47.008 –150.709 –169.513

and Summer 5.275 3.253 6.103 5.626 7.085

precipitation Autumn 55.076 33.083 63.511 44.929 68.425

Total 13.596 30.056 26.539 –88.798 –74.092

Average net revenue per Mu (in 1990Rs) = 102.8 Yuan.

4.2. SPATIAL VARIATION OF SEASONAL IMPACTS

The broad effects given above suggest that Chinese aggregate agriculture is notat risk due to climate change. However, it does not protect local areas, as Chinahas so many different climate types. So, the regional variations in temperature andprecipitation impacts in each season are now discussed. Figures 1–8 exhibit theregional distribution of net revenue changes per Mu in different seasons under thescenario of H2-gs.

In winter, all parts of north China have a negative impact from increased tem-perature, but the southern parts of China benefit from warming winter (Figure 1).This could be because in the most of northern China, the warmer winter is not goodfor winter crops (like wheat). It may cause crop disease and cause wheat to pitchits root to deep soil or reduce soil moisture. The warmer winter in the southern partmay be beneficial for the farmers to produce more fruits and vegetables. Springtemperature effects are negative to almost all of China (Figure 2). The most nega-tive are for the South, the North China Plain and the Yangtze River Delta. This maybe because in the north part of China, a warmer spring may cause wheat growingtoo quickly and intensify spring drought. For the south, since it is warm enough,the temperature increases in spring may be too high for crops growing there. The


Figure 1. Impact of winter temperature in China in the 2050s.

Figure 2. Impact of spring temperature in China in the 2050s.

140 HUI LIU ET AL.

Figure 3. Impact of summer temperature in China in the 2050s.

Figure 4. Impact of autumn temperature in China in the 2050s.


Figure 5. Impact of winter precipitation in China in the 2050s.

Figure 6. Impact of spring precipitation in China in the 2050s.

142 HUI LIU ET AL.

Figure 7. Impact of summer precipitation in China in the 2050s.

Figure 8. Impact of autumn precipitation in China in the 2050s.


distribution of summer temperature effects is mostly neutral. But the most positiveareas are concentrated in the North China Plain and the Yangtze River Delta, andthe most negative areas are in the southern parts of China (Figure 3). This could bebecause the regional disparity of summer temperature is very little in China. Thedistribution of autumn temperature effects is uniformly beneficial (Figure 4), sincea warmer harvest season is expected to facilitate expeditious crop harvesting. Themost beneficial areas are the Yangtze River Delta and the North China Plain.

The increase in winter precipitation has negative effects in the Southwest, theNorth China Plain and the southern part of the Northeast, whereas, the South, theSoutheast, and the middle reaches of the Yangtze River would benefit from theincrease in winter precipitation (Figure 5). The negative effects of spring precipi-tation increasing are concentrated greatly in the Southwest, the Loess Plateau, andthe middle and lower reaches of the Yangtze River. However, it would be bene-ficial for the North China Plain, the northern part of the Northeast, and Xinjiangautonomous region (Figure 6). The increase of summer precipitation has negativeeffects in the south part of China, especially in the middle and the lower reaches ofthe Yangtze River, the South and the Southwest. But the North China Plain wouldbe benefited (Figure 7). Increase of autumn precipitation has different impact indifferent part of China. It is negative in most part of the South, the Southwest, andthe middle reaches of the Yangtze River, but positive in the southern parts of theNorth China Plain, the Northwest, and the Yangtze River Delta (Figure 8).

4.3. REGIONAL DISTRIBUTION OF ANNUAL TOTAL IMPACTS

The county-level changes in temperature and precipitation based on the runs ofthree GCMs and five scenarios mentioned in Section 3 are applied for each countyto simulate their impacts on China’s agriculture.

Table VIII shows the regional distribution of the aggregate effects of climatechange on China’s agricultural net revenue for different scenarios in the year of2050s.

It can be seen from Table VIII that different climate scenarios will have differentimpacts on China’s agriculture. Most scenarios except CG-gs and CG-gg wouldhave an overall positive effect on China’s agriculture. That is because under mostscenarios, temperature and precipitation would increase simultaneously, which isespecially beneficial for arid and semi-arid agriculture in the temperate-zone ofChina. But under the scenario of CG-gs, when temperature increases, precipitationwould decrease, which will lead to more evaporation and the shortage of waterresource that is harmful for China’s agriculture in most areas. Under the scenarioof CG-gg, the increase of temperature is much higher than that of other scenarios(see Table IV), which causes a very large negative effect in spring and dominatesthe total effects in a year.

On the other hand, different regions have different reactions to climate changes.The overall regional impacts of temperature and precipitation under different sce-

144 HUI LIU ET AL.

Table VIII

Possible impacts due to different climate scenarios in 2050s (billion Yuan)

Region Changes of net revenue

E4-gg CG-gg CG-gs H2-gs H2-gx

(3.1 ◦C, 10.4%) (4.7 ◦C, 4.7%) (3.2 ◦C, –6.2%) (1.3 ◦C, 0.2%) (2.5 ◦C, 10.4%)

North 8.820 –57.307 –57.611 19.122 18.133

(18.16) (–117.91) (–118.62) (39.37) (37.33)

Northeast –1.213 –4.298 –3.635 –0.340 –1.075

(–8.61) (–30.50) (–25.9) (–2.41) (–7.63)

East 4.859 –8.333 –9.520 13.306 9.192

(19.25) (–33.01) (–37.71) (52.71) (36.41)

Central 1.190 –3.289 –4.653 1.548 1.883

(13.89) (–38.36) (–54.27) (18.05) (21.97)

South 1.8761 0.633 1.316 2.087 2.406

(15.76) (5.31) (11.05) (17.53) (20.20)

Southwest 0.275 –7.386 –8.915 0.416 –0.628

(1.81) (–48.54) (–58.58) (2.73) (–4.13)

Northwest –0.271 –8.561 –6.559 0.678 0.794

(–4.27) (–134.98) (–103.36) (10.68) (12.51)

Plateau 0.007 –0.018 –0.011 0.004 0.0138

(6.97) (–17.36) (–10.77) (4.29) (13.49)

Total 15.543 –88.559 –89.587 36.821 30.718

(11.95) (–68.09) (–68.88) (28.31) (23.62)

a Total net revenue in 1990 = 130.05 billion Yuan.b Numbers in parentheses are % change in net revenue.

narios are portrayed in Figures 9–13. From a comparison of Figures 9–13, it can beseen that the Southeast and the northern part of the Northeast always benefit fromclimate change. But the southern part of the Northeast, the Southwest (excludingthe Sichuan basin) and most parts of the Northwest would have negative effects.For the North, the Central, the East, the Plateau and the Sichuan basin there are dif-ferent results from different scenarios. Most scenarios indicate that climate changewould be harmful for the southern part of the South, such as the Hainan province,but beneficial for the North, the East, the Central and the Sichuan basin.


Figure 9. Changes of Net Revenue under the scenario of ECHAM4-gg in 2050s.

Figure 10. Changes of Net Revenue under the scenario of CGCM1-gg in 2050s.

146 HUI LIU ET AL.

Figure 11. Changes of Net Revenue under the scenario of CGCM1-gs in 2050s.

Figure 12. Changes of Net Revenue under the scenario of HadCM2-gs in 2050s.


Figure 13. Changes of Net Revenue under the scenario of HadCM2-gx in 2050s.

5. Conclusions and Discussions

The Ricardian approach is a feasible alternative method to provide an assessment ofthe economic impacts of climate change on agriculture. But it needs to be modifiedin China according to the database and characteristics of the country’s agriculture.

The Ricardian approach is not offered as a replacement for other methods,such as the production function approach, but instead as a complement for cross-checking each other. Besides, when using the Ricardian approach in China, we facethe difficulties of how to measure input price, wage, and marginal contributions offamily labor and animal work. These prices are crucial for the calculation of netrevenues. Therefore, other alternatives are necessary for cross checking each other.

Findings from the study indicate that the agricultural impacts of climate changein China are uncertain. The total average impact may be positive or negative de-pending on the climate scenarios. But most scenarios show that climate changewill have an overall positive impact on China’s agriculture. Impacts also varyboth quantitatively and qualitatively by region and season. They are positive inthe middle and the east regions of China in most scenarios, but negative in the westregions (including the Southwest and the Northwest) in some scenarios. As to theseasonal impacts, the spring effect is the most negative, whereas the autumn effectis the most positive.

As any new technique, there are still some problems, which need to be studiedfurther.

148 HUI LIU ET AL.

The effects of CO2 fertilization should be included, for some studies indicatethat this may produce a significant increase in yield. This effect would likely bealso offset somewhat by damages due to other fossil fuel emission (e.g., sulfates).

This paper only examined the existing farm and did not explore the possibilitythat climate may affect whether land is farmed or not. As for analyzing the overallimpact of climate change on China’s agriculture, it also needs to analyze climateeffects on the fraction of land farmed.

This analysis does not adequately capture potential damages that might becaused by an increase in flooding and other extreme events.

References

Ashton, B., Hill, K., Piazza, A., and Zeitz, R.: 1984: ‘Famine in China, 1958–1961’,́ PopulationDevelop. Rev. 10 (4), 613–615.

Cambridge History of China: 1987, Vol. 14, Cambridge University Press, Cambridge, U.K.Darwin, R. F.: 1999a, ‘A FARMer’s View of the Ricardian Approach to Measuring Effects of Climate

Change on Agriculture’, Clim. Change 41 (3–4), 371–411.Darwin, R. F.: 1999b, ‘The Impact of Global Warming on Agriculture: A Ricardian Analysis:

Comment’, Amer. Econ. Rev. 49 (4), 1049–1052.Deng, G.: 1993, ‘The Impact of Climate Change on China’s Agriculture’, Beijing Sci. Tech. Press,

263–312.Dinar, A., Mendelsohn, R., Evenson, R., Parikh, J., Sanghi, A., Kumar, K., McKinsey, J., and Lon-

ergan, S.: 1998, Measuring the Impact of Climate Change on Indian Agriculture, World BankTechnical Paper No. 402, World Bank, Washington, D.C.

Gao, S., Ding, Y., Zhao, Z., and Pan, Ya.: 1993, ‘The Possible Green House Impact of At-mospheric CO2 Content Increasing on the Agriculture Production in the Future in China’,Scientia Atomospherica Sinica 17 (5), 584–591.

Gong, Z., Zhou, H., Shi, X., and Luo, G.: 1999, Soil of China, Introduction to the Legend of theSoil Map of China, Academia Sinica, Food and Agriculture Organization of the United Nations,pp. 1–40.

Mendelsohn, R. and Neumann, J.: 1999, The Economic Impact of Climate Change on the UnitedStates Economy, Cambridge University Press, Cambridge, U.K.

Mendelsohn, R., Nordhaus, W. D., and Shaw, D.: 1994, ‘The Impact of Global Warming onAgriculture: A Ricardian Analysis’, Amer. Econ. Rev. 84 (4), 753–771.

Mendelsohn, R., Nordhaus, W. D., and Shaw, D.: 1996, ‘Climate Impacts on Aggregate Farm Value:Accounting for Adaptation’, Agric. For. Meteorol. 80 (1), 55–56.

Schlenker, W., Hanemann, M., and Fisher, A. C.: 2002, The Impact of Global Warming on U.S.Agriculture: An Econometric Analysis, Department of Agricultural and Resource Economics andPolicy Working Paper 936, University of California, Berkeley, October 9, 2002.

Tang, G., Li, X., Fischer, G., and Prieler, S.: 2000, ‘Climate Change and its Impacts on China’sAgriculture’, ACTA Geographica Sinica 55 (2), 129–138.

Ying, H.: 1995, ‘The Possible Impacts of Climate Change on the Major Grain Yield in LiaoningProvince’, China Agric. Meteorol. 16 (3), 5–8.

(Received 18 February 2001; in revised form 1 October 2003)

study on the impacts of climate change on china’s ... · climate impacts on production and land...

Documents