monitoring the vegetation health over india during contrasting monsoon years using satellite remote...

15
ORIGINAL PAPER Monitoring the vegetation health over India during contrasting monsoon years using satellite remote sensing indices Arnab Kundu 1 & Suneet Dwivedi 1,2 & Dipanwita Dutta 3 Received: 2 September 2014 /Accepted: 7 October 2015 /Published online: 24 February 2016 # Saudi Society for Geosciences 2016 Abstract The detection and monitoring of drought-related vegetation stress over a large spatial area have become possi- ble with the use of satellite-based remote sensing indices, namely, vegetation condition index (VCI) and temperature condition index (TCI). In particular, the water (precipita- tion)-related moisture stress during drought may be deter- mined using the VCI, while the temperature-related stress using the TCI. An attempt is made here to investigate and demonstrate the importance of these indices over India during the contrasting monsoon years, 2009, 2010, and 2013, termed as meteorological drought, wet, and normal monsoon years, respectively. The overall health of the vegetation during these years is compared using the vegetation health index (VHI). The advantage of VHI over the VCI and TCI is also shown. An assessment of drought over India is then made using the combined information of VCI, TCI, and VHI. The occurrence of vegetative drought over Rajasthan, Gujrat, and Andhra Pradesh is confirmed using drought assessment index, which shows very low value (well below 40) during 2009 over these regions. The area-averaged time series indices as well as spa- tial maps over the state of Uttar Pradesh show higher thermal stress and poor vegetation health during 2009 as compared to 2010 and 2013. The standardized precipitation index (SPI) and standardized water-level index (SWI) are used to validate the results obtained using the remote sensing indices. Keywords VCI . TCI . VHI . SPI . SWI . Drought monitoring . Monsoon . Remote sensing Introduction Drought occurs when unusually dry weather conditions with broad spatiotemporal coverage result into abnormal changes in vegetation cover of the area. Drought is an insidious, slow- onset natural hazard that produces a complex web of impacts that ripple through many sectors of the economy (Wilhite et al. 2007). When the precipitation over land in some years is much less than normal, the occurrence of meteorological as well as vegetative drought conditions over India become very likely. The drought incidences over India have great socio-economic impact over the region (Dutta et al. 2015). It directly (reduction in food production and agricultural economy) and indirectly (reduction in energy/power generation, decreased water supply, decreased industrial growth, in- creased inflation, etc.) affects the lives of the billions of people living in the Indian subcontinent. It is, therefore, important to understand and take into account the consequences of drought on the vegetation conditions leading to drought preparedness (Patel and Yadav 2015). The techniques of remote sensing have served as a useful tool for this purpose. Remote-sensing-based vegetation indi- ces have been widely used by several researchers for various studies, for example, for the mapping of agricultural fields, rainfall monitoring, estimation of the weather impacts, crop yield, pasture and biomass production, investigation of drought conditions, and determination of the strength of the vegetation (Tucker et al. 1982; Justice et al. 1985; Hielkema * Suneet Dwivedi [email protected] 1 K Banerjee Centre of Atmospheric and Ocean Studies, University of Allahabad, Allahabad, UP 211002, India 2 M N Saha Centre of Space Studies, University of Allahabad, Allahabad, UP 211002, India 3 Department of Remote Sensing and GIS, Vidyasagar University, Midnapur, West Bengal, India Arab J Geosci (2016) 9: 144 DOI 10.1007/s12517-015-2185-9

Upload: bhu-in

Post on 03-Dec-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

ORIGINAL PAPER

Monitoring the vegetation health over India during contrastingmonsoon years using satellite remote sensing indices

Arnab Kundu1& Suneet Dwivedi1,2 & Dipanwita Dutta3

Received: 2 September 2014 /Accepted: 7 October 2015 /Published online: 24 February 2016# Saudi Society for Geosciences 2016

Abstract The detection and monitoring of drought-relatedvegetation stress over a large spatial area have become possi-ble with the use of satellite-based remote sensing indices,namely, vegetation condition index (VCI) and temperaturecondition index (TCI). In particular, the water (precipita-tion)-related moisture stress during drought may be deter-mined using the VCI, while the temperature-related stressusing the TCI. An attempt is made here to investigate anddemonstrate the importance of these indices over India duringthe contrasting monsoon years, 2009, 2010, and 2013, termedas meteorological drought, wet, and normal monsoon years,respectively. The overall health of the vegetation during theseyears is compared using the vegetation health index (VHI).The advantage of VHI over the VCI and TCI is also shown.An assessment of drought over India is then made using thecombined information of VCI, TCI, and VHI. The occurrenceof vegetative drought over Rajasthan, Gujrat, and AndhraPradesh is confirmed using drought assessment index, whichshows very low value (well below 40) during 2009 over theseregions. The area-averaged time series indices as well as spa-tial maps over the state of Uttar Pradesh show higher thermalstress and poor vegetation health during 2009 as compared to2010 and 2013. The standardized precipitation index (SPI)

and standardized water-level index (SWI) are used to validatethe results obtained using the remote sensing indices.

Keywords VCI . TCI . VHI . SPI . SWI . Droughtmonitoring .Monsoon . Remote sensing

Introduction

Drought occurs when unusually dry weather conditions withbroad spatiotemporal coverage result into abnormal changesin vegetation cover of the area. Drought is an insidious, slow-onset natural hazard that produces a complex web of impactsthat ripple throughmany sectors of the economy (Wilhite et al.2007).When the precipitation over land in some years is muchless than normal, the occurrence of meteorological as well asvegetative drought conditions over India become very likely.The drought incidences over India have great socio-economicimpact over the region (Dutta et al. 2015). It directly(reduction in food production and agricultural economy)and indirectly (reduction in energy/power generation,decreased water supply, decreased industrial growth, in-creased inflation, etc.) affects the lives of the billions of peopleliving in the Indian subcontinent. It is, therefore, important tounderstand and take into account the consequences of droughton the vegetation conditions leading to drought preparedness(Patel and Yadav 2015).

The techniques of remote sensing have served as a usefultool for this purpose. Remote-sensing-based vegetation indi-ces have been widely used by several researchers for variousstudies, for example, for the mapping of agricultural fields,rainfall monitoring, estimation of the weather impacts, cropyield, pasture and biomass production, investigation ofdrought conditions, and determination of the strength of thevegetation (Tucker et al. 1982; Justice et al. 1985; Hielkema

* Suneet [email protected]

1 K Banerjee Centre of Atmospheric and Ocean Studies, University ofAllahabad, Allahabad, UP 211002, India

2 M N Saha Centre of Space Studies, University of Allahabad,Allahabad, UP 211002, India

3 Department of Remote Sensing and GIS, Vidyasagar University,Midnapur, West Bengal, India

Arab J Geosci (2016) 9: 144DOI 10.1007/s12517-015-2185-9

et al. 1986; Kogan 1987a, b, 1990, 1995, 1998; Dabrowska-Zielinska et al. 2002; Singh et al. 2003; Bhuiyan et al. 2006,Bhuiyan and Kogan 2010). There have also been several stud-ies recently on drought monitoring, risk assessment, and pre-diction (Patel et al. 2007; Kumar et al. 2009; Dutta et al. 2013;Belal et al. 2014). The stressed vegetation has a higher reflec-tance than healthy vegetation in the visible (0.4–0.7 μm) re-gion and lower reflectance in the near-infrared (0.7–1.1 μm)region of the electromagnetic spectrum. Vegetation indicestake the advantage of this differential response in the visibleand near-infrared regions of the spectrum (http://www.dsc.nrsc.gov.in/DSC/Drought).

The normalized difference vegetation index (NDVI) servesas the basic metric for mapping the vegetative drought. TheNDVI may be used for exact monitoring of the continentalland cover, vegetation classification, and vegetation phenolo-gy (Tucker et al. 1982; Tarpley et al. 1984; Justice et al. 1985).It is also effective for monitoring rainfall and drought; estimat-ing net primary production of vegetation, crop growth condi-tions, and crop yields; and detecting weather impacts and oth-er events important for agriculture, ecology, and economics(Kogan 1987a; Dabrowska-Zielinska et al. 2002; Dutta et al.2013). The NDVI is computed as the ratio of responses innear-infrared and visible bands of the National Oceanic andAtmospheric Administration (NOAA) Advanced Very HighResolution Radiometer (AVHRR) remote sensing detectors(Gutman 1991). The NOAA-AVHRR is a cross-track scan-ning system with five spectral bands [Ch 1 visible; Ch 2near-infrared; Ch 3–5 thermal] (for detailed specificationincluding wavelength range and resolution of spectral bands,please refer to Table 1). It has a resolution of 1.1 km and afrequency of earth scans twice per day. The global vegetationindex (GVI) data with parameters, namely, radiance, NDVI,

satellites, and Sun angles is produced by sampling and map-ping of 4 km daily radiance, measured from NOAA polarorbiting satellites to a 16-km map. The large NDVI refers tothe healthy and dense vegetation, whereas the rock and baresoil-covered areas show near-zero NDVI values since theyhave similar reflectance in the visible and near-infrared bands.The cloud, water, and snow show negative NDVI values.

The vegetation condition index (VCI) is a NDVI-derivedindex and separates the short-term weather-related NDVI fluc-tuations from long-term ecosystem changes (Kogan 1990,1995). Kogan (1995) has developed the temperature conditionindex (TCI) using the thermal bands of the NOAA-AVHRR todetermine the temperature-related vegetation stress as well asstresses caused by the excessive wetness. The vegetationhealth index (VHI) has been developed using the VCI andTCI and is found to be more effective compared to other indi-ces in monitoring vegetative drought (Kogan 1990, 2001;Singh et al. 2003). The VCI, TCI, and VHI are used togetherto make a good assessment of the vegetative drought. Theseindices depend on the region (topography) aswell as prevailingweather and ecological conditions over the region of interest.There have been several studies in the recent past employingthe remote-sensing-based indices for drought and vegetativehealth assessment over India. Singh et al. (2003) demonstratedthe application of NDVI, VCI, and TCI for drought monitoringin India. Dutta et al. (2015) made an assessment of the droughtover the Rajasthan state of India using the VCI and standard-ized precipitation index (SPI). Gopinath et al. (2015) usedgeospatial techniques for drought monitoring and analyzedvegetation health from MODIS-Terra satellite-derived prod-ucts. The monitoring of the spatio-temporal pattern of droughtstress using integrated drought index over the Bundelkhandregion of India was performed by Patel and Yadav (2015).An agricultural drought was predicted using the NDVI in theeastern part of Rajasthan, India by Dutta et al. (2013). Patel etal. (2012) used vegetation temperature condition index (VTCI)from Terra/MODIS satellite data for monitoring of droughtover Gujarat, India.Murthy et al. (2010) showed the operation-al use of NDVI for detection and monitoring of drought overIndia. Kumar et al. (2009) used the SPI for drought intensityassessment. Bhuiyan et al. (2006) investigated drought dynam-ics in the Aravalli region of Rajasthan, India using differentground-based as well as remote sensing indices.

However, none of the abovementioned studies has shownanalysis performed over densely populated Gangetic plains.Moreover, the examples of a comparison of these indices dur-ing the contrasting monsoon years (of normal and very lowrainfall) are also a few. An analysis considering these points is,however, extremely useful and desirable keeping in mind theincreased frequency of extreme (rainfall and drought) eventsin the recent years in the context of climate change (IPCCAssessment Report 5; http://www.ipcc.ch/report/ar5). One ofthe main objectives of the present paper is, therefore, to

Table 1 Wavelength range and resolution of different bands of theNOAA-AVHRR detectors

Band Wavelength region (μm) Resolution (km)

AVHRR/2 (NOAA-7 through 14)

1 0.58–0.68 (red) 1.1

2 0.725–1.10 (near-IR) 1.1

3 3.55–3.93 (high-temp TIR) 1.1

4 10.3–11.3 (TIR) 1.1

5 11.5–12.5 (TIR) 1.1

AVHRR/3 (NOAA-15 through 18)

1 0.58–0.68 (red) 1.1

2 0.73–0.98 (near-IR) 1.1

3a 1.58–1.63 (mid-IR) 1.1

3b 3.54–3.87 (high-temp TIR) 1.1

4 10.3–11.3 (TIR) 1.1

5 11.5–12.4 (TIR) 1.1

TIR thermal infrared

144 Page 2 of 15 Arab J Geosci (2016) 9: 144

demonstrate the usefulness of VCI, TCI, and VHI forvegetative drought monitoring over India with specialemphasis over Gangetic plains covering the Uttar Pradesh.For this purpose, we discuss these indices during contrastingmonsoon years of 2009, 2010, and 2013 that are termed asmeteorological drought, wet, and normal monsoon years,respectively. Another objective is to showcase the usefulness

of the drought assessment maps in drought monitoring. Theresults obtained using the remote-sensing-based drought mon-itoring indices are also validated against the meteorologicaland hydrological drought monitoring indices, namely, SPIand standardized water level index (SWI), respectively.

The remote-sensing-based vegetation indices (TCI, VCI,VHI), SPI, and SWI are briefly described in BMethods and

Fig. 1 August–September climatological rainfall during 1998–2013

Arab J Geosci (2016) 9: 144 Page 3 of 15 144

data used^ section. BResults^ section gives the results follow-ed by BDiscussion^ and BConclusions^ sections.

Methods and data used

Temperature condition index

The TCI represents the relative change in thermal condition interms of brightness temperature obtained from the thermalbands of the NOAA-AVHRR. Slight changes in vegetationhealth due to thermal stress, in particular, could be monitoredusing the analysis of the TCI data (Kogan 1995, 2001, 2002).The TCI may be expressed as follows:

TCI ¼ 100 BTmax−BTð Þ= BTmax−BTminð Þ½ � ð1Þ

where brightness temperature (BT), BTmin, and BTmax are theseasonal average of weekly brightness temperature and itsmulti-year absolute minimum and maximum value, respec-tively. Here we use NOAA-AVHRR 16 km, 7-day compositeTCI over India for the year 2009 and 2013 as a proxy forthermal condition expressed as an anomaly relative to 25-year climatology estimated based on biophysical and ecosys-tem laws. The TCI is based on 10.3–11.3 μm AVHRR’s radi-ance measurements converted to BT using a look-up table,and a nonlinear correction was applied following Weinreb

Fig. 2 August–September climatological rainfall during 1998–2013 over the Gangetic plains covering Uttar Pradesh, Bihar, and West Bengal

�Fig. 3 a Temperature condition index (TCI) during the month of August(weeks 31–34) for the drought year 2009 (upper panel), wet year 2010(middle panel), and normal monsoon year 2013 (lower panel). b TCIduring the month of September (weeks 35–39) for the drought year2009 (upper panel), wet year 2010 (middle panel), and normalmonsoon year 2013 (lower panel)

144 Page 4 of 15 Arab J Geosci (2016) 9: 144

Arab J Geosci (2016) 9: 144 Page 5 of 15 144

144 Page 6 of 15 Arab J Geosci (2016) 9: 144

et al. (1990), which was improved by completely removinghigh-frequency noise (Kogan 1997).

Vegetation condition index

The VCI (given in percentage, %) rescales vegetation dynam-ics between 0 and 100 to reflect the relative changes in thevegetation condition from extremely bad to optimal (Kogan1995; Kogan et al. 2003). The VCI is derived using NDVI asfollows:

VCI ¼ 100 NDVI−NDVIminð Þ= NDVImax−NDVIminð Þ½ � ð2Þ

where NDVI, NDVImin, and NDVImax are the seasonal aver-age of smoothed (no-noise) weekly NDVI and its multi-yearabsolute minimum and maximum values, respectively. Herewe use NOAA-AVHRR 16 km, 7-day composite VCI overIndia for the year 2009 and 2013 as a proxy for moisturecondition expressed as NDVI anomaly relative to 25-year cli-matology estimated based on bio-physical and ecosystemlaws.

Vegetation health index

The VHI is said to represent the overall vegetation healthusing a combined estimation of moisture and thermal condi-tions. It is computed using VCI and TCI as follows:

VHI ¼ α VCIþ 1−αð Þ TCI ð3Þ

where α is a coefficient determining contribution of the twoindices and is generally taken as 0.5. Here we use NOAA-AVHRR 16 km, 7-day composite VHI over India for the year2009 and 2013 as a proxy for characterizing vegetation health.The five classes of the VHI used for classifying the vegetativedrought are as follows (Kogan 2002):

Drought classes VHI

Extreme drought <10

Severe drought ≥10 and <20

Moderate drought ≥20 and <30

Mild drought ≥30 and <40

No drought ≥40

Thus, lower the value of VHI, greater will be the intensityof drought.

Standardized precipitation index

The SPI is defined as follows:

SPI ¼ Pij−Pim

� �=σ ð4Þ

where,Pij is the seasonal precipitation at the ith rain gauge stationand jth observation, Pim is the long-term seasonal mean and σ isthe standard deviation. In other words, the SPI at a location iscomputed by dividing the difference between the seasonal pre-cipitation at a station and its long-term seasonal mean by thestandard deviation of the seasonal precipitation. McKee et al.(1993, 1995) proposed the SPI as a drought monitoring index.The index takes into account the anomalous and extreme precip-itation values. The high-resolution (0.1 × 0.1 degree) daily pre-cipitation data used for the analysis are procured from theClimate Prediction Center (CPC), NOAAwebsite: http://www.cpc.ncep.noaa.gov/products/fews/SASIA/rfe.shtml). To takeinto account the skewness of the precipitation data, the data arenormalized using gamma function before computing the SPI.The calculation of SPI involves fitting a gamma probabilitydensity function to a given frequency distribution ofprecipitation totals for a station (Bhuiyan et al. 2006). For thepurpose of this study, we used following four classes of SPI:

Drought classes SPI

Severe drought < −1.5Moderate drought < −1.0Mild drought <0

No drought >0

Standardized water level index

SWI is defined as follows:

SWI ¼ WLij−WLim

� �=σ ð5Þ

where WLij is the monthly water level of the ith district and jthobservation (year), WLim is the monthly mean water level of ithdistrict, andσ is the standard deviation. Any decrease in thewatertable of a location may be known using the SWI. The index,therefore, serves as a drought monitor and an indirect measureof the groundwater recharge. Since groundwater level is mea-sured from ground surface down into observation wells, positiveanomalies correspond to water stress and negative anomaliesrepresent no drought condition (Bhuiyan et al. 2006).

We obtained the groundwater level data for the month ofAugust 2009, 2010, and 2013 from the Central Ground WaterBoard (CGWB), Ministry of Water Resources, Govt. of Indiaat website http://gis2.nic.in/cgwb/Gemsdata.aspx. The datafor the month of September was, however, not available.

�Fig. 4 a Vegetation condition index (VCI) during the month of August(weeks 31–34) for the drought year 2009 (upper panel), wet year 2010(middle panel), and normal monsoon year 2013 (lower panel). b VCIduring the month of September (weeks 35–39) for the drought year2009 (upper panel), wet year 2010 (middle panel), and normalmonsoon year 2013 (lower panel)

Arab J Geosci (2016) 9: 144 Page 7 of 15 144

144 Page 8 of 15 Arab J Geosci (2016) 9: 144

Results

The production of the Kharif crop (monsoon crop) heavilydepends on the amount and timing of the water from the rain-fall. The August and September months of the year are impor-tant months in this context. To identify the regions over India[8°4′N–37°6′N; 68°7′E–97°25′E] with distinct rainfall vari-ability, we show in Fig. 1 the map of the climatological rainfallaveraged during these months spanning the years 1998 to2013. The regions with high and low rainfall are clearly dis-tinguishable in Fig. 1. The Western Ghats, northeastern states,parts of Orissa, Uttarakhand, Panjab, and Himanchal Pradeshreceive more rainfall on an average, whereas the states ofRajasthan, Jammu and Kashmir, Gujrat, Andhra Pradesh,and Tamil Nadu receive less rainfall. Moderate (mix ofheavy to low) rainfall occurs on an average over theGangetic plains covering Uttar Pradesh, Bihar, andWest Bengal (Fig. 2). It is clear from Fig. 2 that thenorth-central and eastern Uttar Pradesh, eastern Bihar,and northern West Bengal are high rainfall regions inGangetic plains, whereas the western Uttar Pradesh and south-west Bihar are low rainfall regions.

In an effort to understand the advantage of TCI for droughtmonitoring, we make a comparative analysis of the TCI of themeteorological drought year 2009 with the normal monsoonyear 2013 and wet year 2010 for the months of August andSeptember (weeks 31–39). We show in Fig. 3a the August 5–26 weekly TCI maps for the year 2009 (upper panel), 2010(middle), and 2013 (lower panel). The acuteness of droughtincreases with the increase in brightness temperature. The lowvalues of TCI suggest increase in brightness temperature(Eq. 1), and hence point towards drought-like conditions. Itis clear from Fig. 3a that the TCI values during 2009 are verylow over almost all parts of India as compared to 2010 and2013. Highly significant changes between the years 2009 with2010 and 2013 are observed over Jammu and Kashmir,Panjab, Rajasthan, and central and southern Indian statesclearly demonstrating the usefulness of TCI in investigatingthe vegetative drought. Similar analysis is also made for themonth of September 2009, 2010, and 2013 (Fig. 3b). Thedifferences between corresponding upper, middle, andlower panels in Fig. 3b are clearly seen over theGangetic plains also. The TCI values during the year2010 and 2013 are much higher as compared to 2009.While highly significant TCI differences between droughtyear 2009 and the wet year 2010 are seen throughout the

September month, in the case of normal monsoon year2013, the remarkable differences are seen only during theweeks of September 23 and September 30.

To make a comparison of the vegetation condition as aconsequence of the rainfall-driven moisture availability overIndia during the drought year 2009, wet year 2010, and nor-mal monsoon year 2013, we show in Fig. 4a, b the spatio-temporal VCI maps for the months of August and September,respectively. The higher values of VCI suggest favourablemoisture condition, whereas lower values suggest stressedcondition resulting to decreased vegetation. It is clear fromthe figure that the vegetation around the desert areas (inRajasthan) shows particularly strong VCI response. The VCIvalues in these regions are much higher in 2010 and 2013 ascompared to 2009. The same holds good for most of the areasin the Gangetic plains. It is interesting to note that during thewet year 2010, the VCI values are very high over almost theentire India thus confirming increased vegetation as a result ofheavy rainfall and abundance of moisture availability overIndia. On the other hand, it is remarkable to observe that theUttar Pradesh in 2010 and entire central India and most partsof the south India show lowVCI values (stressed condition) in2013 as compared to 2009. However, it is to be noted thatvegetative drought is very sensitive to regular availability ofmoisture. Therefore, frequency and interval of consecutiverainfall events are more important than total seasonal rainfall,particularly for the agricultural crops (Bhuiyan 2008). Thus,although the total seasonal monsoon rainfall was low in theyear 2009, the vegetation moisture was out-of-stress in mostparts of the India. Opposite is the case for the year 2013 overthe Central Indian region and 2010 over the Uttar Pradeshregion. This may be due to heavy rainfall or persistent cloud-iness leading to erroneous values.

We see that the vegetative drought monitoring in terms ofVCI and TCI shows contrasting results over some parts ofIndia. It is, therefore, useful to employ such an index whichtakes into account both the VCI and TCI i.e. moisture as wellas thermal stress. This is done using the VHI for the overalldrought monitoring. In Fig. 5a, we show the VHI for the year2009 (top panel), 2010 (middle panel), and 2013 (bottom pan-el) for the month of August. The higher the value of VHI, thebetter is the vegetation condition (BVegetation health index^section). It may be noted from Fig. 5a that most of the regionsover India show low tomediumVHI values (in the range of 6–36) during August 2009. For example, very low VHI (<10)over Tamil Nadu and Andra Pradesh represent extremedrought, while the VHI values over Rajasthan and some partsof Uttar Pradesh, Madhya Pradesh and Bihar (in the range of10–30) suggest severe to moderate drought conditions overthese regions duringAugust 2009. The values are significantlyhigh during the same period (i.e. August) for the year 2010and 2013. The highest differences between the VHI values ofthe year 2009 with 2010 and 2013 are observed over the

�Fig. 5 a Vegetation health index (VHI) during the month of August(weeks 31–34) for the drought year 2009 (upper panel), wet year 2010(middle panel), and normal monsoon year 2013 (lower panel). b VHIduring the month of September (weeks 35–39) for the drought year 2009(upper panel), wet year 2010 (middle panel), and normal monsoon year2013 (lower panel)

Arab J Geosci (2016) 9: 144 Page 9 of 15 144

144 Page 10 of 15 Arab J Geosci (2016) 9: 144

northwestern parts (including areas with desert vegetation)and southeastern parts of the India. The VHI values in theseregions (in the range of 60–100) represent no drought condi-tion during August 2010 and 2013. The wet year 2010 showshighest VHI values, as expected. The VHI values for themonth of September in Fig. 5b show similar results. Duringthe year 2013, the better vegetation health (higher VHI) isobserved over the northwestern and southeastern Indianstates; however, most of the regions over central India andparts of coastal Andhra Pradesh and Orissa show some-what lower values of VHI (in the range of 30–40) ascompared to the year 2009. These regions, therefore,suffered very mild drought during 2013. The VHIvalues are quite high all over the India, with remarkablyhigh values over the western and central Indian statesduring September month of the wet year 2010. It may, there-fore, be argued from the August–September VHI maps of theyears 2009, 2010, and 2013 that the vegetative drought oc-curred during the year 2009 over the states of Rajasthan,Gujarat, and Andhra Pradesh, since the VHI differences arethe highest over these regions.

To put the results in perspective, we show thedrought assessment map based on the VHI, VCI, andTCI. A drought condition is declared if these values areless than 40. The categorization of drought for this pur-pose is done as follows: ‘exceptional’ for indices be-tween 0 and 5, ‘extreme’ between 6 and 15, ‘severe’between 16 and 25, ‘moderate’ between 26 and 35, and‘abnormally dry conditions’ between 35 and 40. Thedrought maps are shown in Fig. 6a, b for the month ofAugust and September, respectively. The states of Rajasthanand Andhra Pradesh are distinctly represented as highdrought regions in these maps during the year 2009.The low to medium vegetative drought-like conditionswere prevailing over rest of the country during the year2009. However, during the year 2013, except statesalong east coast and some parts of central-east India,which were marked with low drought conditions, otherparts of the country were drought-free. Similarly, thedrought assessment maps of 2010 clearly suggest thatalmost the entire India was drought-free during thisyear, with only few states along north-eastern coast showinglow drought conditions.

Discussion

The state of Uttar Pradesh [23°52′N–31°28′N; 77°3′E–84°39′E]in the Gangetic plains is one of the world’s most densely popu-lated states, which is highly vulnerable to vegetative drought.We demonstrate the area-averaged drought monitoring indicesover the state of Uttar Pradesh in Fig. 7. The highVCI values (inthe range of 50–70%) of the years 2009, 2010, and 2013 duringAugust–September suggest favourable moisture conditions overUttar Pradesh. There is not much difference between the VCIvalues of 2009 and 2013. However, it is interesting to note theVCI variability during the weeks 36–39. The VCI values of2009 and 2013 during these weeks show opposite nature thanthat observed during weeks 31–35. For example, during theweeks 31–35, the VCI for 2013 is higher than 2009 values,whereas during the weeks 36–39, the VCI of 2009 is higherthan 2013 values. The VCI values of 2010, on the other hand,

�Fig. 6 a Drought assessment map during the month of August (weeks31–34) for the year 2009 (upper panel), 2010 (middle panel), and 2013(lower panel). The abnormally dry conditions, moderate drought, severedrought, and extreme and exceptional drought are shown with 0, 1, 2, and3 values on the colour scale, respectively. b Drought assessment mapduring the month of September (weeks 35–39) for the year 2009 (upperpanel), 2010 (middle panel), and 2013 (lower panel). The abnormally dryconditions, moderate drought, severe drought, and extreme andexceptional drought are shown with 0, 1, 2, and 3 values on the colourscale, respectively

Fig. 7 The VCI (upper panel), TCI (middle panel), and VHI (lowerpanel) averaged over the state of Uttar Pradesh during the year 2009(blue) and 2013 (red)

Arab J Geosci (2016) 9: 144 Page 11 of 15 144

are lower than 2009 and 2013 over the Uttar Pradesh duringweeks 31–37. Even though the result is quite surprising for awet year, but as explained earlier in BResults^ section, this maybe due to erroneous remotely sensed values as a result of persis-tent and widespread clouds associated with heavy rainfall.

The TCI values of 2009 are, however, much smaller thancorresponding TCI values of 2010 and 2013 suggesting strongthermal stress over Uttar Pradesh in 2009. The difference be-tween TCI values of 2009 and 2013 is relatively small duringweeks 34–36. The TCI difference between these 2 years grad-ually increases during weeks 36–39 and decreases from week31–35. On the other hand, the TCI values of wet year 2010 areobviously quite high as compared to drought year 2009 duringAugust and September months. The TCI values of 2010 arealso much higher as compared to corresponding values ofnormal monsoon year 2013 during weeks 34–38.

TheVHI values of 2009 are also smaller than 2010 and 2013,as expected. The August 2009 VHI values indicate abnormallydry conditions over the state of Uttar Pradesh though the situa-tion improves a little during September. The VHI values of 2010are lower than 2013 during August (weeks 31–34) but become alittle higher during later weeks in September. The VHI values of2010, which are nearly 60 % all through September month(weeks 36–39), indicate quite good vegetation health during thatyear over the state of Uttar Pradesh.

To further examine the vegetative health over the state ofUttar Pradesh, we show in Fig. 8 the maps of VHI duringAugust and September of 2009, 2010, and 2013. The VHIvalues are very low in most parts of the Uttar Pradesh duringAugust 2009. The month of August 2009 is thus marked withstrong vegetative drought over the Uttar Pradesh. InSeptember 2009, the VHI values are relatively higher (indi-cating no drought) in most parts of central Uttar Pradesh,whereas the values are low (indicating drought) in southernand northern Uttar Pradesh. The scenario is somewhat differ-ent in the wet year 2010 and normal monsoon year 2013. Wenotice very high VHI during August–September 2010 in mostplaces of the Uttar Pradesh. For example, barring a few placesin central and eastern Uttar Pradesh during August 2010, therest of the Uttar Pradesh is showing no drought-likeconditions. The month of September 2010 shows thehighest VHI (no drought) over the Uttar Pradesh (alsoseen in Fig. 7 lower panel). In addition to this, most ofthe regions of Uttar Pradesh (except southwest andnorthwest regions along with some districts of centralUttar Pradesh) show high to very high VHI values duringAugust 2013, thus representing no drought condition.Similarly, the September 2013 VHI values are also high inmost parts of the Uttar Pradesh, except in the northwest andsouthwest portion.

Fig. 8 The mean VHI over the state of Uttar Pradesh during the August (upper panel) and September (lower panel) months of 2009, 2010, and 2013

144 Page 12 of 15 Arab J Geosci (2016) 9: 144

To demonstrate the robustness and significance of ourstudy, we make validation of our results obtained from theremote sensing estimates against the ground truth i.e. real-time in situ observations. The meteorological and hydrologi-cal drought monitoring indices, namely, SPI and SWI respec-tively are used for this purpose. The SPI has been a widelyused index for drought monitoring and agricultural health as-sessment (Bhuiyan et al. 2006; Dutta et al. 2013, 2015; Kumaret al. 2009; Patel et al. 2007; Sahoo et al. 2015; Shah et al.2015). We show in Fig. 9 the spatial maps of SPI during

August and September months of 2009, 2010, and 2013 overthe state of Uttar Pradesh. It is clear from the figure that 2009was a meteorological drought year for the state of UttarPradesh since barring a few districts, the SPI of all other re-gions indicates drought. This confirms our similar argumentobtained using VHI (Fig. 8). Quite remarkably, most of theregions with low VHI in Fig. 8 are also the regions of low SPIin Fig. 9, thus establishing the robustness of remote sensingindices for drought monitoring. The SPI values of August2010 show mild drought over some parts of eastern and

Fig. 9 SPI over the state of Uttar Pradesh during the August (upper panel) and September (lower panel) months of 2009, 2010, and 2013

SPI

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

2009

Aug

201

ust Sep

0

tember

2013

Fig. 10 Area-averaged SPIvalues over the state of UttarPradesh during the August (blue)and September (red) months of2009, 2010, and 2013

Arab J Geosci (2016) 9: 144 Page 13 of 15 144

western Uttar Pradesh, whereas mild drought occurred oversome districts of west and west-central Uttar Pradesh duringSeptember 2010. Similarly, the eastern Uttar Pradesh (and fewother districts) suffered mild drought during August–September 2013. The overall SPI values of the year 2010and 2013, however, clearly put them under the category ofwet and normal monsoon years, respectively. To further clar-ify this point, we show in Fig. 10 the area-averaged SPI valuesover the Uttar Pradesh during 2009, 2010, and 2013. As ex-pected, out of these years, the highest SPI value is obtainedduring 2010 and lowest during 2009.

The SWI is yet another important index of drought moni-toring utilizing in situ observations (Bhuiyan et al. 2006;Mishra and Nagarajan 2013; Sahoo et al. 2015). The SWIspatial maps over the Uttar Pradesh during August 2009,2010, and 2013 are shown in Fig. 11 (September data wasnot available, hence not shown). The figure clearly shows thatthe year 2009 was a hydrological drought year. The SWIvalues are less than zero over all parts of the Uttar Pradeshduring 2009. On the other hand, except for some of the dis-tricts over southeast Uttar Pradesh, nearly all other regionsshow positive SWI i.e. no drought-like conditions during2010. Similarly, with the exception of some districts of north-west and north-central Uttar Pradesh, which suffered hydro-logical drought during 2013, all other regions were drought-free. The results using the SPI and SWI are, therefore, inconformity with those obtained using the remote sensingdrought monitoring indices.

Conclusions

The remote-sensing-based vegetation health indices, namely,TCI, VCI, and VHI are important for drought detection andmapping. These are also useful tool for estimating and fore-casting the crop condition (losses) and anticipated yield over a

region. We used these indices to monitor vegetative droughtover India during the August–September months of the Kharifseason of the contrasting monsoon years 2009, 2010, and2013. It is found that the TCI values during the wet monsoonyear 2010 and normal monsoon year 2013 are much higher ascompared to the drought year 2009. In other words, the year2009 shows much higher thermal stress compared to 2010 and2013. The VCI-based drought monitoring suggests that thevegetation condition was poor along the desert areas inRajasthan with high moisture stress during 2009. The otherareas were relatively free from moisture-related stress duringthe drought year 2009 possibly due to regular availability ofmoisture, even though the total seasonal monsoon rainfall waslow. The state of Uttar Pradesh in the year 2010 and entirecentral India and most parts of south India, however, showlow VCI values in 2013 as compared to 2009. This may bedue to the fact that when there is excessive soil moisture due toheavy rainfall or persistent cloudiness in some regions, theNDVI is depressed and as a result the VCI values becomelow, which can be interpreted erroneously as drought in thoseregions. This seems to happen over central and south Indiaduring 2013 and over the Uttar Pradesh in 2010. In such cases,rather than using VCI and TCI alone for drought monitoring,it is desirable to use the VHI that does their combined estima-tion for distinguishing drought events. It may be concludedfrom the August–September VHI maps that the vegetativedrought over the states of Rajasthan, Gujrat, and AndhraPradesh occurred during the year 2009. We also used thecombined drought assessment map based on VHI, VCI, andTCI to demonstrate vegetative drought over India. The indiceswell below 40 during 2009 over parts of Rajasthan andAndhra Pradesh indicate intense vegetation stress leading tovegetative drought in these regions. The analysis carried outover the state of Uttar Pradesh confirms poor overall vegeta-tion health during the year 2009. The results obtained usingthe remote sensing indices are also validated against the

Fig. 11 SWI over the state of Uttar Pradesh during the August month of 2009, 2010, and 2013

144 Page 14 of 15 Arab J Geosci (2016) 9: 144

meteorological SPI and hydrological SWI indices. In future,the analysis shall be extended for a longer period for examin-ing the vegetation health in the context of climate change.

Acknowledgments The authors thank all the three anonymous re-viewers for their valuable comments and suggestions to improve thequality of the manuscript. The authors would also like to thank theDepartment of Science and Technology (DST), Govt. of India, for pro-viding financial help in the form of research project. AK thanks DST forproviding research fellowship. Acknowledgements are also due to theNOAA for making available the AVHRR-based satellite images usedfor this study.

References

Belal AA, El-Ramady HR,Mohamed ES, Saleh AM (2014) Drought riskassessment using remote sensing and GIS techniques. Arab J Geosci7:35–53

Bhuiyan C (2008) Desert vegetation during droughts: response and sen-sitivity. Int Arch Photogramm Remote Sens Spat Inf Sci 37(B8):907–912

Bhuiyan C, Kogan FN (2010) Monsoon variation and vegetative droughtpatterns in the luni basin in the rain-shadow zone. Int J Remote Sens31:3223–3242

Bhuiyan C, Singh RP, Kogan FN (2006)Monitoring drought dynamics inthe Aravalli Terrain (India) using different indices based on groundand remote sensing data. Int J Appl Earth Obs 8:289–303

Dabrowska-Zielinska K, Kogan F, Ciolkosz A, Gruszczynska M,Kowalik W (2002) Modelling of crop growth conditions and cropyield in Poland using AVHRR-based indices. Int J Remote Sens 23:1109–1123

Dutta D, Kundu A, Patel NR (2013) Predicting agricultural drought ineastern Rajasthan of India using NDVI and standardized precipita-tion index. Geocarto Ints 28:192–209

Dutta D, Kundu A, Patel NR, Saha SK, Siddiqui AR (2015) Assessmentof agricultural drought in Rajasthan (India) using remote sensingderived vegetation condition index (VCI) and standardized precipi-tation index (SPI). Egypt J Remote Sens Space Sci. doi:10.1016/j.ejrs.2015.03.006

Gopinath G, Ambili GK, Gregory SJ, Anusha CK (2015) Drought riskmapping of south-western state in the Indian peninsula–a web basedapplication. J Environ Manag. doi:10.1016/j.jenvman.2014.12.040

Gutman GG (1991) Vegetation indices fromAVHRR data: an update andfuture prospects. Remote Sens Environ 35:121–136

Hielkema JU, Prince SD, Astle WL (1986) Rainfall and vegetation mon-itoring in the Savanna zone of Democratic Republic Sudan usingNOAA Advanced Very High Resolution Radiometer. Int J RemoteSens 7:1499–1514

Justice CO, Townshend JRG, Holben BN, Tucker CJ (1985) Analysis ofthe phenology of global vegetation using meteorological satellitedata. Int J Remote Sens 6:1271–1318

Kogan FN (1987a) Vegetation index for areal analysis of crop conditions.In: Proceedings of 18th conference on agricultural & forest meteo-rology. AMS, Indiana, USA, pp. 103–106

Kogan FN (1987b) On using smoothed vegetation time-series for identi-fying near-optimal climate conditions. In: Proceedings of the 10thconference on probability and statistics. AMS, Edmonton, Canada,pp. 81–83

Kogan FN (1990) Remote sensing of weather impacts on vegetation innon-homogeneous areas. Int J Remote Sens 11:1405–1419

Kogan FN (1995) Application of vegetation index and brightness tem-perature for drought detection. Adv Space Res 15:91–100

Kogan FN (1997) Global drought watch from space. Bull Am MeteorolSoc 78:621–636

Kogan FN (1998) Global drought and flood watch from NOAA polar-orbiting satellites. Adv Space Res 21(3):477–480

Kogan FN (2001) Operational space technology for global vegetationassessment. Bull Am Meteorol Soc 82:1949–1964

Kogan FN (2002) World droughts in the new millennium from AVHRR-based vegetation health indices. EOS Trans AmGeophys Union 83:562–563

Kogan FN, Gitelson A, Edige Z, Spivak L, Lebed L (2003) AVHRR-based spectral vegetation index for quantitative assessment of veg-etation state and productivity: calibration and validation.Photogramm Eng Remote Sens 69:899–906

Kumar MN, Murthy CS, Sesha Sai MVR, Roy PS (2009) On the use ofstandardized precipitation index (SPI) for drought intensity assess-ment. Meteorol Appl 16:381–389

McKee TB, Doesken NJ, Kleist J (1993) Proceedings of the 8th confer-ence on applied climatology. In: The relationship of drought fre-quency and duration to time scales. AMS, Boston, pp. 179–184

McKee TB, Doesken NJ, Kleist J (1995) Drought monitoring with mul-tiple time scales. Proceedings of the. In: 9th conference on appliedclimatology, Dallas, TX, AMS, pp, pp. 233–236

Mishra SS, Nagarajan R (2013) Hydrological drought assessment in Telriver basin using standardized water level index (SWI) and GISbased interpolation techniques. Int J Concept Mech Civil Eng 01:01–04

Murthy CS, Chakravorty A, Sesha Sai MVR, Roy PS (2010) Spatio-temporal analysis of the droughts of kharif 2009 and 2002. CurrSci 100:1786–1788

Patel NR, Yadav K (2015) Monitoring spatio-temporal pattern of droughtstress using integrated drought index over bundelkhand region,India. Nat Hazards 77:663–677

Patel NR, Chopra P, Dadhwal VK (2007) Analyzing spatial patterns ofmeteorological drought using standardized precipitation index.Meteorol Appl 14:329–336

Patel NR, Parida BR, Venus V, Saha SK, Dadhwal VK (2012) Analysis ofagricultural drought using vegetation temperature condition index(VTCI) from terra/MODIS satellite data. EnvironMonit Assess 184:7153–7163

Sahoo RN, Dutta D, Khanna M, Kumar N, Bandyopadhyay SK (2015)Drought assessment in the Dhar and Mewat Districts of India usingmeteorological, hydrological and remote-sensing derived indices.Nat Hazards 77:733–751

Shah R, Bharadiya N, Manekar V (2015) Drought index computationusing standardized precipitation index (SPI) method for Surat dis-trict, Gujarat. Aquat Proc 4:1243–1249

Singh RP, Roy S, Kogan F (2003) Vegetation and temperature conditionindices from NOAA AVHRR data for drought monitoring overIndia. Int J Remote Sensing 24:4393–4402

Tarpley JD, Schnieder SR, Money RL (1984) Global vegetation indicesfrom NOAA-7 meteorological satellite. J Clim Appl Meteorol 23:4491–4503

Tucker CJ, Gatlin J, Schnieder SR, Kuchinos MA (1982) Monitoringlarge scale vegetation dynamics in the Nile delta and river valleyfrom NOAA-AVHRR data. Proceedings of the Conference onRemote Sensing of Arid and Semi-Arid Lands, Cairo, pp. 973–977

Weinreb MP, Hamilton G, Brown S (1990) Nonlinearity correction incalibration of the Advanced Very High Resolution Radiometer in-frared channels. J Geophys Res 95:7381–7388

Wilhite DA, SvobodaMD, Hayes MJ (2007) Understanding the compleximpacts of drought: a key to enhancing drought mitigation and pre-paredness. Water Resour Manag 21:763–774

Arab J Geosci (2016) 9: 144 Page 15 of 15 144