Monitoring agricultural vulnerability using NDVI time series

Monitoring agricultural vulnerability using NDVI time series

SHARE

 

A study to evaluate agricultural vulnerability at district –level in rainfed agro-ecological regions in India, is being conducted at the Central Research Institute for Dryland Agriculture (CRIDA) at Hyderabad under the National Initiative on Climate Resilient Agriculture programme of ICAR. Long-term NDVI time-series data is being used to assess agricultural vulnerability. As variations in NDVI would indicate impact of climate change on vegetation growth and vigour, it could be used as an indicator to study agricultural vulnerability. Based on coefficient of variation (CV) in NDVI, vulnerable districts were identified in order to develop climate resilient technologies for coping with climate change and adapting to it. NDVI data products based on NOAA-AVHRR (8km) data (1982-2006) and MODIS-TERRA (250m) NDVI data product (2001 – 2011) were used for the study.

Objective
The main objective of the study was to understand variability in ground vegetation also termed surface greenness as indicated by NDVI based on NOAA-AVHRR and MODIS-TERRA time-series datasets and to examine correlation between NDVI variability and Standard Precipitation Index (SPI) instead of actual daily rainfall data, in order to, understand the impact of extreme weather events, viz., droughts, floods, heat and cold waves, cyclones, untimely rains, etc.

Methodology
Over 27 GB of NDVI data from NOAA-AVHRR and MODIS – TERRA was downloaded from GIMMS and GLCF website respectively for analysis. Temporal analysis of Max NDVI was used to identify agricultural vulnerability in the country. The volume of data was huge and the entire work was time consuming. ArcGIS (Ver. 10) and ERDAS Imagine (Ver. 2011) were used for carrying out the work. Standard Precipitation Index (SPI) instead of actual rainfall was used to corroborate extreme events that impact vegetation growth in natural forest, open scrub, agricultural land and plantations. NDVI derived from 2-band information (Red and Near-infra Red) of multi-spectral imagery of AVHRR and from MODIS dataset were analysed and coefficient of variation (CV) of Max. NDVI from 15-day composite for the total length of study period was used to assess agricultural vulnerability to climate change.

NOAA-AVHRR dataset also called the GIMMS data and MODIS-TERRA dataset were downloaded and pre-processed using Savtizky – Golay filter. To study NDVI variations, only Max NDVI value were used. Coefficient of variation of Max. NDVI has been used to assess agricultural vulnerability by ICARDA for the Sub-Saharan region (Celis et al, 2007). Standard Precipitation Index which represents Normal rainfall for a given location, obtained from accounting for standard deviation (SD) of mean (x ¯) of variability of actual rainfall for any location, was derived from 10 *10 grid rainfall data for India obtained from India Meteorological Department (IMD, 2007) and then interpolated to match spatial resolution of AVHRR and MODIS datasets, in order to analyse variations in NDVI. AVHRR NDVI (8km) dataset was found to be suitable for state-level analysis while MODIS (250m) NDVI data was suitable for district – level mapping. Zonal Statistics were calculated using Spatial Analyst Tool in ArcGIS (ver.10.0) to determine variations in NDVI at regional level in the country. SPI indicated periods when extreme weather events like drought or flood affected the NDVI. National Agricultural Statistics pertaining to agricultural production, yield and net sown area data were analysed to corroborate the results obtained from analysis of NDVI variations. Impact of climatic variability on human and livestock population was studied to determine agricultural vulnerability in India.

Results
In order to assess variations in NDVI annually, the 15-day AVHRR (8km) composites were stacked for each year and one Max NDVI image was generated as indicated in Figure1. MODIS based NDVI product with a 250m resolution was analysed in a similar manner as indicated in Figure 2. SPI was estimated from actual rainfall data and then Kriged for the NDVI map resolution as required by AVHRR or MODIS dataset (Figure 3). This was followed by a trend analysis to ascertain variations in NDVI and SPI in the country during the study period. Based on AVHRR-NDVI data, analysis indicated that during 1982-2006, over 92.98 million ha area in the country experienced a decreasing trend in NDVI while in 25.2 million ha there was no change in NDVI. In 183.96 million ha in India an increasing trend in NDVI was recorded. With reference to a spatial context, the decreasing NDVI trend was noticed in the Western Ghats, Orissa and Chattisgarh region, NE states and the Shivalik hills in Lower Himalayas in Himachal Pradesh and southern Kashmir region. This is a matter for concern as these regions are rich in biodiversity and are source of origin of many rivers in India. More recently in the last decade, i.e., 2001- 2011, the MODIS NDVI data indicates that surface cover has improved comparatively. Study indicated that in 56 districts covering 30.93 million ha, a decrease in vegetative cover was registered. In 41 districts covering 22.25 million ha, there was no perceptible change in NDVI. In 457 districts covering over 249 million ha, a positive trend in NDVI was registered (Figure 4 & 5). Trend in SPI (Figure 6) indicated that while Deccan region, West Bengal and Bihar and parts of NE states besides western Rajasthan and western J&K were receiving more rainfall, large parts of the country including the Indo-Gangetic Plains and Arunachal Pradesh were drying up. This would be disastrous for the country. Spatially positive trend was seen in 162 districts covering 91.14 million ha. Desiccation was seen in 278 districts accounting for 139.96 million ha. No change in NDVI status was seen in 111 districts covering 69.67 million ha. In one district namely Vaishali in Bihar, a positive trend was recorded.


Figure 1: Trend in AVHRR Max. NDVI (1982-2006)

 


Figure 2: Trend in MODIS Max. NDVI (2001 -2011)

 


Figure 3: Trend in Standard Precipitation Index (1982 – 2006)

 


Figure 4: Long-trends in NDVI (1982 – 2006)

 


Figure 5: Trends in NDVI during 2001-2011

 


Figure 6: Trends in SPI in India

Discussion
Based on the trends seen in NDVI and SPI as discussed earlier the agriculturally vulnerable regions were identified. As already mentioned, the ground resolution of AVHRR dataset enabled us to broadly identify the regions in the country that were vulnerable to climate change. The extent of this region has been depicted in Figure 7. District-level vulnerability was mapped using MODIS dataset. The vulnerable districts have been indicated in Figure 8.


Table 1: List of vulnerable districts in India based on variability of MODIS data

 


Figure 8: Agricultural vulnerability at district-level

 


Figure 8: Agricultural vulnerability at district-level

Conclusion
Study indicated that agriculture will be vulnerable in over 81.3 million ha in the arid, semi-arid and dry sub-humid regions in Rajasthan, Gujarat, Marathwada and Vidharbha regions in Maharashtra, parts of Karnataka and Andhra Pradesh where rainfed agriculture is widely practiced. These regions are already challenged and climate change would acerbate the situation. Over 12.1 million ha of Kharif cropland would be mildly affected while 1.81 million ha would be severally affected. In case of Rabi agriculture, 6.86 million ha may be mildly affected while 0.5 million ha would be severely impacted by climate change.

AVHRR and MODIS NDVI time-series data proved useful to understand the trend in climate change and its impact on ground cover vegetation. Study indicated a trend of moderate drying in West Bengal, eastern Bihar and Jharkhand, parts of Vidharba and southern Madhya Pradesh, National Capital Region, southeastern Punjab, southern Himachal Pradesh and south-west Uttarakhand. This is a large region covering densely populated areas which cultivates important cash crop in the country. This information would be useful to develop coping strategies to adapt to climate change.

References

  • CRIDA (2011): CRIDA News Letter. Central Research Institute for Dryland Agriculture, ICAR, Santoshnagar, Hyderabad. 2 p.
  • CRIDA (2012): CRIDA Annual Report 2011-2012. Central Research Institute for Dryland Agriculture, ICAR, Santoshnagar, Hyderabad. 97-99 p.