Evaluation of MODIS and Landsat multiband vegetation indices used for wheat yield estimation in irrigated Indus Basin

Abstract

Crop yield estimation has significant importance for policy makers to make timely dicisions on import/export of particular crop. Traditionally, in Pakistan crop yield estimation is being carried out by Village Master Sampling (VMS) that is laborious and time-consuming. Satellite imagery is also being used as an alternative to estimate vegetation health and yield. Various vegetation indices are being used for the purpose however, their efficiency to estimate yield has not been tested. In this study, a comparison was performed among various satellite-based vegetation indices e.g. Soil Adjusted Vegetation Index (SAVI), Modified Soil Adjusted Vegetation Index (MSAVI) Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), to evaluate most appropriate index that performs better in cropping area of irrigated Indus Basin (a complex basin with spatially heterogeneous land use). A stepwise regression based model was developed for remotely sensed crop (i.e. Wheat) using multi-band MODIS and Landsat 8 products based on Land use and Land cover map developed by Semi-Supervised Classification. The results revealed that SAVI showed a fairly acceptable association with reported yield data as compared to other indices. The correlation coefficient (R2) was estimated at 0.60. Yield estimated by SAVI obtained from Landsat 8 showed good results with R2 and Pearson correlation (r), estimated at 0.74 and 0.88 as compared to SAVI obtained from MODIS with 0.63 and 0.79 respectively. The results support that SAVI vegetation indices is reliable for quick and efficient wheat area mapping under Pakistani’s farm conditions.