Crop Yield Prediction Under Soil Salinity Using Satellite Derived Vegetation Indices

Abstract

Monitoring the crop yield is one of the key factors to define agricultural land management strategies. Recent developments in spatial information technologies enabled cost and energy saving in crop yield prediction. The aim of this paper was to predict yield of the three major crops and yield loss under soil salinity effect which is one of the most important limitation in many Mediterranean countries. Crop yields were estimated using vegetation indices and Stepwise Linear Regression (SLR) derived from Landsat (Thematic Mapper and Enhanced Thematic Mapper) TM/ETM satellite images. Additionally, related crop pattern of the area was mapped using multi-temporal Landsat data set using object based classification. Soil salinity was mapped using radial basis function and field measurements with a Root Mean Square Error (RMSE) accuracy of 0.96 dSm(-1). The predictions were validated using real-time field measurements. Mean percent error (MPE) for wheat, corn and cotton were 7.9%, 8.8% and 6.3% respectively. Crop yield estimates were incorporated with various degrees of soil salinity. Soil salinity ranging between 8 and 10 dSm(-1) resulted yield loss of 55%, 28%, and 15% in corn, wheat and cotton respectively. The highest soil salinity resistance was observed only at cotton in 18 dSm(-1) with 55% yield loss. (C) 2016 Elsevier B.V. All rights reserved.