WHEAT YIELD ESTIMATION USING SATELLITE REMOTE SENSING

Abstract

Estimation of wheat (Triticum aestivum L.) yield prior to harvesting is a challenging task for informed decision making, food security analysis and trading policy formulation. The conventional approach used in Pakistan is based on crop cutting surveys is tedious, time consuming and statistics are often disseminated at macro level. This research was conducted as a pilot study project for wheat yield estimation (Rabi 2008-09) at National Agriculture Research Centre (NARC), Islamabad as study area. The objectives of the study were to estimate the wheat area under cultivation and to estimate the wheat yield using Linear Regression Model (LRM) and Production Efficiency Model (PEM). The PEM estimates were based on (a) absorbed photosynthetically active radiation (APAR) (b) light use efficiency (LUE) and (c) harvest index (HI). For this purpose, temporal SPOT imageries were calibrated and exoatmospheric reflectance data was used to calculate vegetation indices. The Pearson’s coefficient (r) value for March normalized difference vegetation index (NDVI) i.e. flowering stage of wheat and actual yield was 0.716 (n=16). The estimated yield range using LRM between wheat yield and NDVI was 2500-2600 kg ha-1. The average estimated yield and maximum yield using PEM were 1233 kg ha-1 and 1436 kg ha-1, respectively. Validation of LRM and PEM based wheat yield estimates and farmer reported yield, gave high positive correlation i.e. r2= 0.85 and r2= 0.91, respectively. The study results suggest that LRM based on yield and NDVI acquired during the month of March is significant. The PEM yield estimates extend spatial variability at pixel level and result accuracy largely depends upon the quality of input parameters. For agro based economies like Pakistan, comprehensive wheat yield estimation by integrating remote sensing based PEM, supporting GIS layers (land use, soil and elevation etc.) and crop growth simulation models should be the next step.