Estimating maize grain yield from crop biophysical parameters using remote sensing

The overall objective of this investigation was to develop a robust technique to predict maize (Zea mays L.) grain yield that could be applied at a regional level using remote sensing with or without a simple crop growth simulation model. This study evaluated capabilities and limitations of the Moderate Resolution Imaging Spectroradiometer (MODIS) Vegetation Index 250-m and MODIS surface reflectance 500-m products to track and retrieve information over maize fields. Results demonstrated the feasibility of using MODIS data to estimate maize green leaf area index (LAIg). Estimates of maize LAIg obtained from Wide Dynamic Range Vegetation Index using data retrieved from MODIS 250-m products (e.g. MOD13Q1) can be incorporated in crop simulation models to improve LAIg simulations by the Muchow-Sinclair-Bennet (MSB) model reducing the RMSE of LAIg simulations for all years of study under irrigation. However, more accurate estimates of LAIg did not necessarily imply better final yield (FY) predictions in the MSB maize model. The approach of incorporating better LAIg estimates into crop simulation models may not offer a panacea for problem solving; this approach is limited in its ability to simulate other factors influencing crop yields. On the other hand, the approach of relating key crop biophysical parameters at the optimum stage with maize grain final yields is a robust technique to early FY estimation over large areas. Results suggest that estimates of LAI g obtained during the mid-grain filling period can used to detect variability of maize grain yield and this technique offers a rapid and accurate (RMSE \u3c 900 kg ha-1) method to detect FY at county level using MODIS 250-m products