Rice yield forecasting models using satellite imagery in Egypt

Ability to make yield prediction before harvest using satellite remote sensing is important in many aspects of agricultural decision-making. In this study, canopy reflectance band and different band ratios in form of vegetation indices (VI) with leaf area index (LAI) were used to generate remotely sensed pre-harvest empirical rice yield prediction models. LAI measurements, spectral data derived from two SPOT data acquired on August 24, 2008 and August 23, 2009 and observed rice yield were used as main inputs for rice yield modeling. Each remotely sensed factor was used separately and in combination with LAI to generate the models. The results showed that green spectral band, middle infra-red spectral band and green vegetation index (GVI) did not show sufficient capability as rice yield estimators while other inputs such as red spectral band, near infrared spectral band and vegetation indices that are algebraic ratios from these two spectral bands when used separately or in combined with leaf area index (LAI) produced high accurate rice yield estimation models. The validation process was carried out using two statistical tests; standard error of estimate and the correlation coefficient between modeled and predicted yield. The validation results indicated that using normalized difference vegetation index (NDVI) combined with leaf area index (LAI) produced the model with highest accuracy and stability during the two rice seasons. The generated models are applicable 90 days after planting in any similar environmental conditions and agricultural practices.