Main Article Content
In Sri Lanka, a conventional method named the Crop-cut survey is currently used to estimate the seasonal rice production. However, it fails to predict rice yield before the harvest as it is conducted during the harvest. Therefore, this study focuses on testing rice yield estimation models based on satellite remote sensing data. Landsat 8 OLI/TIRS images (30m spatial resolution) from Earth Explorer and 8-day composite images (250 m spatial resolution) from Moderate Resolution Imaging Spectro-radiometer (MODIS) sensor onboard NASA EOS Terra/Aqua satellite from 2014 to 2017 were used. Cultivated paddy lands were identified by land cover classification, using field-training samples and Landsat 8 OLI/TIRS data. In addition, the temporal change of Normalised Differenced Vegetation Index (NDVI) for paddy and forest were analysed to validate the classification. The observed minimum accuracy of the land cover classification, out of the tested four (4) seasons, was 99.4%, and the minimum Kappa coefficient was 0.9916. The correlation between reference net harvested paddy area and identified paddy cultivated area by Landsat 8 was 0.93. Linear and exponential yield forecasting models developed for Kurunegala District were validated and tested, based on NDVI and EVI2 (Enhanced Vegetation Index) indices obtained through MODIS surface reflectance images of Polonnaruwa District. Both NDVI and EVI2-based models derived after about 80-days of transplanting provide more reliable estimations with compared to national statistical records. Nevertheless, the EVI2-based model provides more reliable estimates than the NDVI-based model with 83.7% average accuracy. Therefore, the rice yield can be successfully estimated for each season before one-month to the harvest time using the EVI2-based model in Polonnaruwa district.