As new remote sensing instruments and data become available their utility for improving established terrestrial monitoring tasks need to be evaluated. In this study an empirical, generalized remotely sensed based yield model was developed and successfully applied at
the state level in Kansas using daily, high quality 0.05° NDVI timeseries data to drive the regression model, a percent crop mask as a filter to identify the purest winter wheat pixels, and USDA NASS county crop statistics for model calibration. The model predictions of production in Kansas closely matched the USDA/NASS reported numbers with a 7% error. This empirical regression model that was developed in Kansas was successfully applied directly in Ukraine. The
model forecast winter wheat production in Ukraine six weeks prior to harvest with a 10% error of the official production numbers. In 2009 the model was run in real-time in Ukraine and forecast production within 7% of the official statistics which were released after the harvest. This model is simple, has limited data requirements, and can offer an indication of winter wheat production, shortfalls and surplus prior to harvest in regions where little ground data is available. As the results of this study are promising, the performance of the developed
model should now be assessed in the other main winter wheat growing regions of the world.