Regional Crop Yield Assessment by Combination of a Crop Growth Model and Phenology Information Derived from MODIS

This study presents a method for crop yield assessment at regional scales. It integrates a crop growth model with phenology information derived from MODIS LAI product. This research focused on South Hebei province, an alluvial plain which is a primary winter wheat production region in North China. Previous experiment shows that, although the MODIS LAI product (MOD15A2) was generally underestimated in North China, it resembles the phenological development trend of ground measured LAI. A data assimilation scheme based on SCE (Shuffled Complex Evolution) global optimization algorithm was introduced. Two sets of the phenological transition dates of the winter wheat growing seasons were extracted: one from model simulation, and the other from MODIS LAI products. Then, the initial conditions of SWAP model was re-adjusted by minimizing the difference of the phenological transition days between the two sets based on the SCE algorithm. The results indicated that the method with assimilation have a lower RMSE 5.4% in 2007 and 15.4% in 2008 compare to the method without assimilation, suggesting a moderate improvement achieved by the assimilation method. Furthermore, this method can be also extended by using phenology information derived from MODIS Vis (vegetation index) or other satellite data.