Estimation of crop yield based on weight optimization combination and multi-temporal remote sensing data

Multi-temporal remote sensing data can cover more information related with yield than that of single-temporal, so it is of great significance to explore how to integrate the useful information from multi-temporal remote sensing data for improving the precision of estimating yield. WOC (weight optimization combination) is the algorithm which optimizes the weights of many models to form the combined model with higher precision. Taking the estimation of barley yield as an example in friendship farm, Heilongjiang Province, firstly four different temporal Landsat5 TM images were used to construct the single-temporal estimating models of barley yield, then applying the iteration algorithm of WOC to calculate the weights of the four models formed the new combined model, which was employed to estimate the barley yield finally. The results showed that the combined model based on WOC and multi-temporal remote images displayed better performance, and it was R2 (determinant coefficient) was remarkably improved in comparison with those of the single-temporal models. In addition, analyzing the weight values in the combined model showed that the size of weight of each single model was sensitive to the amount of yield information involved by the corresponding temporal satellite image, and that was of great importance for determining the key temporal satellite images to estimate crop yield.