Cotton yield estimation model based on fusion image from MODIS and Landsat data

Deficiency in the spatiotemporal resolution of remote sensing images limits crop yield estimations at the farm and field scale. High spatiotemporal resolution fusion images blended from moderate resolution imaging spectro-radiometer (MODIS) and Landsat data may alleviate this problem. This paper selected California's San Joaquin Valley Sheely Farm as study area, using the (Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model) ESTARFM/(Flexible Spatiotemporal Data Fusion Model) FSDAF to fusion image, it can enhance the efficiency of the high spatial resolution remote sensing image for yield estimation. And this paper chose MOD09GA and Landsat images to fusion analysis respectively, then on the basis of the study of the original cotton estimation model, the estimation model of two cotton plots in the study area was established, and the evaluation of the accuracy was carried out. The results showed that: 1) compared with FSDAF model, the ESTARFM model and the fusion image result of ESTARFM and accuracy for the yield estimation was higher, indicating that the two reference Landsat images were superior than the single image for fusion; In addition, from the results of the study, the fusion effect of ESTARFM is better than that of FSDAF. The ESTARFM model referred to the crop growth in the early and middle stage of cotton growth. The fusion result was better than FSDAF model, and FSDAF only referred to the high precision of a reference point in the middle of crop growth. 2) the correlation coefficient between predicted NDVI by the fusion image and cotton yield of A block is higher than that of block B, and the shape of the plot and the type of crop around the block had influence on the fusion result. 3) the predicted NDVI mean value of the three-phase in the flower-blooming stage of cotton growth by fusion images was used for the cotton yield estimation model, and the results showed that it had a higher accuracy than the single stage. And the scatter diagram is more concentrated, which proves the time-series vegetation index played an important role on the crop yield estimation, so fusion images are more important; 4) the accuracy of fusion image estimation is slightly lower than that of Landsat TM image, but the accuracy is above 0.60, and the predicted NDVI of fusion images can also be used for the yield prediction model as the estimated production factor. This study predicts crop yields using fusion images at the field scale, which can be used as a reference for studying vegetation monitoring using remote sensing at the field scale.