Retrieving the spatial-explicit harvest index for winter wheat from NDVI time series data

Crop harvest index (HI) is critical to crop yield estimation and prediction. In this paper, a novel technique to extract spatial-explicit HI information for winter wheat from remotely sensed data was developed. Based on the understanding of the intrinsic features of HI and crop phenology expressed by NDVI time series data, a derivative index HINDVI-SUM for the inversion of HI from remotely sensed data was devised. HINDVI-SUM was defined as the ratio between accumulated 10-day NDVI for reproductive and vegetative stages of winter wheat. Then the relationship between HINDVI-SUM and the in-situ HI of winter wheat was tested in the study area in North China Plain. Finally, the established relationship to retrieve the spatial-explicit crop harvest index information from remote sensing data was applied in the study area. After validation, the accuracy of the retrieved HI of winter wheat was satisfactory. The mean relative error of the retrieved crop HI was only 2.40% and RMSE was only 0.02. It was shown that the method developed to construct HINDVI-SUM and extract the harvest index for winter wheat from MODIS NDVI time-series data was feasible and reliable.