Optimizing soybean harvest date using HJ-1 satellite imagery

Abstract

Satellite imagery provides contiguous spatial coverage of a field and can be used to measure crop and soil attributes. In the last decade, remote sensing has been proved useful in guiding field management such as sowing, irrigation and fertilization, yet its potential in optimizing large-area crop harvest has not been explored. With a limited number of observations from Hongxing Farm in Northeast China, this paper first analyzes the influence of harvest date on yield. Optimal harvest date (OHD) was estimated for 41 sites in total based on maximum yield. Then the indicating ability of seven satellite-derived indices at different maturing stages for OHD was analyzed. These indices included NDVI (normalized difference vegetation index), soil adjusted vegetation index (SAVI), enhanced vegetation index (EVI), green normalized difference vegetation index (GNDVI), visible atmospherically resistant index (VARI), chlorophyll vegetation index (CVI) and normalized difference water index (NDWI). The analysis showed that there was a continuous increase of yield before maximum yields were reached and then a rather fast decrease occurred. All seven indices acquired from HJ-1 satellite are capable of predicting soybean OHD; EVI and NDWI performed better than other indices with higher correlation coefficients. The highest correlation coefficients between remote sensing indices and observed optimal harvest date were 0.723 and 0.720 for EVI and NDWI respectively. The temporal variation of correlation coefficient between seven indices and observed OHD suggests that the best time for OHD prediction is the period 2–3 weeks earlier than the general harvest. Stepwise regression analysis was undertaken and optimum regression expressions were constructed, taking optimal soybean harvest date as the dependent variable and remote sensed indices (EVI, NDWI and NDVI at suitable dates) as independent variables. An optimal soybean harvest date map of Hongxing Farm was produced with average standard error of 1.15 days. This knowledge may be of use when determining the best times to harvest soybean to obtain a higher yield.