Remote Sensing Estimation Of Winter Wheat Water Demand Based On AquaCrop Model

Winter wheat, the main food crops in north China, is one of high water-intensive crops. The balance between water-saving irrigation and yield production as well as the mechanism of irrigation amount and yield response is the foundation of reasonable irrigation decision for winter wheat. In this study, the field experiment data of winter wheat in National Precision Agriculture Experimental Base of Xiaotangshan town in 2013/2014 and 2014/2015 are used to evaluate sensitive parameters of Aqua Crop model through EFAST(Extended Fourier Amplitude Sensitivity) method. And then, the localized parameters of Beijing area are calibrated to simulate biomass and yield of the winter wheat. Finally, effective irrigation and water demand of winter wheat is assessed by the localized Aqua Crop model combined with remote sensing data. The main contents and conclusions are as follows:(1) EFAST analysis shows that wp(water productivity normalized Water Production sensitive index of 0.93), sen(period from sowing to start senescence, sensitive index of 0.41), stbio(minimum growing degrees for full biomass production, sensitive index of 0.14) and cc(crop coefficient, sensitive index of 0.13) are most sensitive parameters for biomass simulation. The parameters Sen, flo( period from sowing to flowering, sensitive index of 0.45), hi(reference harvest index, sensitive index of 0.39) and mat(total length of crop cycle from sowing to maturity, sensitive index of 0.24) had strong sensitivities to wheat yield simulation.(2) The sensitive parameters were adjusted and localized by trail-and-error method using the field experiment data of 2014/2015. We got the localized model with the absolute coefficient R~2 of 0.96 and RMSE of 4.8% to canopy coverage. The absolute coefficient R~2 and RMSE was 0.95 and 1.29 t/ha, respectively. It indicated that the model had the better suitability in Beijing area.(3) Four growing period(jointing, booting, anthesis and filling stage) of GF-1WFV images in 2015 are used to evaluate the ground biomass content. Five vegetation indices such as NDVI, EVI and TVI are used to the Partial Least Squares Regression Model. Compared the predicted value with the ground measuring value, the regression model had good precision with calibrationR~2 of 0.8073 and RMSE of 1.66 t/ha, and simulation validation R~2 0.5297 and RMSE 3.02 t/ha.(4) Particle Swarm Optimization algorithm method was introduced to the assimilation algorithm between remote sensing and crop growth model. The field wheat biomass image used to the state variable drove the model to simulate the real effective irrigation according to the current biomass. The irrigation supplement was made out based on the better growing biomass to provide irrigation decision making support reference.