基于集合卡尔曼滤波的PyWOFOST模型在东北玉米估产中的适用性验证

以叶面积指数(LAI)为结合点,引入基于集合卡尔曼滤波(Ensemble Kalman Filter,EnKF)的作物模型-遥感信息耦合模型PyWOFOST,利用气象数据、农业气象记录观测表数据及MODIS LAI数据检验PyWOFOST模型在东北玉米种植区的适用性,并选取在研究区内均匀分布、覆盖所有玉米品种且具有有效MODIS LAI数据的16个玉米农气站点,模拟该模型在不同的TSUM1(出苗-开花期积温)不确定性水平下各站点的玉米产量及LAI。结果表明,与WOFOST模型相比,PyWOFOST模型对LAI和产量的模拟能力都有极大提高。当TSUM1的不确定性为0、10、20、30℃时,PyWOFOST模拟的产量平均误差分别为10.32%、9.25%、7.31%和8.49%,均较未同化LAI观测数据的WOFOST模拟的产量平均误差(10.55%)低;同化后模拟LAI与同化前模拟LAI相比,其轨迹更接近实测值,更符合玉米的生长发育趋势,表明基于EnKF的PyWOFOST模型在东北玉米种植区有较好的适用性。

Applicability of PyWOFOST Model Based on Ensemble Kalman Filter in Simulating Maize Yield in Northeast China

PyWOFOST model based on Ensemble Kalman Filter （EnKF） was introduced to assimilate LAI into crop model. Meteorological data, agro-meteorological data and MODIS LAI data were used to test the applicability of PyWO- FOST model in simulating maize yield in Northeast China. 16 agro-meteorological stations with effective MODIS LAI data which distributed evenly and contained all maize varieties in study area were selected to model maize LAI and yield on each stations at different levels of uncertainty of TSUM1 （Thermal time from emergence to anthesis）. The re- sult showed that, compared with WOFOST model, the PyWOFOST model greatly improved in simulating LAI and yield of maize. The mean errors of maize yield simulated by PyWOFOST were 10. 32% ,9.25% ,7.31% and 8.49% at the uncertainty of TSUM1 with 0,10,20,30~C respectively which all were lower than the mean error of maize yield （ 10. 55% ） simulated by WOFOST without assimilating LAI. The trajectory of LAI simulated by PyWOFOST which was more in line with maize growth and development trends was closer to observed LAI than LAI simulated by WO- FOST. Therefore, the PyWOFOST based on EnKF was applicable for yield simulation of maize in Northeast China.