由于初始土壤水分、灌溉量等变量的空间分布不易获得，区域尺度水分胁迫条件下作物生长模拟存在一定难度。本文在WOFOST模型本地化和区域化的基础上，采用调控型方法，重点探讨了利用MODIS数据反演的地表蒸散在大范围内估算土壤水分平衡过程中的参数或变量初始值，以实现水分胁迫条件下作物模型区域模拟的可行性。2002年模拟结果显示，引入遥感信息优化获得初始土壤有效含水量、返青期生物量及抽穗期灌溉量后，土壤水分的模拟效果得到改善；32个农业气象试验站点模拟产量的相对均方根误差（RRMSE）由0.63降至0.20；华北冬小麦模拟产量的空间分布与实际产量分布更加接近，产量低估的情况得到较好改善；河北、河南、山东3省平均产量的模拟误差分别为－4.9%、4.3%和8.6%。初步结果表明，结合卫星遥感信息通过优化方法在大范围内估算作物模型的相关参变量，以实现水分胁迫条件下作物模型的区域应用是行之有效的。
Accurate crop growth monitoring and yield forecasting are significant to food security and sustainable development of agriculture. However, Regional crop growth simulation under water stress faces the difficulties in determining the spatial distribution of some model parameters and initial conditions, such as initial available soil water and irrigation. It appears to be a big potential in this field to couple remote sensing data with crop model. In this paper, we proposed a way of combining evapotranspiration derived from satellite remote sensing data with crop grow simulation model (WOFOST) under water stress. Some modifications of WOFOST model were performed with field experimental data to make it applicable in North China Plain. The combination method was first applied to simulate the growth, development and yield formation processes for winter wheat at two sites, Tai’an and Zhengzhou, during the growing season from 2001 to 2002. According to the results of sensitivity analysis, the initial available soil water was chosen to be recalibrated by observed evapotranspiration derived from MODIS data based on SEBS model (Surface Energy Balance System). Also the biomass at reviving and irrigation at heading stage were selected to re-estimated by observed SAVI and evapotranspiration, considering over-winter process and the importance of irrigation on winter wheat yield formation in North China. The difference between observed and simulated evapotranspiration/SAVI was minimized by re-initializing/re-parameterizing three chosen initial conditions/parameters with an optimization program (FSEOPT). The estimated values of initial available soil water and irrigation showed good agreement with observations at the two sites. And the relative errors of simulated dry matter weight of gross above-ground and storage organ were reduced also. On the basis of the regionalization of weather data, model parameters, and initial conditions, we used this method to estimate winter wheat yields in North China during the growing season from 2001 to 2002 at the scale of 0.25 degrees, especially for Henan, Hebei, and Shandong provinces. It was showed that both soil water estimates and final winter wheat production estimates were consistent with ground measurements since the initial available soil water, biomass at reviving and irrigation at heading stage were recalibrated by remote sensing data. The relative root mean square error (RRMSE) decreased from 0.63 to 0.20 for the yield from 32 experimental sites, which distribute uniformly in North China Plain. Also the aggregated yields for three provinces were improved, with relative errors －4.9%, 4.5%, and 8.6%, respectively. These results illustrated that the evapotranspiration derived from MODIS data could be used to improve the winter wheat yield estimate under water stress on a regional scale. Further study should focus on better understanding of processes, error accumulation, and improvement on validation of both evaportranspiration derived from MODIS data and simulated yields for winter wheat.
全 文 ：