基于作物模型与叶面积指数遥感影像同化的区域单产估测研究

Assimilating remotely sensed LAI into GIS-based EPIC model for yield assessment on regional scale

通过对作物光合、呼吸、蒸腾、营养等一系列生理生化过程的定量模拟,作物生长模型已经被成功应用于田间尺度的作物单产研究。为了进一步将作物模型扩展应用于区域尺度,提高区域作物单产的模拟精度,该文探讨了将作物模型与多时相叶面积指数(LAI)遥感影像同化以改善区域单产估测的方法。研究首先通过地理信息系统将美国农业部开发的"考虑气候的作物环境决策模型"——EPIC模型,扩展为空间模型。然后,通过基于Landsat TM影像差值植被指数DVI与田间观测叶面积指数构建的最优回归模型,反演了研究区域的多时相叶面积指数影像。最后通过优化算法实现了空间EPIC模型与影像信息的同化,并将系统应用于河北石家庄地区2004年冬小麦的单产估测。结果表明,通过数据同化校正部分关键参数后的空间作物模型的单产模拟精度得到有效提高,但要达到业务运行精度仍有待进一步改善。

Many crop growth models have been used successfully for simulating the physiological development, growth, and yield of a crop on field scale. To upscale crop models to large areas and to improve accuracy of yield assessment on regional scale, this article describes the development of a methodology for assimilating remotely sensed LAI into crop growth model. In this study, the Environmental Policy Integrated Climate（EPIC） model from United States Department of Agriculture （USDA） is integrated with Geographical Information System （GIS） by a Loose Coupling Approach, firstly. Then, the empirical Leaf Area Index（LAI） maps, which are retrieved from multi-temporal Landsat TM images by Regression Analysis Procedure, are assimilated with the GIS-based EPIC model by an optimization algorithm and Lookup-Table method. Some key parameters for LAI simulation in EPIC model, such as DMLA（the maximum potential LAI） and DLAI（the fraction of growing season when leaf area started declining）, are recalibrated through data assimilation. Finally, the methodology is applied during the 2004 crop season in 28 counties of North China Plain to simulate the yield of winter wheat. The result of yield assessment indicates that the data assimilation has significantly improved the accuracy of yield simulation by the GIS-based crop growth model on regional scale. The correlation coefficients （R2） between statistic yield and simulated yield at county level have been changed from 0.12 to 0.51.