基于历史样本投影的四维变分陆面数据同化方法及其初步应用

Historical-Sample-Projection Four-Dimensional Variational Land Surface Data Assimilation and Its Preliminary Application

法面临着计算量上的挑战。本研究将一种历史样本投影的四维变分同化方法(Historical-Sample-Projection4DVar,简写为HSP-4DVar)应用于陆面数据同化,建立起CoLM陆面模型的HSP-4DVar系统。相比其他四维变分同化方法,HSP-4DVar的分析值是显式求解,不需要编写和使用伴随模式,从而大大节省了计算量,是一种易于实现的同化方案。通过同化56个月的土壤湿度观测数据表明,新的陆面同化系统不仅省时,而且能够有效吸取观测信息,使得同化后的均方根误差显著降低,各层土壤湿度模拟都有所改善,陆表1000mm层的改善最为明显。

Data assimilation has been successfully applied in atmospheric,oceanic,and land surface models.However,the four-dimensional variational(4DVar)assimilation system demands great computational costs.The authors introduced a new Historical-Sample-Projection data assimilation scheme(HSP-4DVar),and accomplished the HSP-4DVar land surface data assimilation system based on the Common Land Model(CoLM).As a scheme which requires no adjoint models,HSP-4DVar can be directly solved and easily realized,therefore avoids high computational costs.The land surface data assimilation system was used to assimilate the soil moisture data for 56 months.After assimilation,the overall root-mean-square error was significantly reduced with improved simulations,especially the simulation for the top 1000 mm layer.