一种新的资料同化方法

A NEW DATA ASSIMILATION APPROACH

中文关键词:  映射观测，变分资料同化，省时性

英文关键词:Mapped observation, Variational data assimilation, Timesaving.

中文摘要:

      为寻求一种快速有效的四维变分资料同化（英文缩写4DVar）作了有意义的尝试，提出了映射观测的新概念和反向四维变分资料同化的新思路，并以此为基础建立了三维变分映射资料同化（英文缩写为3DVM：3-Dimensional Variational data assimilation of Mapped observation）。该方法与传统的四维变分资料同化一样，不仅考虑了模式的动力和物理约束，使得同化后的初值与模式协调，而且通过模式方程对同化窗口中不同时刻的观测资料作了最佳拟合。与传统四维变分同化方法不同的是，由3DVM得到的初值不在同化窗口的始端，而在窗口的末端。正是所求初值时刻的改变，使得该方法的计算代价大大减少，几乎与三维变分资料同化（英文缩写3DVar）相当，这实际上是用3DVar的代价实现了4DVar的功能。同时，由于3DVM不再需要切线性和伴随近似来计算代价函数的梯度也提高了同化的精度。对具体的台风个例（Dan）用AMSU-A反演的温度场进行变分同化模拟试验，发现3DVM能比传统4DVar产生更好的初值，而且所花计算时间只需4DVar的1/7。

英文摘要:

      A significant attempt for designing a timesaving and efficient 4-dimensional variational data assimilation (4DVar) was made, and a new data assimilation approach called“3-Dimensional Variational data assimilation of Mapped observation (3DVM)” was proposed, based on the new concept of mapped observationand the new idea of backward 4DVar. Like the available 4DVar, 3DVM produces an optimal initial condition (IC) that is consistent with the prediction model due to the inclusion of dynamical and physical constraints of the model and is best fitting to the observations in the assimilation window through the model solution trajectory. Different from the 4DVar, the IC derived from 3DVM is not located at the beginning but the end of the assimilation window. It is the change of the IC time that makes the computing cost of the new approach greatly reduced. Actually, 3DVM costs almost the same as the 3-Dimensional Variational data assimilation (3DVar) does, but performs as same as the 4DVar does. Especially, it is able to improve the assimilation accuracy because it does not need the tangent linear and adjoint approximations for calculating the gradient of cost function anymore. The new approach produced better IC for 72- hour simulation of TY9914 (Dan) than 4DVar does, by assimilating the three-dimensional fields of temperature retrieved from the Advanced Microwave Sounding Unit-A (AMSU-A) observations. Meanwhile, it takes only 1/7 of the computing costs the 4DVar requires for the same initialization with the same retrieved data.