Improved Weather Prediction by Adaptive Data Assimilation

数值天气预报的改进主要集中在两个基本方面：模式模拟的改进和初始条件的改进。初始条件是利用数据同化技术从观测数据库导出的。资料同化需要所讨论的全球大气变量的观测误差统计以及预报误差统计。本文将重点研究与亚网格湍流相关的测量采样方式所导致的总观测误差部分。来自20公里水平分辨率数值天气预报模式档案的数据将用于研究湍流对各种测量系统误差统计变率的贡献，以期用状态依赖的“自适应”方法取代目前假设恒定观测误差统计的技术。

Improvements in numerical weather prediction are focused on two basic areas: improvements in the model simulations and improvements in the initial conditions. Initial conditions are derived from the observational database using data assimilation techniques. Data assimilation requires observational error statistics of the global atmospheric variables in question as well as forecast error statistics. This research will focus on the part of the total observation error that is due to the sampling pattern of the measurement related to the sub-grid turbulence. Data from a 20 km horizontal resolution numerical weather prediction model archive will be used to study the contribution of turbulence to the variability of the error statistics for various measurement systems with a view to replace current techniques that assume constant observation error statistics with "adaptive" ones that are state-dependent.