Targeted observations for radar data assimilation

Next generation weather radars, such as phased array systems will likely have a great operational flexibility. We want to take advantage of this flexibility by targeting radar observations to optimize the acquired information. From the perspective of data assimilation, targeted radar observation is the method to find the optimal observations required to obtain the smallest uncertainties in analysis and forecast. In this research, some simple experiments are performed to examine the impact of targeted radar observation on analysis only. Kalman filter equations are used to calculate the total variance of analysis as a measurement of how well the assimilated observations contribute to the data assimilation system. From these experiments, the optimal observation quantity and optimal observation density are found under different assumptions. Under the assumption that the observation error structure is known exactly, as many as possible observations are optimal, which is quite obvious because more information is beneficial for data assimilation. However, if the observation error structure is not exactly known, providing as many as possible observations is not the best choice. The optimal number of observations depends mostly on the forecast error correlation distance when it is smaller than the observation error correlation distance. Otherwise, the optimal observation quantity depends on both observation and forecast error correlation distances. What's more, the experiments prove that the optimal observation density exists and is a function of both forecast and observation error structures.