Representing model error and observation error uncertainty for data assimilation of polarimetric radar measurements

点击翻译按钮获取中文摘要

Data assimilation compares each observation of a variable with a prior estimate of the variable taken from a discrete dynamical model, in order to deduce a revised estimate on the model grid. This process requires knowledge of, or assumptions on, the error characteristics or uncertainty properties of the observed values and prior estimates. One of the difficulties encountered in this process is that the discrete geophysical model is not able to represent all the physical processes, nor all of the spatial and temporal scales, of the observed geophysical state and that additional approximations are needed to represent the equivalent of any observation. These uncertainties need to be accounted for in data assimilation algorithms through the model error and observation error statistics. Only this way, we can optimally initialize convection-permitting models with polarimetric radar measurements. The goal of this proposal is to find an optimal approach for assimilating polarimetric radar measurements with an ensemble Kalman filter. To this end, strategies of perturbing hydrometeors during data assimilation to improve their forecast accuracy will be investigated; we will explore benefits of correlated observation error statistics for polarimetric radar measurements. Finally, representation error part of the observation error will be parameterized using high-resolution simulations and included in data assimilation algorithm. The aim of the proposal is to address the objective: "generation of precipitation system analyses by assimilation of polarimetric radar observations into atmospheric models for weather forecasting".