Error modeling of simulated reflectivity observations for ensemble Kalman filter assimilation of convective storms

Abstract

The impact of two different ways of modeling errors in simulated radar reflectivity data for observing system simulation experiments (OSSEs) with an ensemble Kalman filter is investigated. An error model different from the one used in earlier studies is introduced, and it specifies relative Gaussian-distributed errors in the linear domain of the equivalent radar reflectivity factor. This model is consistent with the processes of error propagation in real radar data. When the error variances specified in the filter and in the data are consistently smaller or larger, the analysis is more accurate, but when these values do not match, poorer analyses result. Such behaviors agree with expectation but are not observed when errors are directly added to the reflectivity in the log domain. These results point to the importance of properly modeling observation errors in OSSEs when the observation operator is nonlinear. Copyright 2007 by the American Geophysical Union.