Abstract
A non-Gaussian probability distribution function (PDF) and a new displacement correction method using PDF pseudo-regimes for precipitation were introduced to the dual-scale neighboring ensemble-based variational assimilation scheme (EnVar) to achieve improved assimilation of all-sky microwave imager (MWI) brightness temperatures (TBs) into a cloud-resolving model (CRM).
We evaluated the fits of the precipitation forecast perturbations of various disturbances with the existing non-Gaussian PDF models and selected a mixed lognormal distribution for the precipitation PDF model. Then, we introduced rain-free and rainy PDF regimes to EnVar. We developed a new method for correcting precipitation displacement that introduces pseudo rain-free, rainy, and heavy-rain regimes and approximated their PDFs as regional averages of the PDFs around the target point. We estimated the horizontal scales of averaging based on the similarity of precipitation forecast perturbations. These methods improved the bias and normality of TB differences between observation and the first guess.
We conducted assimilation experiments using all-sky MWI TB observations for Typhoon Etau (T1518). Results show that the precipitation analysis using the EnVar employed in this study was more similar to the global satellite map for precipitation (GSMaP) retrievals than those using a conventional EnVar. The introduction of a mixed lognormal PDF strengthened the precipitation analysis of heavy-rain areas around the typhoon and near a front. Using PDF pseudo-regimes considerably reduced the precipitation displacement error of the analysis. The EnVar employed in this study improved the CRM forecasts for precipitation distribution up to 12 h and the typhoon position and central surface pressure for > 24 h. The forecast analysis cycle of EnVar improved the CRM forecasts for heavy rain around the typhoon center up to 6 h and a heavy-rain band associated with the typhoon for > 24 h when compared with the EnVar using a single-time TB observation.
