Implementation of 1D+4D‐Var assimilation of precipitation‐affected microwave radiances at ECMWF. I: 1D‐Var

Abstract This paper presents the operational implementation of a 1D+4D‐Var assimilation system of rain‐affected satellite observations at the European Centre for Medium‐Range Weather Forecasts. The first part describes the methodology and performance analysis of the 1D‐Var retrieval scheme in clouds and precipitation that uses Special Sensor Microwave/Imager microwave radiance observations for the estimation of total‐column water vapour. The second part shows the global and long‐term impact of these observations on both model 4D‐Var analyses and medium‐range forecasts. The 1D‐Var scheme employs a complex observation operator that consists of linearized moist physics parametrization schemes and a multiple‐scattering radiative‐transfer model. The observation operator shows rather linear behaviour in most situations except in the presence of very intense precipitation suggesting a possible use even for a direct assimilation of radiances in 4D‐Var. A bias correction and observation‐error estimation method were implemented and indicate stable error behaviour. The 1D‐Var algorithm quality control shows the largest failure number in areas with mostly frozen precipitation where the Special Sensor Microwave/Imager channels have little sensitivity to changes in hydrometeor contents. From test analyses on a global scale, a small moisture increase was computed that was greatest in dry subtropical areas. Large‐scale and convective precipitation were increased similarly but showed a significantly different geographical distribution. The large‐scale precipitation scheme has a stronger sensitivity to moisture changes and therefore moisture increments mainly affect stratiform precipitation distributions. While the global mean moisture fields are only weakly affected by the assimilation of rain‐affected observations, the impact on local systems may be quite large. The forecast of synoptic system development through the 4D‐Var analysis can be significant. Copyright © 2006 Royal Meteorological Society