Lessons learnt from the 1D+4D-Var assimilation of rain and cloud affected SSM/I observations at ECMWF

Lessons learnt from the 1D+4D-Var assimilation of rain and cloud affected SSM/I observations at ECMWF
Date Published
EUMETSAT/ECMWF Fellowship Programme Research Reports
Document Number
Philippe Lopez
Event Series/Collection
EUMETSAT/ECMWF Fellowship Programme
Abstract Cloud and rain affected observations from SSM/I have been assimilated operationally at ECMWF since June 2005, using a 1D+4D-Var method. This paper examines the performance of the system, using departure statistics, forecast verification, observing system experiments (OSEs) and comparisons to independent rainfall observations from the precipitation radar on the Tropical Rainfall Measuring Mission (TRMM). Some improvements to the system are also described. Cloud and rain affected SSM/I observations are shown to benefit forecast scores, principally in the tropics, and they benefit not just relative humidity but also vector wind. However, there are many areas where the system could still be improved. First, in the 1D+4D-Var method used here, only the TCWV amount retrieved by 1D-Var is passed to 4D-Var for assimilation. Cloud and rainfall amounts are not passed into the 4D-Var analyses. It appears that the forecast benefits come from the ability to assimilate total column water vapour in cloudy and rainy areas, which are not otherwise observed by satellites. The system is not yet making use of the observational information on cloud and rain. Second, the simplified moist physics operators used in the 1D-Var retrieval are shown to produce excessive amounts of rain compared to observations. A theme underlying all this work is the inability of the first guess to correctly predict the location and intensity of rainy and cloudy areas observed by SSM/I. This means that in 1D-Var, rain will often have to be completely removed in one retrieval, and completely added in another. Because of this, when observations are assimilated in separate observationally-determined `clear-sky' and `rainy' streams, a large sampling bias can be created. Our recommendation is that in future all satellite observations with a rain and cloud information content are assimilated in a single stream, including cloudy and rainy radiative transfer when required.
URL https://www.ecmwf.int/en/elibrary/74555-lessons-learnt-1d4d-var-assimilation-rain-and-cloud-affected-ssmi-observations