Introduction
The rain retrieval algorithm used within the Megha-Tropiques framework is called BRAIN and is described in more details Viltard et al. (2006). The general principle of the algorithm is somewhat similar to the Goddard Profiling algorithm (GProf) in the sense that a Bayesian/Monte-Carlo scheme is used to retrieve the precipitation field from the MADRAS measured brightness temperatures.
General Principle
To do so, BRAIN relying on a retrieval database that contains a set of all possible (in theory) rain-TBs pair and the retrieved rain is given by the weighted average of the Rs of the neighboring points in the TB space. The weights affected to each point of the database is proportional to its distance from the measured set of TBs in the TB space. BRAIN will retrieve on a pixel-by-pixel basis an estimate of the instantaneous surface rain ( in mm.hr-1) and the associated "precipitation profile". This precipitation profile is made of 28 layers equally spaced of 0.5km between 0 and 10km and 1km between 10 and 18 km. For each layer 4 species are retrieved (in g.m-3): the ice cloud content; the liquid cloud content; the ice precipitation content and the liquid precipitation content. In addition, the surface rain due to convective precipitation (mm.hr-1) is also given for each pixel which allows to compute a "convective" rain fraction. The nominal resolution of BRAIN depends on the considered instrument and is 20km in the particular case of MADRAS.
Retrieval Database Specifics
As mentioned previously these retrieved variable represent the most "probable" solution, in the Bayesian sense, that was computed from the retrieval database. Hence, the surface rain, the 4 variables of the profile and the convective fraction are stored in the retrieval database along with the vector of brightness temperatures associated. The specificity of BRAIN lies actually in the database that was originally built from TRMM’s PR profiles and TMI brightness temperatures. Since the PR does not give a very detailed profile of the ice phase and no information on the cloud, this rain profile is completed with ice and cloud content from a pool of meso-scale cloud resolving model simulations. The retrieved profile is thus a mix of two sources of data, the ice phase and cloud content from model and the rain profile from PR. This is why the rain statistics from BRAIN can be found to be correlated to the PR rain statistics.
Land-Ocean-Coast
BRAIN will give a rain rate over both oceanic and land surfaces but the user should be aware of the fact that the retrieval over each surface is not performed in exactly the same way since over land only the channels above 80GHz are used while all the channels are used over ocean. This might lead in some cases to possible discontinuities in the rain field along the coastal regions or in the transition between ocean and land. This artifact is unfortunately inherent to the use of passive microwave measurements to retrieve rain.
Validation
The validation activities are an ongoing effort nevertheless first results of comparison over West Africa in the framework of AMMA show that an accuracy on the surface rain intensity of about 40% is reached over land and about 20% over ocean. These numbers are preliminary and an error model is being developed and should be made available asap.The retrieved profiles even if their consistency has been tested should be considered as experimental products which validation is still ongoing.
Contact
Any questions should be directed to N. Viltard (nicolas.viltard-at-latmos.ipsl.fr).
Reference
Viltard, N., C. Burlaud, and C.D. Kummerow, 2006: Rain Retrieval from TMI Brightness Temperature Measurements Using a TRMM PR-Based Database. J. Appl. Meteor. Climatol., 45, 455-466.