Premium
Comparison of TMI rainfall estimates and their impact on 4D‐Var assimilation
Author(s) -
Marécal Virginie,
Mahfouf Jeanfrançois,
Bauer Peter
Publication year - 2002
Publication title -
quarterly journal of the royal meteorological society
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.744
H-Index - 143
eISSN - 1477-870X
pISSN - 0035-9009
DOI - 10.1256/qj.01.195
Subject(s) - data assimilation , environmental science , precipitation , rain rate , meteorology , quantitative precipitation forecast , climatology , assimilation (phonology) , radar , computer science , geography , geology , telecommunications , linguistics , philosophy
The objectives of this paper are to perform a comparison between three rainfall‐rate estimates from Tropical Rainfall Measuring Mission Microwave Imager (TMI) data and to study their impact in the European Centre for Medium‐Range Weather Forecasts (ECMWF) four‐dimensional variational (4D‐Var) assimilation system. The three algorithms for rainfall estimation considered are: Precipitation Radar Adjusted TMI Estimation of Rainfall (PATER), Bayesian Algorithm for Microwave‐based Precipitation Retrieval (BAMPR‐P), and the National Aeronautics and Space Administration operational algorithm 2A12 level 5. Results from the comparison show that BAMPR‐P and 2A12 retrievals provide on average higher rain rates by about a factor of 1.5 with respect to PATER, mainly because of the different spatial resolutions and rain detection methods used in the three algorithms. The three TMI products are then compared to the rainfall rates from the ECMWF model. Globally, the rain occurrence simulated by the ECMWF model is higher than in TMI estimates. The model also produces lower rainfall rates in precipitating systems. PATER retrievals are generally closer to the model than those from other algorithms. Three 15‐day assimilation experiments (‘Rain‐PATER’, ‘Rain‐BAMPR‐P’ and ‘Rain‐2A12’) were run using the three TMI rainfall‐rate estimates in the ECMWF 4D‐Var assimilation system. Fairly small differences were found between the global analyses and forecasts from all experiments and a control experiment without rainfall‐data assimilation. This is explained by the small number of TMI rain observations which are used per assimilation cycle compared to the other types of observational data. The local impact on analyses was studied in two cases, namely a tropical cyclone (‘Bonnie’) and a midlatitude front propagating into the subtropics. For these two cases, the differences between the three TMI products and their associated errors lead to significant changes in the humidity and wind analyses. Copyright © 2002 Royal Meteorological Society.