%0 Conference Proceedings
%4 sid.inpe.br/mtc-m19@80/2010/07.21.16.50
%2 sid.inpe.br/mtc-m19@80/2010/07.21.16.50.01
%T optimization firefly method for weighted ensemble of convective parameterizations: part II: sensitivity experiment using TRMM satellite data
%D 2010
%A Santos, Ariane Frassoni dos,
%A Freitas, Saulo Ribeiro de,
%A Luz, Eduardo Favero Pacheco da,
%A Velho, Haroldo Fraga de Campos,
%A Gan, Manoel Alonso,
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@electronicmailaddress ariane.frassoni@cptec.inpe.br
%@electronicmailaddress saulo.freitas@cptec.inpe.br
%@electronicmailaddress eduardo.luz@lac.inpe.br
%@electronicmailaddress haroldo@lac.inpe.br
%@electronicmailaddress manoel.gan@cptec.inpe.br
%B The Meeting of the Americas.
%C Foz do Iguaçu
%8 8-12 Aug. 2010
%I AGU
%S Posters
%K numerical solutions, convective processes.
%X The model simulation of a classical case of South Atlantic Convergence Zone (SACZ) occurred on 21 March 2004 using the Coupled Chemistry-Aerosol-Tracer Transport-Brazilian developments on the Regional Atmospheric Modeling System (CCATT-BRAMS) was performed. The convective parameterization scheme of Grell and Dévényi is used to represent clouds and their interaction with the large scale environment. The Grell-Dévényis method considers an ensemble of several methodologies of cloud parameterizations. The model was run 6 times with different choice of parameterizations. Five experiments used only one type of parameterization mentioned above and provide five different responses for the rainfall. The 6th experiment ran with all parameterizations, but the precipitation field was computed by an average among the members of the ensemble. The 5 simulated precipitation fields were used as the direct problem (see companion paper), and the precipitation field estimated by the Tropical Rainfall Measuring Mission (TRMM) satellite as the observed data. The inverse problem is solved as an optimization problem with regularization operator of Tikhonov of zero order. For determining the best combination of the ensemble members reproducing the TRMM measurement, the quadratic difference between simulation fields and the observed data ( , where W= {w1,..., w5} is the weight vector to be calculated) should be minimized. The objective function is optimized by the Firefly (FA) method. Each firefly represents a five dimension candidate solution, and the brightest firefly represents the best set of weights to weight the schemes of convective parameterization. Some tests were carried out with FA optimizer to set up the parameters of the algorithm. The best results were obtained with 5 fireflies and 10 iterations. The retrieved field of precipitation was in agreement with the observed field. We computed the error precipitation field obtained with simple ensemble average and the error to the retrieved precipitation: errors of the ensemble average are greater than the errors of the retrieved precipitation. We expect to employ the method introduced here to improve the simulated precipitation of the CCATT-BRAMS system.
%@language en
%3 Santos_optimization.pdf