%0 Electronic Source
%@holdercode {isadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S}
%@nexthigherunit 8JMKD3MGPCW/3ESGTTP
%@nexthigherunit 8JMKD3MGPCW/3EUPEJL
%@nexthigherunit 8JMKD3MGPCW/444BQ9E
%2 sid.inpe.br/mtc-m19@80/2010/07.12.17.27.42
%@usergroup administrator
%@usergroup estagiario
%4 sid.inpe.br/mtc-m19@80/2010/07.12.17.27
%X Statistical downscaling method (SD) intertwined with computational intelligent techniques for quantitative daily rainfall forecasting of the SACZ-ULCV weather patterns is proposed in this paper. The SD precipitation forecasting is achieved first with the support of artificial neural network (ANN) and latter with fuzzy logic (FL). The SACZ-ULCV occurs when the cloudiness of the upper level cyclonic vortices (ULCV) in the Brazilian Northeast meets the South Atlantic Convergence Zone (SACZ) enhancing convection and cloudiness over the Southeastern region of Brazil. This weather pattern is responsible for severe rainfalls and thunderstorms. Finding out a manner to anticipate the severe rainfall produced by SACZULCV is of vital importance for alerting, or avoiding disasters. The daily surface rainfall of the southeastern, in 12 major urban centers over the state of São Paulo, is the output while the dynamical meteorological variables from ETA regional model are the inputs. The ETA regional model simulates the large scale dynamical and thermodynamical behavior of the SACZ-ULCV weather pattern. For this reason, meteorological variables from ETA model are used to generate statistical downscaling for the periods of occurrence of the SACZ-ULCV in summer from 2000 to 2003. Afterwards, the statistical models are extended to the entire summer period including, thus, other weather patterns, beside of SACZ-ULCV, for comparative analysis. Quantitative daily rainfall forecasting results to events of SACZ-ULCV had their performance improved when was considered only the model training for SACZ-ULCV periods. The results confirm that the rain forecast can be improved when used as predictors dynamical variables obtained from similar weather patterns. On the other hand, the using FL or AAN models were efficient techniques as auxiliary mechanisms for SD. Further, both techniques accomplished better performance when compared to the ETA meteorological model forecasting.
%@mirrorrepository sid.inpe.br/mtc-m19@80/2009/08.21.17.02.53
%T artificial neural network and fuzzy logic statistical downscaling of atmospheric circulation-type specific for rainfall forecasting
%@electronicmailaddress maria.valverde@cptec.inpe.br
%@electronicmailaddress ernesto.araujo@lit.inpe.br
%@electronicmailaddress haroldo@lac.inpe.br
%K fuzzy model, artificial neural network, statistical downscaling, quantitative rainfall forecasting.
%1 submitted
%@group CPT-CPT-INPE-MCT-BR
%@group LIT-LIT-INPE-MCT-BR
%@group LAC-CTE-INPE-MCT-BR
%3 v1.pdf
%C São José dos Campos
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%8 2010-07-13
%@project mudanças climaticas
%I Instituto Nacional de Pesquisas Espaciais
%A Valverde, Maria Cleófe,
%A Araujo, Ernesto,
%A Campos Velho, Haroldo,
%@area MET
%9 On-line