%0 Journal Article
%@nexthigherunit 8JMKD3MGPCW/43SQKNE
%@holdercode {isadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S}
%@archivingpolicy denypublisher denyfinaldraft24
%@issn 0022-1694
%@resumeid
%@resumeid 8JMKD3MGP5W/3C9JHC3
%@resumeid 8JMKD3MGP5W/3C9JHUB
%@usergroup administrator
%@usergroup fabia
%3 Ramirez_Artificial neural network.pdf.pdf
%@dissemination WEBSCI; PORTALCAPES; MGA; COMPENDEX.
%@secondarykey INPE-12422-PRE/7726
%A Valverde Ramirez, Maria Cleofe,
%A Campos Velho, Haroldo Fraga de,
%A Ferreira, Nelson Jesus,
%T Artificial neural network technique for rainfall forecasting applied to the Sao Paulo region
%B Journal of Hydrology
%8 Jan.
%D 2005
%V 301
%N 1-4
%P 146-162
%2 cptec.inpe.br/walmeida/2005/02.23.10.20.04
%4 cptec.inpe.br/walmeida/2005/02.23.10.20
%K artificial neural network, statistical rainfall forecast,  multiple linear regression,  Regional ETA model.
%X An artificial neural network (ANN) technique is used to construct a nonlinear mapping between output data from a regional ETA model ran at the Center for Weather Forecasts and Climate Studies/National Institute for Space Research/Brazil, and surface rainfall data for the region of São Paulo State, Brazil. The objective is to generate site-specific quantitative forecasts of daily rainfall. The test was performed for six locations in São Paulo State during the austral summer and winter of the 1997/2002 period. The analysis was made using a feedforward neural network and resilient propagation learning algorithm. Meteorological variables from the ETA model (potential temperature, vertical component of the wind, specific humidity, air temperature, precipitable water, relative vorticity and moisture divergence flux) are used as input data to the trained networks, which generate rainfall forecast for the next time step. Additionally, predictions with a multiple linear regression model were compared to those of ANN. In order to evaluate the rainfall forecast skill over the studied region a statistical analysis was performed. The results show that ANN forecasts were superior to the ones obtained by the linear regression model thus revealing a great potential for an operational suite.
%@language English
%@copyholder SID/SCD
%@e-mailaddress atus@cptec.inpe.br
%@secondarytype PRE PI
%@area MET
%@group DOP-INPE-MCT-BR
%@affiliation CPTEC-INPE-Cachoeira Paulista-12630-000-SP-Brasil