An analysis of the seasonal precipitation forecasts in South America using wavelets

Authors

  • Luciano Ponzi Pezzi,

    Corresponding author
    1. Center for Weather Forecast and Climate Studies (CPTEC), National Institute for Space Research (INPE), Cachoeira Paulista, SP, Brazil
    • Center for Weather Forecast and Climate Studies (CPTEC/INPE), Rodovia Presidente Dutra Km 40, Cachoeira Paulista, SP, 12630-000, Brazil.
    Search for more papers by this author
  • Mary Toshie Kayano

    1. Center for Weather Forecast and Climate Studies (CPTEC), National Institute for Space Research (INPE), Cachoeira Paulista, SP, Brazil
    Search for more papers by this author

Abstract

A post-processing technique was applied to statistically correct the seasonal rainfall forecasts over South America (SA). The aim of this work was to reduce errors in the seasonal climate simulations obtained from the Centro de Previsão de Tempo e Estudos Climáticos (CPTEC) atmospheric general circulation model (AGCM) which was run with different deep cumulus convection parameterizations. One of the main contributions of this study is the discussion of the super-ensemble approach to reduce errors in the seasonal rainfall prediction for SA. A novel aspect here is the use of the wavelet technique to compare forecast and observed time series by investigating their time-frequency structures. This methodology has not yet been applied to super-ensemble model validations. The statistical algorithm used in the super-ensemble technique was based on the linear multiple regression method. The time series of the super-ensemble forecast (FCT), arithmetic averaged forecast (MEM) and individual model forecasts and the observed (OBS) ones for selected areas of SA were compared by calculating the root mean square errors (RMSEs) and by applying the wavelet technique on these time series. In general, for the analysed areas we obtained a super-ensemble skill superior to that for the MEM. The wavelet analysis proved to be very useful to compare forecast and observed time series. In fact, differences and similarities among the time series such as the dominant scale of variability and the time location of the largest variances in the time series were detected with the wavelet analyses. Copyright © 2008 Royal Meteorological Society

Ancillary

Advertisement