Dynamical Downscaling of Seasonal Climate Prediction over Nordeste Brazil with ECHAM3 and NCEP's Regional Spectral Models at IRI

View More View Less
  • 1 International Research Institute for Climate Prediction, Lamont-Doherty Earth Observatory of Columbia University, Palisades, New York
© Get Permissions
Restricted access

This study presents an evaluation of a seasonal climate forecast done with the International Research Institute for Climate Prediction (IRI) dynamical forecast system (regional model nested into a general circulation model) over northern South America for January–April 1999, encompassing the rainy season over Brazil's Nordeste. The one-way nesting is one in two tiers: first the NCEP's Regional Spectral Model (RSM) runs with an 80-km grid mesh forced by the ECHAM3 atmospheric general circulation model (AGCM) outputs; then the RSM runs with a finer grid mesh (20 km) forced by the forecasts generated by the RSM-80. An ensemble of three realizations is done. Lower boundary conditions over the oceans for both ECHAM and RSM model runs are sea surface temperature forecasts over the tropical oceans. Soil moisture is initialized by ECHAM's inputs.

The rainfall forecasts generated by the regional model are compared with those of the AGCM and observations. It is shown that the regional model at 80-km resolution improves upon the AGCM rainfall forecast, reducing both seasonal bias and root-mean-square error. On the other hand, the RSM-20 forecasts presented larger errors, with spatial patterns that resemble those of local topography. The better forecast of the position and width of the intertropical convergence zone (ITCZ) over the tropical Atlantic by the RSM-80 model is one of the principal reasons for better-forecast scores of the RSM-80 relative to the AGCM. The regional model improved the spatial as well as the temporal details of rainfall distribution, and also presenting the minimum spread among the ensemble members. The statistics of synoptic-scale weather variability on seasonal timescales were best forecast with the regional 80-km model over the Nordeste. The possibility of forecasting the frequency distribution of dry and wet spells within the rainy season is encouraging.

*On leave of absence from Centro de Previsão de Tempo e Estudos Climáticos, Instituto Nacional de Pesquisas Espaciais, Cachoeira Paulista, São Paulo, Brazil.

Corresponding author address: Dr. Paulo Nobre, International Research Institute for Climate Prediction, Lamont-Doherty Earth Observatory of Columbia University, 61 Route 9W, IRI Monell, Bldg., Palisades, NY 10964-8000. E-mail: pnobre@cptec.inpe.br

This study presents an evaluation of a seasonal climate forecast done with the International Research Institute for Climate Prediction (IRI) dynamical forecast system (regional model nested into a general circulation model) over northern South America for January–April 1999, encompassing the rainy season over Brazil's Nordeste. The one-way nesting is one in two tiers: first the NCEP's Regional Spectral Model (RSM) runs with an 80-km grid mesh forced by the ECHAM3 atmospheric general circulation model (AGCM) outputs; then the RSM runs with a finer grid mesh (20 km) forced by the forecasts generated by the RSM-80. An ensemble of three realizations is done. Lower boundary conditions over the oceans for both ECHAM and RSM model runs are sea surface temperature forecasts over the tropical oceans. Soil moisture is initialized by ECHAM's inputs.

The rainfall forecasts generated by the regional model are compared with those of the AGCM and observations. It is shown that the regional model at 80-km resolution improves upon the AGCM rainfall forecast, reducing both seasonal bias and root-mean-square error. On the other hand, the RSM-20 forecasts presented larger errors, with spatial patterns that resemble those of local topography. The better forecast of the position and width of the intertropical convergence zone (ITCZ) over the tropical Atlantic by the RSM-80 model is one of the principal reasons for better-forecast scores of the RSM-80 relative to the AGCM. The regional model improved the spatial as well as the temporal details of rainfall distribution, and also presenting the minimum spread among the ensemble members. The statistics of synoptic-scale weather variability on seasonal timescales were best forecast with the regional 80-km model over the Nordeste. The possibility of forecasting the frequency distribution of dry and wet spells within the rainy season is encouraging.

*On leave of absence from Centro de Previsão de Tempo e Estudos Climáticos, Instituto Nacional de Pesquisas Espaciais, Cachoeira Paulista, São Paulo, Brazil.

Corresponding author address: Dr. Paulo Nobre, International Research Institute for Climate Prediction, Lamont-Doherty Earth Observatory of Columbia University, 61 Route 9W, IRI Monell, Bldg., Palisades, NY 10964-8000. E-mail: pnobre@cptec.inpe.br
Save