Contribution of
Centro de Previsão de Tempo e Estudos Climáticos (CPTEC)
Instituto Nacional de Pesquisas Espaciais (INPE)
12630-000 Cachoeira Paulista, SP, Brazil
Web site: www.cptec.inpe.br

The Center for Weather Forecasts and Climate Studies (CPTEC) of the National Institute for Space Research (INPE), located in Cachoeira Paulista, São Paulo, Brazil, develops, produces and disseminates weather forecasts as well as seasonal climate forecasts since early 1995. This Center is part of the research network of the Ministry of Science and Technology (MCT) of Brazil. CPTEC/INPE is a leader in operational meteorology and climate research in South America. The Center relies on a staff of about 20 scientists at Ph.D. level and 25 meteorologists, and several staff members in the fields of meteorology, climatology, oceanography, computer sciences and systems engineering. It is a regional focal point for numerical seasonal climate forecast for Brazil and the rest of the continent. Currently, the principal users of CPTEC/INPE weather and climate products are research groups, universities, federal and state government agencies, meteorological and hydrological services, the media (TV, radio, newspapers), hydropower, agricultural, industry, and tourism sectors, as well as the private sector. Postgraduate training is part of the activities of CPTEC/INPE, besides on-the-job training for operational meteorologists and climatologists. The Center coordinates the Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). It is involved in other international and regional programmes such as Variability of the American Monsoon System (VAMOS), PIRATA, and CLIVAR. Relevant international collaborations in seasonal climate prediction have been established with ECMWF, COLA, IRI, and University of Buenos Aires, Argentina.

CPTEC/INPE has two NEC supercomputers: SX3 (peak performance of 3.2 MFlops) and SX4 (peak performance of 16 MFlops). From 1995 to 1997 the global seasonal and weather forecasts were performed using the SX3 supercomputer. Since 1998 both forecast systems have migrated to the SX4 supercomputer. This improvement in the computer capability made possible the application of a larger number of integrations for climate prediction experiments, increasing the confidence of the results with the analyses of the ensemble average and also making possible various analyses of member variability.

The model used for seasonal climate predictions is the COLA-CPTEC Atmospheric Global Circulation Model (AGCM) which was originally derived from the NCEP model by COLA. The same model is used at CPTEC for medium-range NWP. Another model is also used in daily weather forecast, the ETA limited area model adapted from NCEP ETA model for a domain covering South America and part of the adjacent oceans. This model is been tested as a downscaling tool for regional climate predictions. CPTEC/INPE developed also a statistical model, the "Sistema de Modelagem Estatística dos Oceanos (SIMOC)", based on canonical correlations using Pacific and Atlantic SST's as predictors of seasonal rainfall anomalies for many areas of South America (Northeast Brazil, Southeastern South America, etc). This model is also used to predict SST fields in the Tropical Atlantic Ocean. The predicted SST fields are introduced as boundary conditions for the CPTEC/COLA AGCM to produce seasonal climate forecasts.

Monthly to seasonal numerical predictions have been performed at CPTEC every month since January 1995 using the COLA-CPTEC AGCM. Spectral horizontal resolution T62 and 28 levels in the vertical are used in the climate forecast runs. Seasonal predictions at CPTEC were derived from 1995 through 1998 from results of an ensemble of 4 members, using 4 consecutive days as initial conditions, running the model on the NEC SX3 supercomputer. In 1998 there was an increase in the number of members, with the predictions based on an ensemble of 25 integrations using the NEC SX4 supercomputer. The application of boundary conditions also changed. Previously, persisted SST anomalies were used as lower boundary conditions. Currently, some form of predicted SST fields is preferred. The predicted Tropical Pacific SST fields are obtained from NCEP coupled atmospheric-ocean model. Predicted Tropical Atlantic SST fields are derived from the statistical SIMOC model and used when the forecasting skill of such model is high. Otherwise, persisted SST anomalies are used for the Tropical Atlantic. Several simulations experiments with the AGCM have been performed to assess the model response related to interannual variability, mainly to El Niño and La Niña conditions in the Pacific Ocean and to the dipole SST pattern in the Tropical Atlantic Ocean. Seasonal climate forecasts are continuously verified against observations, mostly for South America. Preliminary studies indicated higher forecasting skills for Northeast Brazil, parts of Amazonia, and Southeastern South America (southern Brazil, Uruguay, Paraguay and central-northeastern Argentina). Such forecasts, albeit of an experimental nature, have been successfully used on a number of applications.

A 10-year (January 1982-December 1991) ensemble simulation of 9 integrations with different initial conditions using observed monthly SST and the model at T62L28 resolution is in course. It will allow assessing model climatology, interannual variability, and anomaly patterns in comparison to observations. That will form the basis for establishing the model's skill for seasonal climate predictions. That simulation will be extended for 10 more years subsequently. In order to improve the climate prediction, some experiments are being developed changing the convective scheme of parameterization from the Kuo scheme to the Arakawa-Schubert scheme. Another development is to change the monthly climatological soil moisture, used in the current version, to a quasi real time soil moisture from ECMWF analyses. A better use of the supercomputer processors with a paralelization of the model code is in progress, and will be used to reduce the computer time for simulations.

Recently the MOM Ocean model has been coupled to CPTEC-COLA AGCM at CPTEC. Currently, the coupled ocean-atmosphere model is being tested, including testing the ocean data assimilation scheme implemented. Experiments in seasonal climate forecasts using the coupled model will be performed. Assimilation of oceanic surface and subsurface data from the PIRATA Array is being planned.