UTILIZAÇÃO DE PREVISÕES DE PRECIPITAÇÃO DE MODELOS ATMOSFÉRICOS WRF, GFS E GEFS NA BACIA HIDROGRÁFICA DO RIO AVE (PORTUGAL) PARA GESTÃO OPERACIONAL DE UM SISTEMA DE DRENAGEM

Os sistemas de previsao e alerta utilizados na gestao de recursos hidricos e operacao de sistemas de drenagem tiveram desenvolvimentos significativos nos ultimos anos. Esses desenvolvimentos resultaram da disponibilidade de informacoes meteorologicas em tempo real, em particular de medicoes por sensores em satelites, medicao atraves de radar meteorologico e de previsoes de modelos atmosfericos para diferentes horizontes temporais. Todos os modelos de previsao ambiental sao incertos e essa incerteza e variavel no tempo e no espaco. Este trabalho tem como objetivo apresentar os resultados da avaliacao da evolucao do erro associado a diferentes previsoes de curto prazo. A plataforma Delft-FEWS ( Flood Early Warning System) foi utilizada para proceder a importacao e processamento de dados de observacoes e previsoes disponiveis para a bacia do rio Ave, localizada no norte de Portugal. Os dados meteorologicos medidos foram obtidos no Sistema Nacional de Informacoes de Recursos Hidricos (SNIRH), em quatro estacoes meteorologicas instaladas na bacia em estudo e dados de refletividade medidos pelo radar meteorologico operado pela Meteogalicia. As previsoes avaliadas correspondem as precipitacoes simuladas por modelos atmosfericos desenvolvidos pela National Oceanic and Atmospheric Administration (NOAA ) e Meteogalicia, nomeadamente os modelos Global Forecast System (GFS) , Global Ensemble Forecast System (GEFS) e Weather Research and Forecasting , (WRF) operado pela Meteogalicia. A incerteza associada as precipitacoes previstas foi avaliada considerando horizontes de previsao de um a quatro dias. Os melhores resultados foram obtidos para o modelo WRF durante eventos de precipitacao ocorridos entre janeiro de 2017 e maio de 2018 e apresentaram medias de erros relativos que variaram entre 7% (um dia de previsao) e 29% (quatro dias). O sistema implementado permite, assim, do ponto de vista operacional, antecipar com antecedencia de dois dias eventos extremos. USE OF PRECIPITATION FORECASTS FROM WRF, GFS AND GEFS ATMOSPHERIC MODELS AT RIVER AVE BASIN (PORTUGAL) FOR OPERATIONAL MANAGEMENT OF A DRAINAGE SYSTEM ABSTRACT The forecasting and warning systems used in water resources management and drainage systems operation have had significant developments in recent years. These developments resulted from the availability of meteorological information in real time, in particular from measurements by sensors in satellites, measurement through meteorological radar and forecasts of atmospheric models for different time horizons. All environmental forecasting models are uncertain and this uncertainty varies over time and space. This work aims to present the results of the evaluation of the evolution of the error associated with different short-term forecasts. The Delft-FEWS (Flood Early Warning System) platform was used to import and process observation and forecast data available for the river Ave basin, located in northern Portugal. The measured meteorological data were obtained from the National Water Resources Information System (SNIRH), at four new meteorological stations installed in the basin and radar reflectivity data measured by the meteorological radar operated by Meteogalicia. The forecasts evaluated correspond to the rainfall simulated by atmospheric models developed by the National Oceanic and Atmospheric Administration (NOAA) and Meteogalicia, namely the Global Forecast System (GFS), Global Ensemble Forecast System (GEFS) and Weather Research and Forecasting (WRF) model operated by Meteogalicia. The uncertainty associated with the predicted rainfall was evaluated considering forecast horizons of one to four days. The best results were obtained for the WRF model during precipitation events that occurred between January 2017 and May 2018 and presented average relative errors that varied between 7% (one forecast day) and 29% (four days). The implemented system thus allows, from an operational point of view, to forecast extreme events in advance of two days.