Validação da previsão de visibilidade calculada a partir de dados do modelo WRF para a Baía de Guanabara

Abstract

This study validated horizontal visibility forecasts produced by the Weather Research and Forecasting (WRF) model using the Forecast Systems Laboratory (FSL) estimator, by comparing the simulations with METAR observations from the Santos Dumont and Galeão aerodromes for the period of August 08/01/2024 to 07/31/2025. Forecasts and observations were analyzed hour- by-hour and evaluated using statistical metrics and the range classification proposed in NAVEMARINST No. 10-11A. For Santos Dumont, the annual root mean square error (RMSE) was 2.48 km, bias (BIAS) was 0.86 km (a tendency to overestimate), the false alarm ratio (FAR) = 0.82, and the probability of detection (POD) = 0.25, with 64.47% of forecasts classified as correct. For Galeão, the annual RMSE was 1.90 km, bias −0.14 km, and the correct classification reached 76.64% (a low number of fog events was observed in the series, which affected FAR and POD). The results indicated that the model, in the studied configuration, shows satisfactory performance for horizontal visibility forecasts in general terms, but presents limitations when values are below 1 km, manifested as false alarms and errors in intensity. It was concluded that the evaluated configuration is promising for operational visibility forecasting but requires improvements, such as the, inclusion of observed sea surface temperature data, tuning of parameterizations, and analysis of a larger set of events to consolidate the validation.