Aplicabilidade da plataforma Mydewetra para a previsão da precipitação induzida por ciclones tropicais em Moçambique: caso do ciclone Freddy (2023)

Abstract

In recent years, Mozambique has been affected by tropical cyclones, which cause property destruction and loss of life. The impacts of tropical cyclones can be minimized by improving weather forecasts. Forecasts of good quality can enable early action, ensuring that threatened populations can prepare for the possible impacts of cyclones. For this reason, it is important to ensure that this type of event is predicted with an accuracy that allows disaster risk reduction. As a measure to improve weather forecasts in Mozambique, the myDewetra platform was introduced in 2021. This is a platform made up of two weather forecasting models: GFS and ECMWF. These two global models allow meteorological forecasting and have already been exhaustively evaluated for most of the ocean basins in the Northern Hemisphere, but little research has been done for the Southern Hemisphere, mainly in the Southwest Indian Ocean (SOI) basin. The evaluation of the meteorological model allows us to understand how suitable it is to make forecasts in a given ocean basin. Due to the lack of data from the ECMWF model, this study evaluated only the ability of the GFS model to predict the 24-hour accumulated precipitation associated with tropical cyclones using Cyclone Freddy (2023) as a case study. This cyclone was the first and only cyclone that was used to test the myDewetra platform until the date of this research. Since the amount and distribution of precipitation can be strongly affected by the trajectory and intensity of the tropical cyclone, there was a need to also evaluate the ability of the GFS model to predict the trajectory and intensity of Cyclone Freddy. The prediction of trajectory and intensity were evaluated by calculating the prediction error. The precipitation forecast was evaluated by calculating the root mean squared error (RMSE – The Root Mean Squared Error). The results show that the GFS model is capable of forecasting the location through which the cyclone will pass, but it is not capable of forecast the exact moment of the cyclone's passage. As a result, cyclone trajectory predictions are prone to false alarms. Furthermore, trajectory forecast showed a dependency on lead time, i.e., the longer the lead time, the lower the prediction performance. Regarding intensity predictions, the results show that the intensity prediction does not depend on the lead time. The model overestimates the intensity when the system is over the Earth's surface and underestimates the intensity when the system is over the waters of the Mozambique Channel. Accumulated precipitation forecasts showed that the model has poor performance in predicting extreme precipitation associated with Tropical Cyclone Freddy. Furthermore, the accumulated precipitation forecasts are dependent on the lead time, being better for the 24h lead time and worse for longer lead times (48h and 72h).