Short-Term Precipitation Forecasting Based on BU-LSTM Model in Dublin City, Ireland

Accurate short-term precipitation forecasting is essential for mitigating the impacts of weather on daily activities and infrastructure. However, enhancing the precision of such forecasts remains a persistent challenge. In this study, we introduce an innovative prediction model, termed BU-LSTM, which leverages the architectural advantages of bidirectional long short-term memory (BI-LSTM) networks while incorporating uncertainty estimation into its predictive framework. This dual capability enables BU-LSTM to effectively capture bidirectional temporal dependencies and provide robust uncertainty estimates, thus improving its capacity to model complex rainfall dynamics. In addition, we studied the interdependence between meteorological variables in the ERA5 reanalysis dataset in the Dublin area and designed comparative algorithm experiments. The results indicate that the BU-LSTM model outperforms other advanced models in terms of root mean square error (RMSE) and coefficient of determination <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\mathbf{R}^{\mathbf{2}})$</tex>. Its RMSE is only 0.019 and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{R}^{\mathbf{2}}$</tex> is 0.761. This significantly improves the accuracy of short-term precipitation forecasts in the Dublin area, providing new research insights for the development of deep learning in the field of precipitation forecasting.