Meteorological Data for Kampala, Entebbe, Masindi and Namulonge observation Stations

This study explores the application of theLong Short-Term Memory (LSTM) neural network architectureto develop a machine learning model capable of predictingthe onset, cessation, and duration of rainfall seasons. Climateobservation data from weather stations in Entebbe, Kampala,Kawanda, Namulonge, and Masindi were used. The datasetincludes 12,025 daily records from 1991 to 2021, consisting ofsurface pressure, temperature, wind speed, relative humidity, andrainfall amounts. Two main prediction tasks were addressed:binary classification of daily rainfall occurrence (rainfall or norainfall), and regression-based prediction of rainfall amounts.The data were split into 80% for training and 20% for testingacross both tasks. Classification model performance metricsincluded accuracy, precision, recall, specificity, F1 score, andMatthews correlation coefficient (MCC), while the regressionmodel was evaluated using mean absolute error (MAE). Theclassification model achieved an accuracy of 0.915, precisionof 0.9435, recall of 0.9166, specificity of 0.864, F1 score of0.9299, and MCC of 0.7653. The developed seasonality modelwas validated using the March\u2013April\u2013May 2025 rainfall seasonand demonstrated the capability to detect and predict seasonaltransitions. These findings hold significant potential for theadvancement of early warning systems aimed at forecastingrainfall patterns, which are critical for managing weather-relatedhazards and reducing risks to life and property.Index Terms\u2014Long Short-Term Memory (LSTM), RainfallOnset, Rainfall Cessation, Stacked LSTM, Numerical WeatherPrediction System (NWPS), Matthews Correlation Coefficient(MCC), Mean Absolute Error (MAE), Relative Humidity (RH)