Abstract:
Drought remains one of the most devastating environmental hazards, significantly affecting agricultural productivity, water availability, and socio-economic stability, particularly in semi-arid regions like Machakos County, Kenya. This study proposes a hybrid drought forecasting model that integrates Facebook’s Prophet Time series algorithm with Long Short-Term Memory (LSTM) neural networks to enhance the accuracy and reliability of drought prediction. Using historical climate data — including rainfall, temperature, humidity, and soil moisture — sourced from the Visual Crossing Weather Data platform, the model captures complex temporal patterns and non-linear dependencies. The research demonstrates the limitations of traditional models such as ARIMA and SARIMA and highlights the advantages of combining Prophet’s seasonality modeling with the temporal depth of LSTM networks. Evaluation metrics such as RMSE, MAE, and R² are used to validate the model's performance. This approach contributes to early warning systems and decision-making processes for drought management in Kenya and similar semi-arid regions.
