Development of Hybrid ARIMA-LSTM and Prophet-LSTM Models for Air Quality Forecasting in the Tangier Region

Accurately predicting tropospheric ozone (O3) levels is essential for improving air quality in Tangier, given its significance as a major atmospheric pollutant. In this study, we propose a hybrid forecasting approach that combines ARIMA-LSTM and Prophet-LSTM models to improve both the accuracy and robustness of ozone concentration predictions. The method leverages the strengths of ARIMA and Prophet in capturing linear trends and seasonal variations, while leveraging the ability of LSTM networks to model nonlinear dynamics and long-term dependencies. Comparative analyses show that the Prophet-LSTM combination delivers the most reliable performance, providing improved forecasting accuracy. This hybrid model is particularly effective in identifying pollution peaks, with an agreement index reaching d=0.99.