Arima-Based Modeling and Forecasting of Monthly Temperature Variability in Burkina Faso Under Climate Change Scenarios up to 2050

---

Résumé

In a world facing climate change, temperature estimates are crucial for Sahelian
countries like Burkina Faso, where agricultural production is highly dependent on weather
conditions. This research applies the ARIMA (AutoRegressive Integrated Moving Average)
model to predict monthly temperature variations between 2025 and 2050, using a time se
ries spanning 41 years (1984–2024). The differentiated data were validated by stationarity
tests (ADF), and the use of SARIMA models allowed for the capture of seasonality. The mod
ifications performed well, with MAPE typically below 2% and RMSE frequently below 0.5°C.
The results indicate a clear increase in temperatures. In January, the historical average of
24.2°C could rise to 27.0°C in 2050, representing an increase of +2.8°C. An increase of
+0.7°C is predicted for February (from 27.1°C to 27.8°C). The most notable increases are
expected for March and April, where average temperatures could exceed 33.5°C by 2050.
Generally, a temperature increase of +2.5°C to +3°C is expected during key months, accom
panied by increased heat stress during the dry season. These forecasts are consistent with
Sahelian trends identified by various sources and highlight the suitability of the ARIMA
model for climate forecasting. It is crucial to incorporate these forecasts into agriculture,
water, and health policies to strengthen Burkina Faso's resilience to climate change.

Mots-clés

climate forecasting, MAPE, heat stress, Burkina Faso, climate modeling