Time Series Modelling and Forecasting of Temperature and Rainfall in Nigeria Using ARIMA Models

Abstract

Background: Temperature and rainfall are among the most important indicators of climate change because they directly influence water resources, agricultural productivity, ecosystem functioning, and human well-being. Although numerous studies have investigated historical climatic trends in Nigeria, comparative long-term forecasting across major ecological zones remains limited. Consequently, knowledge of how future climatic trajectories may differ among contrasting environmental regions remains incomplete, constraining evidence-based climate adaptation and environmental management planning. Objectives: This study evaluated long-term temperature and rainfall dynamics across three contrasting ecological zones of Nigeria and assessed the capability of ARIMA models to forecast future climatic conditions. The study further examined spatial differences in projected climate responses and tested hypotheses regarding regional climate variability and model suitability. Methods: Observed temperature and rainfall records from the Nigerian Meteorological Agency covering the period 1985–2023 were obtained for Sokoto (Sahel savanna zone), Abuja (Guinea savanna zone), and Port Harcourt (coastal rainforest zone). Following quality control and pre-processing, annual temperature and rainfall series were developed and analysed using the Box–Jenkins ARIMA framework. Model identification was performed using autocorrelation and partial autocorrelation diagnostics, and ARIMA (1,1,3) was selected as the optimal forecasting model. Model adequacy was evaluated using stationary R², Mean Absolute Percentage Error (MAPE), residual diagnostics, and significance testing. Forecasts were generated for 2024–2050, while long-term monotonic trends were assessed using Kendall's tau-b correlation analysis. Results: The forecasts revealed statistically significant warming trends across all three ecological zones through 2050. Temperature exhibited strong positive temporal associations at Sokoto station (τ_b = 0.884), Abuja station (τ_b = 0.539), and Port Harcourt station (τ_b = 0.914) (p < 0.001). Rainfall projections demonstrated substantial spatial variability. Increasing rainfall trends were observed at Sokoto station (τ_b = 0.572, p < 0.001) and Port Harcourt station (τ_b = 0.673, p < 0.001), whereas Abuja station showed a significant decreasing rainfall trend (τ_b = −0.519, p < 0.001). Model performance statistics indicated acceptable predictive capability, with low temperature forecasting errors and satisfactory diagnostic results. The findings reveal spatially heterogeneous climate-change responses among Nigeria's major ecological zones and confirm the operational usefulness of ARIMA forecasting under data-limited conditions. Conclusion: The study confirmed significant future warming across all investigated ecological zones, while rainfall trajectories varied considerably among regions. All proposed hypotheses were supported. The findings fill an important gap in comparative climate forecasting across Nigeria's major ecological systems and provide evidence that regional climate adaptation strategies should account for substantial spatial differences in future climatic change.