Characterization of Rainfall Regimes in the Middle Belt States of Nigeria Using a 3-State Hidden Markov Model

Abstract

This study applies a 3-state Hidden Markov Model (HMM) to 30 years daily rainfall data for the core seven Middle Belt states (Benue, Plateau, Niger, Kogi, Nasarawa, Kwara, FCT) in Nigeria with the aim of identifying underlying rainfall regimes. Rainfall in Nigeria exhibits considerable spatial and temporal variability, particularly across the Middle Belt states, which play a vital role in agriculture and water resources. Rainfall data was categorized into seven classes: No rain, Very Light rain, Light rain, Moderate rain, Heavy rain, Very Heavy rain and Extreme rain. The rainfall regimes are classified as Dry, Moderate and Wet using the Hidden Markov Model. The results show that Dry rainfall regime (State 1) is dominated by Classes 1 (No rain) and 2 (Very Light rain), Moderate rainfall regime (State 2) corresponds mainly to Classes 3 (Light rain) and 4 (Moderate rain) and Wet rainfall regime (State 3) is associated with Classes 5(Heavy rain), 6(Very Heavy rain), and 7 (Extreme rain). Transition probabilities (ranges from 0.89-0.96) from moderate to dry regime and wet to moderate regime is very high over Benue, Niger, Kogi, Nasarawa, Kwara and FCT. The frequency of the hidden states indicates that the dry rainfall regime predominates over most days in Benue, Niger, Nasarawa, and the Federal Capital Territory. Using the seven rainfall amount classes, the HMM was able to correctly identify three hidden states that corresponded to the Dry, Moderate, and Wet rainfall regimes. This information is crucial for describing the climate, planning agricultural projects, and managing water resources in the Middle Belt. However, the classification of rainfall into discrete classes, while useful for interpretation and modelling simplicity, may reduce sensitivity to subtle variations in rainfall intensity. Further studies should integrate additional atmospheric variables such as temperature, humidity, and wind patterns into multi-variable Hidden Markov Models to enhance understanding of rainfall-driving mechanisms in the Middle Belt. Extending the analysis to climate change projections would also help assess future shifts in rainfall regime persistence and transition behavior.