Predictive Modelling of Groundnut Yield Using Climate Variables in Katsina State

Groundnut is a major cash and food security crop in Katsina State, and developing predictive models linking climate variables to yield performance supports agricultural planning and early warning in a region vulnerable to weather variability. This study developed yield prediction models for groundnut production across 12 Local Government Areas in Katsina State, North West Nigeria, using 2012 to 2022 historical data. Seasonal rainfall totals, onset and cessation dates, maximum and minimum temperatures, and soil type were merged with official crop yield statistics from the Katsina State Agricultural Development Programme. Linear regression, support vector regression, and random forest models were compared. Random forest achieved R-squared of 0.79 and RMSE of 0.19 tonnes per hectare, outperforming regression baselines. Early rainfall onset (before 20 May) was the strongest positive predictor of high yield. Temperatures exceeding 38 degrees Celsius during flowering reduced yield by an estimated 18 percent. The model identified 2018 as the worst climate-yield season in the study period, corresponding to a short rainy season. Long-range forecast data from NIMET were incorporated to produce six-month ahead yield outlooks with acceptable accuracy. Limitations include unofficial supplemental irrigation being uncaptured in climate-yield relationships. Recommendations include deploying this forecast system through Katsina ADP extension services, integrating NIMET seasonal forecast advisories into planting schedules, and establishing automated weather stations in all 12 LGAs to improve input data quality.

Keywords: groundnut yield prediction, climate variables, random forest, Katsina State, agricultural forecasting