go-v.orgdoi

Machine Learning Framework for Smart Irrigation Water Demand and Pumping Energy Cost Estimation in Data-Scarce Agricultural Regions

Lotachukwu Azubuike, Omolola A. Ogbolumani · Journal of Digital Food Energy & Water Systems · 2026

  • DOI 10.36615/83fa8a36
  • Publisher University of Johannesburg
  • Type article
  • Issued 2026

Not cached yet. This record was fetched live for this request and is stored after the lookup. Reopening the page should serve the cached copy.

Journal of Digital Food Energy & Water Systems · University of Johannesburg · article · 2026

Abstract

Irrigation in semi-arid northern Nigeria is constrained by variable rainfall, water scarcity, rising pumping costs and limited field measurements, creating a need for decision-support methods that jointly represent water and energy implications. This study develops and evaluates a machine learning framework that combines FAO Irrigation and Drainage Paper No. 56 (FAO-56) evapotranspiration, effective rainfall and deterministic pumping energy calculations to estimate seasonal irrigation demand and associated energy costs in Kano, Kaduna and Katsina States.

Monthly climate data for 2015–2025 were obtained from the National Aeronautics and Space Administration Prediction of Worldwide Energy Resources (NASA POWER) and combined with United States Department of Agriculture (USDA) crop-calendar information for maize, millet, rice, sorghum, soybean, wheat and peanut. Seasonal irrigation requirements were derived from FAO-56 Penman–Monteith reference evapotranspiration, stage-specific crop coefficients, and effective rainfall, yielding 222 crop–state–season observations. A Random Forest regressor was evaluated against Linear Regression and Gradient Boosting using five-fold cross-validation.

Random Forest achieved a mean Coefficient of determination (R²) of 0.928 ± 0.024 and Mean Absolute Error (MAE) of 37.50 ± 4.77 mm; Gradient Boosting produced lower MAE (34.36 ± 4.01 mm), while Random Forest provided more directly interpretable feature importance information. Relative humidity, seasonal reference evapotranspiration and wind speed were the dominant predictors. Mean irrigation requirements ranged from 63.4 mm for maize to 625.2 mm for wheat. Under a 10 m lift, 70% pump efficiency, and 206.80 Nigerian Naira per kilowatt-hour (NGN 206.80 kWh−1), corresponding pumping energy ranged from 24.7 to 243.4 kWh ha−1.

The framework provides a transparent, transferable basis for irrigation-water allocation, pumping-energy planning and tariff sensitivity analysis in data-scarce semi-arid regions where local primary data are not easily accessible.

Open the resolved locator · 10.36615/83fa8a36

Metadata licence: OpenAlex metadata (CC0)