Explainable machine learning reveals water-related drivers of sub-field dryland wheat yield variability

Authors: LAMICHHANE, MANOJ - South Dakota State University; MEHAN, SUSHANT - South Dakota State University; Mankin, Kyle; TROOIEN, TODD - South Dakota State University; MAIMAITIJIANG, MAITINIYAZI - South Dakota State University; REKABDARKOLAEE, HOSSEIN - South Dakota State University

Submitted to: European Journal of Agronomy
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/3/2026
Publication Date: 8/6/2026
Citation: Lamichhane, M., Mehan, S., Mankin, K.R., Trooien, T., Maimaitijiang, M., Rekabdarkolaee, H.M. 2026. Explainable machine learning reveals water-related drivers of sub-field dryland wheat yield variability. European Journal of Agronomy. 118. Article e128296. https://doi.org/10.1016/j.eja.2026.128296.
DOI: https://doi.org/10.1016/j.eja.2026.128296

Interpretive Summary: If we could predict wheat yield early in the season, farmers could make management decisions (such as split-applied fertilizer) tailored to real yield goals, instead of wishful targets. We used detailed (15-ft) yield data from 6 dryland Colorado wheat fields over 6 years and various machine learning models. We could explain more than 80% of yield variability with just a few inputs: water availability (actual evapotranspiration was best, but on-site precipitation and soil moisture also performed well), measured soil characteristics (such as surface clay and other properties), nutrient management zone, and topography (elevation). We had similar yield prediction when we removed detailed soil data and kept generalized nutrient management zones in the model, to focus only on data easily obtained from drone or satellite sensors. We still need to test this model on other research-station and on-farm fields to see if the relationships hold. After that, this or similar models could fine-tune real-time farmer management decisions and increase profits.

Technical Abstract: Understanding the drivers of within-field yield variability is essential for precision management in dryland agroecosystems where water limitation governs productivity. This study identifies and quantifies controls on sub-field-scale wheat yield variability using an interpretable machine-learning framework. Eighteen dryland wheat fields in northeastern Colorado (2019–2024) were grouped into aspirational (ASP; no-till, diversified rotation with spatial nitrogen zones) and business-as-usual (BAU; reduced tillage, wheat–fallow rotation) systems. Extreme gradient boosting (XGBoost) models were evaluated under eight feature combinations that systematically included precipitation (P), soil moisture (SM), actual evapotranspiration (ETa), soil properties, and topography. Under random partitioning, models explained over 80% of yield variability in both systems. Performance remained stable even when high-resolution soil properties were excluded, indicating that much of sub-field variability can be captured using spatially explicit water indicators and terrain data. Under field-level holdout, models combining ETa and SM retained strong predictive skill, explaining ~78% of yield variability in ASP and ~55% in BAU fields. SHAP analysis showed that early-season water-related variables governed yield variability, with ETa emerging as the most spatially informative indicator of water limitation. Nitrogen-zone classifications were important in ASP fields, likely reflecting structured yield-potential gradients embedded within management zoning rather than isolated fertilizer-rate effects. Elevation acted as a secondary but persistent control. These findings demonstrate that explainable machine learning can provide actionable insight into water–management interactions driving yield variability in semi-arid, dryland wheat systems.