Estimating virtual water content and yield of wheat using machine learning tools

•AI models predict wheat virtual water content and yield without crop coefficient values.•Linear Regression achieves 98 % accuracy for water content and yield prediction.•April rainfall days identified as key predictor for wheat water use and yield.•AI approach enhances water management in regions with limited data availability.•Simplified model equations developed for virtual water content and yield estimation. The global escalation of water demand has led to significant depletion of water resources, making virtual water content (VWC) and yield assessment crucial for agricultural water management. Traditional calculations heavily rely on pre-determined crop coefficient (Kc) values and extensive datasets, presenting three major challenges: limited data availability in many regions, inaccurate reflection of local conditions through standardized values, and inability to capture spatial–temporal variations in water use patterns. Our study addresses these challenges by developing an innovative machine learning (ML) framework that eliminates Kc dependency while maintaining high prediction accuracy. Our approach combines climate, soil and agronomic variables collected from 81 Turkish provinces (2008–2019) to develop Linear Regression (LR) and Random Forest (RFR) models. Comparative analysis revealed LR’s superior performance, achieving high accuracy (R2 = 0.98) and consistently low error rates (