Visualizing N–P–K Interactions in Winter Wheat: Yield-Response Surfaces Based on Long-Term Observations

Winter wheat (Triticum aestivum L.) is a strategic crop in Hungary, where increasing interannual weather variability requires nutrient management that supports both yield and stability. This study quantified the effects of fertilization, soil properties, and seasonal weather in a ten-year field experiment (2015–2024) at three contrasting Hungarian sites: Debrecen, Martonvásár, and Iregszemcse. Four fertilization regimes were evaluated in four replications. Grain yield, thousand-kernel weight (TKW), and grain moisture were analysed using analysis of variance (ANOVA), double generalized linear models (DGLMs), and validated regressions generating two- and three-dimensional N–P–K response surfaces. Year, location, and treatment significantly affected grain yield (p ≤ 0.002). Phosphorus had a positive significant effect on grain yield (F = 36.91, p < 0.001) and TKW (F = 18.28, p < 0.001). The separate one-way ANOVA did not detect a significant N effect on grain yield (F = 2.77, p = 0.097), whereas the DGLM estimated a positive association conditional on its selected model structure (p < 0.0001). The fertilized regimes increased adjusted mean yield by 2.13–2.32 t/ha relative to the control. The Environmentally Friendly regime achieved a 2.29 t ha−1 yield increase and showed the largest model-estimated reduction in interannual yield dispersion relative to the control (92%; Regional Genesis, 83%; Balance, 78%). The regression model explained 81.0% of yield variation, with leave-one-out cross-validated R2 = 0.754. The response surfaces showed that the predicted effects of N, P, and K varied with nutrient interactions and site conditions; the highest predicted yields formed a broad N–P–K ridge rather than a single optimum. Mild, wet winters followed by cool springs with moderate precipitation were associated with the most favourable yields. These results support flexible, site-specific nutrient combinations that prioritize yield stability over maximum fertilizer input.