Nonlinear Modeling of Sugarcane Yield Based on Edaphic Attributes and Soil Classes

Abstract Edaphic attributes strongly influence sugarcane yield, yet production environments are commonly classified based on average productivity, which may not fully capture environmental variability. This study evaluated how physicochemical soil attributes and soil classes influence sugarcane productivity and whether integrating nonlinear modeling and multivariate analyses improves the characterization of production environments. Edaphic data obtained from SoilGrids and mean Tons of Stalks per Hectare (TSH) per genotype, environment, and harvest cycle collected between 2018 and 2021 were used. Generalized additive models (GAMs) were fitted to assess the effects of soil classes and continuous soil attributes on productivity, while principal component analysis (PCA) and hierarchical clustering were used to identify edaphic gradients and group environments according to soil similarity. The GAM explained 67% of the variance in TSH and revealed significant nonlinear effects of bulk density, cation exchange capacity (CEC), silt, nitrogen, and organic carbon across soil depths. Soil class also significantly affected productivity, with LATOSSOLO VERMELHO Distrófico (LVd) associated with the highest TSH values. PCA identified two major edaphic gradients related to texture/fertility and surface fertility, and clustering grouped environments into four distinct edaphic profiles. Comparison with the traditional productivity-based classification showed that environments with similar TSH can exhibit markedly different soil characteristics, indicating that average yield alone is insufficient to describe environmental variability. Integrating nonlinear modeling and multivariate soil characterization provides a more informative framework for environmental stratification, supporting genotype recommendation and agronomic management in sugarcane production systems.