Quantile-Dependent Humidity Associations and Localized EVI Non-Linearity in Irrigated Semi-Arid Sugarcane

Microclimatic dynamics in irrigated sugarcane fields of the Brazilian Semi-Arid reflect the interaction among water availability, radiative energy, and vegetation cover. As linear approaches fail to capture asymmetric responses, this study investigated mean air relative humidity (MRH), treated as an integrating state of the soil–canopy–atmosphere system rather than a mere input variable, as a function of solar radiation, precipitation, and MODIS vegetation indices (EVI and NDVI) in Juazeiro, Bahia. Non-linear quantile regression with natural cubic splines and blocked temporal cross-validation were employed. The results demonstrate that the association varies across the distribution: EVI exhibited statistically supported negative curvature and non-linearity at the upper quantile (τ=0.95), identifying inflection points associated with the attenuation of spectral saturation under high biomass. In contrast, NDVI maintained a linear response. Sensitivity analyses confirmed that irrigation management and microclimatic forcings modulate these thresholds. It is concluded that EVI proves to be more sensitive to canopy structural variation in detecting microclimatic inflections, while NDVI maintains a stable response across the distribution. The results support precision agricultural management and climate adaptation in the Semi-Arid region, paving the way for future research utilizing predictive modeling and machine learning to forecast microclimatic impacts.