Field-based assessment and multilinear modelling of infiltration dynamics across heterogeneous zones

DOI registering

Abstract

Accurate estimation of infiltration is critical for hydrological modelling, particularly in regions with heterogeneous physiographic conditions. Conventional field measurements, while reliable, are labour-intensive and spatially constrained, underscoring the need for robust predictive models. This study presents a multiple linear regression (MLR) framework for predicting infiltration rates across diverse terrains in Kerala, India, using both primary field measurements and secondary soil property data. Key infiltration influencing parameters percentage silt, clay, sand, bulk density, initial moisture content, and time were incorporated into the model, with logarithmic transformation applied to linearize the relationship. Model coefficients were derived using the least squares method in IBM SPSS, with statistical significance, multicollinearity, and overall adequacy assessed through variance inflation factor (VIF), coefficient of determination (R²), adjusted R², and standard error of estimate. Calibration was achieved by introducing an infiltration coefficient (K), determined from the ratio of field to predicted infiltration rates, with values ranging from 5,7 to 8,9 across locations. Validation using one-way ANOVA and R² analysis confirmed high predictive accuracy and model robustness. LOOCV sensitivity analysis identified time (T) as the most influential predictor, with soil texture parameters showing comparable effects. The developed MLR model provides a scalable, validated tool for infiltration estimation across varied physiographic conditions.