Modeling Soil Water Dynamics and Hydrogel Doses Optimization Using a Machine Learning Approach: A Case Study on Sandy Clay Loam Soil under Varying Bulk Densities

Climate change has intensified droughts in Brazil, threatening agriculture through altered rainfall and temperature patterns. A promising approach to mitigating the soil water deficit is the addition of biodegradable hydrophilic polymers (hydrogels). However, water dynamics in soil hydrogel systems remain complex and depend on soil type, bulk density, and hydrogel dosage. The hydrophilic properties of the matrix may persist over time, highlighting the importance of hydrogel residual effects. The influence of bulk density on polymer dosage dynamics remains underexplored, and no rapid analytical method currently exists to estimate the dose equivalence of active hydrogels for agricultural practices with accuracy and without excessive time consumption. This study addresses two main goals: (1) to evaluate the effectiveness of hydrogel dosages in sandy clay loam soil at varying densities for enhanced water retention and (2) to develop a machine learning-based analytical method for rapid estimation of active hydrogel doses from short soil water content time series. Results showed a significant increase in soil water retention at the 3 g L^-1 dosage. The proposed method, using a locally weighted regression model, achieved a high correlation (0.875) and low error (0.749 g L^-1) in cross-validation without requiring density information, offering a practical tool for agricultural applications. These findings support the efficient and sustainable use of hydrogels, providing a practical framework that facilitates their management and enables rapid field-scale interventions to improve water use in agriculture.