Researchers used machine-learning techniques to predict soil hydraulic conductivities based on soil properties like texture and water content. Parametric methods like stepwise linear modeling and Lasso regression showed good performance, but nonparametric methods like Gaussian process regression and support vector machine were even more accurate in predicting hydraulic conductivities. These nonparametric methods were able to predict saturated and near-saturated hydraulic conductivities with high accuracy, outperforming the parametric methods. The study found that certain soil properties like soil texture and water content were key predictors of hydraulic conductivities.