The article discusses two methods for predicting soil properties like saturated hydraulic conductivity (Ksat) on a regional scale. One method uses machine learning models trained with local soil data and environmental datasets to indirectly predict Ksat. The other method directly trains machine learning models with soil hydraulic data to predict Ksat. The indirect approach reproduces original data variability but has uncertainties in Ksat predictions. The direct approach shows similarities to global Ksat maps but reduces small-scale variability. Overall, the XGBoost machine learning model performs better than the feed-forward neural network in both approaches. The study helps fill gaps in soil data for Austria but emphasizes the need for more field data on Ksat.