Amir Verdi
Direct measurement of the soil water retention curve (WRC) and hydraulic conductivity curve (HCC) is expensive and time-consuming. Pedotransfer functions (PTFs) are statistical models that estimate soil hydraulic properties using basic soil data that are typically available or easier to measure. This paper introduces KURVE, a novel web-based tool that uses a recently published HYPROP-WP4C-based German reference dataset and a k -nearest neighbor approach to estimate complete WRCs and HCCs. KURVE selects nearest soils from the reference dataset based on basic soil properties, including soil texture, bulk density (BD), and organic carbon content (OCC), and estimates target-soil hydraulic curves by averaging fitted van Genuchten (VG) and Peters–Durner–Iden (VG-PDI) hydraulic functions from the selected neighbors. The performance of KURVE was evaluated using two independent datasets, a large international dataset and a smaller regional dataset from California, USA, and was compared with the widely used Rosetta and Rosetta3 PTF models. KURVE showed promising performance for estimating both curves. For WRC prediction, KURVE achieved RMSE values as low as 0.061 cm 3 cm⁻ 3 for the international dataset and 0.039 cm 3 cm⁻ 3 for the California dataset, which was comparable to or slightly better than Rosetta-based models. For HCC prediction, KURVE substantially outperformed Rosetta-based models, with log(K) RMSE values of 0.594–0.610 for the international dataset and 0.560–0.643 for the California dataset when soil texture and BD were used as inputs. The results also showed that adding BD to soil texture improved model performance, whereas adding OCC as an additional predictor did not provide consistent improvement. KURVE is freely available at kurve-ptf.streamlit.app and provides the scientific community with an accessible tool for applications that require soil hydraulic information.