Core Ideas: Accurate Ks and K10 were obtained using water content data at several matric potentials. Robust K _s and K ₁₀ prediction by machine‐learning methods confirmed by bootstrapping. Gaussian process regression predicted K _s and K ₁₀ with minimum number of predictors. Parametric and nonparametric supervised machine learning techniques were used to estimate saturated and near‐saturated hydraulic conductivities ( K _s and K ₁₀ , respectively) from easily measurable soil properties including the name of the pedological horizon (HOR), soil texture (sand, silt, and clay), organic matter (OM), bulk density (BD), and water contents (θ _pF1 , θ _pF2 , θ _pF3 , and θ _pF4.2 ) measured at four different matric heads (−10, −100, −1000, and −15,848 cm, respectively). Using a stepwise linear model (SWLM) and the Lasso regression as parametric methods with 316 data in training and 135 data in the testing phase, four pedotransfer functions (PTFs) were obtained in which water contents for both methods play an important role compared with other variables. The SWLM showed better performance than Lasso in the testing phase for log( K _s ) and log( K ₁₀ ) prediction, with RMSE values of 0.666 and 0.551 cm d ⁻¹ and R ² of 0.26 and 0.65. Nonparametric supervised machine learning methods trained and tested with a similar data set significantly improved the accuracy of K _s prediction, with R ² of 0.52, 0.36, and 0.53 for Gaussian process regression (GPR), support vector machine (SVM), and ensemble (ENS) methods in the testing stage. These methods also described 74.9, 66.7, and 72.5% of the variation of log( K ₁₀ ). Bootstrapping validated the strong performance of nonparametric techniques. The feature selection capability of GPR determined that instead of using a model with all predictors, HOR, silt, θ _pF1 , and θ _pF3 are sufficient for the prediction of log( K _s ), while HOR, silt, and OM can predict log( K ₁₀ ) as accurate as the comprehensive model with all variables.

Flash floods are a significant threat to arid and semi-arid regions, causing considerable loss of life and damage, including roads, bridges, check dams and dikes, reservoir filling, and mudslides in populated areas as well as agricultural fields. Flood risk is a complex process linked to numerous morphological, pedological, geological, anthropic, and climatic factors. In arid environments such as where Bayech basin is located in southwestern Tunisia, the hydrometric data are insufficient due to the absence of measuring points. Using the hybrid fuzzy Analytical Hierarchy Process (F-AHP) and the frequency ratio statistical methods, this study aims to map flooding risks in an ungauged basin that is extremely prone to flooding. Data related to soil texture, slope, land use, altitude, rainfall, drainage density, and distance from the river were used in the risk analysis. The obtained flood risk maps from both F-AHP and FR models were validated on the basis of the Receiver Operating Characteristic (ROC), the Area Under the Curve (AUC), and the inventory map. Results revealed that areas of high and very high susceptibility to flooding are mainly located in the downstream part of the basin, where the town of Gafsa is located. Other upstream sites are also at risk. In this basin, slope is predominantly behind runoff accumulation, whereas soil type plays a major role in amplifying waterproofing and therefore overflow. The results derived from both methods clearly demonstrate a viable and efficient assessment in flood-prone areas. The F-AHP and FR methods have ROC values of 95% and 97%, respectively. Considering these results in the decision-making process, these outputs would enable the implementation of the necessary measures to mitigate flood risk impacts ensure sustainable development along with an effective management in Tunisian arid environments, for the well-being of local communities at risk.