Water resource management scenarios have become more crucial for arid to semi-arid regions. Their application prerequisites rigorous hydrological modelling approaches since data are usually exposed to uncertainties and inaccuracies. In this work, Soil Water Assessment Tool (SWAT), an open source semi-distributed, continuous-time, process-based physical hydrological model is used to model hydrological processes and eventually calculate groundwater recharge estimations in Seybouse basin, Northeast Algeria. The model uses estimated rainfall to calibrate the model with observed discharge from hydrometric stations. Model calibration and validation are performed over four hydrometric stations located in the basin. Uncertainty analysis and sensitivity analysis supported the calibration period. SUFI-2 algorithm is used for uncertainty estimations along with a global sensitivity analysis prior to calibration simulations. Simulated flood hydrographs showed generally good accuracy with few misfits on the peaks. The model obtained satisfactory and consistent calibration and validation results for which the Nash score varied from 0.5 to 0.7 for calibration and from −0.1 to 0.6 for validation and R² from 0.6 to 0.7 for calibration and 0.03 to 0.8 for validation. Moreover, estimated water budget values show strong similarities with the observed values found in the literature. The present work shows that the rigorously calibrated and validated SWAT model can simulate hydrological processes as well as major high and low flows using estimated rainfall data.

Leakages from damaged or deteriorated buried pipes in urban water distribution networks may cause significant socio-economic and environmental impacts, such as depletion of water resources and sinkhole events. Sinkholes are often caused by internal erosion and fluidization of the soil surrounding leaking pipes, with the formation of soil cavities that may eventually collapse. This in turn causes road disruption and building foundation damage, with possible victims. While the loss of precious water resources is a well-known problem, less attention has been paid to anthropogenic sinkhole events generated by leakages in water distribution systems. With a view to improving urban smart resilience and sustainability of urban areas, this study introduces an innovative framework to localize leakages based on a Machine learning model (for the training and evaluation of candidate sets of pressure sensors) and a Genetic algorithm (for the optimal sensor set positioning) with the goal of detecting and mitigating potential hydrogeological urban disruption due to water leakage in the most sensitive/critical locations. The application of the methodology on a synthetic case study from literature and a real-world case scenario shows that the methodology also contributes to reducing the depletion of water resources.