SWAT (Soil and Water Assessment Tool) is a continuous-time, semi-distributed, river basin model widely used to evaluate the effects of alternative management decisions on water resources. This study examines the application of the SWAT model for streamflow simulation in an experimental basin with mixed-land-use characteristics (i.e., urban/peri-urban) using daily and hourly rainfall observations. The main objective of the present study was to investigate the influence of rainfall resolution on model performance to analyze the mechanisms governing surface runoff at the catchment scale. The model was calibrated for 2018 and validated for 2019 using the Sequential Uncertainty Fitting (SUFI-2) algorithm in the SWAT-CUP program. Daily surface runoff was estimated using the Curve Number method, and hourly surface runoff was estimated using the Green–Ampt and Mein–Larson method. A sensitivity analysis conducted in this study showed that the parameters related to groundwater flow were more sensitive for daily time intervals, and channel-routing parameters were more influential for hourly time intervals. Model performance statistics and graphical techniques indicated that the daily model performed better than the subdaily model (daily model, with NSE = 0.86, R² = 0.87, and PBIAS = 4.2 %; subdaily model with NSE = 0.6, R² = 0.63, and PBIAS = 11.7 %). The Curve Number method produced higher discharge peaks than the Green–Ampt and Mein–Larson method and better estimated the observed values. Overall, the general agreement between observations and simulations in both models suggests that the SWAT model appears to be a reliable tool to predict discharge in a mixed-land-use basin with high complexity and spatial distribution of input data.

Machine learning (ML) algorithms are extensively used with outstanding prediction accuracy. However, in some cases, their overfitting capabilities, along with inadvertent biases, might produce overly optimistic results. Spatial data are a special kind of data that could introduce biases to ML due to their intrinsic spatial autocorrelation. To address this issue, a special resampling method has emerged called spatial cross-validation (SCV). The purpose of this study was to evaluate the performance of SCV compared with conventional random cross-validation (CCV) used in most ML studies. Multiple ML models were created with CCV and SCV to predict groundwater electrical conductivity (EC) with data (A) from Rhodope, Greece, in the summer of 2020; (B) from the same area but at a different time (summer 2019); and (C) from a new area (the Salento peninsula, Italy). The results showed that the SCV provides ML models with superior generalization capabilities and, hence, better prediction results in new unknown data. The SCV seems to be able to capture the spatial patterns in the data while also reducing the over-optimism bias that is often associated with CCV methods. Based on the results, SCV could be applied with ML in studies that use spatial data.

In this study, monthly streamflow and satellite-based actual evapotranspiration data (AET) were used to evaluate the Soil and Water Assessment Tool (SWAT) model for the calibration of an experimental sub-basin with mixed land-use characteristics in Athens, Greece. Three calibration scenarios were performed using streamflow (i.e., single variable), AET (i.e., single variable), and streamflow–AET data together (i.e., multi-variable) to provide insights into how different calibration scenarios affect the hydrological processes of a catchment with complex land use characteristics. The actual evapotranspiration data were obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS). The calibration was achieved with the use of the SUFI-2 algorithm in the SWAT-CUP program. The results suggested that the single variable calibrations showed moderately better performance than the multi-variable calibration. However, the multi-variable calibration scenario displayed acceptable outcomes for both streamflow and actual evapotranspiration and indicated reasonably good streamflow estimations (NSE = 0.70; R² = 0.86; PBIAS = 6.1%). The model under-predicted AET in all calibration scenarios during the dry season compared to MODIS satellite-based AET. Overall, this study demonstrated that satellite-based AET data, together with streamflow data, can enhance model performance and be a good choice for watersheds lacking sufficient spatial data and observations.