Stable and accurate reconstruction of pollutant transport is a crucial and challenging problem, including the inverse problem of identifying pollution sources and physical coefficients and the forward problem of inferring pollutant transport. Governed by advection, diffusion, and reaction processes, this transport phenomenon can be represented by the advection–diffusion–reaction (ADR) equation. In this paper, the physics-informed neural networks (PINNs) are applied to solve the forward and inverse ADR problems. To further enhance the stability and accuracy of the original PINN, two improvements are developed. The first adjusts the orthogonal grid (OG) point selection method and the other suggests adding an additional regulation function, namely, first derivative constraint (FDC). The new method is referred to as OG-PINN with FDC. To verify the effectiveness of the proposed method, five forward and inverse ADR problems are solved, and the results are compared with the analytical and reference solutions. For forward problems, the improved method can solve various ADR problems accurately and stably. For inverse problems, the ability of the OG-PINN for model parameter learning and initial distribution prediction is demonstrated and analyzed. The former gives the missed physical information in the ADR equation from the data, and the latter is used to trace the source of pollutants. The proposed method is quantitatively reliable for investigating various advection–diffusion–reaction processes.

A proper understanding of hydrological processes is vital for water resource assessment, management, and conservation at a local, national, and global scale. The role of hydrological models is critically important in rarely studied ungauged catchments including of Kobo-Golina, in the Danakil basin of Ethiopia. The main objective of this research is to model the hydrology of the Kobo-Golina catchment using the completely restructured SWAT (SWAT+) model. Validated reanalysis river flow from the Global Flood Awareness System (GloFAS) and actual evapotranspiration (AET) from Moderate Resolution Imaging Spectroradiometer (MODIS) were used for single and multi-variable calibration. It is found that the multi-variable calibration scenario reasonably attained the minimum satisfactory performance limit for both variables (NSE = 0.67, R² = 0.68, PBias = −9.68%, and RSR = 0.57 for calibration of GloFAS flow; and NSE = 0.56, R² = 0.63, RSR = 0.66, PBias = 3.86 for calibration of MODIS AET). The model simulation showed that evapotranspiration accounts for 47% of the input water while surface runoff, lateral flow, and groundwater recharge account for 30%, 1.53%, and 21.4%, respectively. The simulated mean annual streamflow at the Basin outlet is 10.6 m³/s. The monthly low flow occurs in June with a median flow of 1.43 m³/s and a coefficient of dispersion of 0.67. High flows occur in August, with a median flow of 16.55 m³/s and a coefficient of dispersion of 1.55. The spatial distribution of simulated runoff was depicted as being higher in the floodplains and along the riparian/drainage lines, whereas upland areas showed lower runoff. The maximum monthly recharge occurs in September with a recharge value of 78.2 mm. The findings of the study suggested that both surface water harvesting and groundwater exploitation can be sought in floodplain areas while conserving the uplands.