<p>Modeling river runoff, including addressing gaps in observations and ensuring the continuity of hydrological data series in catchment areas with limited hydrological knowledge, such as the Orontes River in the Syrian Arab Republic, is a critical issue. This is particularly important for the calculation and justification of hydraulic structures and the development of efficient measures for water resource management. The study aimed to model the surface runoff in the Upper Orontes basin using statistical analysis and machine learning methods under conditions of insufficient input data. It also assessed the impact of the modeling process on water balance calculations. River runoff was calculated using autoregressive integrated moving average (ARIMA/SARIMA) models, artificial neural networks, fuzzy logic models, and a physical-mathematical rainfall-runoff model (MIKE11 NAM). Comparing the efficiency of river runoff reproduction by various methods revealed some advantages of artificial neural networks, which enabled their use in filling gaps in the time series of observations of river runoff observations in the Upper Orontes section, followed by the calculation of the current water economy balance. The results showed that the water economy balance in a catastrophically low-water year changed toward a deeper deficit, which required comprehensive water management measures and management decisions in the catchment area under study.</p>