Development of a Hybrid Hydrological Modeling Approach for Runoff Estimation Using SCS-CN and Random Forest in the Wadi Ar Rahma Basin

Abstract

Abstract

This study aims to develop a hybrid hydrological modeling framework that integrates the Soil Conservation Service Curve Number (SCS-CN) model with machine learning techniques to improve the accuracy of runoff estimation in the Wadi Ar Rahma basin, which covers an area of 43.6 km². The study is based on the spatial analysis of key environmental variables influencing the basin’s hydrological response, including land use, hydrological soil groups, and slope characteristics. The results indicate that the dominant soils in the basin belong to hydrological group (B), characterized as moderately permeable, reddish-brown soils of moderate to shallow depth, composed mainly of silt and gravel. Terrain analysis further reveals that elevations range between 107 and 187 meters above sea level, while slope gradients vary from 0 to 12.21°, reflecting relatively moderate topographic conditions that influence both the velocity and spatial distribution of surface runoff.

The conventional SCS-CN model was first applied, followed by the development of an enhanced model using the Random Forest algorithm to recalibrate curve number values based on environmental variables. Comparative results between the two models show spatial differences ranging from −21 to +19, indicating a significant improvement in representing the spatial variability of hydrological response when using the hybrid model. The findings also show that 62.2% of the basin area falls within the moderate response category (27.1 km²), while 26.4% corresponds to low response (11.5 km²), and 11.4% to high response (5 km²).

The results confirm that integrating machine learning techniques with the traditional SCS-CN model enhances the accuracy of hydrological response representation in drainage basins by incorporating the spatial influence of topographic and environmental variables into the modeling process. This approach demonstrates strong potential for supporting runoff estimation studies and water resources management planning, particularly in flood-prone regions.