<p>The soil conservation service curve number (SCS-CN) method, developed by the United States Department of Agriculture (USDA), is widely used to estimate direct runoff from watersheds for specific rainfall events. However, the current SCS-CN method has shown limited accuracy in reliably estimating observed rainfall-runoff relationships, especially for long-duration hydrologic simulations. To address this, efforts have been made to enhance the performance of the existing SCS-CN method. This study introduces SCS-CN-inspired models for simulating long-duration rainfall-generated runoff across various time scales, including monthly, bi-monthly, seasonal, and annual. The performance of these proposed models (M3–M8) is evaluated in comparison with the existing SCS-CN models (M1 and M2) and the Mishra and Singh model (M9) from a prior study (J Hydrol Eng 4(3):257–264, 1999). The improved models (M3–M8) are developed by integrating two approaches: a simplified long-duration water balance equation and the proportionality hypothesis of the SCS-CN method. To assess these models, rainfall-runoff data from four distinct agro-climatic river basins in Ethiopia are used. Model performance is quantitatively measured using three statistical indices: Nash–Sutcliffe efficiency (NSE), root mean square error (RSR), and percent bias (PBIAS). Additionally, the runoff residuals between the top two performing models are tested with a two-tailed, paired Student’s t-test at a significance level of 0.05. The results indicate that the proposed SCS-CN-inspired models (M3–M8) generally outperform existing models, showing the higher NSE values and, lower RSR and PBIAS values. When applied to observed datasets across various time scales, the proposed models perform well in both calibration and validation phases for all watersheds, underscoring their effectiveness in practical field applications.</p>