Precision vs. Conservatism in Initial Abstraction Ratio: Implications for Hydrologic Design

For several decades, the initial abstraction ratio (λ) in the Curve Number method has been a point of debate. The method traditionally uses a fixed value of λ = 0.20, but growing evidence shows that lowering it to 0.05 improves runoff predictions under certain conditions. This study supports the idea that accuracy should take precedence over conservatism when selecting λ for hydrologic design. Using rainfall-runoff information from 1,358 US watersheds, λ was calibrated under two decision rules: (1) precision, minimizing the hybrid Kling-Gupta efficiency objective function, and (2) conservatism, maximizing simulated runoff. These calibrations were tested across different rainfall scenarios (P > 0 mm and P > 25.4 mm), methods (Least Squares Error and Asymptotic), and return periods (2, 10, 50, and 100 years). Switching from a precision-based to a conservative λ increased mean design runoff by 8–18%, with the largest jumps seen in LSE calibrations at longer return periods. Using precision-based λ values leads to runoff estimates that better reflect actual watershed behavior, helping avoid overdesign and supporting more cost-effective, climate-resilient stormwater infrastructure.