Hydrologic Conservatism as a Rationale for Selecting NRCS Initial Abstraction Ratio

We employed a dataset of 1,229 rainfall-runoff events in 47 watersheds in Texas to explore the performance of the Natural Resources Conservation Service (NRCS) Curve Number Method in determining peak discharges given varying initial abstraction ratios ($\lambda$). For each watershed, a curve number was calibrated from rainfall-runoff events using the least-squares approach, and the standard error ratio was used to evaluate calibration performance. Curve numbers and standard error ratios were compared for both $\lambda =0.20$ and $\lambda =0.05$. Hypothesis testing found no significant difference in the standard error ratios within the five Texas regions studied, highlighting the need for some alternative criteria on which to select the most appropriate $\lambda$ value for modeling and design. An equation was developed that computes equivalent curve number between $\lambda =0.20$ and $\lambda =0.05$. Finally, we presented a method to predict whether small or large $\lambda$ will estimate larger runoff and recommended a design approach that selects the $\lambda$ value such that the larger runoff is estimated, leading to the more conservative hydrologic design.