Comparison of storm runoff models for a small watershed in an urban metropolitan area, Albuquerque, New Mexico

In order to comply with a current U.S. Environmental Protection Agency watershed-based National Pollutant Discharge Elimination System permit, the City of Albuquerque required a better understanding of the rainfall-runoff processes in its small urban watersheds. That requirement prompted the initiation of the assessment of three existing watershed models that were developed to simulate those processes. Three existing rainfall-runoff modeling software packages—Hydrologic Engineering Center Hydrologic Modeling System (HEC-HMS) (using two sets of methods), Program for Predicting Polluting Particle Passage Through Pits, Puddles, and Ponds (P8), and Arid-Lands Hydrologic Model (AHYMO)—were compared to determine which provided the best balance of accuracy and usability for simulating storm runoff in small watersheds in the urbanized area of Albuquerque, New Mexico. Additionally, results of this study could help inform model users who have interest in simulating storm runoff in similar urban areas throughout the United States. Each model was used to simulate storm runoff in the Hahn Arroyo watershed, an urbanized watershed with concrete-lined arroyo channels in the northeastern quadrant of Albuquerque that exhibits flashy, monsoonal-driven storm runoff. Model results were compared to observed discharge data, according to literature-recommended performance measures and performance evaluation criteria. The HEC-HMS model using the Soil Conservation Service (SCS) curve number (CN) and SCS unit hydrograph methods ranked the highest when averaging the individual performance measures (Nash-Sutcliffe Efficiency, percent bias, and coefficient of determination) rankings together across the hourly calibration and validation periods, followed by P8, which was tied with the HEC-HMS initial and constant approach. For daily rankings using the same rank-averaging approach, the HEC-HMS CN-based model and P8 were tied for the highest ranking, followed by the HEC-HMS initial and constant approach. Alternatively, rating performance using validation period results as an indication of the expected confidence in forecasted results for future conditions, the P8 model performed best for both hourly and daily time-steps, followed by the HEC-HMS CN-based model and the HEC-HMS initial and constant-based model. However, based on the literature performance evaluation criteria, the HEC-HMS and P8 models overall had marginally satisfactory performance only for operation at the daily time-step. Direct comparison of the HEC-HMS and P8 models to the AHYMO is difficult, given the different performance assessment criteria used to assess these models separately in this study, as recommended by the literature. The AHYMO results generally lacked precision, given the wide range in the performance assessment values across events in percent error in peak discharge, difference in timing of peak discharge, percent error in total runoff volume, and difference in duration of event relative to observed data. For some events, however, the AHYMO results were fairly accurate, and AHYMO was likely a good predictor of the timing of storm runoff and the shape of the hydrograph. This study did not assess the results for all potential applications of the models in the Albuquerque urbanized area. Further study may be required to assess the model performance capabilities in other modeling applications.