Development and comparison of integrated river/reservoir models in the Apalachicola–Chattahoochee–Flint basin, USA

Abstract

This paper describes the development of a stakeholder-derived, water system model for the Apalachicola–Chattahoochee–Flint (ACF) basin (ACF-STELLA) and directly compares simulated daily outputs with a more complex model (HEC-ResSim) used by the U.S. Army Corps of Engineers to formally evaluate alternative basin management options. The two models were compared using 70 years of daily output (1939–2008; n = 25,668) for eight different ACF sites: five flow stations and three reservoir elevations. The comparison between models showed a strong match (p < 0.01 rejection significance) between the daily outputs for six of the eight sites, with median Nash–Sutcliffe coefficient of efficiency ranging from 0.732 to 0.979. In the two sites where daily comparisons were less successful, additional analysis was conducted to explore where simulated results diverged. At the Lake Lanier outflow site, a 7-day moving average comparison provided a successful match with a Nash-Sutcliffe coefficient of efficiency of 0.788 (p < 0.01 rejection significance). At the Walter F. George Lake elevation site, comparisons showed that a primary source of disagreement stemmed from a period where HEC-ResSim model outputs were significantly greater than historically observed reservoir elevations. Given the satisfactory model comparison as well as the significantly increased simulation speeds of ACF-STELLA, it was concluded that the ACF-STELLA model could be a useful tool in water policy planning activities to explore alternative basin management scenarios for expanded simulation by the more complex HEC-ResSim model. The use of multiple models in a river basin should not necessarily be interpreted as competitive, but instead can be seen as part of a collaborative effort to empower all stakeholders to systematically engage with their limited water resources.