Improving hydrological model optimization for riverine species. (September 2017)

• Record Type: Journal Article
• Authors: Kiesel, Jens; Guse, Björn; Pfannerstill, Matthias; Kakouei, Karan; Jähnig, Sonja C.; Fohrer, Nicola
• Is Part Of: Ecological indicators. Volume 80(2017)
• Page Start: 376
• Page End: 385
• Publication Date: 2017-09
• Subjects: Hydrological indicators -- Flow preferences -- Optimization -- Trade-off -- Ecohydrology -- Riverine species; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2017.04.032
• Languages: English
• ISSNs: 1470-160X

Abstract

Abstract: Hydrological indicators such as duration, frequency, magnitude, rate and timing of flow events are used to evaluate and, in models, predict the response of aquatic species to hydrological conditions. Hydrological models are generally optimized without considering specific species preferences by minimizing overall differences between observed and simulated flows using common hydrological objective functions. We hypothesize that hydrological models optimized to implicitly consider species' flow preferences will yield more reliable predictors. To test this hypothesis, we developed a flexible evaluation method in which the hydrological model was optimized for different objective functions. We tested this concept for benthic invertebrates and selected seven species-relevant hydrological indicators as well as four objective functions based on commonly used hydrological performance criteria. Model parameterizations for these cases were assessed on their ability to reproduce an indicator-based performance criterion developed by applying feature scaling to the indicators. The results show that three indicator-based objective functions performed up to 14% better than the standard hydrological performance criterion KGE. When optimizing for individual or multiple indicators, we found that it is important to consider that other indicators are compromised. An evaluation of this trade-off showed a considerable range in indicator values and implausible indicator depictions. We conclude that optimizing hydrological models to depict species preferences most effectively requires consideration of different objective functions that are based on hydrological indicators. Doing so, instead of simply optimizing to standard hydrological objective functions, can yield a rewarding result in terms of a more species-tailored model parameterization and, ultimately, a better prediction of aquatic species.