A multi-objective assessment of alternate conceptual ecohydrological models

Abstract

A merging of a conceptual hydrological model with two vegetation models is performed to improve the ability to simultaneously predict catchment scale streamflow and vegetation dynamics (represented by the Leaf Area Index, LAI). A modeling study is performed across 27 catchments of 90–1600 km2 in the Murray–Darling Basin in Australia. Validation results from the modeling exercise show that the merged ecohydrological models were capable of improving streamflow prediction compared to hydrological models alone, while also providing as good estimates of LAI as dynamic vegetation models alone. It was shown that a single-objective optimization could independently produce good estimates of streamflow and LAI, but the other un-calibrated predicted outcome (LAI if streamflow was the focus of the optimization and vice versa) was consistently compromised. In essence, single-objective optimization has limited capacity to represent the multi-response dynamics in conceptual ecohydrological models. However, using multi-objective optimization, good predictions for both streamflow and LAI are obtained. Our results illustrate that the multi-objective optimization provides a balanced solution for multivariate responses and gives better representation of streamflow and LAI dynamics. It is suggested that further development of this approach in terms of conceptual model design and optimization techniques could lead to greatly improved ecohydrological modeling and applications.