Increasing parameter certainty and data utility through multi-objective calibration of a spatially distributed temperature and solute model

摘要

To support the goal of distributed hydrologic and instream model predictionsbased on physical processes, we explore multi-dimensional parameterizationdetermined by a broad set of observations. We present a systematic approachto using various data types at spatially distributed locations to decreaseparameter bounds sampled within calibration algorithms that ultimatelyprovide information regarding the extent of individual processes representedwithin the model structure. Through the use of a simulation matrix,parameter sets are first locally optimized by fitting the respective data atone or two locations and then the best results are selected to resolve whichparameter sets perform best at all locations, or globally. This approach isillustrated using the Two-Zone Temperature and Solute (TZTS) model for acase study in the Virgin River, Utah, USA, where temperature and solutetracer data were collected at multiple locations and zones within the riverthat represent the fate and transport of both heat and solute through thestudy reach. The result was a narrowed parameter space and increasedparameter certainty which, based on our results, would not have been assuccessful if only single objective algorithms were used. We also found thatthe global optimum is best defined by multiple spatially distributed localoptima, which supports the hypothesis that there is a discrete and narrowlybounded parameter range that represents the processes controlling the dominanthydrologic responses. Further, we illustrate that the optimization processitself can be used to determine which observed responses and locations aremost useful for estimating the parameters that result in a global fit toguide future data collection efforts.