Parallel multi-objective calibration of a component-based river temperature model

Many recent modeling efforts have employed component-based modeling frameworks to take advantage of the flexibility they provide in representing systems more holistically. Despite the benefits that are driving this adoption, conducing model parameter estimation, uncertainty analysis, sensitivity assessment, and other simulations of this nature within component-based modeling frameworks has remained unexplored. Using a multi-objective calibration of a component-based river temperature model, we illustrate how the component cloning and parallel model execution interfaces we have implemented in the HydroCouple framework support such simulations. The river temperature model calibration application we present involves a heavily human mediated 13.6 km section of the Logan River in Utah, USA with limited information regarding variable inflows. Due to the flexibility in the modeling and calibration framework, results from the calibration effort were successful with root mean square errors of 0.4-0.7 degrees C and provided insights on mechanisms controlling river temperatures.