Kasprzyk, Joseph R听1听;听Stewart, Jenna R听2听;听Livneh, Ben听3听;听Rajagopalan, Balaji听4
1听麻豆影院
2听麻豆影院
3听麻豆影院
4听麻豆影院
Both conceptual and physically-based hydrologic models contain parameters that must be identified and calibrated to improve the fit between simulated and observed streamflow. Single objective calibration, using numerical objective functions such as Nash Sutcliffe Efficiency (NSE), provides a general means to fit simulated and observed streamflow, but the NSE is often biased to improving simulations of high flows at the expense of matching lower flows. Multiple objective calibration uses additional objectives beyond the NSE within the calibration procedure, to improve the holistic model fit, both for low and high flows. The calibration approach is carried out using Multi-objective Evolutionary Algorithms (MOEAs) and complemented with numerical exploration of parameter sensitivity such as random parameter sampling and global sensitivity analysis. This presentation will briefly review the literature within multi-objective model calibration and suggest a suite of interesting objective functions. We then present a series of results on how to build an effective model calibration using MOEAs and the Variable Infiltration Capacity (VIC) hydrologic model, in support of an interdisciplinary project that seeks to improve the prediction of hydrologic fluxes and sediment within forested water supply catchments.