The calibration and uncertainty evaluation of spatially distributed hydrological models

The availability of spatially distributed information, from remote sensing and Geographic Information Systems (GIS), has allowed for the development and implementation of spatially distributed hydrologic models. In particular, remotely sensed distributed snow data sets and precipitation forcing from radar information have allowed us to conduct various studies about snow modeling, snow calibration, and snow effects on runoff. The snow information is very important as a water source, especially in the snowy mountainous regions of the western United States. In this study, we calibrate, evaluate and diagnose the National Weather Service Office of Hydrology HL-RDHM model, a spatially distributed hydrological model to investigate both snow and runoff information over the Durango river basin, which is a mountainous snow-dominated area. For the calibration and evaluation of the HL-RDHM model, we employ overall basin runoff discharge Q1, upstream sub-basin runoff discharge Q2, snow water equivalent and snow cover data in situ and remotely sensed from USGS, SNOTEL and NSIDC as observations, respectively. The snow cover extent is also used as an observation. Through the calibrations and evaluations of HL-RDHM, this study investigates the effect of the additional snow information on runoff simulations only; and on both runoff and snow simulation together; and contrasts the model performance attained when using single- or multi-criteria calibrations. We explore the advantages and disadvantages of using shape-matching error functions such as Hausdorff and Earth Movers' Distance (EMD) in the calibration procedures. Additionally, we seek to establish an appropriate level of model spatial distribution (model complexity) based on the quality of the calibrated model performances. Finally, through parameter estimations, we seek to decide the constrained parameter ranges and parameter uncertainty for the HL-RDHM.;We showed that snow simulations are improved with both single- and multi-criteria calibrations using either traditional or shape-matching error functions. The snow information is very useful to calibrate and evaluate the hydrologic model for snow and runoff information. The multi-criteria calibrations reveal better performances for simultaneously improving overall and sub-basin runoff discharges based on snow information only. The use of shape-matching error functions shows several advantages for model performances: the use of non-commensurate observations, and constrained parameter estimations. In general, after calibration, a distributed model (multi signatures) yields a better performance of snow and runoff than a single signature model, for the case study. Lastly, the shape-matching error functions are more effective in constraining the parameter estimations into physically plausible ranges for the HL-RDHM model.