Uncertainty Evaluation and Appropriate Complexity for the NOAA NWS Distribution Hydrologic Model (RDHM) in a Snow Driven Basin Using a Multiple Criteria Approach
Location
Eccles Conference Center
Event Website
http://water.usu.edu/
Start Date
3-30-2011 10:45 AM
End Date
3-30-2011 10:50 AM
Description
With the increasing availability of remote sensing and GIS data, the use and development of spatially distributed hydrologic modeling has been significantly advanced in recent years. Moreover, the interest in the uncertainty associated with spatially-distributed hydrologic model predictions has increased as well, because the value of such predicted values are limited, if a reasonable estimation of the epistemic (systemic) uncertainty is not provided. Herein, we have applied the NOAA National Weather Service Distributed Hydrologic Model (RDHM) over the Durango River basin in Colorado and assess the model prediction uncertainty arising from parameter estimation. The Multi-Objective Shuffled Complex Evolution Metropolis (MOSCEM) algorithm is used to estimate the parameter estimation uncertainty, in conjunction with a variety of variables such as snow water equivalent, snow cover, and two different discharge gauge observations - one gauge is at the outlet of the basin while the other is internal. In addition, we evaluate the appropriateness of the a priori parameter estimation procedure of Koren et al. (2006) using a variety of traditional criteria or error functions. In particular, given the distributed nature of the model and the availability of remotely sensed distributed observations, we introduce a novel approach for the use of two shape matching or similarity functions: the Hausdorff distance and the Earth Movers distance. We find a high degree of usefulness for this type of error functions. Lastly, we attempt to address the important question of the appropriate degree of distribution (complexity) of the model based on the model performance and the estimates of epistemic uncertainty.
Uncertainty Evaluation and Appropriate Complexity for the NOAA NWS Distribution Hydrologic Model (RDHM) in a Snow Driven Basin Using a Multiple Criteria Approach
Eccles Conference Center
With the increasing availability of remote sensing and GIS data, the use and development of spatially distributed hydrologic modeling has been significantly advanced in recent years. Moreover, the interest in the uncertainty associated with spatially-distributed hydrologic model predictions has increased as well, because the value of such predicted values are limited, if a reasonable estimation of the epistemic (systemic) uncertainty is not provided. Herein, we have applied the NOAA National Weather Service Distributed Hydrologic Model (RDHM) over the Durango River basin in Colorado and assess the model prediction uncertainty arising from parameter estimation. The Multi-Objective Shuffled Complex Evolution Metropolis (MOSCEM) algorithm is used to estimate the parameter estimation uncertainty, in conjunction with a variety of variables such as snow water equivalent, snow cover, and two different discharge gauge observations - one gauge is at the outlet of the basin while the other is internal. In addition, we evaluate the appropriateness of the a priori parameter estimation procedure of Koren et al. (2006) using a variety of traditional criteria or error functions. In particular, given the distributed nature of the model and the availability of remotely sensed distributed observations, we introduce a novel approach for the use of two shape matching or similarity functions: the Hausdorff distance and the Earth Movers distance. We find a high degree of usefulness for this type of error functions. Lastly, we attempt to address the important question of the appropriate degree of distribution (complexity) of the model based on the model performance and the estimates of epistemic uncertainty.
https://digitalcommons.usu.edu/runoff/2011/Posters/23