Evaluating the Performance of the Generalized Likelihood Uncertainty Estimation Approach on Predictive Uncertainty under Different Sampling and Behavioral Threshold Considerations
Abstract
Distributed hydrologic model predictions are affected by errors inherent in the numerical model structure, parameters and input. Thus, uncertainty analyses are being relied upon more and more in the hydrologic modeling practice to evaluate and quantify the effect of these errors on model-results. The Generalized Likelihood Uncertainty Estimation (GLUE) method is one such approach that is being increasingly utilized for the evaluation of predictive uncertainty in distributed hydrologic models. The GLUE approach requires the selection of an appropriate likelihood measure as well as a suitable behavioral threshold value. These selections have a direct impact on the uncertainty estimates obtained through the use of this method. In addition, these estimates are influenced by the sampling approach, the parameter ranges as well as the sampling distributions chosen for parameter ensemble generation. The number of realizations used in the Monte Carlo simulation may also have an effect on the uncertainty results. There are many other considerations as well. The manifestations of several of these subjective choices or considerations on the predictive uncertainty of simulated stage results are investigated in this study by using data from the Everglades National Park (ENP) watershed in Florida, USA. The corresponding Monte Carlo simulation is performed by using the Regional Simulation Model, a finite volume, distributed-parameter hydrologic model. The values of several model parameters that influence surface-water and groundwater stages in the ENP watershed are perturbed in this study. The ENP watershed, with its unique seasonal stage fluctuations and sheet-flow dynamics, provides a very valuable test bed for evaluating the effect of the aforementioned subjective choices on the model’s predictive uncertainty. The study results show that the underlying assumptions and selections used in the uncertainty analysis approach can have a bearing on uncertainty bounds, and hence, on how the model’s predictive uncertainty is portrayed and interpreted.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2010
- Bibcode:
- 2010AGUFM.H21F1123S
- Keywords:
-
- 1847 HYDROLOGY / Modeling;
- 1850 HYDROLOGY / Overland flow;
- 1873 HYDROLOGY / Uncertainty assessment;
- 1890 HYDROLOGY / Wetlands