scholarly journals Technical note: Method of Morris effectively reduces the computational demands of global sensitivity analysis for distributed watershed models

2013 ◽  
Vol 10 (4) ◽  
pp. 4275-4299 ◽  
Author(s):  
J. D. Herman ◽  
J. B. Kollat ◽  
P. M. Reed ◽  
T. Wagener
Keyword(s):  
Sensitivity Analysis ◽  
Global Sensitivity Analysis ◽  
Hydrologic Model ◽  
Technical Note ◽  
Diagnostic Methods ◽  
Laboratory Research ◽  
Storage Space ◽  
Hydrologic Models ◽  
Global Sensitivity ◽  
Spatially Distributed

Abstract. The increase in spatially distributed hydrologic modeling warrants a corresponding increase in diagnostic methods capable of analyzing complex models with large numbers of parameters. Sobol' sensitivity analysis has proven to be a valuable tool for diagnostic analyses of hydrologic models. However, for many spatially distributed models, the Sobol' method requires a prohibitive number of model evaluations to reliably decompose output variance across the full set of parameters. We investigate the potential of the method of Morris, a screening-based sensitivity approach, to provide results sufficiently similar to those of the Sobol' method at a greatly reduced computational expense. The methods are benchmarked on the Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM) model over a six-month period in the Blue River Watershed, Oklahoma, USA. The Sobol' method required over six million model evaluations to ensure reliable sensitivity indices, corresponding to more than 30 000 computing hours and roughly 180 gigabytes of storage space. We find that the method of Morris is able to correctly identify sensitive and insensitive parameters with 300 times fewer model evaluations, requiring only 100 computing hours and 1 gigabyte of storage space. Method of Morris proves to be a promising diagnostic approach for global sensitivity analysis of highly parameterized, spatially distributed hydrologic models.

scholarly journals Technical Note: Method of Morris effectively reduces the computational demands of global sensitivity analysis for distributed watershed models

2013 ◽  
Vol 17 (7) ◽  
pp. 2893-2903 ◽  
Author(s):  
J. D. Herman ◽  
J. B. Kollat ◽  
P. M. Reed ◽  
T. Wagener
Keyword(s):  
Sensitivity Analysis ◽  
Global Sensitivity Analysis ◽  
Hydrologic Model ◽  
Technical Note ◽  
Diagnostic Methods ◽  
Laboratory Research ◽  
Storage Space ◽  
Hydrologic Models ◽  
Global Sensitivity ◽  
Spatially Distributed

Abstract. The increase in spatially distributed hydrologic modeling warrants a corresponding increase in diagnostic methods capable of analyzing complex models with large numbers of parameters. Sobol' sensitivity analysis has proven to be a valuable tool for diagnostic analyses of hydrologic models. However, for many spatially distributed models, the Sobol' method requires a prohibitive number of model evaluations to reliably decompose output variance across the full set of parameters. We investigate the potential of the method of Morris, a screening-based sensitivity approach, to provide results sufficiently similar to those of the Sobol' method at a greatly reduced computational expense. The methods are benchmarked on the Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM) over a six-month period in the Blue River watershed, Oklahoma, USA. The Sobol' method required over six million model evaluations to ensure reliable sensitivity indices, corresponding to more than 30 000 computing hours and roughly 180 gigabytes of storage space. We find that the method of Morris is able to correctly screen the most and least sensitive parameters with 300 times fewer model evaluations, requiring only 100 computing hours and 1 gigabyte of storage space. The method of Morris proves to be a promising diagnostic approach for global sensitivity analysis of highly parameterized, spatially distributed hydrologic models.

scholarly journals Hierarchical Sensitivity Analysis for Large-scale Process-based Hydrological Modeling with Application in an Amazonian Watershed

2020 ◽  
Author(s):  
Haifan Liu ◽  
Heng Dai ◽  
Jie Niu ◽  
Bill X. Hu ◽  
Han Qiu ◽  
...  
Keyword(s):  
Sensitivity Analysis ◽  
Vadose Zone ◽  
Model Calibration ◽  
Large Scale ◽  
Global Sensitivity Analysis ◽  
Hydrologic Model ◽  
Global Sensitivity ◽  
Uncertainty Sources ◽  
Sensitivity Indices ◽  
Zone Parameters

Abstract. Sensitivity analysis is an effective tool for identifying important uncertainty sources and improving model calibration and predictions, especially for integrated systems with heterogeneous parameter inputs and complex process coevolution. In this work, an advanced hierarchical global sensitivity analysis framework, which integrates the concept of variance-based global sensitivity analysis with a hierarchical uncertainty framework, was implemented to quantitatively analyse several uncertainties associated with a three-dimensional, process-based hydrologic model (PAWS). The uncertainty sources considered include model parameters (three vadose zone parameters, two groundwater parameters, and one overland flow parameter), model structure (different thicknesses to represent unconfined and confined aquifer layers) and climate scenarios. We apply the approach to an ~ 9,000 km2 Amazon catchment modeled at 1 km resolution to provide a demonstration of multiple uncertainty source quantification using a large-scale process-based hydrologic model. The sensitivity indices are assessed based on two important hydrologic outputs: evapotranspiration (ET) and groundwater contribution to streamflow (QG). It was found that, in general, parameters, especially the vadose zone parameters, are the most important uncertainty contributors for all sensitivity indices. In addition, the influence of climate scenarios on ET predictions is also nonignorable. Furthermore, the thickness of the aquifers along the river grid cells is important for both ET and QG. These results can assist in model calibration and provide modelers with a better understanding of the general sources of uncertainty in predictions associated with complex hydrological systems in Amazonia. We demonstrated a pilot example of comprehensive global sensitivity analysis of large-scale, complex hydrological and environmental models in this study. The hierarchical sensitivity analysis methodology used is mathematically rigorous and capable of being implemented in a variety of large-scale hydrological models with various sources of uncertainty.

scholarly journals From maps to movies: high-resolution time-varying sensitivity analysis for spatially distributed watershed models

2013 ◽  
Vol 17 (12) ◽  
pp. 5109-5125 ◽  
Author(s):  
J. D. Herman ◽  
J. B. Kollat ◽  
P. M. Reed ◽  
T. Wagener
Keyword(s):  
Sensitivity Analysis ◽  
High Resolution ◽  
The United States ◽  
Hydrologic Model ◽  
Laboratory Research ◽  
Model Parameters ◽  
Time Varying ◽  
Resolution Time ◽  
Watershed Models ◽  
Spatially Distributed

Abstract. Distributed watershed models are now widely used in practice to simulate runoff responses at high spatial and temporal resolutions. Counter to this purpose, diagnostic analyses of distributed models currently aggregate performance measures in space and/or time and are thus disconnected from the models' operational and scientific goals. To address this disconnect, this study contributes a novel approach for computing and visualizing time-varying global sensitivity indices for spatially distributed model parameters. The high-resolution model diagnostics employ the method of Morris to identify evolving patterns in dominant model processes at sub-daily timescales over a six-month period. The method is demonstrated on the United States National Weather Service's Hydrology Laboratory Research Distributed Hydrologic Model (HL-RDHM) in the Blue River watershed, Oklahoma, USA. Three hydrologic events are selected from within the six-month period to investigate the patterns in spatiotemporal sensitivities that emerge as a function of forcing patterns as well as wet-to-dry transitions. Events with similar magnitudes and durations exhibit significantly different performance controls in space and time, indicating that the diagnostic inferences drawn from representative events will be heavily biased by the a priori selection of those events. By contrast, this study demonstrates high-resolution time-varying sensitivity analysis, requiring no assumptions regarding representative events and allowing modelers to identify transitions between sets of dominant parameters or processes a posteriori. The proposed approach details the dynamics of parameter sensitivity in nearly continuous time, providing critical diagnostic insights into the underlying model processes driving predictions. Furthermore, the approach offers the potential to identify transition points between dominant parameters and processes in the absence of observations, such as under nonstationarity.

scholarly journals Hierarchical Sensitivity Analysis for Large Scale Process-based Hydrological Modeling with Application in an Amazonian Watershed

2019 ◽  
Author(s):  
Haifan Liu ◽  
Heng Dai ◽  
Jie Niu ◽  
Bill X. Hu ◽  
Han Qiu ◽  
...  
Keyword(s):  
Sensitivity Analysis ◽  
Large Scale ◽  
Overland Flow ◽  
Global Sensitivity Analysis ◽  
Hydrologic Model ◽  
Model Parameters ◽  
Hydrological Models ◽  
Global Sensitivity ◽  
Uncertainty Sources ◽  
Sensitivity Indices

Abstract. Sensitivity analysis is an effective tool for identifying important uncertainty sources and improving model calibration and predictions, especially for integrated systems with heterogeneous parameter inputs and complex processes coevolution. In this work, an advanced hierarchical global sensitivity analysis framework, which integrates a hierarchical uncertainty framework and a variance-based global sensitivity analysis, was implemented to quantitatively analyze several uncertainties of a three-dimensional, process-based hydrologic model (PAWS). The uncertainty sources considered include model parameters, model structures (with/without overland flow module), and climate forcing. We apply the approach in a ~ 9000 km2 Amazon catchment modeled at 1 km resolution to provide a demonstration of multiple uncertainty source quantification using a large-scale process-based hydrologic model. The sensitivity indices are assessed based on three important hydrologic outputs: evapotranspiration (ET), ground evaporation (EG), and groundwater contribution to streamflow (QG). It is found that, in general, model parameters (especially those within the streamside model grid cells) are the most important uncertainty contributor for all sensitivity indices. In addition, the overland flow module significantly contributes to model predictive uncertainty. These results can assist model calibration and provide modelers a better understanding of the general sources of uncertainty in predictions of complex hydrological systems in Amazonia. We demonstrated a pilot example for comprehensive global sensitivity analysis of large-scale complex hydrological models in this research. The hierarchical sensitivity analysis methodology used is mathematically rigorous and can be applied to a wide range of large-scale hydrological models with various sources of uncertainty.

Sign in / Sign up

Export Citation Format

Share Document