Skip to main content
ARS Home » Plains Area » Fort Collins, Colorado » Center for Agricultural Resources Research » Water Management and Systems Research » Research » Publications at this Location » Publication #303939
Title: Overview and application of the Model Optimization, Uncertainty, and SEnsitivity Analysis (MOUSE) toolbox

Author
item Ascough Ii, James
item FISCHER, CHRISTIAN - Friedrick-Schiller University
item LIGHTHART, NATHAN - Colorad0 State University
item DAVID, OLAF - Colorad0 State University
item Green, Timothy
item KRALISCH, SVEN - Friedrick-Schiller University

Submitted to: Environmental Modeling International Conference Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 6/14/2014
Publication Date: 6/21/2014
Citation: Ascough II, J.C., Fischer, C., Lighthart, N.P., David, O., Green, T.R., Kralisch, S. 2014. Overview and application of the Model Optimization, Uncertainty, and SEnsitivity Analysis (MOUSE) toolbox. Environmental Modeling International Conference Proceedings. Available: http://www.iemss.org/sites/iemss2014/papers/iemss2014_submission_358.pdf.

Interpretive Summary: This paper provides an overview of the Model Optimization, Uncertainty, and SEnsitivity Analysis (MOUSE) software application, an open-source, Java-based toolbox of visual and numerical analysis components for the evaluation of hydrological and environmental models. MOUSE is based on the OPTAS model calibration system for the Jena Adaptable Modeling System framework, is model-independent, and helps the modeler understand model parameters and overall model behavior, identify and select optimal model parameterizations, and evaluate model uncertainties. MOUSE offers well-established local and global sensitivity analysis methods and employs a meta-model approach which permits simultaneous sensitivity analysis with multiple methods but requires only a single Monte Carlo sampling run. MOUSE also contains efficient and reliable single- and multi-objective optimization algorithms (e.g., SCE and NSGAII) to find optimal parameter sets and uses the Generalized Likelihood Uncertainty Estimation (GLUE) method to quantify model uncertainty. Finally, MOUSE has a robust GUI that: 1) allows the modeler to constrain objective functions for specific time periods or events (e.g., runoff peaks, low flow periods, or hydrograph recession periods); and 2) permits graphical visualization of many of the methods described above in addition to visualization of numerous tools contained in the Monte Carlo Analysis Toolbox (MCAT). In addition to an overview of MOUSE, application to the Root Zone Water Quality Model is presented to further demonstrate the model behavior, optimization, and sensitivity/uncertainty analysis tools described above.

Technical Abstract: For several decades, optimization and sensitivity/uncertainty analysis of environmental models has been the subject of extensive research. Although much progress has been made and sophisticated methods developed, the growing complexity of environmental models to represent real-world systems makes it increasingly difficult to fully comprehend model behavior, sensitivities and uncertainties. This paper provides an overview of the Model Optimization, Uncertainty, and SEnsitivity Analysis (MOUSE) software application, an open-source, Java-based toolbox of visual and numerical analysis components for the evaluation of hydrological and environmental models. MOUSE is based on the OPTAS model calibration system for the Jena Adaptable Modeling System framework, is model-independent, and helps the modeler understand model parameters and overall model behavior, identify and select optimal model parameterizations, and evaluate model uncertainties. MOUSE offers well-established local and global sensitivity analysis methods including Regional Sensitivity Analysis, Sobol’, and Morris Screening; and standard sampling schemes including Latin Hypercube, Uniform Random, and Sobol Sequence. MOUSE employs a meta-model approach which permits simultaneous sensitivity analysis with multiple methods but requires only a single Monte Carlo sampling run. MOUSE also contains efficient and reliable single- and multi-objective optimization algorithms (e.g., SCE and NSGAII) to find optimal parameter sets. MOUSE uses the Generalized Likelihood Uncertainty Estimation (GLUE) method to quantify model uncertainty; however, since GLUE has been criticized for being subjective and not consistent with statistical estimates MOUSE includes alternative uncertainty analysis approaches (e.g., the Gaussian Process Regression method). Finally, MOUSE has a robust GUI that: 1) allows the modeler to constrain objective functions for specific time periods or events (e.g., runoff peaks, low flow periods, or hydrograph recession periods); and 2) permits graphical visualization of many of the methods described above in addition to visualization of numerous tools contained in the Monte Carlo Analysis Toolbox (MCAT) including identifiability plots, dotty plots, and Dynamic Identifiability Analysis (DYNIA). In addition to an overview of MOUSE, application to the Root Zone Water Quality Model is presented to further demonstrate the model behavior, optimization, and sensitivity/uncertainty analysis tools described above.