Geol-5700, Spring 2013:Parameter Selection for Hydrologic Models
Material: Hydrologic models are used to predict water, energy, sediment and other fluxes at or near the surface of the earth. Predictions depend on the parameters used in the models. Unfortunately, selecting appropriate parameters is not straightforward. Issues of scale and model structure usually preclude the use of measured values. Different procedures for selecting model parameters will be reviewed, including a variety of inverse procedures. We will learn how to quantify uncertainty in model predictions related to parameter selection. This class will not consider parameters used in groundwater modeling, as this material is covered elsewhere.
Outline:
1. Overview: Models and parameters.
a. Why model?..... predict fluxes, given some set of conditions. Without some sort of calibration, limited success.
b. Focus in class on rainfall-runoff modeling; but methods and issues related to a broad range of models.
c. Why is calibration necessary? Models are abstractions, scale issues, model structural error, input data
d. Parsimonious models, ‘identifiability’ of parameters, and ‘equifinality’.
2. Objective Functions: measures of fit
a. Different OFs: sensitivity and assumptions
b. Response surface (1-d, 2-d,….., n-d)
c. Residuals: Check if assumptions are valid?
3. Model structure in Rainfall-Runoff Modeling Toolbox
a. Soil moisture accounting
b. Routing
4. Parameters
a. Parameter sensitivity; no point in searching for “all parameters”.
b. Establishing bounds
5. Calibration 1: single-criterion approach
a. Advantages/disadvantages
b. Methods
c. Predicting errors
6. Calibration 2: multi-criteria approach
a. Advantages/disadvantages
b. Concept and methods
c. Predicting errors
7. Calibration 3: GLUE
a. Advantages/disadvantages
b. Concept and approach
c. Predicting errors
8. Regionalization: translation to “ungauged” basins, or other sites (time permitting)
a. Why model?..... predict fluxes, given some set of conditions. Without some sort of calibration, limited success.
b. Focus in class on rainfall-runoff modeling; but methods and issues related to a broad range of models.
c. Why is calibration necessary? Models are abstractions, scale issues, model structural error, input data
d. Parsimonious models, ‘identifiability’ of parameters, and ‘equifinality’.
2. Objective Functions: measures of fit
a. Different OFs: sensitivity and assumptions
b. Response surface (1-d, 2-d,….., n-d)
c. Residuals: Check if assumptions are valid?
3. Model structure in Rainfall-Runoff Modeling Toolbox
a. Soil moisture accounting
b. Routing
4. Parameters
a. Parameter sensitivity; no point in searching for “all parameters”.
b. Establishing bounds
5. Calibration 1: single-criterion approach
a. Advantages/disadvantages
b. Methods
c. Predicting errors
6. Calibration 2: multi-criteria approach
a. Advantages/disadvantages
b. Concept and methods
c. Predicting errors
7. Calibration 3: GLUE
a. Advantages/disadvantages
b. Concept and approach
c. Predicting errors
8. Regionalization: translation to “ungauged” basins, or other sites (time permitting)
Format:
Each class meeting will be divided into two parts
1. Discussion of assigned reading (text and papers)
2. Exercises, designed to have first-hand experience with the topics covered
1. Discussion of assigned reading (text and papers)
2. Exercises, designed to have first-hand experience with the topics covered
Related Texts:
1. Rainfall-Runoff Modelling in Gauged and Ungauged Catchments
By (author): Thorsten Wagener (Pennsylvania State University, USA) By (author): Howard S Wheater (Imperial College London, UK) By (author): Hoshin V Gupta (University of Arizona, USA)
2. Rainfall-runoff modelling [electronic resource] : the primer
By (author): Keith Beven. 2nd edition.
By (author): Thorsten Wagener (Pennsylvania State University, USA) By (author): Howard S Wheater (Imperial College London, UK) By (author): Hoshin V Gupta (University of Arizona, USA)
2. Rainfall-runoff modelling [electronic resource] : the primer
By (author): Keith Beven. 2nd edition.
Papers:
Assigned and pdf’s available on a convenient webpage (TBA)
Models:
http://www3.imperial.ac.uk/ewre/research/software/toolkit