GMS:Parameter Dialog: Difference between revisions

From XMS Wiki
Jump to navigationJump to search
No edit summary
Line 18: Line 18:
After performing an inverse model run, the inverse code writes out a text file containing the set of optimal parameter values corresponding to the minimum calibration error. The next step is to read these values into GMS. This is accomplished by selecting the '''Import Optimal Values''' button.  Once you open the file, the optimal parameter values will be loaded into the starting value field and displayed in the parameter list.
After performing an inverse model run, the inverse code writes out a text file containing the set of optimal parameter values corresponding to the minimum calibration error. The next step is to read these values into GMS. This is accomplished by selecting the '''Import Optimal Values''' button.  Once you open the file, the optimal parameter values will be loaded into the starting value field and displayed in the parameter list.


If any of your parameters use pilot points, importing the optimal values will also import and  create a new data set for the associated 2D scatter point set.
If any of your parameters use pilot points, importing the optimal values will also import and  create a new dataset for the associated 2D scatter point set.


==Spreadsheet==
==Spreadsheet==
Line 62: Line 62:
Select this option to include a multiplier array for RCH and HK parameters.
Select this option to include a multiplier array for RCH and HK parameters.


===Data Set / Folder===
===Dataset / Folder===
Use this button to select a multiplier array of a set of multiplier arrays (stochastic only) by selecting a folder of data sets.
Use this button to select a multiplier array of a set of multiplier arrays (stochastic only) by selecting a folder of datasets.


==Stochastic Options==
==Stochastic Options==

Revision as of 15:29, 19 August 2013

MODFLOW
Pre-processing
MODFLOW Commands
Building a MODFLOW Model
Map to MODFLOW
Calibration
Packages Supported in GMS
Saving a MODFLOW Simulation
Importing MODFLOW Files
Unsupported MODFLOW Features
Run MODFLOW
Post-processing
MODFLOW Display Options
MODFLOW Post-Processing Viewing Options
Reading a MODFLOW Simulation
Tutorials
Packages
Flow: BCF6, HUF, LPF, UPW
Solvers:

DE4, GMG, NWT, PCG,

PCGN, LMG, SIP, SOR,

SMS
Other:

BAS6, BFH, CHD1, CLN,

DRN1, DRT1, EVT1, ETS1,

GAGE, GHB1, GNC, HFB1,

HUF, LAK3, MNW1, MNW2,

OUT1, RCH1, RIV1, SFR2,

STR1, SUB1, SWI2, WEL1,

UZF1

Calibration,
Parameters,
Stochastic Modeling
Calibration
Model Calibration
Automated Parameter Estimation
PEST Dialog
PEST
Run Options
Observations
MODFLOW-USG Observations
Plot Wizard
Calibration Targets
Parameters
Parameters
Parameter Dialog
Pilot Points
Multiplier Arrays for Parameters
Standard MODFLOW Parameters
Stochastic Modeling
Stochastic Modeling
Gaussian Field Generator
Risk Analysis Wizard
T-PROGS

When building a MODFLOW inverse model, the input data must be parameterized. This is accomplished by assigning a set of key values to selected input fields. Once the key values are assigned, the next step is to create a parameter list.

Parameters can also be used to perform forward runs.

See also Standard MODFLOW Parameters.

Creating/Deleting Parameters

The Parameters dialog contains the list of parameters. A new parameter can be created by selecting the New Parameter button. Each parameter that is defined should correspond to a key value that has been defined in the MODFLOW input. A parameter can be removed from the list by selecting the parameter and selecting the Delete Parameter button. The entire list of parameters can be deleted by selecting the Delete All button.

Initialize from Model

In most cases, the fastest and simplest way to create the parameter list is to use the Initialize from Model button. When this button is selected, GMS traverses the MODFLOW input data corresponding to legal parameter values and searches for key values. It is assumed that the key values are entered as negative numbers. When a unique negative number is found, a new parameter is added to the list, and a default name is given to the parameter based on the parameter type.

NOTE: The Initialize from Model command will not search for key values in the Well package. This is because negative pumping rates are perfectly common and do not necessarily correspond to key values. If you wish to define a well Q as a parameter, you should use the New Parameter button and manually create the parameter.

Import Optimal Values

After performing an inverse model run, the inverse code writes out a text file containing the set of optimal parameter values corresponding to the minimum calibration error. The next step is to read these values into GMS. This is accomplished by selecting the Import Optimal Values button. Once you open the file, the optimal parameter values will be loaded into the starting value field and displayed in the parameter list.

If any of your parameters use pilot points, importing the optimal values will also import and create a new dataset for the associated 2D scatter point set.

Spreadsheet

The spreadsheet holds parameters, their types, starting and bounding values and other options that depend on the packages and run options defined for the current MODFLOW simulation in the Global Options\Basic Package.

Parameter Estimation Solve

Toggle on this option to solve for the optimal parameter value using a parameter estimation program like PEST. (See Automated Parameter Estimation)

Name

The name must be unique and limited to 8 characters. The default name should be sufficient. This name will be used to reference the parameter in the MODFLOW solution and output files.

Key

The key value, usually a negative value, must be unique and is used to link the parameter with the data in the MODFLOW input files. The key value should also be entered into the MODFLOW data by directly entering the values using the grid based approach or by using the map module.

Type

GMS supports most of the parameter types that are also supported by the MODFLOW PES process.

Start Value/Pilot Points

This value will be the starting value for inverse modeling, the mean value for stochastic modeling, or the substitute value for forward runs.

For inverse modeling, the closer the starting value is to the "optimal" value, the better the odds that the inverse model will converge and the less time it will take to converge. It is generally not a good idea to give all parameters of a given type (e.g., recharge) a constant value and let the inverse model start from that point. Ideally, field tests, soil types, ground cover, and sound modeling judgement can provide a good set of starting values. It is also a good idea to undergo some manual trial-and-error calibration prior to setting up the inverse code.

Pilot points can be used to define a parameter by selecting the drop-down arrow in this column and selecting the Pilot points option. Pilot points are an alternative to using zonation to define parameter locations. When you use pilot points, you use the Pilot Point Options button to choose a 2D point scatter point set and choose the appropriate interpolation options.

Min Value Max Value

The min and max values will provide the bounds for the parameter, and they must encompass the starting value. The parameter values will be forced between these values during inverse and stochastic modeling. In these cases, the min and max values are just "suggestions" and are used to predict parameter values.

When selecting the min and max values for the parameters, care should be taken not to make the range in values too large. Inverse models are highly senstive to the stability of the underlying model. If MODFLOW does not converge, the invese model will not be able to find a solution. Furthermore, excessive cell drying can cause the inverse model to fail to converge. If the min and/or max values are too extreme, the odds of the MODFLOW model not converging or excessive numbers of cells going dry increases. It is best to select a limited range for each parameter and then to compare the final optimal parameter value to this range. If the optimal value is at either the min or max of the range, the range can be adjusted and the inverse model can be re-launched.

Log xform

This option log transforms the value during prediction process of inverse modeling and the random number generation process of stochastic modeling. The best parameters to log transform are those that can vary by orders of magnitude like hydraulic conductivity.

It is also recommended that you log transform recharge parameters if you are using pilot points for hydraulic conductivity and the hydraulic conductivity parameter is log transformed.

BSCAL

MODFLOW documentation includes:

This value is an alternate scaling factor for the parameter, and always needs to be a positive number. If the parameter value becomes 0.0, which can occur for parameters that are not log transformed, BSCAL is used in the scaling. If the absolute value of the parameter is less than BSCAL, BSCAL is used in the scaling. The best value to use is problem dependant. Good choices are the smallest reasonable value of the parameter or a value two to three orders of magnitude smaller than the value specified by the starting value. If the smallest reasonable value is 0.0, a reasomable non-zero value needs to be used. BSCAL has no effect on the scaled sensitivities for log-transformed parameters.

Multiplier

Select this option to include a multiplier array for RCH and HK parameters.

Dataset / Folder

Use this button to select a multiplier array of a set of multiplier arrays (stochastic only) by selecting a folder of datasets.

Stochastic Options

Standard Deviation
Use this field to specify the standard deviation of a parameter for a stochastic simulation.
Mean value
Use this field to specify the mean of a parameter for a stochastic simulation.
Distribution
When the parameter is stochastic, use this option to choose between a normal or linear distribution. A random number for the parameter is generated using the distribution, the mean value (starting value) and the standard devation.
Std Deviation
The standard deviation is used for a stochastic parameter to generate a random number using the chosen distribution.
Num Segments
When the parameter is stochastic and the stochastic method is Latin Hypercube, the number of segments helps determine how man total MODFLOW runs will be used.

See also