GMS:Multiplier Arrays for Parameters

From XMS Wiki
Jump to navigationJump to search
Stochastic Modeling
Model Calibration
Automated Parameter Estimation
PEST Dialog
Run Options
MODFLOW-USG Observations
Plot Wizard
Calibration Targets
Parameter Dialog
Pilot Points
Multiplier Arrays for Parameters
Standard MODFLOW Parameters
Stochastic Modeling
Stochastic Modeling
Gaussian Field Generator
Risk Analysis Wizard

HK and RCH parameters may be associated with a multiplier array. A dataset is associated with a parameter zone and the starting parameter value is multiplied by the dataset to define the spatial variation of the parameter throughout the zone.

Setting up Multiplier Arrays

The multiplier dataset is associated with the parameter zone in the Parameters dialog by first turning on the Use multiplier arrays toggle, selecting either Single Array or Multiple Array from the pull-down list in the Multiplier column, and then selecting the dataset using the button in the Dataset/Folder column. If the Single Array option is selected, a single dataset should be selected in the Dataset/Folder column. If the Multiple Array option is selected, an entire folder of datasets should be selected. The Multiple Array option is only available for stochastic modeling.

If the Single Array option is selected the array for the parameter will be defined as the starting/mean value multiplied by the multiplication array for all cells associated with the parameter. Multiplier arrays are not available for parameters that use pilot points.

When the Multiple Arrays option is selected, the number of stochastic model runs will be a function of how many datasets are in the folder selected by the user for the multipliers.

The Parameters dialog showing the Multiple Arrays option

Guassian Field Based Monte Carlo Simulations

The Multiple Array option for user-defined multiplier arrays can be used in combination with the new Gaussian Field Generator to perform a Monte Carlo simulation. This is accomplished by first generating a set of Gaussian fields using the FIELDGEN code and then associating the resulting folder of datasets with a parameter as described above. A stochastic simulation can then be performed using the Parameter Randomization method. For each model instance, GMS loads a new instance of the multiplier array into the model input. A default parameter value of 1.0 is generally recommended in such instances.