Once MODFLOW you have your MODFLOW model functional, you will move on to calibrate the model

GWV offers tools for trial and error as well as automated calibration

We will use 16 head target locations

The data could be entered one at a time through the interface

But we will import text file containing the target data

Select Add->Import->Targets

Select targets.dat

Click OK

In this case, the default values for file format are correct, so look them over to become aware of the options/issues,then

Select OK

16 targets will appear on the plan view as small blue dots

6 in layer 1, 5 in layer 2, and 5 in layer 3

You may edit target information in Analytic Element mode by double-clicking on a target symbol

Model results must be imported again to compute the calibration statistics

Select Plot->Import Results

Click OK

Select Plot->Calibration->Statistics/Plots…

Select the type of targets to use in the calculation and ranges of layers

Click the Statistics button and review the items in the list

Click OK

Click the Plot Observed vs. Simulated button

Targets are color coded by layer

Notice the strong bias such that higher heads are simulated too low

Notice all of the residuals are positive

To post residuals on a contour map

Select Plot->Calibration->Post Residuals

Check Post Residuals

Change the font size by selecting Plot->Calibration->Target Options

Click the font button to change the size or font

Select View->Refresh to redraw

Turn off the grid display by selecting Plot->What to Display

Uncheck the option for finite-difference grid

For layer one:


Sensitivity analysis is the process whereby model parameters or boundary conditions are altered slightly and the effect on model calibration statistics is observed

You choose a parameter type, the number of simulations, and a parameter multiplier for each simulation

GWV then runs MODFLOW the desired number of times and produces a sensitivity plot

Calibration statistics versus parameter multiplier are displayed to show Select Model->AutoSensitivitySetup

Select K_x as the parameter to vary

Keep the default of Zone number 1 and 5 simulations

Click Multipliers and Enter 0.2, 0.5, 1.0, 1.5, and 2.0 for simulations 1-5

OK

Select Models->RunAutoSensitivity

Yes to view results

View the sum of squared residuals, residual mean, residual standard deviation, average drawdown, and total flux to a designated boundary condition reach ... sum of squared residuals are shown

Notice a K_x value of half the current value gives a much better calibration

GWV also supports AUTOMATED CALIBRATION through PEST, UCODE, and MODFLOWP

Select Model->Calibration

Select the parameters that you would like to estimate

We will estimate horizontal hydraulic conductivity in zone 1 (which is the whole model in this case)

Select Model->Calibration->Parameters and enter the data shown on the following dialog

GWV estimates a multiplier on the parameter value so the multiplier of 1.0 in column 4 above means that we will start with our initial guess (in this case 100 ft/d).

The minimum and maximum values constrain the parameter value. These are also multipliers.

Now select Model->GVCalibration->Options and change the number of iterations from 25 to 5

Click OK

Select Model->GVCalibration->Start Calibration

MODFLOW will run a number of times

If the model does not converge on some executions, allow it to continue anyway

If you are asked, say YES to look at the calibration summary each time

If you are asked, Continue to save parameters for next iteration and continue to next iteration until the values no longer change and the optimal solution has been found

When the calibration is complete

Notice how the value of K_x has changed (the multiplier reflects this)

Notice that the optimal value (0.5) was found quickly and most of the effort was necessary

Notice how the statistics have changed

If you like the results of the calibration run,

Select Model->GVCalibration->EndCalibration if you haven't reached the last iteration

Select Model->GVCalibration->Update Databases

This command takes the results of the calibration and puts the new values in the data base

Rerun the model for one simulation to make sure your MODFLOW files contain the new values

View the reduced and balanced residuals

Automated calibration does not always go smoothly

Estimate as few parameters as possible and only sensitive parameters

If GWV Calibration does not work try one of the more robust calibration tools such UCODE, PEST, or MODFLOWP.