Stochastic Simulations - Big Valley
For this exercise we will revisit the Big Valley model and perform a stochastic simulation.
Do the following:
- Download and unzip stochastic.zip if necessary. Open start.gpr
- Change the MODFLOW Run Option in the Global/Basic Package dialog to Stochastic Simulation
- Check the settings in the Stochastic Options dialog.
- Parameter Randomization
- Go to Parameters dialog
- Randomize both the K fields and the recharge fields.
- Make sure that Log Xform is turned ON for K, but OFF for recharge.
- Change the standard deviation for K fields to 0.3 log cycles.
- Change the standard deviation for recharge fields to 0.00001.
- Change to random sampling
- 50 instances
- Save and run as montecarlo-1
- Browse through the folder of solutions. Click on individual solution folders to update contours.
- Right-click on folder of solutions and select Statistical Analysis. View mean head, etc.
- Review Latin Hypercube distributions/options in the Parameters dialog. No need to run another model.
- Gaussian Simulation
- Create a dummy, empty scatter point set. Rename it to gauss points.
- Select the Interpolation|Gaussian Simulation Options command.
- Log interpolation
- 50 realizations (this number is extremely low - for demo purposes only)
- Truncate 0.001 to 200
- Mean = 0.4
- Variogram
- Range = 700-1000
- Contribution = 0.3
- Nugget = 0
- Select the Run Gaussian Simulation command.
- Turn on log contouring
- Contouring Options.
- Use the Populate Values button and turn on the Log Scale option.
- Run MODFLOW
- Change all K polygons to -100 key value
- Map -> MODFLOW
- Delete all of the HK_XXX parameters except for HK_100
- Turn on Use multiplier arrays option.
- Change Multiplier option for HK_100 to Mult. Arrays. Make sure you have NOT selected Latin Hypercube or you won't be able to see this option.
- Use the Data Set(s) button to select gaussian folder. Note run configuration at bottom.
- Save and run as montecarlo-2.gpr.
- Reset contour options and browse through folder of solutions.
- Probabilistic Capture Zone Analysis
- Right-click on one of the stochastic solutions and select Risk Analysis.
- Probabilitistic capture zone analysis.
- Next.
- Stop in cells with weak sinks.
- Review placement options. One particle per cell is faster, but more particles can give you a cleaner capture zone.
- Weight results based on residual error.
- Finish.
- Select each of the three data sets generated to view. Each one represents probability of capture for a particular well.
- Change contour options
- Contour specified range.
- Change lower limit to 0.1 and turn off the Fill below option.
Note: If you get an error message when running the gaussian simulator, it could be a permissions problem when running the gaussian simulator on CAEDM. It tries to write a temp scratch file and it can't. To solve the problem, copy the FIELDGEN.EXE file to a local drive and change the path to the file using the Edit|Preferences dialog. You can see the current location of FIELDGEN.EXE using the Edit|Preferences dialog.