Wednesday, May 14, 2008

Advanced vs. Standard Versions in Vistas

A common question I get about Groundwater Vistas is "what is the difference between advanced and standard versions"? Here is a quick summary of features found in the advanced and enterprise versions that are not in the standard version:

- Recharge/ET memory compression
- Support for the HUF Package (Hydrogeologic Unit Flow)
- Support for SWIFT
- Monte Carlo modeling for analysis of uncertainty
- Pest's SVD Assist automation
- Pest's Null Space Monte Carlo technique

Recharge and ET memory compression is useful when you are running a large model that has lots of stress periods. Normally GV allocates enough memory to hold all recharge and ET values for all stress periods. As models get larger and more transient, that presents a problem whereby there is just not enough memory to hold it all. In the advanced version, when GV sees this problem, it uses a compression algorithm (like WinZip) to compress the arrays. They are still stored in memory, just in compressed form.

Pest's SVD Assist is a technique that allows for much faster run times when you are trying to estimate a large number of parameters. It also removes the problem of trying to estimate insensitive parameters. Parameters are grouped into "super" parameters that are linear combination of the real parameters. Only super parameters are estimated. This usually results in about a ten-fold decrease in the number of parameters to estimate. In GV, we have made the use of this complex technique relatively simple.

Pest's Null Space Monte Carlo technique builds on the SVD Assist calibration technique. It allows you to create multiple realizations of your model that are all calibrated to a desired level of calibration quality. In the past, calibrating each realization of a monte carlo simulation was not feasible due to long model run times. Special features in the Pest technique, along with the use of SVD Assist, allows you to calibrate many realizations in a relatively short period of time. Like SVD Assist, GV has made the use of this technique straightforward.

The Enterprise version of Vistas is the same as Advanced but also adds Remote Model Launch (RML) and GW3D for 3D visualization of model results.

Monday, May 05, 2008

Note to Self - USGS Documents have Version Numbers

I noted in a previous post that SFR2 (Streamflow-Routing with unsaturated flow beneath streams) is now supported in Groundwater Vistas. I did not realize at the time that I was working from the original SFR2 manual and it had changed subsequently. Many thanks to Dave Prudic for helping me figure this out. Turns out that USGS publications are not necessarily static - they have version numbering (or at least the capability for version numbering). The current SFR2 manual is called Version 1.10, April 2006.

Anyway, I noticed that something was wrong when trying to import an existing SFR2 input file into Groundwater Vistas and the format did not match the manual I was working from. This problem would also affect files created by Groundwater Vistas for SFR2 (SFR1 would not be affected and this is the default setting in GV). However, the created files would not run so if your model was affected by this problem you would know it.

Version 5.17 Build 5 fixes this problem in SFR2.

Friday, May 02, 2008

MF2K5 V 1.05.00 in GV

I just uploaded Groundwater Vistas Version 5.17 Build 4. This version supports the new version of MODFLOW2005 Version 1.05.00 released earlier this week. If you are using the UZF Package, you will see one new variable called SURFDEP which is the average undulation depth at land surface and effects recharge. The default value is GV is 0.0. There is also a new option to apply recharge directly to the water table. I do not have support for that option in Vistas yet so if you need it let me know.

There is also a new variable in the PCG2 solver that allows a separate damping factor for transient stress periods when a run contains both steady and transient stress periods. I do not yet have that hooked up either.

Another addition to Vistas is post-processing for Null Space Monte Carlo runs. Testing has gone very well for this new Pest procedure. It is truly amazing what it can do. Kudos to John Doherty for yet another significant contribution to our science.

Thursday, May 01, 2008

Null Space Monte Carlo - Done!

Groundwater Vistas 5.17 Build 3 contains a working Null Space Monte Carlo procedure. This was developed by John Doherty in Pest Version 11.5. It allows you to calibrate any number of realizations based on the same base calibration. The assumption is that there are lots of parameters in the calibration (preferably pilot points) and that you used SVD-Assist for the calibration. I have tested this in GV for a synthetic 3D model using pilot points for Kx and Kz. While it is a simple model, there were over 400 pilot points and so the characteristics are fairly representative of a real model.

The next step is to work on the post-processing of the calibrated realizations and finally to incorporate this into the Monte Carlo simulation and post-processing that has been in GV for years.

USGS Has Been Busy

The USGS has released numerous versions of MODFLOW recently. The most recent is an update for MODFLOW2005 (version 1.05.00). The updated the model to the same version as in GSFLOW (described later) and fixed some minor bugs. Most important to me is they added back the DDREFERENCE flag in output control that they removed when going from MODFLOW2000 to MODFLOW2005. That undocumented flag is very nice because it allows drawdown to be computed from the results of any particular time step. If you use Groundwater Vistas, you can access this flag through the new custom output control.

SEAWAT Version 4 was released last month. The major new feature is the ability to simulate heat transport.

The Conduit Flow Process (CFP) was released recently. It allows for the simulation of karst features (conduits) in MODFLOW2005.

GSFLOW is a new surface water/groundwater model. The groundwater part is MODFLOW2005 and the surface water is PRMS (Precipitation-Runoff Modeling System).

None of these new versions are in Groundwater Vistas yet. The new MODFLOW2005 and SEAWAT will be added first, followed by CFP, and finally GSFLOW. The latter will take quite a while to implement.