This document summarizes a panel discussion on the past, present, and future of inverse modeling and uncertainty quantification in groundwater modeling. [1] Inverse modeling methods have been developed since the 1980s but are not yet standard practice, with only 20-50% of modelers reported to use them. [2] While inverse modeling and uncertainty quantification provide benefits, their use has been limited when not required by regulators. [3] For these methods to see more widespread future application, explicit requirements by regulators may be needed to outweigh concerns about costs and resources required.
2. Discussion Points
Where have we been?
Inverse methods in model calibration
Uncertainty for use in decision making
Observations
3. Model Calibration Benefits from Using Inverse
Methods
A brief, incomplete time line:
Carrera and Neuman (1986) – theoretical development
Cooley and Naff (1990) – theoretical development
Hill (1992) – MODFLOWP
Doherty: MODINV (1990) and PEST (1994)
UCODE (1998)
1st PEST User’s Conference (2009)
Persuasive case made by Poeter and Hill (1997) Inverse
Models: A Necessary Next Step in Ground-Water
Modeling
4. A qualified success for inverse methods?
Inherently systematic approaches leads to streamlined analysis
And probably technically better results
≡
But Carrera et al. (2005) Inverse Problem in Hydrogeology don’t think it is
standard practice
A survey of 252 groundwater model users found that 20-50% used
inverse methods [Ginn et al., 2007]
Negative reasons include cost (!), need, regulatory acceptance
The linking of the methods to commonly used tools is key
http://cee.engr.ucdavis.edu/faculty/ginn/
5. Uncertain about Uncertainty
Monte Carlo versions of MODFLOW, MODPATH, and MT3DMS
developed and GUI implemented into Groundwater Vistas by 1998
A few projects used, but no commercial demand
In a special issue Stochastic
Environmental Research and Risk
Assessment (2004) considered the
lack of uncertainty quantification (UQ)
in practice
UQ is extensively used at the Nevada
Test Site because it is required by the
regulatory authority
6. A qualified failure for UQ?
Driven only by regulatory need
Consider A Comprehensive Strategy
of Hydrogeologic Modeling and
Uncertainty Analysis for Nuclear
Facilities and Sites by Neuman and
Weirenga, which includes:
Uncertainty in conceptualization, model
development, and predictive uncertainty
Underground Test Area subproject
considers all these things because
the regulator requires it
7. Observations
There is always a gap between theory and
application
Solutions in search of problems are not useful
Future routine use of UQ is unlikely Characterizing Uncertainty in GW
Will only be used when explicitly required 10.0
Modeling
9.0
Academic (11/16)
Government (7/7)
Importance to Clients
8.0 Industry (9/12)
Average
As a business the trade off is between 7.0
6.0
investment cost versus return 5.0
4.0
3.0
As a person the trade off is between my 2.0
1.0
personal time and my enthusiasm for my 0.0
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
profession Importance to Respondent
http://cee.engr.ucdavis.edu/faculty/ginn/