Data and Model-Driven Decision Support for Environmental ...
Transcript of Data and Model-Driven Decision Support for Environmental ...
![Page 1: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/1.jpg)
Data and Model-Driven Decision Support for Environmental Management of a Chromium Plume at Los Alamos National Laboratory (LANL) - 13264
Velimir V. Vesselinov (“monty”)1 [email protected], Danny Katzman2, David Broxton1, Kay Birdsell1, Steven Reneau1, David Vaniman3, Pat Longmire3, June Fabryka-Martin3, Jeff Heikoop3, Mei Ding3, Don Hickmott3, Elaine Jacobs3, Tim Goering2, Dylan Harp1, Phoolendra Mishra1
1 Computational Earth Sciences, Earth and Environmental Sciences, 2 Environmental Programs, 3 Earth Systems Observations, Earth and Environmental Sciences, Los Alamos National Laboratory (LANL), Los Alamos, NM
Waste Management Symposium 2013 ER Challenges: Alternative Approaches for Achieving End State (Session 109)
February 28, 2013, Phoenix, AZ
LA-UR-13-21534
![Page 2: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/2.jpg)
Outline
Model-based Decision Support
Deterministic, Probabilistic vs Non-Probabilistic Decision Methods
Information Gap (info-gap) Decision Theory
Decision Support for Chromium contamination site @ LANL o Site conceptual model
o Model-based decision analyses
o Monitoring network design
o Additional activities related to contaminant remediation
MADS: Model Analyses & Decision Support Open source computational framework http://mads.lanl.gov
Decision Support in ASCEM (Advanced Subsurface Computing for Environmental Management) | http://ascemdoe.org
![Page 3: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/3.jpg)
Model-based Decision Support provides decision makers with model analysis of decision scenarios:
evaluation, ranking and optimization of alternative decision scenarios
takes into account site data and knowledge including existing uncertainties (uncertainties in conceptualization, model parameters, and model predictions)
Decision metric(s): e.g. contaminant concentration or environmental risk at a point of compliance, etc.
Decision goal(s): e.g. no exceedance of MCL’s, dose limits, or risk levels at compliance points
Decision scenarios: combinations of predefined activities to achieve the decision goal(s)
![Page 4: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/4.jpg)
Model-based Decision Support (cont.)
Activities: o data acquisition campaigns o field/lab experiments o monitoring o remediation
Activities are analyzed in terms of their impact on decision making process (decision uncertainties)
Decision uncertainties: uncertainties associated with selection of optimal decision scenarios, or performance of specific decision scenarios
The Game: Decision maker vs. Nature
Important:
Additional activities are selected only to reduce decision uncertainties
Additional activities are not selected to reduce model or parameter uncertainties (unconstrained problem).
![Page 5: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/5.jpg)
Decision Methods
Deterministic methods (“traditional” performance assessment): a single model simulation representing worst-case or expected (“best” estimated) system behavior
![Page 6: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/6.jpg)
Decision Methods
Deterministic methods (“traditional” performance assessment): a single model simulation representing worst-case or expected (“best” estimated) system behavior
Probabilistic methods (Bayesian techniques, GoldSim): analyses based on a series of model simulations capturing expected probabilistic uncertainties (Monte Carlo, Markov Chain Monte Carlo, Null Space Monte Carlo, etc.)
![Page 7: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/7.jpg)
Decision Methods
Deterministic methods (“traditional” performance assessment): a single model simulation representing worst-case or expected (“best” estimated) system behavior
Probabilistic methods (Bayesian techniques, GoldSim): analyses based on a series of model simulations capturing expected probabilistic uncertainties (Monte Carlo, Markov Chain Monte Carlo, Null Space Monte Carlo, etc.)
Non-probabilistic methods: analyses based on a series of model simulations representing unknown uncertainties (Minimax/Maximin Theory, Information Gap Decision Theory, etc.)
![Page 8: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/8.jpg)
Non-Probabilistic Decision Methods
Lack of knowledge or information precludes decision analyses requiring probabilistic distributions (e.g. Bayesian approaches)
o probability distributions cannot be defined (!)
o uniform distributions frequently applied instead (causing biased decision analyses)
Severe uncertainties can have important impact in the decision analyses
o heavy tails: non-Gaussian distributions will infinite variances
o black swans: low probability events in distribution tails with significant decision impacts
o dragon kings: unexpected high probability events in the distribution tails
Non-probabilistic decision methods can be applied to effectively incorporate lack of knowledge and severe uncertainties in decision making process
Non-Probabilistic and Probabilistic methods can be coupled (e.g. unknown probability distribution parameters can be a subject of non-probabilistic analysis, e.g. info-gap)
![Page 9: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/9.jpg)
Information Gap Decision Theory
Non-probabilistic methodology for comparison of alternative decision scenarios
Decision uncertainty is bounded by robustness and opportuness functions
Robustness function (immunity to failure)
Opportuness function (immunity to windfall)
Ben-Haim (2006). Info-gap decision theory: decisions under severe uncertainty. Academic Press.
Information Gap Decision Theory @ http://mads.lanl.gov
![Page 10: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/10.jpg)
Example analyses: • Deterministic • Probabilistic (Bayesian) • Non-probabilistic (Info-Gap)
![Page 11: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/11.jpg)
Deterministic analysis
”Worst case” or “Best estimate” parameter set
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
MCL
”Worst case” or “Best estimate”
prediction
0.1 1 10 100
![Page 12: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/12.jpg)
Bayesian (probabilistic) analysis
“Best” parameter set
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
Pro
bab
ility
MCL “Best” prediction
0.1 1 10 100
![Page 13: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/13.jpg)
Bayesian (probabilistic) analysis
“Best” parameter set
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
Pro
bab
ility
MCL “Best” prediction
0.1 1 10 100
![Page 14: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/14.jpg)
Bayesian (probabilistic) analysis
“Best” parameter set
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
Pro
bab
ility
MCL “Best” prediction
0.1 1 10 100
![Page 15: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/15.jpg)
Bayesian (probabilistic) analysis
“Best” parameter set
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
Pro
bab
ility
MCL “Best” prediction
0.1 1 10 100
… the challenge is in tail characterization
![Page 16: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/16.jpg)
Info-Gap (non-probabilistic) analysis
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
MCL Nominal prediction
0.1 1 10 100
Nominal parameter set
1
2
3
4
Info
-gap
un
cert
ain
ty m
etri
c α
![Page 17: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/17.jpg)
Info-Gap (non-probabilistic) analysis
Nested parameter sets
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
MCL Nominal prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etri
c α
α1 α2 α3 α4
info-gap uncertainty metric (horizon of unknown uncertainty) = α α1 < α2 < α3 < α4 … < α∞
![Page 18: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/18.jpg)
Info-Gap (non-probabilistic) analysis
Nested parameter sets
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
MCL Nominal prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etri
c α
α1 α2 α3 α4
info-gap uncertainty metric (horizon of unknown uncertainty) = α α1 < α2 < α3 < α4 … < α∞
min{α1: C}
min{α2: C}
min{α3: C}
max{α1: C}
max{α2: C}
max{α3: C}
![Page 19: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/19.jpg)
Info-Gap (non-probabilistic) analysis
Nested parameter sets
Parameter 1
Par
amet
er 2
Contaminant concentration [ppb]
MCL Nominal prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etri
c α
α1 α2 α3 α4
info-gap uncertainty metric (horizon of unknown uncertainty) = α α1 < α2 < α3 < α4 … < α∞
Decision uncertainty
min{α1: C}
min{α2: C}
min{α3: C}
max{α1: C}
max{α2: C}
max{α3: C}
… the challenge is in finding min/max’s
![Page 20: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/20.jpg)
Info-Gap Analysis: Decision selection based on robustness
Contaminant concentration [ppb]
MCL Nominal
prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etr
ic α
0
![Page 21: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/21.jpg)
Info-Gap Analysis: Decision selection based on robustness
Contaminant concentration [ppb]
MCL Nominal
prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etr
ic α
“Strong” robustness
“Weak” robustness
0
![Page 22: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/22.jpg)
Info-Gap Analysis: Decision selection based on robustness
Contaminant concentration [ppb]
MCL Nominal
prediction
0.1 1 10 100
1
2
3
4
Info
-gap
un
cert
ain
ty m
etr
ic α
“Strong” robustness
“Weak” robustness
0
![Page 23: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/23.jpg)
Info-Gap Analysis: Decision selection based on opportuness
Contaminant concentration [ppb]
MCL
Nominal prediction
0.1 1 10 100
Info
-gap
un
cert
ain
ty m
etr
ic α
0
1
2
3
4
![Page 24: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/24.jpg)
Info-Gap Analysis: Decision selection based on opportuness
Contaminant concentration [ppb]
MCL
Nominal prediction
0.1 1 10 100
Info
-gap
un
cert
ain
ty m
etr
ic α
“Weak” opportuness
“Strong” opportuness 0
1
2
3
4
![Page 25: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/25.jpg)
Info-Gap Analysis: Decision selection based on opportuness
Contaminant concentration [ppb]
MCL
Nominal prediction
0.1 1 10 100
Info
-gap
un
cert
ain
ty m
etr
ic α
“Weak” opportuness
“Strong” opportuness 0
1
2
3
4
![Page 26: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/26.jpg)
Info-Gap Analysis: Synthetic Network Design
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
10 monitoring wells in an aquifer 2 wells detect contaminant concentrations above
MCL (5 ppm) 8 wells detect background concentrations (0.5 ppm)
MCL = 5 Background = 0.5
![Page 27: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/27.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
3 new proposed monitoring well locations (green dots) Which well can be expected to detect contaminants
above MCL?
Info-Gap Analysis: Synthetic Network Design
MCL = 5 Background = 0.5
![Page 28: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/28.jpg)
Info-Gap Analysis: Synthetic Network Design
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
Where is the contaminant source?
MCL = 5 Background = 0.5
![Page 29: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/29.jpg)
Info-Gap Analysis: Synthetic Network Design
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
Where is the contaminant source?
MCL = 5 Background = 0.5
![Page 30: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/30.jpg)
Info-Gap Analysis: Synthetic Network Design
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
Where is the contaminant source?
MCL = 5 Background = 0.5
![Page 31: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/31.jpg)
Info-Gap Analysis: Synthetic Network Design
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
Where is the contaminant source?
MCL = 5 Background = 0.5
![Page 32: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/32.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 33: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/33.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 34: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/34.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 35: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/35.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 36: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/36.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 37: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/37.jpg)
100 40
0.5
0.5
0.5
0.5
0.5
0.5 0.5
0.5
MCL = 5 Background = 0.5
c > 5 (MCL)
Info-Gap Analysis: Synthetic Network Design Multiple plausible plume configurations …
![Page 38: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/38.jpg)
0.5
0.01
0.1
1
10
0.1 1 10
100 40
0.5
0.5
0.5
0.5
0.5 0.5
0.5
Info-Gap Analysis: Network Design
MCL = 5 Background = 0.5
MCL 0.5
Wb Wc Wa
Info
-gap
un
cert
ain
ty m
etri
c α
Predicted concentrations [ppm] vs Info-gap uncertainty
opportuness (Wa, Wb, Wc)
For more information: http://mads.lanl.gov
![Page 39: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/39.jpg)
Chromium plume in the regional aquifer at LANL
![Page 40: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/40.jpg)
Chromium plume in the regional aquifer at LANL
![Page 41: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/41.jpg)
Chromium plume in the regional aquifer at LANL
![Page 42: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/42.jpg)
Chromium plume in the regional aquifer at LANL
![Page 43: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/43.jpg)
Chromium plume in the regional aquifer at LANL
![Page 44: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/44.jpg)
Chromium plume in the regional aquifer at LANL
![Page 45: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/45.jpg)
Chromium plume in the regional aquifer at LANL
Perched zone monitoring wells
![Page 46: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/46.jpg)
Chromium plume in the regional aquifer at LANL
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Two-screen aquifer monitoring wells
![Page 47: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/47.jpg)
Chromium plume in the regional aquifer at LANL
Supply wells
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Two-screen aquifer monitoring wells
![Page 48: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/48.jpg)
Chromium plume in the regional aquifer at LANL
Supply wells
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Cr concentrations (~2012) [ppb] MCL = 50 ppb Background 5-8 ppb
Two-screen aquifer monitoring wells
![Page 49: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/49.jpg)
Chromium plume in the regional aquifer at LANL
Supply wells
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Cr concentrations (~2012) [ppb] MCL = 50 ppb Background 5-8 ppb
Two-screen aquifer monitoring wells
Water-level contours (~2012) [2 ft]
![Page 50: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/50.jpg)
Chromium plume in the regional aquifer at LANL
Supply wells
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Cr concentrations (~2012) [ppb] MCL = 50 ppb Background 5-8 ppb
Two-screen aquifer monitoring wells
Water-level contours (~2012) [2 ft]
![Page 51: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/51.jpg)
Chromium plume in the regional aquifer at LANL
Supply wells
Single-screen aquifer monitoring wells
Perched zone monitoring wells
Cr concentrations (~2012) [ppb] MCL = 50 ppb Background 5-8 ppb
Two-screen aquifer monitoring wells
Water-level contours (~2012) [2 ft]
![Page 52: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/52.jpg)
~54,000 kg of Cr6+ released in Sandia Canyon between 1956 and
1972
Cr6+ detected above MCL (50 ppb; NM standard) in 4 monitoring
wells in the regional aquifer beneath LANL
Cr6+ plume size is about 2 km2 (region above MCL)
Cr6+ plume is located near LANL site boundary
Series of water-supply wells are located nearby
Contaminant source location and mass flux at the top of the
regional aquifer are unknown due to complex 3D pathways
through the vadose zone
Limited remedial options due to aquifer depth (~300 m below the
ground surface) and complexities in the subsurface flow
Current conceptual model for chromium migration in the
subsurface is supported by multiple lines of evidence (hydrogeological, geophysical geophysical, mineralogic, petrographic, and
geochemical studies and model analyses)
LANL Chromium site
![Page 53: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/53.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
![Page 54: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/54.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
![Page 55: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/55.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
![Page 56: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/56.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
![Page 57: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/57.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
![Page 58: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/58.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
![Page 59: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/59.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
![Page 60: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/60.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
Cr6+ ~5,600 kg Cr3+ ~17,000 kg
![Page 61: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/61.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
Cr6+ ~5,600 kg Cr3+ ~17,000 kg
Cr6+ ~230 kg Cr3+ ~2 kg
![Page 62: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/62.jpg)
N
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+ ~54,000 kg
Cr6+ ~5,600 kg Cr3+ ~17,000 kg
Cr6+ ~230 kg Cr3+ ~2 kg
Cr6+ ~3,000 kg Cr3+ ~9,000 kg
![Page 63: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/63.jpg)
N
Cr6+
> 50 ppb
Cr6+
>400 ppb
Water supply well PM-3
Vad
ose
zo
ne
(~3
00
m)
Cr6+
> 100 ppb
Cr6+ ~54,000 kg
Cr6+ ~5,600 kg Cr3+ ~17,000 kg
Cr6+ ~230 kg Cr3+ ~2 kg
Cr6+ ~3,000 kg Cr3+ ~9,000 kg
![Page 64: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/64.jpg)
3D simulation of flow and transport in the vadose zone
![Page 65: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/65.jpg)
3D simulation of flow and transport in the vadose zone
Sandia Canyon
Mortandad Canyon
![Page 66: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/66.jpg)
3D simulation of flow and transport in the vadose zone
Sandia Canyon
Mortandad Canyon
Boreholes
![Page 67: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/67.jpg)
3D simulation of flow and transport in the vadose zone
Sandia Canyon
Mortandad Canyon
Boreholes
Alluvial flow
![Page 68: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/68.jpg)
3D simulation of flow and transport in the vadose zone
Sandia Canyon
Mortandad Canyon
Perched zones
Boreholes
Alluvial flow
![Page 69: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/69.jpg)
3D simulation of flow and transport in the vadose zone
Sandia Canyon
Mortandad Canyon
Perched zones
Laterally diverted flowpaths
Boreholes
Alluvial flow
![Page 70: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/70.jpg)
LANL chromium site
50 ppb
1000 ppb
Model predicted plume shape (~2012) Cr6+ MCL 50 ppb
Sandia Canyon Mortandad Canyon
Vad
ose
zo
ne
(~3
00
m)
Single-screen aquifer monitoring wells
Two-screen aquifer monitoring wells
![Page 71: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/71.jpg)
Plumes represent
Cr6+ > 50 ppb
(NM standard) along
the water table
Due to uncertainties, a series of alternative models (plumes) are plausible
Model predictions are constrained by all the available regional-aquifer data
(hydrogeological and geochemical)
11 out 83 plausible plumes shown
2009 model analyses for source identification / network design
![Page 72: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/72.jpg)
Plausible contaminant-arrival locations (83 out of 551)
Wells R-62, R-61 and R-50 were not drilled yet
2009 model analyses for source identification / network design
![Page 73: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/73.jpg)
2009 model estimate of the plausible Cr6+ [ppb] along the
regional aquifer water table
MCL = 50 ppb
Wells R-62, R-61 and R-50 were not drilled yet Locations of wells R-62, R-61 and R-50 were optimized based on model analyses Observed concentrations at R-62, R-61 and R-50 confirmed model predictions R-43 concentration were at background when the analyses were performed Since 2010, R-43 concentrations are increasing and approaching the model
predicted concentration
![Page 74: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/74.jpg)
2009 model estimate of the plausible Cr6+ [ppb] along the
regional aquifer water table
100
~30-200
MCL = 50 ppb
18
Wells R-62, R-61 and R-50 were not drilled yet Locations of wells R-62, R-61 and R-50 were optimized based on model analyses Observed concentrations at R-62, R-61 and R-50 confirmed model predictions R-43 concentration were at background when the analyses were performed Since 2010, R-43 concentrations are increasing and approaching the model
predicted concentration
![Page 75: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/75.jpg)
2009 model estimate of the plausible Cr6+ [ppb] along the
regional aquifer water table
100
~30-200
MCL = 50 ppb
~40
18
Wells R-62, R-61 and R-50 were not drilled yet Locations of wells R-62, R-61 and R-50 were optimized based on model analyses Observed concentrations at R-62, R-61 and R-50 confirmed model predictions R-43 concentration were at background when the analyses were performed Since 2010, R-43 concentrations are increasing and approaching the model
predicted concentration
![Page 76: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/76.jpg)
2012 model estimate of the plausible Cr6+ [ppb] along the
regional aquifer water table
MCL = 50 ppb
![Page 77: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/77.jpg)
Plausible contaminant-arrival locations (83 out of 551)
2009 model analyses for source identification / network design
![Page 78: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/78.jpg)
o Series of plausible contaminant-arrival locations in a well-constrained
region
o All the obtained solutions (492) are almost equivalent
o Additional analyses are performed considering multiple contaminant
arrival locations
2012 model analyses for source identification / network design
![Page 79: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/79.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Cr6+ mass distribution
![Page 80: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/80.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
There are uncertainties associated with these estimates. For example, source mass may vary between 31,000 and 72,000 kg, and mass in the aquifer may vary between 300 and 3,300 kg.
Cr6+ mass distribution
![Page 81: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/81.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
There are uncertainties associated with these estimates. For example, source mass may vary between 31,000 and 72,000 kg, and mass in the aquifer may vary between 300 and 3,300 kg.
Series of additional activities are identified to reduce decision uncertainties related to contaminant mass distribution (source)
Cr6+ mass distribution
![Page 82: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/82.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 83: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/83.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 84: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/84.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 85: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/85.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 86: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/86.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 87: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/87.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 88: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/88.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 89: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/89.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities
![Page 90: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/90.jpg)
an open-source high-performance computational framework for Model Analyses and Decision Support (MADS)
advanced adaptive computational techniques: o sensitivity analysis (local / global); o uncertainty quantification (local / global); o optimization / calibration / parameter estimation (local / global); o model ranking & selection o decision support (probabilistic / non-probabilistic)
novel robust algorithms o Agent-Based Adaptive Global Uncertainty and Sensitivity (ABAGUS)
Harp & Vesselinov (2012) An agent-based approach to global uncertainty and sensitivity analysis. Computers & Geosciences.
o Adaptive hybrid (local/global) optimization strategy (Squads) Vesselinov & Harp (2012) Adaptive hybrid optimization strategy for calibration and parameter estimation of physical process models. Computers & Geosciences.
internal coupling with analytical 3D contaminant transport solvers external coupling with any process simulator (e.g. ModFlow, FEHM, Amanzi,
PFLOTRAN, STOMP/eSTOMP, TOUGH, TOUGHREACT, …) source code, examples, performance comparisons, and tutorials @
http://mads.lanl.gov MADS tools will be implemented in the ASCEM project
![Page 91: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/91.jpg)
an open-source high-performance computational framework for Model Analyses and Decision Support (MADS)
advanced adaptive computational techniques: o sensitivity analysis (local / global); o uncertainty quantification (local / global); o optimization / calibration / parameter estimation (local / global); o model ranking & selection o decision support (probabilistic / non-probabilistic)
novel robust algorithms o Agent-Based Adaptive Global Uncertainty and Sensitivity (ABAGUS)
Harp & Vesselinov (2012) An agent-based approach to global uncertainty and sensitivity analysis. Computers & Geosciences.
o Adaptive hybrid (local/global) optimization strategy (Squads) Vesselinov & Harp (2012) Adaptive hybrid optimization strategy for calibration and parameter estimation of physical process models. Computers & Geosciences.
internal coupling with analytical 3D contaminant transport solvers external coupling with any process simulator (e.g. ModFlow, FEHM, Amanzi,
PFLOTRAN, STOMP/eSTOMP, TOUGH, TOUGHREACT, …) source code, examples, performance comparisons, and tutorials @
http://mads.lanl.gov MADS tools will be implemented in the ASCEM project
![Page 92: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/92.jpg)
Summary:
Both Non-Probabilistic and Probabilistic uncertainties often exist in decision problems
In the case of probabilistic methods, definition of prior probability distributions for model parameters with unknown/uncertain distribution can produce biased predictions and decision analyses
In the case of non-probabilistic methods, lack of knowledge and severe uncertainties can be captured
Non-probabilistic methodologies have been successfully applied for a series of synthetic and real-world problems, though less often for waste and environmental management
o Harp & Vesselinov (2011). Contaminant remediation decision analysis using information gap theory.
o Vesselinov & Harp (2013). Model-driven decision support for monitoring network design using information gap theory.
MADS provides a computationally efficient framework for decision analyses using non-probabilistic and probabilistic methods ( http://mads.lanl.gov )
![Page 93: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/93.jpg)
Summary:
Current conceptual model for chromium migration in the subsurface is supported by multiple lines of evidence (hydrogeological, geophysical geophysical, mineralogic, petrographic, and geochemical studies and model analyses)
Data- and model-based (systems-based) decision analyses are successfully implemented to progress characterization and performance assessment at the site (monitoring network design, additional characterization activities)
Plume characterization is a challenging and nonunique problem because multiple models are consistent with the site data and conceptual knowledge
Decision analyses are facilitated by implementation of robust techniques and high-performance computing
Activities are currently planned to constrain uncertainties impacting decision analyses:
o aquifer heterogeneity: spatial distribution of low-permeable zones that can act as secondary contaminant sources
o contaminant mass distribution o spatial and temporal distribution of contaminant mass flux to the aquifer o implementation of remedial activities
![Page 94: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/94.jpg)
![Page 95: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/95.jpg)
Chromium plume in the regional aquifer at LANL
Challenges:
define site conceptual model and existing uncertainties: o complex hydrostratigraphy, geochemistry, flow and transport regimes
o data characterized with different support volumes and uncertain due to various factors
o multiple contaminant pathways
o hydrogeological, geophysical geophysical, mineralogic, petrographic, and geochemical studies applied
o current conceptual model is supported by multiple lines of evidence
perform computationally efficient analyses:
o parameter estimation (PE)
o model calibration
o uncertainty quantification (UQ)
o decision support (DS)
high computational demands for model simulations and analyses (requiring utilization of LANL high-performance computing capabilities)
uncertainties associated with application of the remedial options
![Page 96: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/96.jpg)
Model-driven decision support o evaluation and optimization of additional characterization activities
(e.g. field pumping and tracer tests)
o evaluation and optimization of monitoring network design (well
locations)
o evaluation and optimization of remedial activities (ongoing)
Characterization activities: o exploration, analysis & evaluation of alternative conceptual models
o estimation of nature/extent/fate of contaminant plumes (Cr6+, ClO4-)
o source identification (estimating location/flux of contaminant mass
arriving at the top of regional aquifer)
o estimation of vadose zone & aquifer heterogeneity (hydrogeology
and geochemistry)
Work related to LANL Chromium site
![Page 97: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/97.jpg)
mean min max mean min max mean min max
Source 54,000 31,000 72,000 0 0 0 54,000 31,000 72,000 0
Canyon alluvial sediments 18 6 27 17,982 5,694 26,973 18,000 5,700 27,000 99.9
---- Wetland 15 5 23 15,105 4,783 22,657 15,120 4,788 22,680 99.9
---- Downstream sediments 3 1 4 2,877 911 4,316 2,880 912 4,320 99.9
Bandelier 2,625 250 12,750 7,875 750 38,250 10,500 1,000 51,000 75
Puye 3,000 600 15,000 9,000 1,800 45,000 12,000 2,400 60,000 75
Perched zones 230 100 500 0 0 0 230 100 500 0
Lavas 1,750 225 2,250 5,250 675 6,750 7,000 900 9,000 75
Puye 990 250 2,000 2,970 750 6,000 3,960 1,000 8,000 75
Miocene 181 25 1,000 542 75 3,000 722 100 4,000 75
Aquifer 1,100 270 3,300 0 0 0 1,100 270 3,300 0
Total 9,894 1,726 36,827 43,619 9,744 125,973 53,512 11,470 162,800
Estimates of chromium
mass distributionCr6+ [kg] Cr3+ [kg] Cr6+ + Cr3+ [kg] Cr3+/Cr6+
ratio [%]
Estimates of chromium mass distribution in the subsurface including existing uncertainties
![Page 98: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/98.jpg)
Information Gap Decision Theory Nominal (“best”) model prediction intended for decision making
(based on nominal / “best estimates” model parameter set)
Decision metric(s)
Decision goal(s)
Decision scenarios: a series of alternative decisions to compare
Info-Gap Uncertainty Model
Model predictions for each decision scenario constrained by Info-Gap Uncertainty Model
Ben-Haim (2006). Info-gap decision theory: decisions under severe uncertainty. Academic Press.
![Page 99: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/99.jpg)
Info-Gap Analysis: Synthetic Network Design Unknown model parameters (8) characterizing plume size:
o source locations (coordinates x, y)
o source lateral size (xS, yS)
o flow direction
o aquifer dispersivities (longitudinal, horizontal/vertical transverse)
Uncertain concentration observations (calibration targets) (10) due to:
o measurement errors
o uncertain background concentrations
o uncertain local hydrogeological and geochemical conditions
Analytical model of the 3D contaminant flow
Unknown model parameters estimated using inversion
Decision question: which of the new proposed well location has the highest immunity of failure to detect concentrations above MCL (c > 5 ppm) i.e. which well provides the most robust decision to improve the monitoring network
![Page 100: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/100.jpg)
Series of alternative models: different scale and complexity
Espanola basin
LANL site
![Page 101: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/101.jpg)
Series of alternative models: different scale and complexity
Espanola basin
LANL site
LANL site
Chromium site
![Page 102: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/102.jpg)
Series of alternative models: different scale and complexity
Espanola basin
LANL site
LANL site
Chromium site
![Page 103: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/103.jpg)
Series of alternative models: different scale and complexity
Espanola basin
LANL site
LANL site
Chromium site
![Page 104: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/104.jpg)
Series of alternative models: different scale and complexity
Espanola basin
LANL site
LANL site
Chromium site
![Page 105: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/105.jpg)
Regions along the
top the regional
aquifer where the
calculated Cr6+
concentrations
exceed 1500 ppb
based on averaging
of all the acceptable
model solutions
2012 model
analyses
1 arrival location
2 arrival locations
3 arrival locations
![Page 106: Data and Model-Driven Decision Support for Environmental ...](https://reader031.fdocuments.us/reader031/viewer/2022012101/6169ec6411a7b741a34cde24/html5/thumbnails/106.jpg)
Cr3+ [kg] Cr6+ [kg] Estimated mass distribution [kg]
Source
Wetland
Alluvial
Bandelier
Puye
Perched zones
Lavas
Puye
Miocene
Regional aquifer
54,000
15
3
2,625
3,000
230
1,750
990
181
1,100
0
15,105
2,877
7,875
9,000
2
5,250
2,970
542
10
Geochemical lab-scale analyses (cores) • key support for optimizing CME – MNA • attenuation potential • reduction potential
Pumping/tracer tests at existing wells • immediate affect • source removal • capture zone analysis • characterize field–scale hydrogeologic and
geochemical properties • characterize secondary Cr source
Grade Control Structure • immediate effect • stabilize wetland to control Cr, PCBs, and
other
Reduced effluent volume (infiltration) • mid-term effect • reduce flux of secondary Cr source
Groundwater flow & transport modeling • key for interpretation of the collected data • key support for optimizing CME - MNA
Planned activities