You are on page 1of 27

What’s On The Menu?

Finite Difference Finite Element Finite Volume

Wayne Hesch
Waterloo Hydrogeologic
2013 NGWA Summit
Outline
• Numerical Methods
– Pros/Cons of Each Method
– What Factors Go Into Deciding What Method to Use?
• Challenges of Numerical Modeling
• Conceptual Approach to Groundwater Modeling
– Improving model acceptance/credibility
– Communicating models to non-modelers
• Example
Take Home Points
• Don’t count out Finite Volume methods
– MODFLOW-USG is a game-changer; it combines
MODFLOW processes and packages with Finite
Element flexibility
• A Conceptual Approach to Modeling lets you…
– Present clear, concise results to stakeholders
– Evaluate several numerical models/methods
– Identify vulnerabilities, minimize uncertainty
• Find and choose the best numerical model
Numerical Methods - Overview
Finite Difference Finite Element Finite Volume
MODFLOW FEFLOW MODFLOW-USG
Structured, Rectilinear Grids Triangular meshes Unstructured grids
(varying geometries)

Image Citation: http://en.wikipedia.org/wiki/File:Rectilinear_grid.svg


http://w3.antd.nist.gov/wctg/netanal/honeycomb.gif
MODFLOW
Strengths Weaknesses
• Widely used and accepted • Inefficient grid refinement
• Open Source • Not ideal for complex
geology (can have
• Well-documented disconnected cells)
• FD solution is easy to • No local refinement for
understand wells and boundaries
• Mass conservative • Limited representation of
• Numerous packages and anisotropy (conductivity
extensions available: PEST, must be perpendicular to
transport, particle tracking… grid cell faces)
MODFLOW-LGR
Strengths Weaknesses
• Refine only in the area of • No contaminant transport
interest between the parent/local
• Local horizontal and vertical grids
refinement • Only 9 Child Grid Regions –
• Two-way feedback between not suited for refining
the coarse and local grids -- around flow boundaries or
consistent boundary individual wells
conditions along their • File management – multiple
adjoining interface MODFLOW models (one for
• Easy to design local (child) parent and each child
grids grid(s))
FEFLOW
Strengths Weaknesses
• Efficient mesh refinement • Local mass conservation is
around wells and boundaries not guaranteed
• Variable stratigraphy • Discontinuous velocities at
(sloping/pinchout layers) element boundaries
• Versatile discretization of non- =>difficult to determine
rectangular model domains unique pathlines
• Representation of anisotropy
• Proprietary
(each element fully represents
the conductivity tensor) • Steep learning curve
MODFLOW-USG (Finite Volume)
• A new version of MODFLOW that uses
Un-Structured Grids
• Follows a Control Volume Finite Difference
formulation
• Discretize the model domain using triangular,
rectangular, or polygon-shaped geometries
MODFLOW-USG
Strengths Weaknesses
• Flexible Grid Geometry • New!
– Refine around wells and
boundaries • Acceptance / Adoptance?
– Variable stratigraphy
• Not all add-on packages are
• Mass-conservative, robust and
efficient solutions supported (yet)
• Follows MODFLOW notation – Transport, UZF, …
– for most packages • Learning curve?
• Open source
• Efficient vertical gridding (sub-
layering)
• A framework for coupling
other processes (conduits,
fractures via CLN package,....)
What Factors Decide What Tool to Use?
Technical Non-technical
• What processes have to be • Client/regulator
simulated requirements
• Data requirements • Company/organizational
standard
• In-house expertise
• Available Licenses/GUIs
• Cost
CHALLENGES OF NUMERICAL
MODELING
Challenges of Numerical Modeling
• First: Choose the Simulator
– FD, FE, or FV? Which one is best?
– Cannot change simulator mid-stream XS: Deformed Grid

• Second: Design Grid/Mesh


– Size? Type? Spacing?
– Deformed or Non-conforming layers?
XS: Uniform Grid
• Third: Assign parameters the grid/mesh
– If grid is modified, often need to adjust the inputs,
through several iterations… Inefficient!
Consequences of Poor Decisions in
Numerical Modeling
• Wrong simulator
– Reach limits of the code capabilities
– Model instabilities, errors, long runtimes
RESULT: Re-design model in a new code… lost time/costs
• Inappropriate grid design
– Model instabilities, errors, long runtimes
– Inadequate resolution (eg. transport simulations)
RESULT: Re-design the grid and re-assign input
parameters, or re-create the model… lost time/costs
Model reliability/credibility is questioned
CONCEPTUAL APPROACH TO
GROUNDWATER MODELING
Conceptual Model
• Define the Geological Structure, Properties,
and Boundary Conditions with GIS shapes and
attributes
• Simulator-Independent
– Convert to MODFLOW, FEFLOW, or MODFLOW-USG
• Grid-Independent
– Generate multiple numerical grids or meshes
Correlate Numerical Models to Conceptual Model

Conceptual
Model
Unstructured
Grid
Uniform Grid Deformed Grid
Multiple Property Realizations
Conceptual Property Zones

60 cols
60 rows

140 cols
140 rows

3:1 Child
Grid
Refinement
Multiple Boundary Condition Realizations
Conceptual 28 cols
Boundary Conditions 27 rows
3 layers

186 cols
190 rows 85 cols
3 layers 117 rows
3 layers
Benefits of Conceptual Modeling
• The decision of what numerical method to use is
delayed until later in the modeling workflow
• Allows the modeler to generate a range of
numerical models, with different grids/meshes,
for different simulators
– With a purely numerical modeling approach, this is
typically not practical due to the required time/effort,
and constraints on budgets for modeling projects
• The modeler can identify vulnerabilities in each
numerical method and reduce uncertainty of the
assumptions/predictions as a whole
How to improve model acceptance/credibility

• Evaluate several hypotheses


– Different simulators (FD, FE, FV)
Scenario1
– Different grid types/parameters
– Varying boundary condition assumptions
• Present results from a variety of
numerical representations, not just one Scenario2

– Run 10’s (100’s?) of numerical models for


the site (Stochastic Approach)
• Using Cloud resources similar to PEST runs?
Scenario3
Effectively communicating models to stakeholders

• Correlate Conceptual Model to


Numerical Model (2D, 3D, transient)
• Inputs:
– Correlate model layers to cross-
sections/borehole logs
– Gridded property zones to GIS data
– Boundaries to GIS/raster data, transient
parameters
• Outputs:
– Interpret head/flux contours and pathlines
traveling through multiple hydrogeological
formations
Example: Evaluating Several Numerical Methods
Conceptual Model: 2D Example, homogenous unconfined aquifer,
river, two pumping wells, and constant head boundary conditions

Finite Difference Finite Element Finite Volume

MODFLOW FEFLOW MODFLOW-USG


Example: Mass Balance Comparison
MODFLOW-2005-fine USG-fine FEFLOW-fine
2000

1800
Cumulative Mass Balance (m3/d)

1600

1400

1200

1000

800

600

400

200

0
Const. Head River Leakage TOTAL IN Const. Head Wells (OUT) River Leakage TOTAL OUT
(IN) (IN) (OUT) (OUT)
Summary
• Don’t count out Finite Volume methods
• MODFLOW-USG provides capabilities similar
to Finite Element methods, with familiarity of
MODFLOW packages/conventions
• Short-term concerns
– Acceptance/adoptance
– Learning curve
– Lengthy development cycles
Summary
• Conceptual Approach to Groundwater Modeling:
– Overcomes many of the challenges of traditional
numerical modeling
• Allows the modeler to..
– Evaluate several numerical models/methods
– Identify vulnerabilities, minimize uncertainty
– Choose the best numerical model
– Improve credibility and reliability
– Present clear, concise results to stakeholders
Acknowledgements
• Sorab Panday: USG slides/figures
• Serguei Chmakov: SLB, USG grid examples and
benchmarking results
Thank You.
Questions?

You might also like