Conceptual Model Development for MODFLOW or FEFLOW models

FEFLOW Conference
September 2009
Wayne Hesch Schlumberger Water Services

Outline
Introduction
What is a conceptual model

Groundwater modeling workflows

Numerical modeling Conceptual modeling

Benefits of Conceptual Modeling Future Development Questions

Introduction
In order for a groundwater model to be accurate, reliable, and robust, it requires a tremendous amount of information and understanding of the aquifer. The first step in developing a groundwater model, and perhaps the most important, involves the design of a conceptual model Conceptual modeling is often overlooked => modelers constrained by selected simulator, and/or a specific numerical grid or mesh Conceptual modeling can lead to more efficient model development, and opportunity for multiple interpretations and multiple discretizations.

Build a Conceptual Model

a conceptual model is a hydrogeologists mental representation of the groundwater flow system always sketch the system and augment this representation with:
distribution of hydrogeologic layers, location of boundaries, 2D/3D representation of the domain, plan vs. cross-sections, tables of parameter input values,

Conceptual Model: Definitions

A conceptual model is a simplified, high-level representation of the site to be modeled The conceptual model represents our best idea of how the aquifer works. A conceptual model is a basic graphical representation of a complex natural aquifer system that can more easily be adjusted prior to dedicating the effort in developing the numerical model.

Why Create Conceptual Models?

Simplify the field problem Organize field data so that the system can be analyzed more easily The closer the conceptual model approximates the field situation, the more accurate is the numerical model Strive for parsimony simplest is best, but retain enough complexity to adequately reproduce the system behavior Failure of numerical models to make accurate predictions can often be attributed to errors in the conceptual model

Numerical Model Development

the conceptual hydrogeologic model is the most important step in groundwater model process it forms the basis for developing the numerical model an increased level of effort in creating the conceptual model reduces the effort calibrating the numerical model

Conceptual Model

Level of Effort

Numerical Model

Everything should be made as simple as possible, but not simpler. Albert Einstein

Conceptual Model
Developing a good conceptual model requires you to compile detailed information on
geologic formations groundwater flow directions hydrologic boundaries (recharge, rivers, lakes, wetlands, ) hydrogeologic parameters (conductivity, storage, porosity, ) extraction or injection from wells (location, depth, screens, rates), and observations of groundwater head and water quality

Conceptual Model Challenge

vast amounts of data generated from numerous sources in a variety of formats
Field, Analytic, Spatial Eg: GIS, CAD, Gridded files, spreadsheets, databases

added complexity of multiple projects and changing conditions over time determining which data is needed for the groundwater model gathering the required data from other applications in the correct format to import into the modeling software package

The Groundwater Modeling Process

Define Objectives
Conceptual Modeling

Calibrate and Validate Model

No

Collect Data Build Conceptual Model

No

Suitable?
Yes

Sensitivity Analysis
Suitable?
Yes No

Suitable?
Yes

Design Model Grid/Mesh

Numerical Modeling

Predictive Simulations Assign Boundary Conditions

No

Suitable?
Yes

Assign Model Parameters

Post Audit?
After Anderson & Woessner (1982)

Traditional Approach - Numerical Modeling

with numerical modeling designing the grid/mesh is the first step the disadvantages of this approach include:
The correct grid/mesh must be generated before assigning properties, boundaries, wells, etc. If the grid/mesh is modified after other inputs are defined, you will need to check and re-work those input elements, to see that they are still in the appropriate location Generally the input elements are not easily modified, typically you need to delete them and then re-assign them

Numerical Modeling Workflow (FEFLOW)

Input Data
Import shapes, wells, surfaces, XYZ points, cross-sections Digitize new GIS layers

Define SuperElement Mesh

Define 2D Mesh Define Slice Elevations

Define Numerical Model

Develop mesh Define the Property Zones Define Boundaries (rivers, wells,, )

el? od Another m

Define Property Zones

Run Simulation & Analyze Results

Run FEFLOW Check (visualize) results

Define Flow Boundaries

Conceptual Modeling
with conceptual modeling, designing the grid or mesh is the last step Advantages:
define the conceptual model boundary, and model inputs independent of any numerical grid or mesh provides the freedom to design multiple conceptualizations of your site, and easily change your conceptual model define multiple grids or mesh types, each with different resolution and size, and choose the most appropriate one transfer the conceptual model, and the desired numerical grid/mesh, to the numerical model Ability to change the simulator, based on the project needs

all model inputs including properties, wells, and boundary conditions are assigned to the selected grid/mesh automatically resulting MODFLOW or FEFLOW input files are generated

Conceptual Modeling
other advantages:
if you are not happy with the grid/mesh, you can design a new one and re-generate a new numerical model using this new grid this flexibility is not possible with classical numerical modeling, as it would require you to build and manage multiple numerical models

Easily change your model after it is created

raw data are left in tact and grid/mesh-independent Easily expand size of the model domain, vertical discretization, and the model inputs can be easily regenerated from the conceptual objects if the project objectives change, a new numerical model can be easily generated, or existing ones updated, from the conceptual objects it allows for translating the conceptual model to FEFLOW or MODFLOW, with vertical layers that follow the geology or are layer-independent

Workflow: Data Conceptual Numerical

Input Data
Import shapes, wells, surfaces, XYZ points, cross-sections Digitize new GIS layers

Structure

Properties

Boundary Conditions

Define Conceptual Model

Define the Geology: Coverage and Horizons Define the Property Zones Define Boundaries (recharge, pumping wells)

Run Simulation & Analyze Results

Load the files into VMOD/FEFLOW to run the simulation Load results into Hydro GeoBuilder for visualization and interpretation

Numerical Model (MODFLOW/FEFLOW)

Apply a grid/mesh Assign the conceptual model to the grid Create input files for the simulator (MODFLOW/FEFLOW)

Define Model Domain

Define the region where you want to run a model simulation

Finite Differences

Finite Elements

Conceptual Model Structure

define horizons from surfaces horizon truncation rule determines hierarchy; in case of intersections, which will be pushed up/down, or be truncated by surfaces above/below several horizons types accommodate various geological conditions (pinchouts, discontinuous layers)

Conceptual Model: Generating Geologic Model

Define surfaces by interpolating XYZ points from well unit contacts from cross-sections Importing .DEM, .GRD, etc. Convert to horizons

Conceptual Model: Generating Geologic Model

load fence diagrams, cross-sections interpolate contact points to create surfaces

Conceptual Model Structure: Benefits

Model Area
Easily modify the size of the model =>Re-generate superelement mesh and slices =>Re-translate .FEM file input.

Horizons
Use native file formats to define surfaces, and resulting horizons (.XLS, XYZ points, ESRI .GRD, Surfer .GRD, cross-sections) Horizon rules simplifies modeling of complex geology

Conceptual Model: Property Zones

use shapefiles (*.SHP) or CAD polygons to define property zones several methods for defining property zone values:
constant value (by layer) Use shapefile attributes 2D interpolated surface (2D Grid) use 3D Gridded Data

Conceptual Model Properties: Benefits

Flexible units for flow materials Various methods for defining input Not assigned to a mesh/grid
If mesh changes, can easily re-generate FEFLOW input from conceptual model

Conceptual Model: Boundary Conditions

use shapefiles (*.SHP) or CAD polygons/polylines to define boundary geometry and attributes several methods for defining boundary conditions:
constant value use Surface (river stage from DEM) use time schedule use shapefile attributes

Assign values to entire zone or vertices on lines (eg. River gauging stations) Assign geometry to side faces of model domain

Conceptual Model Boundary Conditions: Benefits

Flexible units for flow rates, heads, etc Various methods for defining input Work with combination of data objects and operations minimized pre-processing in GIS Not assigned to a mesh/grid
If mesh changes, can easily re-generate FEFLOW input from conceptual model Can move boundary objects (eg. Groundwater divide)

Pumping wells
Screen locations and pumping rates are mesh-independent: if mesh changes, FEFLOW input can be easily re-generated During translation to .FEM file:
well screens are assigned between appropriate slices flow rates are distributed accordingly for multi-layered wells (no need to assign wells on layer-by-layer basis)

Define Numerical Model

Select simulator and define appropriate grid or mesh MODFLOW
Define horizontal grid resolution, rotation Refine grid, or define local grids Define vertical layers
Use Horizons Independent of geology

FEFLOW
Define superelement mesh Define 2D Horizontal mesh Define 3D Slice elevations
Using Horizons Independent of geology

Benefits of Grid/Mesh Generation

Deformed layer elevations automatically taken from conceptual model Generate model layers independent of the geologic structure Min layer thickness enforced, in pinchout regions Advanced vertical refinement Iterative approach

From Conceptual Model to Multiple Numerical grids with MODFLOW properties

Property zones in the conceptual model

Deformed Grid Layers follow geology Easy, few layers Problems with pinch-outs and cell aspect ratios

Semi-uniform Grid deformed top and bottom layers, uniform in middle Useful for discontinuous layers (common in unconsolidated aquifers)

Uniform Grid Flat layer top/bottoms Fully respects FD assumptions More layers, but useful for transport/density dependent simulations

From Conceptual Model to Multiple Meshes

Property zones in the conceptual model

Deformed Mesh Layers follow geology Easy, few layers Convergence issues with tight geometry/water table fluctuations Semi-uniform deformed top and bottom layers uniform in middle Property upscaling is applied Useful where Deformed mesh fails

Property Translation
With numerical modeling, properties in pinch out layers have to be assigned manually. With conceptual modeling, properties are assigned to 3D Volumes. During translation, for layers that pinch out, the properties are automatically assigned from layers above/below (depending on minimum layer thickness and horizon rules)

Property Upscaling:
Algorithm to Satisfy Darcys Law on Element Level For each finite element
Calculate all property zones intersected by the element (even the thinnest ones are taken into account) Upscale horizontal conductivity using parallel connection rules Upscale vertical conductivity using sequential connection rules using a weighted average of zone values intersected by finite element

Numerical Property Upscaling

Zone lines Grid lines 1 Zone=1 2 3

Zone=2 4 Zone=3

Elements 1, 2, 3 get zone values calculated at their centers. Elements 4, 5, 6 use properties upscaled from all intersected zones (1, 2, and 3)

Conductivity Upscaling

Horizontal conductivity : kh =

k
i =1

Horizontal conductivity : k h h = k h1h1 + k h 2 h2 ( parallel connection)

hi i

( parallel connection)

Vertical conductivity : kv = h N hi i =1 k vi
N

Vertical conductivity : h h1 h2 = + kv kv1 kv 2 h = h1 + h2 ( sequential connection)

( sequential connection)

h = hi
i =1

Simple Budget Analyzer: Comparing Meshes

Deformed Mesh 5 layers

Semi-Uniform Mesh 10 Layers

2.75% difference .more in future work

Future Development
Fully conceptual, simulator-independent approach to building a groundwater model Current implementation supports USGS MODFLOW and FEFLOW FEFLOW: supports 3D mesh design, flow materials, and pumping wells
Future support for Type 1,2,3 boundary conditions

Additional Analytical models Additional Finite Difference/Finite Element models Integration with surface water models Support for Linked simulations using OpenMI technology

Summary
the classical approach to numerical modeling starts with a grid or mesh and then assigns model properties and boundaries
for better local modeling the grid is refined over a number of iterations, which requires you to re-work property zones and boundaries this can be a time-consuming/frustrating process

Summary
A conceptual model improves the efficiencies of these iterations, by housing all data, and providing a visual environment
It helps with the up-front design of the model; more detailed adjustments are done on numerical level It can be considered as the common root for a family of numerical models, so it can also be used as a version control for modeling projects

the use of a conceptual model builder allows you to define mesh and grid-independent model location, flow properties, and boundary conditions
the model grid/mesh is assigned after these have been designed this allows more flexibility in choosing grid orientation and discretization grid refinement is easy to apply to conceptual objects it supports multiple conceptual models for determining the best approach to simulating a specific groundwater environment

Acknowledgments
Co-authors
Serguei Chmakov, Petr Sychev, Collin Tu, Marconi Lima, Schlumberger Water Services

DHI-WASY: Peter Schatzl and Support Team The workflow based approach was strongly motivated by powerful Schlumberger seismic to simulation workflows in the Petrel software (http://www.slb.com/content/services/software/geo/petrel/index.asp?)

References
Anderson, M.P. and W.W. Woessner (1992) Applied Groundwater Modeling: Simulation of Flow and Advective Transport. Academic Press, Inc. New York. Visual MODFLOW 3D-Builder Users Manual: Schlumberger Water Services A New Generation of Waterloo Hydrogeologic Software. MODFLOW and More 2008: Ground Water and Public Policy - Conference Proceedings, Poeter, Hill, & Zheng www.mines.edu/igwmc/ pp. 154-158

http://www.twdb.state.tx.us/gam/GAM_GW_model.htm http://www.ce.utexas.edu/prof/maidment/GISHyd97/gms/gms.htm http://www.indygov.org/

For more information on the OpenMI project, please refer to the extensive OpenMI website at www.openmi.org FEFLOW. FEM File Format

Thank you Questions?