Groundwater Modeling of a Fractured Shale Aquifer

1
Groundwater Modeling of a Fractured Shale Aquifer

• GG 655/CEE 623
• Class project
• Atiim Senthill
• Sanjay Mohanty
• Christian Thielmann

2
Overview

• Description of project site
• Background
• Objectives
• Analytical Approach
• Numerical Approach
• Model setup
• Results
• Failed Attempts (8)
• Drawbacks
• Conclusion

3
Site

• Avtex Fibers Superfund Site in Front Royal, VA

4
Background

• Manufactured rayon, polyester, and polypropylene
  fibers from 1940-1989
• Added to the Superfund Programs National
  Priority List in 1986
• Tons of rayon manufacturing waste and by-products
  improperly disposed of over 220-acres (vadose
  basins)

5
Contaminants

• Long and narrow plume, 400 ft deep along
  structural strike, towards and under the river
• Contaminants of concern
• Carbon Disulfide
• Heavy metals
• Phenol
• PCBs

6
Objectives

• Construct a numerical model (MODFLOW) to simulate
  observed drawdown
• Calibrate the hydraulic parameters of the system
• Approximate site heterogeneity
• Run MODPATH and/or MT3DMS to simulate contaminant
  transport
• Show the required capture zone
• Show remediation timeframe

7
Analytical Model

• Model was developed by Papadopulos (1965)
• Goal to understand effective aquifer behavior
  for purpose of evaluating remediation method
• Assumes aquifer is homogenous, anisotropic and
  confined
• Simplified analytical models will not fully
  describe the local details of the aquifer
  response in this heterogeneous system

8
Numerical Model Approach

• Collection of Hydrogeological information about
  the site
• Conductivity, Porosity, site heterogeneity
• Groundwater elevation, surface elevation
• Drawdown data
• Defining boundary conditions, assumption
• Run Numerical model (MODFLOW)
• Calibration (MF2K PES/PEST)
• Solute transport (MT3DMS/MODPATH)

9
Modeling Concerns

• Complex site heterogeneity
• Horizontal and vertical anisotropy
• Adjacent river
• Insufficient data regarding observation wells and
  pumping well

10
Hydrogeological parameters

• Aquifer thickness 200 ft (61 m)
• Pumping rate 40 gpm
• Storage coefficient 7.3e-05
• Transmissivity 2.1e-04 m²/s along strike
• 1.2e-05 m²/s
  across strike
• Hydr. Conductivity 3.4e-04 cm/s along strike
• 2.0e-05 cm/s
  across strike
• Hydr. gradient 0.03
• Average porosity 0.04

11
Model setup

• 3-dimensional grid in flow direction
• 3 layers with constant layer conductivity
• Horizontal anisotropy
• Vertical anisotropy
• River
• River depth ( 3 5 ft)
• Conductance (2 -10 ft/day)
• Pumping well TW-01 and observation wells (depths
  are best guess)
• Specified head (north-east boundary)

12
Model setup cont.
13
Results (MODFLOW)
Top layer
Intermediate layer
14
Actual drawdown
Top layer
Intermediate layer
15
Model Calibration

• Model was calibrated using the PEST approach
• Assumptions
• River information was correct
• Horizontal and vertical anisotropy values were
  correct
• Defined specified head boundary was out of the
  range of the pumping well
• Depth of observation well

16
Calibration results
17
Calibration results cont.
Head contour (top layer)
18
Solute transport

• Solute transport and capture modeled by MODPATH
• Lack of information made MT3DMS impractical
• Solute transport modeled with and without pumping
  well in place

19
MT3DMS Results
20
MODPATH Results
21
Failed Attempts

• Single layer
• Actual topography/flat topography
• River as specific head boundary
• River as a polygon/arc
• Horizontal isotropy/anisotropy
• MK2K PES process
• Single layer approach was a complete failure with
  no hope of success!
• Cost many nightmares

22
Failed Attempts Cont.

• Multiple layer
• Actual topography/flat topography
• River as specific head boundary
• River as a polygon/arc
• Horizontal isotropy/anisotropy
• MK2K PES process
• Various pumping well Z-values

23
Drawbacks

• Lack of information caused erroneous results
• Software failed to match actual surface and
  groundwater elevations using existing
  interpolation options
• Software appears to be unable to interpret
  realistic river depths

24
Conclusion/Recommendations

• Numerical models require more input information
  compared to analytical models
• MODFLOW requires multiple layers to simulate
  heterogeneous subsurface conditions
• The Z-value of wells can have an impact on the
  results calculated by MODFLOW

25
Conclusion Cont.

• River properties have a significant effect on
  head contours
• High conductance river acts as specific head
  boundary
• Deep river causes flooded cells/ acts as
  specific head boundary
• In this simulation the contaminant plume was
  captured in ? 50years