First International FLAC/DEM Symposium

Determination of Aquifer and Aquitard Parameters from Inverse Modeling

Michael Burlingame, PE Bureau of Design and Construction

New Jersey Department of Environmental Protection

Landfill with Supply and Observation Wells

Landfill Observation Wells

Public Supply Well

Supply and Observation Wells

Landfill

Supply Well

Observation Wells

Conceptual Site Model

Public Supply Well

0 17 29

Observation Wells

Cohansey Aquifer Kirkwood Aquitard Manasquan Aquitard OW-2 OW-1

Depth (m)

78

Mount Laurel Aquifer

113

OW-3

114 m

Groundwater Monitoring Data

68.4 68.0 67.6 67.2 66.8 68.8 68.4 68.0 500 67.6 480 460 440 420 0

Well OW-2

70.2 69.8 69.4 489.6 469.2 448.8 428.4 25

Well OW-3
pump on pump off

10 15 Elapsed Time (hours)

20

Pore Pressure (cm of water)

Well OW-1

Pore Pressure (kPa)

69.4 69.0 68.6

Reverse Water Level Fluctuations - Noordbergum Effect

aquifer

aquitard

pumped aquifer

Shallow aquifer exhibits increasing water level as a lower aquifer is pumped. First documented by A. Verruijt, Delft University, who termed it the Noordbergum Effect after a town in the Netherlands where it was observed.

Reverse Water Level Fluctuations - Rhade Effect

aquifer

aquitard

aquifer after pumping

Aquitard exhibits decreasing water level as a lower aquifer recovers from pumping. First documented by Langguth & Treskatis, who termed it the Rhade Effect after a town in Germany where it was first observed.

FLAC Axisymmetric Grid and Boundary Conditions

Modeling Assumptions
Soil strata are horizontal, isotropic, and saturated. Groundwater viscosity is constant and soil grains are incompressible. Soil porosity and hydraulic conductivity is constant. Hydraulic inefficiencies from well screen and sand pack are negligible. Impermeable, incompressible layer forms the base of the model. Effects of pumping on pore pressures and lateral strains are negligible at the models limits.

Governing Equation - Hydromechanical Formulation

In Biots Theory of 3-D consolidation, pore pressure (P) and volumetric soil strain ( kk ) are coupled or covariant. For a FLAC element:

P K w k 2 kk = P t t n w
where:

kk = + rr + zz
ii =
1 1 P ii kk + 2G 1 + 3K

1 P 1 2 P 2 P 2 P= + r + 2 2 r r r r z 2

i = , r, z

k is hydraulic conductivity, K is drained bulk modulus, is the drained Poisson Ratio, G is the shear modulus, Kw is the bulk modulus of water, n is porosity, w is the unit weight of water, kk is the mean normal stress.

Basic Modeling Strategy

Run sequence: 1) Run in hydraulic mode to establish pore pressure distribution. 2) Run in mechanical mode to develop body forces, then set displacements to zero. 3) Run in coupled, hydromechanical mode until volumetric strains < 10-7. 4) Apply well discharge. 5) Run with Fast-Flow scheme for aquifer and in standard mode for aquitards. 6) Work from most highly stressed to least stressed layer.

Inverse Modeling Strategy

Sensitivity Analysis: Examine effects of soil parameters on pore pressure development to narrow down the number of unknowns. Examine effects of modeling schemes (equilibration time, fast-flow, explicit/implicit, MC/elastic) on pore pressure development.

Model Calibration (trial and error assisted by contouring): Use literature values to bound the variables. Match modeled pore pressure histories to field data. Calibrate each layer then make global runs to adjust for interaction between layers.

Well OW-3 - Mount Laurel Aquifer Curve Matching of Pore Pressure Histories
51 50

Pore Pressure (m of H2O)

49

field data
48 47

sensitive to varying K and k

46 45 44 43 42 0 2 4 6 8 10 12 14 16 18 20 22 24 26

Elapsed Time (hours)

Well OW-2 Manasquan Aquitard Curve Matching of Pore Pressure Histories

7.15

7.10

Pore Pressure (m of H2O)

sensitive to varying K, G, and k

7.05

7.00

6.95

field data
6.90

6.85 0 2 4 6 8 10 12 14 16

Elapsed Time (hrs)

Well OW-1 Kirkwood Aquitard Curve Matching of Pore Pressure Histories

7.00

decreasing k*
6.98

Pore Pressure (m of H2O)

6.96

6.94

decreasing K* and
6.92

6.90

field data
6.88

6.86

* generally true but not always

6.84 0 2 4 6 8 10 12 14

Elapsed Time (hrs)

Manasquan Aquitard Manasquan Aquitard -2 P modeled - Pactual x10 (Pa) Contours of Pmodeled - Pactual x10-2 (Pa) Flow Time = 3000 secs Flow Time = 3000 secs
10 0 10 20 20 40 -10 1e+4 30 10

-10

20

Drained Bulk Bulk Modulus (MPa) Drained Modulus, K (MPa)

0 0 0 0 1e+3 0 0

1e+2 1e-8

1e-7

1e-6

Hydraulic Conductivity, k (cm/sec) (cm/sec) Hydraulic Conductivity

Manasquan Aquitard Manasquan Aquitard Zero Difference Pore Contours at Various Zero Difference PPressure Contours at Various Times Times
8e+3

7e+3
0

DrainedBulk Bulk Modulus, Modulus (MPa) Drained K (MPa)

6e+3
0 0

0 0

5e+3
0 0 0 0 0
0

4e+3

0 0 0 0

3e+3
0

0 0 0

2e+3 1e-8

1e-7

Hydraulic Conductivity, k(cm/sec) (cm/sec) Hydraulic Conductivity

Mount Laurel Aquifer Pore Pressure and Volumetric Strain Histories

Simplified Diffusion Equation for Pumping Test Analysis

Commonly, for pumping test analysis, the change in pore pressure is uncoupled from mechanical strain:
assume.

P kk

or in terms of strain P K k k

so that:

Kw k 1 P P 2 = P n w K t t
rather than Biots more correct formulation:

Kw k kk P 2 P = t t n w

Hydraulic Conductivity of Aquifer by Various Methods

Formation Mt. Laurel

Test Method Pump Test* Pump Test* Pump Test** Pump Test*

Solution Method Hantush & Jacob, 1955 (confined) Hantush, 1961 (semi-confined) Hantush & Jacob, 1955 (confined) FLAC 2D, Ver. 5.1 (fast-flow)

k (cm/sec) -3 2.1x 10 -3 2.3 x 10 -3 2.1 x 10 -3 5.0 x 10

* Kimball, 2006

** Sammon, 1993

Conclusion for aquifer: k (hydromechanical) 2 x k (uncoupled) The difference is due in large part to how volumetric soil strain is handled.

Empirical Determination of Aquifers Shear Modulus

Use Richarts (1977) empirical equation for the small strain shear modulus, Gmax, for clean, round-grained sands as:

( 2 . 17 e )2 kk Gmax = 700 1+ e 3
where:

0.5

e is void ratio, for e < 0.80 or n < 0.44, and for soil shear strains < 10-4
Get kk from the FLAC model and e from laboratory tests.

Aquifer Shear Modulus: Model and Empirical Equation

Formation Mt. Laurel

0.2000

Density n (kN/m3) 0.40 17.28

K (MPa) 335.2

G (FLAC) Gmax (MPa) (MPa) 251.4 207.0

Conclusion for aquifer: G (FLAC Model) is 21% of Gmax (Richart Equation). Very good agreement even though the aquifer response is not very sensitive to and G !

Manasquan Aquitard Pore Pressure and Volumetric Strain Histories

Shallow Kirkwood Aquitard Pore Pressure and Volumetric Strain Histories

Hydraulic Conductivity of Aquitards by Various Methods

k (cm/sec) -7 0.3-2.9 x 10 -7 0.6-3.6 x 10 -7 0.5-1.5 x 10 -9 5.7 x 10 -6 32.0 x 10 -6 7.5 x 10 -6 6.7 x 10

Formation Manasquan

Kirkwood

Test Method Lab Test* ** Lab Test * Lab Test * Pump Test* Lab Test** Lab Test * Pump Test*

Solution Method Falling Head Tests (Darcy) Flex Wall Perm. Tests (Darcy) Consolidation Tests (Terzaghi) FLAC 2D, Ver. 5.1 (double-precision) Constant Head Test (Darcy) Flex Wall Perm. Tests (Darcy) FLAC 2D, Ver. 5.1 (double-precision)

* Kimball, 2006, 2008

** Metcalf & Eddy, 1993

Conclusion for aquitards: k (hydromechanical) < k (Darcy and Terzaghi) by 2 orders of magnitude in the Manasquan Aquitard, but 1 order or less in the Kirkwood Aquitard.

Denison Sampler for Aquitard (Stiff Silt & Clay)

Denison Sample in 0.6 m Long Plastic Tube

Resonant Column Testing - Testing Chamber

Resonant Column Testing Electromagnetic Drive Causes Torsion

Resonant Column Testing Results - Manasquan Aquitard

160

Gmax
140

Shear Modulus, G (MPa)

120

143.6 kPa
100

Gmax
80 60

95.8 kPa

Gmax
40 20 0 1.E-06 1.E-05 1.E-04 1.E-03

47.9 kPa

Shear Strain

Aquitards Shear Moduli: Model and Resonant Column Tests

Aquitard Manasquan Kirkwood

Density K G (FLAC) n (kN/m3) (MPa) (MPa) 0.4945 0.55 11.62 4788.0 52.7 0.4800 0.57 16.34 1197.0 48.4

Gmax * (MPa) 28.3 and 400.0 74.9

* Resonant Column Testing by URS, 2008

Conclusion: G (FLAC Model) is between an order of magnitude and 35% of Gmax (Resonant Column).

Conclusions
IN-SITU AQUITARD PROPERTIES: Modeling of reverse water level fluctuations allows estimation of aquitard properties from a pumping test. TIME CONSUMING AND DIFFICULT: More than 200 runs, each taking more than 8 hours (need faster processors and software). STRAINS: Fully-coupled modeling is more important as soil modulus and permeability decrease. FLAC is able to account for both solid-fluid stresses and strains. APPROXIMATION: Order of magnitude precision is considered possible without perfectly matching the field data. AUTOMATION: Use of calibration codes, such as UCODE, in a FISH subroutine may be practical for aquifers but very difficult for aquitards due to the ill-poised, non-linear response. FLAC TRICK (fully-coupled modeling): Use of FLACs implicit scheme, with a time step as per the FLAC Manual, was up to 4X faster than the explicit scheme without much loss of accuracy.

Nandri Spasibo

Xie xie Efcharisto

Toda Merci

Baniha M goi

Grazie Dziekuje Jag tackar Arigato Gracias Thank You

Hvala vam Acknowledgements:

Danke schn

9 Mr. Juan Salguero, L. Robert Kimball & Associates, designed and conducted the pumping test. 9 Dr. Herb Wang, University of Wisconsin-Madison, provided an independent interpretation of the data. 9 Mr. Greg Thomas, URS, conducted resonant column testing. 9 Dr. Christine Detournay, Itasca Consulting Group, gave helpful guidance on modeling and some words of encouragement. 9 The Symposiums peer review committee made valuable comments.