stochastic inverse modeling of groundwater flow and...

Oct. 2004

Project Overview

Giorgio Amsicora Onnis, Harrie-Jan Hendricks Franssen, Fritz Stauffer and Wolfgang Kinzelbach

Flow and Transport Modeling Results for the Baltenswil Case Study

Due to the large number of unknown parameters and their strong spatial variability, traditional groundwater model calibration is likely to yield non-unique solutions to the inverse problem, and thus unreliable results.We cope with the ill-posedness of the inverse problem through an inverse stochastic approach, considering probability densities of parameters and analyzing statistics of the model results. This allows quantification of uncertainty in the model predictions.

The release of environmental tracers into the atmosphere by nuclear power plants, bomb testing, and industrial processes dates back to the ’50s (Fig.1).

Deterministic Transport ModelingFor transport simulation, one transmissivity field realization was chosen from the ensemble. To account for the lateral inflow to the aquifer, the model domain was expanded (red dashed line in Fig.4). Simulations were performed for 3H, 3He, 85Kr and SF6 (Fig. 5).

Department of Civil, Environmental and Geomatic Engineering

Stochastic Inverse Modeling of Groundwater Flow and EnvironmentalTracer Transport: Baltenswil Case Study (Switzerland)

The number of acceptable solutions – e.g. the set of input parameters capable to reproduce the measured data - can be reduced by conditioning the model parameters. Accounting for prior knowledge, e.g., by introducing environmental tracer data or geophysical data is a possible way to minimize uncertainty in the model predictions. In the present study we are focusing on the implementation of environmental tracer data such as 3H, 3He, 85Kr and SF6 in the flow and transport model. This study is part of a joint project together with EAWAG, Dübendorf and IPB (University of Bern).

Tracer dynamics in the unsaturated zone show different transport timescales for different tracers, resulting in a time lag or delay at the groundwater table . This delay ∆t must be accounted for. It becomes more important with increasing thickness of the vadose zone (Fig.2).

Figure 2: Time delay as a function of vadose zone thickness. Water-bound tracers – like 3H – move by advection following seepage water, gas tracers transport –like 85Kr and SF6 – is diffusion-dominated

The Baltenswil site is situated in the Aathalaquifer, a sandy gravel formation close to Zürich. The modeled area has extensions of 3x3 km2 (Fig.3).

Stochastic Inverse Flow Modeling

Conditioning of the transmissivity field was performed by incorporating available head and transmissivity data using the Sequential Self-Calibrated Method (Gómez-Hernández et al., 1997), implemented in the code INVERTO (Hendricks-Franssen, 2003) (Fig. 4).

Figure 4: Top: Heads and Log10-transmissivity ensemble average. Bottom: Heads and Log10-transmissivity ensemble variance (100 realizations)

Tritium PW Baltenswil - Konstanz_Guttannen input -

05

101520253035

0 5 10 15

Years [1991-2004]

3H [T

U]

calculatedobserved

Figure 3: Baltenswil site

3He PW Baltenswil - Konstanz_Guttannen input -

0

0.20.4

0.6

0.8

11.2

1.4

0 5 10 15

Years [1991-2004]

3He

[TU

]

calculatedobserved

Figure 1: Atmospheric Input Function for 3H, 3He, 85Kr and SF6

3H Input Function - Konstanz [D]

0

20

40

60

80

100

120

140

160

180

Jan 8

0

Jan 8

1

Jan 8

2

Jan 8

3

Jan 8

4

Jan 8

5

Jan 8

6

Jan 8

7

Jan 8

8

Jan 8

9

Jan 9

0

Jan 9

1

Jan 9

2

Jan 9

3

Jan 9

4

Jan 9

5

Jan 9

6

Jan 9

7

Jan 9

8

Jan 9

9

Jan 0

0

Jan 0

1

Years

3H [T

U]

85Kr and SF6 Input Function

0102030405060708090

100

1950 1970 1990 2010

Years

85K

r [dp

m/c

cKr]

0

1

2

3

4

5

6

7

SF6

[ppt

V]

85 Kr - Freiburg [D]SF6 - global

500 1000 1500 2000 2500

500

1000

1500

2000

2500

500 1000 1500 2000 2500

500

1000

1500

2000

2500

500 1000 1500 2000 2500

500

1000

1500

2000

2500

500 1000 1500 2000 2500

500

1000

1500

2000

2500

This work is supported by the Swiss National Science Foundation (SNF), Project no. 2000-067933.02/1

85Kr PW Bruttisellen - Freiburg input -

01020304050607080

0 5 10 15

Years [1991-2004]

85K

r [dp

m/c

cKr]

calculatedobserved

SF6 PW Baltenswil- Global input -

0

1

2

34

5

6

7

0 5 10 15

Years [1991-2004]

SF6

[ppt

V]

calculatedobserved

Environmental Tracers and the Role of the Unsaturated Zone

Conclusions

• The importance of the vadose zone in tracer transport is confirmed. The time delay is a key issue for a correct interpretation of concentration measurements in aquifers

• The saturated zone seems to play a minor role. Dued to the short residence time, no sensitivity to the transmissivity field, to preferential flow paths and to porosity was found

• The choice of the proper local tracer input function is essential for tracer transport simulations.

Figure 5: Modeling results vs. measured data

Transport Timescales in the Vadose Zone

0

10

20

30

40

50

0 2 4 6 8 10 12 14 16 18 20[Years]

Thic

knes

s [m

.]

Tritium85 KrSF6