Modelling Subsurface Flow −

advertisement

Modelling Subsurface Flow

not what you see , but what you imagine

Nils Otto Kitterød

FEMLAB conference, Oslo October 13. 2005

Outline:

• teaching

• how to make COMSOL MULTIPHYSICS behave according to my concepts

• examples

Why focus on groundwater?

Important for society

Groundwater resources of the world

Major part of the world population are depending on groundwater resources

Important for nature unsaturated zone saturated zone

The global water balance

~ 30 times more groundwater than water in the lakes

Groundwater flow is a boundary value problem => measurements

Why focus on groundwater?

Groundwater: the liquid of life!

interface to human society and science groundwater flow concept of geology numerical simulation contaminant transport

Concept of flow: Darcy's experiments (1856):

∆ x

∆ h

Henri-Philibert-Gaspard Darcy

Darcy’s law: Q

A

∆ h

∆ x

Conservation of momentum

A governs the laminar (nonturbulent) flow of fluids in homogeneous, porous media

1. Conservation of momentum: q

= −

K

∇ h,

2. Conservation of mass: where h = p/(

ρ g) + z

∂ (

S

0 h )

∂ t

= − ∇⋅ q

+

N, where S

0 is specific storage [1/L], and N is a sink/source term

3. Constitutive relations: e.g. if S

0 is a function of h

Richards’ equation (not today!) or if

ρ is a function of concentration of solute (later)

Here: Dupuit-Forchheimer assumption:

∂ h

∂ z

= 0, but q z

0 which means 3D

2D

4. Boundary conditions and (if transient) initial conditions

Cross Section of the Gardermoen Delta

m a.s.l.

200

180

160

Delta Foresets

Delta Bottomsets

140

120

Bedrock

3000

West

200

100

West

3000

2000 meters

2000 meters

Delta Topsets

Kettlehole

Ice-contact slope

1:~15

?

?

?

1000

1000 500

East

0

East

0 m a.s.l.

1:1

After K. Tuttle (1997)

From 1,2 we have:

From 3, life gets easier:

For a confined aquifer: m = top – bottom of aquifer

For an unconfined aquifer: m = h – bottom of aquifer

∂ ( S

0 h )

∂ t

= ∇⋅

K

∇ h

+

N.

∂ ( m S

0 h )

∂ t

= ∇⋅ m K

∇ h

+

N, where m is thickness of water saturated zone mS

0

= S [-] is called storage coefficient m K = T [L 2 /T] is called transmissivity

H

1

H

2 top bottom x m = const.

length of aquifer

H

1

H

2 m = h(x,t) bottom x length of aquifer

Implement the Dupuit-Forchheimer assumption:

Remember (1): h(t

0

)

0 if bottom = 0

Transmissivity

Remember (2): Q = Aq (L 3 /T)

By using boundary integration in postprocessing, you get the water flow directly!

Challenge: estimate effective parameters

0 m

T

=

T(z) dz in a way that the response h is reproduced according to the observations

( S h )

t

= ∇ ⋅ (

T

h

) −

N(t)

The Boussinesq equation where: t

T

S h effective transmissivity (L 2 /T) storage coefficient(-) hydraulic head (L) time (T)

N(t) source/sink (L/T), infiltration or pumping

Example: The Gardermoen Delta

15 km

15 km

The inland glacier

~10500 B.P.

(Andersen, B.G.,2000)

Sea level about 200 m higher than present sea level.

Average paleodischarge: ~3000m 2 /s

(Tuttle, K., 1998)

The Gardermoen paleo-delta

The Gardermoen Delta Today:

Trandum Delta

Paleo-portal

Helgebostad Delta

Paleo-portal

Paleo-distribution Channels Show Radial Flow:

11000

10000

9000

8000

7000

6000

5000

4000 x-profile Trandum Delta

Paleo-portal

Helgebostad Delta

Paleo-portal

3000

2000

1000

4000 5000 6000 7000 8000 9000 10000 11000 12000

UTM-East Gardermoen, grw. obs, eu89

” The Gardermoen Doughnut”

Steady state model for R

1

r

l : i) Darcy’s law: a)

Q m r = q = – k b) Q r

= – d( 1 /

2 kh 2 ) dr ii) Balance of mass: dh dr c) Q

Α

= N

π l 2 – N

π r 2 d) Q r

= –

Q

Α

2

π r if m=h, then unconfined aquifer i and ii): d( 1 /

2 kh 2 ) =

N

2 l 2 r dr – r dr h

R

2

2 l h

1

R

1

Repeat for l ≤ r ≤ R

2 and eliminate l to get one closed form expression for h(r)

Plot observations of groundwater head as a function of radial distance from center: h

2

→ r

R

2 l l

N

R

1 h

1 let k = k(r) or H = H(r)

dh 2

=

N l 2 rk dr

– k r dr where k = k

1

– a(r-R

1

)

Two (simple) integrals:

∫ 1 r(k

0

-ar) dr

∫ r

(k

0

-ar) dr

(1)

(2) h

1

R

1 k

1

N

→ l l R

2 k

2 h

2 r

WRR(40) 2004 http://folk.uio.no/nilsotto/PUBL/analytical_gilbert.pdf

Implement the solution as a function in MATLAB: function[h] = funk_h_lin(radius, R1, R2, h1, h2, N, k_1, k_2)

function[h] = funk_h_lin(radius, R1, R2, h1, h2, N, k_1, k_2) minimum(h obs

– h calc

) gives k

1 at R

1 and k

2 at R

2

R

2 h

2 l h

1

R

1

Radial Duit-Forchheimer Solution where k=k(r)

Interpolation Functions:

R

1

R

2 k

1 k

2

Steady State Numerical Solution and

Analytical Solution:

Transient Simulation

Precipitation event

18.00

16.00

14.00

12.00

10.00

8.00

6.00

4.00

2.00

0.00

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 days from 27.09.2000

S=0.01

Precipitation event

18.00

16.00

14.00

12.00

10.00

8.00

6.00

4.00

2.00

0.00

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 days from 27.09.2000

Transient boundary condition:

H

1

(t) = H

1 sin(t/day)

Conclusion:

1) minor changes in h with N(t)

2) not very sensitive to H

1

OK !

But, what about the real aquifer?

190

190

Trandum Delta

Paleo-portal

Helgebostad Delta

Paleo-portal

180

188

188

188

Digitize boundary by using: GINPUT Graphical input from mouse.

[X,Y] = GINPUT(N) gets N points from the current axes and returns the X- and Y-coordinates in length N vectors X and Y.

Spline interpolation

h

2

N h

1 noflux boun dary

Make solid object by: s1 = geomcoerce('solid',{c}); geomplot(s1); and import geometry in Comsol Multiphysics

What about hydraulic conductivity?

Two radial structures with linearly decreasing k superimposed on each other.

What is best average?

Trandum Delta

Paleo-portal

Helgebostad Delta

Paleo-portal

arithmetic mean ?

1/T

1

1/T

2

Glacial portals

geometric mean ?

Glacial portals

harmonic mean ?

1/T

1

1/T

2

Glacial portals

Difference in hydraulic conductivity between harmonic and geometric mean

Hydraulic head

Input: hydraulic head as arithmetic mean between the Trandum and Li deltas arithmetic

Hydraulic head

Input: hydraulic head as geometric mean between the Trandum and Li deltas geometric arithmetic

Hydraulic head

Input: hydraulic head as harmonic mean between the Trandum and Li deltas harmonic geometric arithmetic

Hydraulic head gradient, minimum gradient at ground water divide max. gradient in ravine areas (landslide) and glacial boundary (kettle lake area)

No terminal and railway culvert (drawdown)

Hydraulic head gradient

With terminal and railway culvert (drawdown)

To protect the preservation area, the water balance of the area has to be maintained because the hydraulic gradient is the driving force of the gully processes.

Hydraulic head gradient

With terminal and railway culvert (drawdown) and two injection points of water from the culvert

The C omso l mes hing r outin es m akes l ife m ore p ractic al!

Plot cross-section before and after Airport construction

Note change in head gradient towards ravine area

Important for society Groundwater resources of the world

The majority of the world population depends on groundwater resources

Coastal aquifers : possible contamination of groundwater by seawater intrusion

Density driven flow:

2-way coupling between flow & transport

• Density dependent fluid flow - Darcy’s Law expresses in terms of pressure p :

ρ

[

ξ

( 1

− θ

)

+ ζθ

]

∂ p

∂ t

+ θ

∂ ρ

∂ c compressibility = 0

∂ c

∂ t

+ ∇ ⋅

[

− ρ κ

/

η ∇

( p

+ ρ gD )

]

=

0

• Salt concentration – Saturated solute transport

θ

∂ c

∂ t

+ ∇ ⋅

[

θ

D

∇ c

+ u C

]

=

0

ρ varies with c

ρ = ρ + γ

( c

c

0

)

γ

where:

=

ρ

c s f

ρ

c

0 o from the presentation by

Leigh Soutter

The Henry Problem

Dm = 1.6295 m2/d

Dm = 0.57024 m2/d

Decrease influx of fresh water from 5.702 m 3 /s to 0.05702 m 3 /s

What is the effect of tidal changes of 0.1 times thickness of aquifer?

Time variable boundary conditions

Conclusion

COMSOL MULTIPHYSICS is useful for teaching and business and it’s fun!

Thank you!

Download