Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Parameter estimation for grounding-line transition
Gunter Leguy, Xylar Asay-Davis, William Lipscomb
Los Alamos National Laboratory
February 15th, 2013
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 1 / 20
Introduction
We are exploring how basal physics influences resolution in a 1-Dvertically integrated flowline model.
Several models of the form τb = β2u have been investigated:I Schoof (2007): β2 = - C|u| 1n−1I Pattyn (2006): β2 = - exp[β0(xg − x)]
Goals:
Present a new parametrization of the basal shear stress (in the caseof a Marine ice sheet):
I Physically motivated (ocean connection)I Transitions smoothly between finite basal friction in the ice sheet
and zero basal friction in the ice shelf
Hope for ≈ 1km resolution near the grounding line.
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 2 / 20
Model equations
Valid in both, the ice sheet and the ice shelf:
conservation of mass : Ht + (uH)x = a,
stress− balance equation : τl + τb + τl = 0.
τl =[2A− 1
nH|ux |1n−1|ux
]x,
τd =− ρigHsx ,τb = basal shear stress.
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 3 / 20
Assumption and Boundary conditions
Symmetric ice sheet at the ice divide
u = 0sx = (H − b)x = 0
}at x = 0,
At the grounding line, we have the flotation condition and the balancebetween τl and τd .
H = ρwρib,
2A− 1n |ux |
1n−1ux = 1
2ρig(
1− ρiρw
)H
}at x = xg .
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 4 / 20
Basal stress
We adopt the formulation from Schoof (2005):
τb =− γN
(u
u + λmaxmmax
ANn
) 1n
.
λmax = wavelength of bedrock bumps
mmax = maximum bed obstacle slope
Effective pressure: N ≡ pi − pw
I pi ≡ ρigHI pw?
0 0.5 1 1.5 2−1
0
1
2
3
4
5
6
IceSheetdomain(103km)
Surfaceelevation(km)
0 0.5 1 1.5 2−1
0
1
2
3
4
5
6
IceSheetdomain(103km)Surfaceelevation(km)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 5 / 20
Basal stress (continued)
N(p) = η(p;H, b)ρi g H, η non-dimensional function
η ∈ [0, 1]
p ∈ [0, 1]
We want the following limit for η:
When p = 0, η = 1: no water-pressure support.
When p = 1, η = 0: subglacial water has a free connection to the ocean,Paterson(2010)
At the grounding line, η = 0 (effective pressure goes to zero to allowcontinuity).
One expression with the right limit: N(p) = ρigH(
1− HfH
)p, where
Hf = max(0, ρwρi b)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 6 / 20
Basal stress (continued)Right limits:
I if u � κN(p)n and τb ≈ −γ
κ1/n|u| 1n−1u (frozen ice at the bed)
I if u � κN(p)n and τb ≈ −CN(p) u|u| . (hydrological connection)
Physically motivatedTransitions smoothly from non-zero basal stress in the ice sheet tozero basal sliding in the ice shelf
−60 −40 −20 0
−200
−150
−100
−50
0
−60 −40 −20 0
−200
−150
−100
−50
0
Basal stress term (kPa)
Distance to grounding line (km)
a) p = 0 b) p=1
Figure:LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 7 / 20
Numerics
Moving-grid:
Numerically more accurate
GL must lie on a grid point
Could be used as a benchmark solution for fixed-grid model
Harder to implement in 3-D models
Fixed-grid
Suitable for 3-D model
Constant resolution
Numerically less accurate
GL usually falls between two grid points leading to interpolation error
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 8 / 20
Numerics (continued)
How to compare our results for all p-values?
QualitativeI Schoof (2007) boundary layer solution (in the case of frozen bed
type basal sliding)
QuantitativeI Schoof (2007) boundary layer solution (good approximation for
p=0)I Chebyshev polynomial numerical schemeI Moving grid with high enough resolution used as a benchmark for
fixed-grid model
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 9 / 20
Numerics (end)Experimental set-up:
Moving grid resolution benchmark: 156m ≤ Res ≤ 316 m.
Fixed-grid resolution comparison: 1.6 km and 0.8 km.
We will show results for 3 values of p.
Constant accumulation rate: a = 0.3 m/yr.
MISMIP experiment: neutral equilibrium experiments.
Two different bed topographies (as in Schoof (2007a)).
0 0.5 1 1.5 2−1
0
1
2
3
4
5
6
IceSheetdomain(103km)
Surfaceelevation(km)
0 0.5 1 1.5 2−1
0
1
2
3
4
5
6
IceSheetdomain(103km)
Surfaceelevation(km)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 10 / 20
Results linear bedThe fixed-grid solution is inaccurate in capturing the retreat.
1024
1025
1026
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
analyticnumeric (fix)numeric (moving)
1024
1025
1026
analyticnumeric (fix)numeric (moving)
1/A
Advance Retreat
Grounding lineposition (km)p = 0, Res = 0.8 km
Figure:LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 11 / 20
Results linear bed (continued)MISMIP-type experiments 1 & 2
Ice sheet domain (103 km) 1/A
Gro
un
din
g lin
e p
osi
tion
(1
03
km
)p=0
p=0.5
p=1
Ice s
heet
surf
ace
ele
vati
on
(km
)
Res = 0.8 km
0
2
4
6
1
1.5
2
0
2
4
6
0.5
1
1.5
2
0 0.5 1 1.5 2
0
2
4
6
1023
1024
1025
1026
0.5
1
1.5
fix
moving
moving(advance)
moving(retreat)
fix (advance)
fix (retreat)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 12 / 20
Results linear bed (continued)MISMIP-type experiments 1 & 2 (note the different scales for thedifferent p-values)
0
200
400
0
200
400
0
20
40
60
80
0
20
40
60
80
1023
1024
1025
1026
0
5
10
15
1023
1024
1025
1026
0
5
10
15
advance
Retreat
1/A 1/A
p=0
p=0.5
p=1
Res = 1.6 km Res = 0.8 km
Grounding line position error estimateErr
or
est
imate
(km
)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 13 / 20
Results linear bed (end)
MISMIP-type experiments 1 & 2 (note the different scales for thedifferent p-values)
1/A 1/A
p=0
p=0.5
p=1
Res = 1.6 km Res = 0.8 km
SLR volume error estimate
Err
or
est
imate
(km
2)
1023
1024
1025
1026
0
1000
2000
1023
1024
1025
1026
0
1000
2000
1023
1024
1025
1026
0
100
200
300
1023
1024
1025
1026
0
100
200
300
1023
1024
1025
1026
0
20
40
60
1023
1024
1025
1026
0
20
40
60
advance
retreat
Element ofcomparison: 2850km3 ≈ 7.2 mm.
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 14 / 20
Results poly bedMISMIP-type experiment 3
Ice sheet domain (103 km) 1/A
Gro
un
din
g lin
e p
osi
tion
(1
03
km
)p=0
p=0.5
p=1
Ice s
heet
surf
ace
ele
vati
on
(km
)Res = 0.8 km
0 0.5 1 1.5 2
0
2
4
6
1024
1025
1026
0.5
1
1.5
0 0.5 1 1.5 2
0
2
4
6
1024
1025
1026
0.5
1
1.5
0 0.5 1 1.5 2
0
2
4
6
1024
1025
1026
1027
0.5
1
1.5
fixed
moving
movingadvance
movingretreat
fixedadvance
fixedretreat
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 15 / 20
Results poly bed (continued)MISMIP-type experiment 3 (note the different scales for the differentp-values)
1/A 1/A
p=0
p=0.5
p=1
Res = 1.6 km Res = 0.8 km
Grounding line position error estimateErr
or
est
imate
(km
)
1024
1025
1026
0
50
100
150
1024
1025
1026
0
50
100
150
1024
1025
1026
0
20
40
60
1024
1025
1026
1027
0
5
10
15
20
1024
1025
1026
0
20
40
60
Advance
Retreat
1024
1025
1026
1027
0
5
10
15
20
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 16 / 20
Results poly bed (continued)MISMIP-type experiment 3 (note the different scales for the differentp-values)
1/A 1/A
p=0
p=0.5
p=1
Res = 1.6 km Res = 0.8 km
SLR volume error estimateErr
or
est
imate
(km
2)
1024
1025
1026
0
200
400
600
800
1024
1025
1026
0
200
400
600
800
1024
1025
1026
50
100
150
200
250
1024
1025
1026
50
100
150
200
250
1024
1025
1026
1027
20
40
60
80
100
1024
1025
1026
1027
20
40
60
80
100
Advance
Retreat
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 17 / 20
conclusion
For low p-values the need of high resolution in the vicinity of theGL remains.
Error estimate decreases substantially as the value of p increasesand therefore the required model resolution decreases as pincreases.
p does not play any role in the bulk of the ice sheet. It impacts arelatively small distance (no more than 25 km from the groundingline in our experiments) which is enough to impact the solution.
With p ≈ or = 1, a resolution of ≈ 1km is sufficient to capture thesolution.
Paper in preparation for the cryosphere
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 18 / 20
Future work
3-D implementation
Add the missing physics (lateral drag, buttressing)
Data comparison for admissible values of p (any value besides 0and 1 valid?)
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 19 / 20
LANL (Los Alamos National Laboratory) Parameter estimation for grounding-line transition February 15th, 2013 20 / 20