Fyke and Lipscomb #1

Report
Greenland Ice Sheet model
simulations and validation
Jeremy Fyke, Bill Lipscomb
Los Alamos National Laboratory
Outline
• Simulated Greenland surface mass balance in CESM
• Greenland Ice Sheet model optimization within CESM
framework
• Ongoing development
Background
• The Glimmer Community Ice Sheet Model (GlimmerCISM) has been coupled to version 1.0 of the Community
Earth System Model (CESM 1.0).
– Shallow-ice approximation; Greenland only
– Higher-order ice sheet model (CISM 2.0) to be included in
CESM 1.1 (aiming for Nov. 2012 release)
• The surface mass balance (SMB) of ice sheets is
computed in the Community Land Model (CLM) and
passed to Glimmer-CISM.
– Multiple (~10) glacier elevation classes on CLM’s coarse
grid
– Downscaled and interpolated in z to CISM’s fine grid
Model details
• Fully coupled CESM 1.0 with 0.9 ox 1.25o FV atm/land, 1o ocean
• Focusing on the surface mass balance (accumulation minus
ablation) of the Greenland ice sheet
– SMB(ice+snow) = incoming snow + incoming rain – runoff – sublimation
– Positive ice SMB when snow exceeds max depth (1 m water equivalent)
and turns to ice
– Negative ice SMB when snow depth is zero and bare ice melts
– The SMB of ice (not snow) is passed to the ice sheet model
• Snow and ice physics:
– Liquid water can percolate and refreeze in the snow, but not on bare ice
– Snow albedo follows SNICAR model (depends on snow grain size, solar
angle, etc.)
– Bare ice albedo is prescribed (0.60 visible, 0.40 near IR)
CMIP5 simulations with glacier elevation classes, SMB evolution
Name
Length
Initialization
Pre-industrial
Years 1-100
100-yr IG run (snowpack) + BG1850CN
20th century
1850-2005
from year 100 of Pre-industrial
21st century (RCP8.5) 2005-2100
from year 2005 of 20th century
Pre-industrial
1850
1940
2000
2100
400 Gt/yr
400 Gt/yr
SMB = 0
SMB = 0
• Lower SMB in the 1940s than in the 1990s and 2000s
• Negative SMB in several years after 2060
Greenland SMB, downscaled to 5 km
Pre-industrial (80-99)
20th-century (1980-1999)
RCP8.5 (2080-2099)
452 ± 91
421 ± 107
61 ± 142
SMB (Gt/yr)
Red = net
accumulation
Blue = net
melting
kg m-2 yr-1
• 1980-99 ablation rates are higher than pre-industrial in N & NE
• The equilibrium line rises by ~500 m by end of 21st century
• It reaches almost 2000 m in the NE and southern half of E margin
• High snowfall rates help to keep equilibrium line low in NW and mid-W margins
SMB, comparison with RACMO (at 5 km res)
SMB (Gt/yr)
1958-2007 (plot 1958-2005)
RACMO
409±106
469±41
• Good match in ablation zones
• Accumulation rates are overestimated in the interior and underestimated in
the SE (smoother orography in CESM)
• Snowfall local maxima along W coast and impact on melt (via albedo) are
well captured
Temperature and SMB: 1850-2005
JJA mean temperature over ice sheet
Precip
-5o
Melt
Pi
-10o
1850
Runoff SMB
2005
1850
• Warm period during 1930s and 1940s, with high melt
• Precipitation rates are higher in the 1990s
• High SMB following Pinatubo (Pi) eruption in 1991
2005
Temperature anomalies: 2080-99 minus 1980-99
annual
JJA
• MOC reduction reduces warming SE of Greenland
• JJA increase is highest
• In ice-free regions to N & E, in part due to stronger sea ice losses
(>40%) along the coast
• In the interior of the ice sheet, which remains below melting point
SMB (Gt/yr): 1980-2100
Blue = Precip
Red = Melting
Green = Runoff
Black = net SMB
SMB = 0
1980
• Precipitation increases with time
• Melt and runoff increase by a larger amount
• SMB is negative for the first time around 2030
2100
Summary: Greenland SMB
• The SMB scheme works well. Greenland’s simulated 20th
century surface mass balance and trends are in good
agreement with RACMO, a state-of-the-art regional model
(with differences due to smoother CESM topography).
• During the 21st century simulation, the SMB decreases from
~400 Gt/yr to near zero.
• Greenland average warming in the 21st century is roughly
equal to global average warming. There is more warming in
the North and East (less summer sea ice) than in the
Southeast (reduced MOC).
Ice sheets in RASM
• Coupling to CISM is included in the current version of the
CESM coupler; should not be hard to include in RASM.
• The coupler requires the ice-sheet surface mass balance
in multiple elevation classes from the land model. Next
step is to implement a similar scheme in VIC.
• How much code can be reused from CLM?
Greenland Ice Sheet (GIS) optimization
• Will be necessary for GIS in RASM
• Carried out in support of SeaRise: model
intercomparison project to assess range of
modelled ice sheet responses to idealized climate
perturbations (Δclimate, Δdynamics)
• Initial state of ice sheet should reflect observed
ice sheet: exercise in rapid (1 month turnaround)
model optimization
• Tool: Latin Hypercube Sampling of uncertain
parameter space
Optimization approach
• Generate 100 GIS realizations with LHS-determined random
combinations of:
– Ice sheet enhancement factors
– Basal sliding coefficients
– Geothermal heat fluxes
• Compare equilibrium state (after 9 kyr simulation) to
observed GIS state for:
–
–
–
–
–
Ice volume error
Ice area error
RMSE of ice surface elevation
Maximum ice elevation error
Summit horizontal offset error
• Rank models by ‘worst diagnostic ranking’ to get best allaround GIS realization
Optimization approach
9000 years
today
future
Optimization results: volume evolution
Optimization results: example GIS
model-observed elevation differences
Optimization results: rankings for all
diagnostics
Optimization results: dependence of
diagnostics on LHS parameters
Optimization results: top-performing
ice sheet model realizations
Ice sheet spinup issues
• Spinup/optimization issues to work on:
– Thermal timescale of ice sheet (thus, ice viscosity)
is 105 years – analogous to spinning up the deep
ocean (but worse!)
– How to spin up a GIS model, using forcing that is
continuous between past and future, that
captures transient thermal and geometric state of
ice sheet?
– LHS ensemble limited to sampling internal ice
sheet parameters
Conclusions
• LHS sampling provides a fast way to determine
optimal initial state for GIS models within a
climate model framework
• Flow factor exerts major control on ice sheet
optimization in CISM
• Similar optimization technique will be necessary
to optimize the GIS under RASM forcing
• RASM surface mass balance field (reflected in
long-term GIS spinup geometry) will be sensitive
indicator of regional atmospheric model biases
Ongoing development
• New ice-sheet dynamical cores
1. Payne-Price: 3D higher-order, finite difference, structured
grid, Trilinos solvers
2. BISICLES: Vertically integrated higher-order, finite volume,
Chombo adaptive mesh refinement software
3. FELIX: Full-Stokes/higher-order, finite element,
unstructured variable-resolution mesh (MPAS framework),
Trilinos solvers
• BISICLES and FELIX will be further developed under a
new 5-year DOE SciDAC project, Predicting Ice Sheet
and Climate Evolution at Extreme Scales (PISCEES).
Ongoing development
• Improved physics parameterizations
– Subglacial hydrology and basal sliding (S. Price, M. Hoffman)
– Calving (based on Potsdam-PIK)
• Two-way coupling with land model
– Requires dynamic landunits (glaciers  vegetation)
– May not be important on decadal time scales
• Coupling with ocean model
– POP2X simulates ocean circulation beneath ice shelves (X.
Asay-Davis); will be applied to Antarctica
– May not be practical for RASM in near term; Greenland fjords
require very high resolution (~1 km)
Extra slides
SMB trend 1958-2005 (kg m-2 yr-2)
• Negative trend in ablation zones
• Positive trend in the Southeast,
due to increasing precipitation
• Consistent with RACMO results
and altimetry measurements
Terms
of
SMB
SMB(ice) + MB(snow) = SNOW + RAIN - RU - SUB
Units: Gt per year
RACMO 1958-2007
CLM 1980-1999
Diff CLM-RACMO
SMB (net)
469
403 ± 106
-66
MB (snow)
-5
SNOW
697
742 ± 82
+45
RAIN
46
135 ± 23
+89
PRECIP
743
877 ± 98
+134
RUNOFF
248
425
+177
SUBLIMATION
26
54 ± 3
+28
RU = ME + RAIN - RF = AVLIQWATER - RF
Units: Gt per year
RACMO 1958-2007
MELT (only snow)
1980-1999
430 ± 67
MELT (snow + ice)
404
530 ± 109
+126
MELT+RAIN
450
665 ± 117
+215
REFREEZING
202 (45% of
ME+RAIN)
240 ± 27
(36% of ME+RAIN)
+38
Terms of SMB: 1980-1999
SMB
Melt
Runoff
Rain
• Runoff = Melt + Rain - Refreezing > 0 in the interior of the ice sheet, where
all available liquid water should refreeze
• In CLM, rain is overestimated in ice sheet interior (and rain cannot
refreeze if snow thickness = 1 m w.e.)
21st century temperature increase (ref: 1980-1990)
global
Greenland + ocean
Greenland ice sheet
JJA
annual
Temperature anomalies for 2080-2099
region
Annual (st. dev.)
Global
3.6 (0.3)
Greenland ice sheet
3.8 (0.6)
Greenland region
3.5 (0.5)
Summer (st. dev.)
3.5 (0.8)
Terms of SMB: RCP8.5
SMB(ice) + MB(snow) = SNOW + RAIN - RU - SUB
Units: Gt per year
1980-1999
2080-2099
SMB-net
403 ± 106
12 ± 148
MB (snow)
-5
-1
SNOW
742 ± 82
807 ± 74
RAIN
135 ± 23
279 ± 45
PRECIP
877 ± 98
1086 ± 105
RUNOFF
425
1018 ± 167
SUBLIMATION
54 ± 3
57 ± 5
RU = ME + RAIN - RF = AVLIQWATER - RF
Units: Gt per year
1980-1999
2080-2099
MELT (only snow)
430 ± 67
624 ± 65
MELT (snow + ice)
530 ± 109
1040 ± 160
MELT+RAIN
665 ± 118
1320 ± 187
REFREEZING
240 ± 27
(36% of ME+RAIN)
301 ± 27
(23% of ME+RAIN)
Seasonal cycle of melt
Solid black line =
Ice melt, 1980-1999
Solid red line =
Ice melt, 2080-2099
Dotted black line =
Snow melt, 1980-1999
Dotted red line =
Snow melt, 2080-2099
J F M A M J
J A S O N D
• Length of snow melt season does not change (melt season begins in April)
• Ice begins to melt ~15 days earlier and melts for ~15 days more in late
September

similar documents