1. Interannual and decadal variations of ice shelves
using multi-mission satellite radar altimetry,
and links with oceanic and atmospheric forcings
Fernando S. Paolo
Scripps Oceanography
Committee
PhD defense Sep 2, 2015
University of California, San Diego
Helen A. Fricker, Chair
Sarah Gille
Falko Kuester
Jean-Bernard Minster
Laurie Padman
David T. Sandwell
2. Dissertation structure
Ch1. Introduction
Ch3. Trend analysis
Ch4. Variability analysis
Ch2. Time-series construction
In revision for Remote Sens. Environ.
Published in Science 2015
In preparation for publication
3. The Antarctic ice sheet interacts
with the ocean through the ice shelves
Modified from NASA
4. The ice in Antarctica flows
just like rivers do on continents
NASA, Rignot et al. 2011
East
West
6. Ice shelves restrain the flow
from the ice sheet interior to the ocean
Modified from Hughes 2011
gravitational
driving stress
Plan view of an ice shelf
highs
7. Large portions of the Antarctic ice sheet
are prone to instability
Vaughan and Arthern 2007, Schoof 2007
Grounded below sea level
(prone to instability)
q
H, grounding-line thickness
∝
ax
ax-δ
q
q+Δ
8. Future ice-sheet contribution
to sea-level rise is (highly) uncertain
IPCC AR5, Velicogna andWahr 2013
Total SLR
Thermal
expansion
Mass
increase
Greenland
Antarctica
9. Summary
The ice sheets are currently
loosing mass at an accelerated rate
Large portions of the Antarctic
ice sheet are prone to instability
Ice-shelf loss ➔ increased ice
discharge ➔ sea-level rise
10. Scientific questions
How have ice shelves varied over time?
What are the spatial patterns?
What are the links to climate variability?
How widespread are the changes?
11. Three satellite radar altimetry missions:
18 years of continuous data (1994-2012)
= − ( − )
ERS-1
ERS-2
Envisat
12. Detecting changes in the vertical component
is challenging over floating ice
∂
∂
= ∂ ∆ − ∂ ρ−
+ ∂ ρ−
+ (ρ−
− ρ−
) ( ˙ + ˙ + ∇ · v)
changes in
ocean height
changes in
ocean density
changes in
firn density
changes in
surface mass
changes in
basal mass
changes in
velocity field
ice-ocean
density contrast
Shepherd et al. 2004; Padman et al. 2012
13. We averaged height measurements
over 30-km grid cells and 3-month bins
Paolo et al. 2015b
We stack several time series to
smooth out incoherent signal
14. We estimated trends using
polynomial regularized regression
ˆ( ) = β + β + β + β
Minimizing
Subject to
Fit
(bias)
(variance)
|β |
( − ˆ )
Lasso regularization:
Tibshirani 1996
Cross validation:
15. Efron andTibshirani 1993
We estimated uncertainties by
bootstrapping the residuals of the fit
∗
( ) = ˆ( ) + ε∗
( )
ε( ) = ( ) − ˆ( )Residual of the fit
Bootstrap sample
We performed a total of
1,330,000 sets of calculations
Combined error σ =
(σ∗) + (σ )
Trend fit
Derivative
95% CI of the fit
16. Paolo et al. 2015b
Over 1600 time series
We constructed time series and maps
of ice-shelf height change and acceleration
17. Paolo et al. 2015a
18 years of changes
show a clear spatial pattern:
West ice shelves are thinning fast
East ice shelves not so much
18% volume loss
in less than
2 decades
19. Paolo et al. 2015a, 2015b
West Antarctic ice shelves:
Volume-loss rate increased by
~70% from the 1990s to 2000s
East Antarctic ice shelves:
Earlier increase in volume
ceased in the 2000s
All Antarctic ice shelves:
Volume-loss rate accelerated
-25 km3/yr ➔ -310 km3/yr
20. Schoof et al. 2010, Paolo et al. 2015a
We observe faster ice-shelf melt rates
near the grounding lines
CDW
21. Summary
At current rates some ice shelves
may disappear within this century
Ice shelves are decaying fast,
leading to Antarctic mass-loss increase
Enhanced inflow of warm CDW
is melting West Antarctica
22. Short observational records
with different scales in time
with large errors (noisy)
and many simultaneous time series
Motivation
Given
Can we distinguish between regular deterministic
behavior (cycles) and irregular behavior (noise)?
Question
23. Multivariate Singular Spectrum Analysis
identifies common oscillatory modes
Vautard et al. 1992, Golyandina et al. 2001, Ghil et al. 2002,
Time
Multivariatedataset
Time
Reconstructedcomponent
Window
Rank
EigenvectorEigenvalue
Signal
Noise
24. Paolo et al. in prep.
140 time series
There is statistically significant energy
at the interannual band in AS
25. f = 0.22 ➞ T ≈ 4.5 years
Paolo et al. in prep.
Time window
9 years
Time span
18 years
We identified an interannual oscillation
in Amundsen Sea ice-shelf height
26. NOAA
?
ENSO is the strongest natural climate
fluctuation at interannual time scales
Southern Oscillation Index (SOI)
27. Paolo et al. in prep.
Time window
6 years
Time span
18 years
Low and hight frequency modes of ENSO
are identified in the SOI series
T ≈ 4.5 years
T ≈ 2.5 years
f = 0.22, 0.40
28. Paolo et al. in prep.
Low-freq mode of ENSO
Ice-shelf height variability
El Nino events
Interannual ice-shelf height in Amundsen
is strongly correlated with ENSO
29. Ok, but does this make sense?
ρ 0
(+∆SST)
ρ 0
(−∆SST)
SOI−SST Correlation (ρ)
Riffenburgh 2007, Kwok and Comiso 2002, Cullather et al. 1996
Higher
moisture
convergence
Higher
snowfall
along the coast
Higher
cyclonic
activity
During an El Nino event:
Lower
temperature
along the coast
30. Summary
There is statistically significant
interannual variability in AS height
This variability is strongly correlated
with El Nino-Southern Oscillation
First direct observational evidence
of the ENSO-AIS teleconnection?
33. Circumpolar Deep Water melts
the ice shelves from below
Jenkins et al. 2010, Jacobs et al. 2011
Temperature SalinityPine Island
Ice Shelf
34. As the ice shelves thin,
so does the adjacent grounded ice
Pritchard et al. 2012
Grounded-ice
thinning Ice-shelf thinning
35. As the ice shelves thin,
so does the adjacent grounded ice
Pritchard et al. 2012
Grounded-ice
thinning Ice-shelf thinning
36. As the ice shelf is removed
the glaciers behind speed up
Rignot et al. 2004, Scambos et al. 2004
Before collapse After collapse
Horizontal velocity (InSAR) Horizontal velocity (InSAR)Flow rate (Landsat)
37. The geometry of the bed constrains
the stability of a marine ice sheet
Vaughan and Arthern 2007, Joughin and Alley 2011, Schoof 2007
=
Stable Unstable
Retrograde bed slope
q
ax
Grounding-line thickness H
(steady state)
(at the GL)∝
38. time 1 time 2
H1
H2
∆ = ( − ρ /ρ ) ∆ ≈ ∆
ρ
ρ
∆
39. Paolo et al. 2015a
x is time (1994 to 2012)
y is thickness change (m)
rates are in (m/decade)
Short records do not capture the trend
40. Morris andVaughan 2003, Paolo et al. 2015a
−9℃ isotherm
moving southward
Limit of ice-shelf viability
appears to be moving southward?
42. Paolo et al. in prep.
There is statistically significant energy
at the interannual band
The interannual component explains
a larger portion of the total variance
Interannual
Annual