This document discusses using radar altimetry data from multiple satellite missions spanning 1992-2012 to analyze changes in Antarctic ice shelves. It presents three key challenges: integrating data from different radar altimetry systems; accounting for time-varying surface properties; and applying spatial and temporal corrections. The goal is to capture ice shelf variability at spatial scales of 20-30 km over 20+ years. Results show high spatial and temporal elevation change variability, with coherent changes tracked around the Antarctic coast. Validation with ICESat data is also discussed.

2. Large scale studies on ice shelves
ERS-1/2 1992-2001 ERS-2/Envisat 1994-2008 ICESat 2003-2008
Zwally et al., 2005 Shepherd et al., 2010 Pritchard et al., 2012
Duration 9 years 14 years 5 years
Spat. Res. 100 km One value per ice shelf 30 km
Time Res. 3 months 35 days (!) 1-2 years
This study: ERS-1/ERS-2/Envisat 1992-2012 2
*

3. The need for multi-mission RA
Long vs short records in detecting climate trends
How long?
Decadal records
(20+ years)
Interannual and
decadal variability
unexplored
Fricker and Padman, 2012
Our goal → to capture the variability in space and time on the
ice shelf spatial scales: 20+ years / 20-30 km 3
*

4. Penetration depth (backscatter)
Penetration depth:
! Water ! "(mm)
! Wet snow ! O(cm)
! Dry snow ! O(m)
! And varies with time
A
Radar
penetrates
into firn layer
B
4
*

5. Constructing time series of dh
Similar (but not the same) method as
Davis & Segura (2001), Zwally et al. (2005), Khvorostovsky (2011).
5
*

6. Averaging in time and space
1 vs 3-month averages
less crossovers
per bin → larger
error bars
improved signal-to-
noise ratio and no
gaps
20-30 km bins
6
*

7. Crossing all possible time combinations
Now we have one time series per reference time: t1, t2, t3, ...
These are elevation changes with respect to different epochs
7
*

8. One grid per time combination
Crossovers t1
~ 1500 grids
t2
t3
8
*

9. Multi-referenced time series
At every individual grid-cell we have now several time series
outliers
1) To align we use average of the offset for overlap period only
2) Then we frequency-weighted average the aligned time series 9
*

10. Cross-calibration of average TS
Cross-calibration is done using overlap periods between missions
dh = elevation change dAGC = backscatter power change
ERS-1 ERS-2 Envisat
10
*

11. Backscatter correction (approach 1)
By correlating absolute values ! dAGC x dh
11
Wingham et al., 1998; Davis & Ferguson, 2004; Zwally et al., 2005

12. Backscatter correction (approach 1)
By correlating absolute values ! dAGC x dh

13. Backscatter correction (approach 2)
By correlating differences ! diff(dAGC) x diff(dh)

14. Backscatter correction (approach 3)
By time variable correlation ! R(t) and S(t)
14

15. Backscatter correction (approach 3)
By time variable correlation ! R(t) and S(t)
Correlation
Sensitivity
15

16. Correlation and Sensitivity maps
FRIS
(1)
(2)
16

17. Correlation and Sensitivity maps
(1)
ROSS
(2)
17

18. Now some results
18

19. High spatial and temporal variability
20-year trend in elevation change
(original grid)
19

20. High spatial and temporal variability
20-year trend in elevation change
(original grid)
20
Obs: 2001 was chosen to avoid a big calving event

21. 20-year trend in
elevation change
(original grid)
Obs: 2001
was
chosen to
avoid a
big calving
event 21

22. High interannual variability
22

23. Coherent changes?
Tracking coherent events around the Antarctic margin

24. How well do we know what RA is measuring?
24
*

25. Envisat vs ICESat
We follow
the ICESat
campaigns
25
*

26. Envisat vs ICESat
*

27. Conclusions
!
Multi-mission RA can be used to construct continuous
long records with their variability content
!
There is a lot of variability both in time and space
!
Variability is the key to understand forcings and climate-
induced changes (ocean and atmospheric circulation)
!
Relative error (precision) vs absolute error (penetration)
!
Different b/s approaches yield different results?
!
How can we validate b/s correction when there are so little
ground truthing data and in practice:
27

28. We thanks
!
NASA NESSF Fellowship
!
Jay Zwally & Jairo Santana (NASA/GSFC)
!
Curt Davis (UM) & Duncan Wingham (UCL)
!
NASA grants NNX06AD40G and NNX10AG19G
!
ESA for ERS-1, ERS-2 and Envisat altimeters!
!
San Diego Super Computer Center
!
Geir Moholdt
!
Python and Open Source
fpaolo@ucsd.edu
28

29. Antarctic ice shelf mask
A reliable and complete ice shelf mask is a problem
So we (Geir Moholdt) created our own using all data available: MOA (Scambos
et al. 2007), ASAID (Bindschadler et al. 2011), InSAR (Rignot et al. 2011),
ICESat (Fricker/Brunt et al. 2006-10)
29
*

30. Backup slide
30

31. Challenges of multi-mission integration
!
Differences between missions:
- RA systems, orbit configurations, time spans...
!
Radar interaction with time variable surface properties
!
Spatial and temporal dependent corrections:
- Ocean tides (for high lat)
- Atm pressure (IBE)
- Surface density (firn densification)
- Penetration depth (backscatter)
31

32. How to reduce the noise?
!
Due to hydrostatic equilibrium the altimeter only see 10%
of the grounded ice signal (in elevation change)
!
So to increase signal-to-noise ratio → requires lots of
averaging both in time and space
32
*

33. The uncertainty
!
How well do we know the error?
!
What do error bars in the time series actually represent?
After all the averaging a mean error is: ± 5-20 cm over 20-30 km
!
What about the uncertainty in penetration depth?
O(m/cm)
33