Decadal Changes in the Polar Sea Ice Cover

1
Decadal Changes in the Polar Sea Ice Cover

• Josefino C. Comiso
• NASA Goddard Space Flight Center, Code 614.1
• Greenbelt, MD 20771
• IGOS/CliC Cryosphere Theme Workshop, Kananaskis,
  Alberta, CA, 2-4 March 2005

2
Why Study Changes in theSea Ice Cover

• Amplification of climate change signal due to
  ice-albedo feedback
• Good insulator keeps polar ocean warm
• Redistributes salt- forms in one place, melts in
  another
• Initiates convection
• Forms high salinity and dense bottom water
• Affects weather
• The Arctic perennial ice cover is observed to be
  rapidly declining

3
Sea Ice Parameters

• Ice Concentration needs a good definition
• Ice Extent/Area depends on IC/resolution
• Ice Type new, young, FY, MY
• Average Ice Thickness Seasonal, Perennial
• Snow Cover Characteristics thickness,
  granularity, layering, wetness
• Ice Temperature Surface and snow/ice interface

4
Ice Concentration

• Must provide scientific information about the ice
  cover, e.g., compactness, thickness, etc
• Must not be binary ice vs no ice
• This means, we should not threshold on grease
  ice
• Must be consistently defined regardless of
  sensor, i.e., passive, infrared, visible
• Must enable first order estimate of volume and
  useful for heat flux calculations
• Must take advantage of detailed ice cover
  information that satellite data provides

5
Cluster Plots illustrating the basic differences
between NT1 and Bootstrap results
6
High Resolution AMSR vs Landsat

• Higher spatial details can be inferred from
  AMSR-E data, especially at 89 GHz
• AMSR-E data at 6.25 km resolution captures many
  of the spatial features from a high resolution
  visible channel
• The 12.5 km data show some details but the 25 km
  data smear out much of the features.

7
Divergence in the vicinity of Novaya Zemlya
IslandDivergence and polynyas must be revealed
in IC maps
8
Large leads in AlaskaIC maps must show large
cracks in the ice
9
The statistical error depends on the standard
deviation with respect to the line AD (in Blue)
which has been evaluated to be around 1.5 K.
This means the limit in the ice concentration
error is about 2.
10
Combining Satellite Data Records

• Need consistency checks from one sensor to
  another ESMR-SMMR-SSMI-AMSR
• Match TBs, ICs, ice edge, land mask, land/ocean
  mask, open ocean mask
• Account for time difference (AM vs PM sensors)
• Account for differences in resolution
• Account for differences in radiometric accuracy

11
Antarctic ice anomaliesSMMR (red) to SSM/I
(black) toAMSR (green)
12
Arctic Ice AnomaliesSMMR (red) to SSM/I (black)
toAMSR (green)
13
AMSR Ice edge12.5 km resolution

• High resolution data provide a better definition
  of the ice edge.
• With AMSR data, all channels provide consistent
  ice edge information.
• Some discrepancies between AMSR and SSM/I IC ice
  edge location is observed.

14
Record Length

• What is the minimum data record that must be
  available for trend analysis to be credible?
• Is it okay to combine data from different
  sources?
• How many cycles are needed for harmonic analysis?
• How do we evaluate the trend results in terms of
  known data errors?

15
Sensitivity of trends to record length and
climate cycles

• Data with different record lengths (rl) provide
  different trends but those with rl less than 15
  years may not provide meaningful trends.
• A negative trend is inferred from about 65 years
  of data while spectral analysis show
  periodicities at 12 and 33 years.

16
Total and Regional Ice Extents in the Arctic
17
The Ross versus Bel-Amundsen SeasWhen combining
satellite data with in situ data with longer
records, only relevant local changes must be
considered
18
Sources of Errors in Total Ice Areas

• Consistency in time series Connecting data from
  different satellites with different resolutions
  and calibration is not trivial. One year of
  overlap is highly desirable.
• Geolocation and side lobe effects.
• A one pixel error in the ice edge is possible.
• Ice Concentration (IC) Retrieval
• 1K error in tie point causes 1 error in IC
• Open Ocean Masking-due to extreme weather wind
• Land Masking-algorithm is suitable for oceans
  only
• Land/Ocean Boundary Masking
• Many ice and surface types

19
Satellite sea ice extent record

• SMMR and SSMI records provide different trends
  for the ice extent.
• In the Arctic, the trends are similar and
  slightly higher than average. In the Antarctic,
  the trends have opposite sign and significantly
  different from average.

20
Satellite sea ice area anomalies/trends

• Ice area distributions are similar to those of
  ice extent.
• In the Arctic, the trends for SMMR and SSMI are
  more consistent with average values. In the
  Antarctic, the differences are not as marked.

21
Ice Edge Errors

• Ice edge is difficult to monitor because of high
  temporal and spatial variability.
• SAR data provides much higher resolution but may
  sometimes miss the ice edge because of similar
  backscatter as open water.
• Ship data at x (in red) show actual location of
  the ice edge.

22
Sensitivity of 1 pixel mismatch at the ice edge
to estimated trend in ice extent in the NH

• Consistent mismatch for all months from 1978
  through 2001 yields insignificant difference in
  the NH.
• A mismatch during SMMR years but not during SSM/I
  years or vice-versa shows large change in trend
  estimates.
• Matching of results during periods of overlap is
  very important for accurate trend analysis.

23
Sensitivity of 1 pixel mismatch at the ice edge
to estimated trend in ice extent in the SH

• Trends in ice extent in the Southern Oceans is
  negligible when ice edge is defined as usual.
• Mismatch during SMMR or SSM/I years produced
  significant trends with opposite signs.
• Consistency and accuracy in the identification of
  the ice edge is very important.

24
On September 11, 2002, the Arctic Perennial Ice
Cover was at its lowest extent during satellite
era. AMSR-E is consistent with SSM/I and will
provide continuity and more accurate data in the
series.
25
Sensor TB and IC spatial consistency

• Differences in TBs are mainly in open ocean
  regions where weather effects are apparent.
• The changes are mainly caused by differences in
  revisit times over the polar regions.
• Despite bias and a slight change in TB
  calibration, the derived ice concentrations are
  basically identical.

26
Arctic Ice Cover during Minimum Extent1979-2003
27
Arctic Perennial Ice Cover, 1979-2004
28
Updated Perennial Ice Trends

• Overall, the ice cover in the Northern Hemisphere
  is declining but mainly due to that of summer ice
• The perennial ice cover continues to decline
  rapidly

29
Summary and Conclusions

• We can derive ice concentration, extent, and ice
  area consistently using satellite data.
• Ice concentration should provide scientific
  information about temporal changes in the ice
  cover.
• Trend results depend on length of record. Need
  about at least 20 years to be credible.
  Combining satellite with older in situ
  observation is tricky.
• Trend errors include use of multiple sensors with
  different resolution, visit time, and
  performance.
• We are beginning to get reasonable assessments of
  thickness, type, as well as snow and surface ice
  temperatures. But data records of these
  parameters are still too short for decadal
  studies.

30
End of Presentation