GCOS SST

1
GCOS SSTSI Working Group

• Chairs
• Nick Rayner
• Tom Smith
• Executive committee
• Ken Casey
• Elisabeth Kent
• Alexey Kaplan
• Craig Donlon
• Ed Harrison
• Dick Reynolds

Sea ice subcommittee Søren Andersen, DMI/EUMETSAT
OSISAF Florence Fetterer Walt Meier, NSIDC Tony
Worby Steve Ackley, ASPeCt Per Gloersen, NASA
GSFC Cathleen Geiger, CRREL/U. Delaware John
Stark, UK Met Office Vasily Smolyanitsky,
AARI/JCOMM ETSI Pablo Clemente-Colon, NIC Mark
Drinkwater, ESA Stefan Kern Dirk Notz, U.
Hamburg Jinro Ukita, Chiba University
2
Outline

• Background and summary of objectives
• Organisation
• Motivation and mission
• Illustrative examples
• Proposed list of activities
• Status
• Events
• Reports
• Intercomparison
• Outlook
• Recent relevant results

3
Organisation
4
Motivation and mission

• Motivation from a climate data user perspective
• Uneven data coverage as well as inconsistencies
  between different passive microwave sea ice
  timeseries and ice charts have been identified by
  climate community constructing longterm SSTSI
  analyses

• No uncertainty estimates in existing ice
  analyses.
• An overwhelming number of data sets/algorithms
  with poorly known relative and absolute skills.
• For microwave data Principal homogeneity issues
  with short overlap periods and different equator
  crossing times.
• The popular NSIDC gridded dataset has only 16
  days of overlap between DMSP F8 and F11
• SMMR and SSM/I overlap is 22 days
• For ice charts Changes in analysis capabilities
  and practices are not always well documented
• Major challenge to reconcile ice chart,
  observation and satellite records.
• The Antarctic has significantly less historical
  coverage than the Arctic

Mission To provide analysis and recommendations
on long term consistent sea ice fields with
uncertainty estimates for use in SST SI
analyses. Initially focus on ice concentration
but consider ice thickness as methods and data
sets mature.
5
Rapid changes in MY sea ice
Nghiem et al., 2006
6
Arctic trends
Analysis by Kaleschke, presented in Andersen et
al. 2006 (b). Merged ESMR SSM/I Cavalieri et
al.,2003 with NSIDC sea ice index. Recent
developments have no correlation with AO, but
Maslanik et al. (2007) suggest atm. circulation
may still be significant
Meier et al. Ann. Glaciol. (46) 2007 Confirms
the uniqueness of recent decline in ice coverage,
while illustrating the sensitivity of trends to
period and length of time series.
7
Trends in Antarctic

• NSIDC sea ice indicator

May be due to increased snow ice formation with
increased Precipitation. (Ice mass growth is a
balance between insulation and snow ice), Powell
et al., 2005.
8
NICs Sea Ice Climatology
1976
1993
Example from Partington et al showing change in
chart detail , 1976 to 1993 Over time Increase
in quality, quantity, and effective resolution of
charts
Courtesy Florence Fetterer, NSIDC
9
Differences in trends
Arctic area trends from 7 Microwave algorithms
Ice chart and Microwave trends
1987-2004
-42611
-58111
-33230
Nearly a factor 2 between lowest and highest
estimate
1991-2004
-42611
Courtesy Nick Rayner, Hadley Centre
10
Some blind spots, the Antarctic
HadISST extents (Rayner et al, 2003)

• Available time series
• Arctic has systematic observations back to late
  1800s
• Antarctic data is extremely scattered and spotty
  up to satellite era
• GDSIDB Arctic 1901-1997
• GDSIDB Antarctic 1973-1991

Gaps
German Atlas
Russian Atlas
11
Focused activities

• Systematic and comprehensive intercomparison of
  sea ice analyses
• Help characterize important temporal and
  spatially distributed differences
• Provide starting points for further investigation
  and cooperation
• Help evaluate effects of different assumptions
• Provide a step towards meaningful accuracy and
  uncertainty estimates
• Error estimates
• Develop and promote methods for standardized
  error estimates for both ice charts and satellite
  analyses
• Data
• Provide overview of the numerous data sets
  available
• Promote easy and standardized access to field and
  ship observations
• Examine data gaps and, if possible, recommend
  mitigation actions

12
Implementation
The full level of activity will depend on level
of community commitment and funding. However,
some initial activities are already committed to

• Demonstration of intercomparison of a limited set
  of IC products
• Products repository in common simple netCDF
  format
• Develop statistics and comparison procedures
  (initial short list exists)
• Presentation of analyses and download of data
• Ice chart based ice edge uncertainty project
  (NSF, Geiger)
• Outcome of IICWG 2006
• First step towards a more informed use of ice
  chart information in climate science
• Not to forget, progress through cooperation
• Structuring and collection of in situ
  observations and use of ice chart data for ice
  thickness in ASPeCt
• Structuring, collection and documentation through
  IICWG, ETSI
• Scientific coordination through CliC/WCRP and
  others.

13
Status

• Overview of events
• Documents
• Intercomparison

14
Recent events

• Group meetings
• Reformed in Exeter October 2005 with a wide
  initial representation in sst and the decision to
  form a specific subgroup on sea ice
• Inaugural sea ice meeting in Boulder, March 2006,
  established initial plans
• Participation in IICWG in Helsinki, September
  2006, featuring a dedicated morning session in
  plenary and several break outs. Resulting in an
  IICWG supported proposal to NSF for error
  analysis of ice chart based ice edge estimates.
• Interactions with other groups and bodies
• Presentation at Antarctic Sea Ice thickness
  Workshop, Hobart (July 2006)
• IGOS-Cryo meeting, Noordwijk (October 2006)
• JCOMM-ETSI meeting, Geneva (March 2007), plans
  for a common ice analyst workshop in 2008.
• The SI group is leaving the formative phase,
  entering ops.

15
Reports
16
Intercomparison

• Intercomparison is not a new concept in sea ice
  retrieval.
• Bad news
• It can be implemented in many ways.
• Small details and assumptions (e.g. spatial and
  temporal coverage, products considered,
  filtering/averaging) may affect results.
• Interpretation is often difficult (e.g. in terms
  of which product is more correct)
• Good news
• Comparison of products efficiently reveals
  signatures of underlying algorithm and system
  sensitivities. Links well with efforts to
  understand the underlying processes (e.g.
  radiative/microphysical, retrieval/classification)
  
• Experience from existing intercomparison
  exercises may help identify best practices and
  limitations.

17
Systematic intercomparison
netcdf gdsidb_blended-ps dimensions ni
304 nj 448 time 588
variables float lon(nj, ni)
float lat(nj, ni) short time(time)
timeunits "Months since Reference_time"
short ct(time, nj, ni)
ctlong_name "Ice concentration"
ctunits ""
ct_Fillvalue 165s float pix_area(nj,
ni) pix_arealong_name "Pixel
area" pix_areaunits "km2"
// global attributes
Reference_time 1950.
Data_set_name "GDSIDB blended analysis
1950-1998" Version "1.00"

• In practice the idea is
• Get data in common simple, self-contained format,
  example
• Compute statistics and intercomparison products
• Make comparison products and data sets easily
  available
• The hard part Interpret the results

18
Example problem Regridding

• Nearest neighbour regridding of GDSIDB product
  from 0.25? geographical to 25 km polar stereo
  projection
• 2.5 difference in extent trend (-13.9 to -14.2
  x 103 km2/year)
• 6.5 difference in area trend (-8.9 to -9.5 x
  103 km2/year)

Original
Regridded
Extent
Area
19
Development snapshot

• Intercomparison implemented in Python using
  matplotlib, numpy and Nio packages.
• Possible evolution to web application via
  CherryPy and TurboGears
• Data sets HadISST, NASA/Team, Bootstrap, GDSIDB
• Initial list of products
• Linear trends of monthly mean values of sea ice
  extent and area results in a measure of the
  spread in estimated retreat or increase in the
  sea ice cover. Taking one product as reference
  can be useful.
• Maps of linear trend in concentration or sea ice
  persistence provide the spatial structure of
  differences in estimated sea ice trends.
• Per pixel range of concentration based on several
  products or maps of anomaly with respect to
  wintertime average ice concentration provide
  spatial structure of single algorithm results.
• Maps of differences between algorithms on various
  time scales provide the spatial structure of
  inter-algorithm differences.

Intended to augment SST intercomparison system at
NODC
20
Outlook

• IPY
• Snapshot activities will make large quantities of
  high resolution EO data available and facilitate
  comprehensive validation activities
• Field activities will help in developing and
  validating models that relate snow and ice
  parameters to emissivity
• Massive deployment of buoys will at least help
  relating observed signals to meteorological
  conditions and ice drift patterns
• The systematic data management will make the
  application of the data possible.
• New ice thickness measurement capabilities
• Space based altimeters are becoming operational
  (IceSAT and follow-on, Cryosat-II, Sentinel-3)
• Development of AUV systems is picking up speed,
  in part thanks to IPY
• The ASPeCt data set is close to 25000
  observations spanning more than 2 decades in the
  Antarctic and still growing
• General movement towards sustained operational
  satellite programmes, e.g. ESA Sentinels, (JAXA
  GCOM?) also a movement towards free access to
  SAR data.
• SMMRSSM/I is approaching 30 years of continuous
  operation - but there will most likely be a gap
  in AMSR class passive microwave observations and
  Seawinds scatterometer continuity is very
  unlikely.

21
Conclusions

• The group is coming out of its formative phase
• Group members span ice charting, in-situ,
  satellite observations and to some extent
  modelling disciplines
• Activities and organisation have been outlined
  and agreed in a white-paper document
• Many of the defined activities have basic support
  from existing activities of members.
• An expanded set of activities, including
  standardisation of error estimates,
  interpretation of differences and thorough
  investigations of ice chart records, may require
  external funding, some initiatives have started
  already.
• The group is seen to benefit from and add to the
  momentum of existing groups like ASPeCt and IICWG
• In the longer term, it is hoped that the group,
  in cooperation with the ice charting community,
  may help lessen the gap between ice chart and
  satellite observations to achieve longer data
  sets of sea ice concentration with improved
  confidence.

22
Examples of recent results by group members
23
ASPeCt Tracks of 83 good voyages
Worby et al., 2007
24

Annual mean ice thickness including ridges
and open water 2.5 x 2.5º grid
Worby et al., 2007
25
AMSR-E v ASPeCt snow thickness data
Worby et al., 2007
26
Arctic Sea Ice from Passive Microwave (PM)
Sensors and Operational Ice Charts
From Meier, Fetterer, Fowler, Fall AGU 2006
27
MEMLSI simulations of ice concentration
NASA Team sensitive to layer contrast. Comiso
frequency moderately sensitive to
scattering. Near 90 GHz moderately sensitive to
deep scattering, sensitive to layer contrast.
Upper snow-layer density 100-410kg/m3 Above ice
correlation length 0.14-0.32mm
NASA Team
Comiso frequency
Near 90GHz
Scattering
Layering
Tonboe et al. 2006
28
Winter concentration anomalies
31 Oct 2000 31 Mar 2001
CP
NT2
BRI
Polarisation
NT
N90
CF
Spectral Gradient
Ice/snow Layering
Less Ice/snow Layering
Andersen et al. 2007
29
Backup material
30
Trends
31
Satellite trend

• Analysis by Kaleschke, presented in Andersen et
  al. 2006 (b).
• Merged ESMR SSM/I Cavalieri et al.,2003 with
  NSIDC sea ice index

32
Rapid changes in MY sea ice
Nghiem et al., 2006
33
Temporal analysis of ice index data variability
and linear trends in August for Eurasian
(1900-2003) and Canadian Arctic (1968-2004)
Greenland, Barents, Kara
Laptevs, Eastern-Siberian, Chukcha
Western CA (blue) Eastern CA (green) East Coast
(red) (April) Hudson Bay (light blue)
From V. Karklin, Z.Gudkovich, I.Frolov,
V.Smolyanitsky, J.Falkingham Interannual
variability of Eurasian and Canadian Arctic sea
ice in the 20th century and expectations for the
21st century. JCOMM-II Scientific Conference
Operational Oceanography and Marine Meteorology
for the 21st Century. Septembr 15-17th, Halifax,
Nova Scotia, Canada.
34
Wavelet analysis of sea ice extent variations for
Eurasian Arctic Seas (based on 1900-2003 period)
and Canadian Arctic Seas (based on 1968-2004
period) in August (red more ice, blue less
ice)
Greenland
Barents Kara
Laptev
Eastern-Siberian Chukcha
Western CA Eastern CA
Hudson Bay
35
Trends in Antarctic

• NSIDC sea ice indicator

May be due to increased snow ice formation with
increased Precipitation. (Ice mass growth is a
balance between insulation and snow ice), Powell
et al., 2005.
36
Differences
37
NICs Sea Ice Climatology
Courtesy Florence Fetterer, NSIDC
in 1996/1997 , NIC - Transitioned to digital
imagery (OLS/AVHRR) and digital analysis in GIS
format Started using SAR data in tactically
significant areas Now, NIC uses Quicksat to
compensate for deficiencies in SSM/I
38
Chart vs. NASA/TEAM SSM/I
Agnew Howell, 2003
39
Differences in trends
Arctic ice area trends from 7 Microwave
algorithms
1987-2004
1991-2004
Andersen et al., 2007
40
Error sources
41
Atmospheric
Water vapour
Wind
Cloud liquid water
Andersen et al., 2006
42
Atmospherically induced stdev
Bristol
Bootstrap
43
Same with clouds
Bristol
Bootstrap
44
Empirical fit

• Collocation of large number of SSM/I passes gives
  following relation (Schyberg)
• s0.040.07(C(1-C)/0.25)

Schyberg
Combined
Atmosphere
Ice tiepoint
45
MEMLSI simulations of snow effects
MEMLSI emissivity model setup to explore effects
of layering and coarse grains in the snow layers
above sea ice.
1
2
Layer 1 snow density 100-410kg/m3, varying layer
contrast btw. layers 1 and 2 Layer 3 correlation
length (grain size) 0.14-0.32mm, simulating
effect of depth hoar
NASA Team
Comiso frequency
Near 90GHz
Scattering
Layering
46
Footprint convolution errors
Deconvoluted
Reference
True low res.
Combined convolution and simple averaging
Limaye et al. (2006) Particularly important in
gradient areas, i.e. ice edge and coast.
Additional errors come from mixing of Channels
with different footprint size
47
In summer Melt ponds and others
In summer, melting snow and ice forms open
freshwater ponds on top of the ice that, in the
microwave regime, can be distinguished from the
open ocean only by salinity. Additional
complications occur e.g. due to refreezing of the
ponds. Sea ice topography exerts an important
control on the spatial extent and depth of the
ponds. Spatial coverage can range from 20-50
Russ Hopcroft, from the University of Alaska
Fairbanks, takes a sip from one of the many
meltponds scattered around the ice. (Photo
courtesy of Ian MacDonald.) From NOAA Ocean
Explorer site
48
Applications/impacts
49
Impact in climate models

• Singarayer et al. (2005) Fig.2 Fig.3
• Sea ice uncertainty is large but ocean - atm.
  gradient is small during summer.
• Other contributions Rinke etal. (2006)
  Influence of sea ice on atmosphere.
• Parkinson et al. Impact on climate models (2001)

50
Assimilation over sea ice
Estimation of the surface emissivity
51
Positive impactHirlam AMSU-A assimilation
experiment over sea ice, 24 hour forecasts
Black Control run Green Experiment
Courtesy H. Schyberg, met.no
52
Ice mass
53
Ice mass is the important quantity
IceSAT thickness Kwok et al.,2006