Please support this Forum and Neven's Blog

Author Topic: Piomas vs. CryoSat  (Read 79766 times)

ChrisReynolds

  • ASIF Upper Class
  • Posts: 1711
    • View Profile
    • Dosbat
Re: Piomas vs. CryoSat
« Reply #100 on: October 12, 2013, 10:33:09 PM »
Damn, that's bad news.

Thanks Wipneus.

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #101 on: October 13, 2013, 12:32:48 AM »
According to this message Cryosat resumed its measurements. Haven't found any information whether there is any additional impact.

I had a look at the snow thickness maps used to calculate thickness. It seems snow cover depends on ice age, I have no idea why. Otherwise same ice age gives same snow height for 2012 and 2013. So, the missing thing is real snow data. Has anyone an idea besides GFS or ECMWF? I believe there is somewhere data - isn't PIOMAS using that to estimate flux and isolation?

ChrisReynolds

  • ASIF Upper Class
  • Posts: 1711
    • View Profile
    • Dosbat
Re: Piomas vs. CryoSat
« Reply #102 on: October 13, 2013, 08:01:24 AM »
Arcticio,

I must admit I was anticipating something more serious. I've got too used to satellites failing.

I haven't a clue about snow thickness matching ice age. PIOMAS uses NCEP/NCAR for snow depth, I think this is calculated from precipitation through the winter.

Steven

  • ASIF Citizen
  • Posts: 329
    • View Profile
Re: Piomas vs. CryoSat
« Reply #103 on: October 13, 2013, 03:50:20 PM »
The user guide at the meereisportal website gives a bit more explanation:

There are promising efforts to use passive microwave to estimate snow depth with different approaches, but the current processors uses snow climatology instead of remote-sensing data.
 

Warren et al., 1999 established this climatology with results from drifting station mainly on multi-year sea ice collected over the past decades.  But since the Arctic Ocean shows a significant higher fraction of first-year sea ice, we follow the approach proposed by Kurtz et al., 2011 and multiply the climatological snow depth values over first year ice with a factor of 0.5.  Note: This approach is identical to Laxon et al., 2013.  The climatology is given as a fit function for each month and only valid for the central Arctic Ocean.  Significant error may occur if this fit is used also for region farther south (e.g. Baffin Bay), however this is included in the data product for the reason of completeness.


Emphasis mine.

http://www.meereisportal.de/fileadmin/user_upload/Pictures/Meereisbeobachtung/cryoSat-2/AWI_CryoSat-2_Documentation_20130624.pdf

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #104 on: October 13, 2013, 11:35:58 PM »
Chris, me too as I saw the headline. ESA's stream of information could be better, however it tells they found something, now I wait for confirmation data is actually dropping in again.

Steven, good finding. I might have an explanation. Assuming all snow is gone in September and after a given cut-off date there is no FYI, I would say FYI simply starts later in the freezing season to accumulate snow. Whether 0.5 is close to reality is another question. What bothers me most is because of the snow climatology the 'near-real-time' aspect of Cryosat is gone. A trend can only be visible after years. At least within the range of snow min and max.

Hopefully detecting real snow height from space makes some progress soon. Having an exact measurement in March/April is really worth the effort.

Btw. the laser technology used by IceSat doesn't penetrate snow at all, it always returns freeboard including snow.

dlen

  • ASIF Lurker
  • Posts: 44
    • View Profile
Re: Piomas vs. CryoSat
« Reply #105 on: October 15, 2013, 01:00:47 AM »
I wonder if it would not have been wiser to build in a laser altimeter instead of this radar, which measures the height of a plane somewhere within the snow layer, as I read somewhere.
So absolute values from CryoSat2 have to be taken with great care , it seems to me.

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #106 on: October 15, 2013, 09:31:17 AM »
Well, at least in Winter CryoSat works better (theoretically), because the radar can penetrate down to the ice surface. So, having both technologies ready at the same time should give exact snow height at least within a relevant time span. During the CryoSat Cal/Val campaign NASA and ESA cooperated fruitfully, hopefully that continues.

Actually, what's needed is kind of SnowSat :). May be SMOS can jump in, because snow has effects on the polarization: Snow thickness retrieval over thick Arctic sea ice using SMOS satellite data. I wonder whether the data of 3 satellites from 2 nations flying in an orbit of some hundreds of kilometers can be synced down to centimeters to measure real snow thickness? Sounds like an exciting challenge.


anonymous

  • Guest
Re: Piomas vs. CryoSat vs. IceSat
« Reply #108 on: October 16, 2013, 02:57:30 PM »
diablobanquisa, nice, was just about to link this paper here. It gives a very readable overview about the uncertainties of sea ice thickness retrieval. Not only snow cover, but sea ice density is a major unknown. The authors claim the huge difference between IceSat and the first CryoSat results may be explained by differently assumed density or more exactly the proportion of MYI/FYI, instead of melting.


dlen

  • ASIF Lurker
  • Posts: 44
    • View Profile
Re: Piomas vs. CryoSat
« Reply #109 on: October 19, 2013, 04:14:47 PM »
Concerning freeboard measurement I found the following quote (http://www.meereisportal.de/fileadmin/user_upload/Pictures/Meereisbeobachtung/cryoSat-2/AWI_CryoSat-2_Documentation_20130624.pdf, contains an extensive treatment of uncertainties)

At dry and cold conditions, the main reflector of Ku-Band signal of SIRAL should theoretically be the snow-ice interface. However results from validation experiments (Hendricks et al., 2010, Willatt et al., 2010, Willatt et al., 2011) have shown that the retracked elevation is often within the snow layer or sometimes the air-snow interface, e.g. by ground radar observations or comparisons between airborne laser and radar altimeter. Since validation data regarding penetration are sparse at basin scale, we apply a simplified parameterization which is taking into account seasonal changes.Therefore we introduce an additional penetration factor P, which describes the penetration of the radar reflection horizon. For the entire Arctic we assume a maximum  penetration of 23 cm from November till April, 11 cm for October and May, and 0 cm from June till September. The penetration factor is set to not exceed local snow depth.


Interpretation of the SIRAL height data is by no means trivial.
No I know, why the first significant sea ice volume paper took 2 years after commisioning of the satellite.

From the discussion there, one can guess the one-point-uncertainty as sigma ~~ 0.3 - 0.4 m or so.
The mean uncertainty is of course much lower. But the use of arctic-wide factors is prone to introduce systematic errors - any error there adds directly to the mean error.

Another source can be found here: http://chantier-arctique.lebonforum.com/t140-estimating-sea-ice-thickness-from-its-freeboard.
« Last Edit: October 19, 2013, 04:59:15 PM by dlen »

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #110 on: October 27, 2013, 10:31:06 AM »
Re. snow. There is a NASA Snow Working Group, specialized on remote sensing. Found a recent presentation.

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #111 on: November 28, 2013, 04:58:13 PM »
Will Laxon, S. et al.

Laxon, S. et al. "CryoSat-2 estimates of Arctic sea ice thickness and volume." Geophysical Research Letters, 28 January 2013
http://www.personal.soton.ac.uk/pgc1g08/grl50193.pdf

update their numbers?
Any idea on when such information will be released?
Will small gap in data matter?

nukefix

  • ASIF Citizen
  • Posts: 354
    • View Profile
Re: Piomas vs. CryoSat
« Reply #112 on: November 28, 2013, 05:34:31 PM »
I wonder if it would not have been wiser to build in a laser altimeter instead of this radar, which measures the height of a plane somewhere within the snow layer, as I read somewhere.
So absolute values from CryoSat2 have to be taken with great care , it seems to me.
Lasers work well until they fail, which has been a big problem with IceSat. The big benefit of a radar altimeter is that it does not require cloud-free conditions..

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #113 on: November 28, 2013, 07:37:10 PM »
Will Laxon, S. et al. update their numbers?
Any idea on when such information will be released?
Will small gap in data matter?

CPOM lost both scientists, Laxon and Giles, due to different accidents earlier this year. Apparently AWI took over and released monthly netCDF files, read about here.

Monthly maps till March 2013 are here: http://www.arctic.io/explorer/16C9//4-N90-E0/

I expect new data after winter, may be April.

R Tilling

  • ASIF Lurker
  • Posts: 1
    • View Profile
Re: Piomas vs. CryoSat
« Reply #114 on: December 04, 2013, 03:19:59 PM »
Hi all,

I just wanted to respond to some of the points on this thread. Firstly although we lost our colleagues earlier this year, CPOM UCL are still working on CryoSat data. This is totally separate from the product that AWI has released, which has nothing to do with UCL.

We're constantly updating our estimates of sea ice volume with the newest data, and this will be presented at AGU next week by myself, so come along! The intention is then to publish the results early next year.

Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 3272
    • View Profile
    • Arctic Sea Ice Blog
Re: Piomas vs. CryoSat
« Reply #115 on: December 04, 2013, 05:52:13 PM »
Thanks for this information, Rachel. Good to see that you guys over at CPOM UCL are still working with CryoSat data. Please, let me know when results are released and I'll mention it over on the Arctic Sea Ice Blog.
Il faut cultiver notre jardin

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #116 on: December 10, 2013, 04:50:35 PM »
AGU session C54A-04.
Trends in Arctic Sea Ice Volume 2010-2013 from CryoSat-2
Rachel Tilling, ...

Satellite records show a decline in Arctic sea ice extent over the past three decades with a record minimum in September 2012, and results from the Pan-Arctic Ice-Ocean Modelling and Assimilation System (PIOMAS) suggest that this has been accompanied by a reduction in volume. We use three years of measurements recorded by the European Space Agency CryoSat-2 (CS-2) mission, validated with in situ data, to generate estimates of seasonal variations and inter-annual trends in Arctic sea ice volume between 2010 and 2013. The CS-2 estimates of sea ice thickness agree with in situ estimates derived from upward looking sonar measurements of ice draught and airborne measurements of ice thickness and freeboard to within 0.1 metres. Prior to the record minimum in summer 2012, autumn and winter Arctic sea ice volume had fallen by ~1300 km3 relative to the previous year. Using the full 3-year period of CS-2 observations, we estimate that winter Arctic sea ice volume has decreased by ~700 km3/yr since 2010, approximately twice the average rate since 1980 as predicted by the PIOMAS.


http://fallmeeting.agu.org/2013/scientific-program-2/

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #117 on: December 10, 2013, 05:09:33 PM »
Perhaps also of interest

   
C54A-07. Uncertainties in Arctic sea ice thickness and volume: New estimates and implications for trends

Arctic sea ice area, thickness and volume have decreased in the last decades, but these estimates rely on a number of geophysical parameters which introduce large uncertainties. In our study we quantify the contributions of sea ice density, snow load, and area uncertainties on both sea ice thickness and volume estimates. Sea ice freeboard retrievals from the laser altimeter onboard ICESat are used along with snow depth, sea ice density and area derived from remote sensed measurements, or based on climatological in-situ data.

Our results show that the estimates of sea ice thickness and volume are slightly influenced by uncertainties in sea ice area, while the choice of the ice density value has a major impact. Temporal and spatial patterns of snow depth become particularly important when deriving information about trends and inter-annual variability of sea ice thickness and volume. We find that the order of magnitude of sea ice volume uncertainties is about 10% of the total Arctic sea ice volume and is larger than those reported earlier, while trends within the ICESat period (2003 - 2008) are similar. Using a consistent choice of sea-ice density for the ICESat and CryoSat-2 (2010 - 2011) data we obtain a less dramatic sea ice loss than reported previously.


ChrisReynolds

  • ASIF Upper Class
  • Posts: 1711
    • View Profile
    • Dosbat
Re: Piomas vs. CryoSat
« Reply #118 on: December 11, 2013, 09:56:33 PM »
Thanks Crandles, very useful.

Good to see you back.

anonymous

  • Guest
Re: Piomas vs. CryoSat
« Reply #119 on: December 16, 2013, 10:50:03 AM »
BBC reports CryoSat estimates for October 2013: 9000km³, last year was 6000km³. Apparently a lot of MYI survived in place, north of Canada.


Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 3272
    • View Profile
    • Arctic Sea Ice Blog
Re: Piomas vs. CryoSat
« Reply #120 on: December 16, 2013, 10:17:44 PM »
Blog post up on the ASIB.
Il faut cultiver notre jardin

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #121 on: January 25, 2014, 04:31:26 PM »
In the Cryosphere, an interesting discussion paper is published on a improved retrieval algorithm for Cryosat-2. The reported results (in the abstract) are impressing: freeboard and ice thickness have much better agreement with IceBridge data. In ice thickness the differences are reduced from 1.351 m down to 0.182 m.

An improved CryoSat-2 sea ice freeboard and thickness retrieval algorithm through the use of waveform fitting

N. T. Kurtz1, N. Galin2,3, and M. Studinger1
1Cryospheric Sciences Laboratory, NASA Goddard Space Flight Center, Greenbelt, MD, USA
2Earth System Science Interdisciplinary Center (ESSIC), University of Maryland, College Park, MD, USA
3NOAA, Silver Spring, MD, USA

Abstract. We develop an empirical model capable of simulating the mean echo power cross product of CryoSat-2 SAR and SARIn mode waveforms over sea ice covered regions. The model simulations are used to show the importance of variations in the radar backscatter coefficient with incidence angle and surface roughness for the retrieval of surface elevation of both sea ice floes and leads. The numerical model is used to fit CryoSat-2 waveforms to enable retrieval of surface elevation through the use of look-up tables and a bounded trust region Newton least squares fitting approach. The use of a model to fit returns from sea ice regions offers advantages over currently used threshold retracking methods which are here shown to be sensitive to the combined effect of bandwidth limited range resolution and surface roughness variations. Laxon et al. (2013) have compared ice thickness results from CryoSat-2 and IceBridge, and found good agreement, however consistent assumptions about the snow depth and density of sea ice were not used in the comparisons. To address this issue, we directly compare ice freeboard and thickness retrievals from the waveform fitting and threshold tracker methods of CryoSat-2 to Operation IceBridge data using a consistent set of parameterizations. For three IceBridge campaign periods from March 2011–2013, mean differences (CryoSat-2 – IceBridge) of 0.144 m and 1.351 m are respectively found between the freeboard and thickness retrievals using a 50% sea ice floe threshold retracker, while mean differences of 0.019 m and 0.182 m are found when using the waveform fitting method. This suggests the waveform fitting technique is capable of better reconciling the sea ice thickness data record from laser and radar altimetry data sets through the usage of consistent physical assumptions.

Citation: Kurtz, N. T., Galin, N., and Studinger, M.: An improved CryoSat-2 sea ice freeboard and thickness retrieval algorithm through the use of waveform fitting, The Cryosphere Discuss., 8, 721-768, doi:10.5194/tcd-8-721-2014, 2014.

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #122 on: January 26, 2014, 10:26:49 AM »
Some snippets from the discussion paper's conclusions.

The reduced differences are indeed large, all pointing at an overestimating of ice thickness:

Through comparison with Operation IceBridge data for
the 2011–2013 campaigns, this study has shown that fitting of the CryoSat-2 level 1B
waveforms using a physical model can be used to obtain improved results over the em-
pirical lead and threshold tracker (ELTF) methods which are similar to those used by
Laxon et al. (2013). The ELTF method was found to have respective mean freeboard
differences (CryoSat-2 – IceBridge) of 15.4 cm, 15.9 cm, and 11.9 cm and mean sea
ice thickness differences of 144.2 cm, 149.3 cm, and 111.9 cm. The mean freeboard
differences for the waveform fitting method were 2.2 cm, 2.5 cm, and 1.1 cm, and the
25 mean sea ice thickness differences were 20.6 cm, 23.3 cm, and 10.6 cm.

I am actually amazed about the differences. There is no way I can reconcile these with the uncertainty figures mentioned in Laxon et al.:

: Laxon et al.
Finally we compare CS-2, gathered between the 10th of March and 09th of April 2011, with ice thickness estimates computed using laser altimeter freeboard measurements from the NASA Operation IceBridge (OIB) mission gathered between the 16th and 28th of March 2011. CryoSat data acquired over the same dates in 2012 are also compared with IceBridge data gathered between the 12th of March and the 2nd of April 2012. We compare this data by gridding both CS-2 and the airborne data onto the same 0.4 latitude x 4 longitude grid as used for the EM comparison. The correlation (figure 3c) is lower (R = 0.608) than either the EM or ULS and the mean difference is -0.048±0.723 m.


Back to the TCD paper, the authors discuss remaining possible enhancements, further reducing uncertainties.
About snow thickness they say:

Lastly, an
evaluation of the IceBridge snow depth measurements needs to be done to improve
basin-wide snow depth on sea ice estimates. This has been done for a single season
of data (Kurtz and Farrell, 2011) compared to the snow depth climatology of Warren
et al. (1999), and for passive microwave retrievals of snow depth on sea ice for first year
ice (Brucker and Markus, 2013). The focus of a future study will be to utilize existing
observations to improve estimates of snow depth on sea ice to be used in the retrieval
of sea ice thickness from the CryoSat-2 time series.

Finally they are very satisfied with Cryosat's data:

Overall, this study has further demonstrated the capabilities of CryoSat-2 for the
retrieval of sea ice freeboard and thickness. The advantage of the retrieval processes
used in this study is that they are compatible with the laser altimetry record and show
that the two records can be reconciled to produce a more complete time series of sea
ice volume change. This has distinct advantages for the expected launch of the ICESat-
2 laser altimeter mission in 2017. The lifetime of CryoSat-2 is expected to overlap with
the ICESat-2 mission, as is the new Sentinel-3 radar altimeter mission. The combined
satellite radar and laser altimetry data provided by these missions will thus provide
unmatched information on the state of the Arctic sea ice cover.

Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 3272
    • View Profile
    • Arctic Sea Ice Blog
Re: Piomas vs. CryoSat
« Reply #123 on: January 26, 2014, 10:54:25 AM »
Thanks for this, Wipneus.

The ELTF method was found to have respective mean freeboard
differences (CryoSat-2 – IceBridge) of 15.4 cm, 15.9 cm, and 11.9 cm and mean sea
ice thickness differences of 144.2 cm, 149.3 cm, and 111.9 cm.

What are they saying here exactly? That CryoSat2 data so far has overestimated average ice thickness by over 1 m?
Il faut cultiver notre jardin

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #124 on: January 26, 2014, 11:49:16 AM »
Thanks for this, Wipneus.

The ELTF method was found to have respective mean freeboard
differences (CryoSat-2 – IceBridge) of 15.4 cm, 15.9 cm, and 11.9 cm and mean sea
ice thickness differences of 144.2 cm, 149.3 cm, and 111.9 cm.

What are they saying here exactly? That CryoSat2 data so far has overestimated average ice thickness by over 1 m?

Yes, according to the ELTF method. And that method is supposed to be similar to those used by Laxon et al.  The "similarity" is detailed a bit, but not completely:

We note that several differ-
ences are present between the freeboard retrieval used by Laxon et al. (2013) and the
ELTF method used here. The primary difference is that Laxon et al. (2013) subtracted
a bias from the sea ice lead elevations by taking the difference between returns from
the ocean when sea ice is not present and returns from leads in the nearby ice pack.
This was done following Giles et al. (2007), but was not done in the ELTF freeboard
retrievals. Additional differences include (but are not limited to) the exact definition of
the first peak, as well as the use of a mean sea surface height data set in place of
the EGM08 geoid. Therefore, the comparisons done in this study are similar, but not
exact reproductions of methodologies. The purpose of the comparison is to highlight
the physical basis between differences in the retracking methods

ChrisReynolds

  • ASIF Upper Class
  • Posts: 1711
    • View Profile
    • Dosbat
Re: Piomas vs. CryoSat
« Reply #125 on: January 26, 2014, 11:49:39 AM »
Neven,

From my reading, not necesarily, because the CS2 data used different assumptions for different regions. Those figures are the result if consistent assumptions are used. So the aim of the paper seems to be to develop a less subjective method.

Thanks for posting the link Wipneus.

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #126 on: January 26, 2014, 10:28:03 PM »
Thanks for this, Wipneus.

The ELTF method was found to have respective mean freeboard
differences (CryoSat-2 – IceBridge) of 15.4 cm, 15.9 cm, and 11.9 cm and mean sea
ice thickness differences of 144.2 cm, 149.3 cm, and 111.9 cm.

What are they saying here exactly? That CryoSat2 data so far has overestimated average ice thickness by over 1 m?

I think not quite: To me 'difference' is an always positive quantity not a vector which has direction. So if 52% of time CS2 is over 1m thicker than reality and 48% of time it is over 1m thinner than reality then the mean difference is over 1m however in this case the volume only has to be adjusted by 4% of area * over 1m thickness difference. That is a lot less than Neven is suggesting.

jdallen

  • ASIF Upper Class
  • Posts: 2333
    • View Profile
Re: Piomas vs. CryoSat
« Reply #127 on: January 27, 2014, 03:56:34 AM »
What are they saying here exactly? That CryoSat2 data so far has overestimated average ice thickness by over 1 m?
I think not quite: To me 'difference' is an always positive quantity not a vector which has direction. So if 52% of time CS2 is over 1m thicker than reality and 48% of time it is over 1m thinner than reality then the mean difference is over 1m however in this case the volume only has to be adjusted by 4% of area * over 1m thickness difference. That is a lot less than Neven is suggesting.
As you say, crandles.  The wag in me though, judging from the ELTF variability, want's to chime in to suggest that in many cases/areas, the best that could be said of the ELTF data is, "there is ice here" vs "there is thicker ice here".  It makes the spot data provided by buoys valuable for adjusting thickness estimates. 

I will be very interested to see how the new methodology changes our understanding of the pack's 'topography'.

It strikes me... is there a chance they may be able to run the new algorithm against previous data?  That could be revealing as well!
This space for Rent.

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #128 on: January 27, 2014, 08:00:33 AM »

I think not quite: To me 'difference' is an always positive quantity not a vector which has direction. So if 52% of time CS2 is over 1m thicker than reality and 48% of time it is over 1m thinner than reality then the mean difference is over 1m however in this case the volume only has to be adjusted by 4% of area * over 1m thickness difference. That is a lot less than Neven is suggesting.

Apart from positive quantities and vectors there are some curious things, sometimes called "negative quantities".


And no, "difference" is not open to interpretation:

(...) mean freeboard differences (CryoSat-2 – IceBridge) of (...)


When IceBridge is larger, the difference will be negative.

Your intepretation would be a ver strange one, seldom used in science. Usually they use someting like RMS, it used al the time in the Laxon et al. paper.

If still in doubt, just look at the graphs in this paper.

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #129 on: January 27, 2014, 08:36:02 AM »
Attached is fig 12, freeboard differences. Multiply by 9 to get proximate thickness diffs.


crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #130 on: January 27, 2014, 12:32:50 PM »
no, "difference" is not open to interpretation:

(...) mean freeboard differences (CryoSat-2 – IceBridge) of (...)



Yes you are right, sorry about that. Comparing Fig 9 and 11 the freeboard tracker similar to Laxon et el is giving larger thicknesses everywhere.


There are considerable areas with freeboard of near 0.5m in March 2011 and 2012 (but not 2013??) (Fig 11) which translates to 4.5m thick ice. However Chris Reynolds has shown there isn't much ice over 2m thick in PIOMAS after 2010:



Laxon et al show cryosat2 volume of about 16K Km^3 versus piomas 14K Km^3

Does this suggest that the bias correction (or other differences) done in Laxon et al (but not in the freeboard tracker in this Kurtz et al paper) is rather large? Does this make the differences quoted by the waveform method is paper seem small when compared to the large differences?

Should the waveform method differences be compared to Laxon et al uncertainties? Or does it require something more sophisticated like use half the icebridge data to apply bias corrections to Laxon et al method and Kurtz et al waveform method then see which does best compared to the other half of the icebridge data?

Or is the paper doing a fair comparison, 'before bias correction' against 'before bias correction'?
« Last Edit: January 27, 2014, 12:43:17 PM by crandles »

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #131 on: January 27, 2014, 01:36:20 PM »
As I read it, they carefully suggest that the previous results were effectively biased by selection of ice density and snow thickness:

The focus of this study is to develop a new method for the retrieval of sea ice free
board from CryoSat-2 data. We demonstrate that this method is consistent with inde-
pendent measurements from airborne laser and radar altimetry data sets from NASA’s
Operation IceBridge mission to retrieve sea ice thickness which eliminates the need to
utilize different ice density and snow depth values as an effective bias correction.

I have quoted the snow somewhere before. About ice density:

These quantities are due to environmental processes and should be applied in a con-
sistent manner in the retrieval of sea ice thickness regardless of which instrument is
used. In the case of sea ice density, previous studies have utilized a wide range of
values, which will result in large differences between data sets if the same physical
assumptions are used. For example, in the study by Kwok et al. (2009) an ice-density
of 925 kgm−3 was used, Kurtz et al. (2011) used a value of 915 kgm−3, while Laxon
et al. (2013) used an estimate of 917 kgm−3 for first year ice and 882 kgm−3 for mul-
tiyear ice. In these studies, the range of sea ice density values for multiyear ice is
particularly large at 43 kgm−3. For a typical multiyear sea ice floe with 60 cm of snow-
ice freeboard and 35 cm of snow depth, the sea ice thickness estimate differs by 1.1 m
within this range of ice densities.

Yet the results of these studies agree quite well despite these large differences in density. This leads logically to the freeboard measurements:

Despite the large-scale mean agreement of the sea
ice thickness data sets described in previous studies, this discrepancy in physical as-
sumptions points to the source of the differences as being due to potential biases in
the freeboard and snow depth data sets used. This large discrepancy underscores the
need to establish a set of consistent physical constants for use in the retrieval of sea
ice thickness from satellite radar and laser altimetry data

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #132 on: January 27, 2014, 02:03:57 PM »
Laxon et al:
Nevertheless, comparison of CS-2 measurements with three independent in situ data sets reveals differences of less than 0.1 m in thickness when averaged on a large scale, or over a full winter growth season.

Kurtz et al:
the mean sea ice thickness differences were 20.6 cm, 23.3 cm, and 10.6 cm.
...
A bias of 1.9 cm was found in the waveform fitting method freeboard retrievals compared to the IceBridge data, this bias is consistent with the estimated range bias due to off-nadir ranging of lead points shown by Armitage and Davidson (2014).

How to compare? Is it
20.6-1.9*9=  3.5cm
23.3-1.9*9=  6.2cm
10.6-1.9*9= -6.5cm

These are all <0.1m so the errors are of similar magnitude?




crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #133 on: January 27, 2014, 02:31:09 PM »
For example, in the study by Kwok et al. (2009) an ice-density
of 925 kgm−3 was used, Kurtz et al. (2011) used a value of 915 kgm−3, while Laxon
et al. (2013) used an estimate of 917 kgm−3 for first year ice and 882 kgm−3 for mul-
tiyear ice. In these studies, the range of sea ice density values for multiyear ice is
particularly large at 43 kgm−3. For a typical multiyear sea ice floe with 60 cm of snow-
ice freeboard and 35 cm of snow depth, the sea ice thickness estimate differs by 1.1 m
within this range of ice densities.

Yet the results of these studies agree quite well despite these large differences in density. This leads logically to the freeboard measurements:

Despite the large-scale mean agreement of the sea
ice thickness data sets described in previous studies, this discrepancy in physical as-
sumptions points to the source of the differences as being due to potential biases in
the freeboard and snow depth data sets used. This large discrepancy underscores the
need to establish a set of consistent physical constants for use in the retrieval of sea
ice thickness from satellite radar and laser altimetry data

If Laxon et al get difference of <0.1m then couldn't an alternative explanation be that as ice gets thinner while snow remains similar (possibly increases slightly) the density that should be used changes over time such that 925kgm-3 is a sensible value for 2009, 915 sensible for 2011, while in 2013 a more sophisticated 917 / 882 split for FYI / MYI is sensible.

With a large change of 1.1m in ice thicknesses possible then I can see that this effectively is a bias correction for Laxon et al method. If ice to snow thickness ratio is changing then requiring the same density to be used does not seem sensible, if I have understood correctly. IOW, the set of consistent physical constants that needs to be established may need to change over time.

I am still left feeling I don't know if the method is an improvement or not.

It seems that Cryosat2 volume is likely to be fairly similar to what has been previously reported (not 1m thinner).

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #134 on: January 27, 2014, 04:40:46 PM »
Would a comment on the paper something like the following be sensible?

Laxon et al 2013 find "comparison of CS-2 measurements with three independent in situ data sets reveals differences of less than 0.1 m in thickness when averaged on a large scale, or over a full winter growth season". Comparison with this needs to be much more prominent than with your empirical lead and threshold tracker (ELTF), which you have found to be badly wrong.

In the introduction you assert
"Differences in sea ice thickness estimates from altimetry data arise in particular to the use of different density values and snow depth estimates which are used in the retrieval of sea ice thickness. These quantities are due to environmental processes and should be applied in a consistent manner in the retrieval of sea ice thickness regardless of which instrument is
used. In the case of sea ice density, previous studies have utilized a wide range of
values, which will result in large differences between data sets if the same physical assumptions are used. For example, in the study by Kwok et al. (2009) an ice-density
of 925 kgm−3 was used, Kurtz et al. (2011) used a value of 915 kgm−3, while Laxon
et al. (2013) used an estimate of 917 kgm−3 for first year ice and 882 kgm−3 for multiyear
ice. In these studies, the range of sea ice density values for multiyear ice is
particularly large at 43 kgm−3 5 . For a typical multiyear sea ice floe with 60 cm of snowice
freeboard and 35 cm of snow depth, the sea ice thickness estimate differs by 1.1m
within this range of ice densities. Despite the large-scale mean agreement of the sea
ice thickness data sets described in previous studies, this discrepancy in physical assumptions
points to the source of the differences as being due to potential biases in the freeboard and snow depth data sets used. This large discrepancy underscores the need to establish a set of consistent physical constants for use in the retrieval of sea ice thickness from satellite radar and laser altimetry data."

and your method "eliminates the need to utilize different ice density and snow depth values as an effective bias correction."

Pointing out that different assumptions is an effective bias correction and that you would prefer to eliminate this is sensible. However I wonder if the paper would be improved by considering whether the density assumption needs to change with the year. Presumably as ice is getting thinner it may be correct for the density value to be declining with time. Would adding that considering whether the density assumptions can be varied over time as a function of PIOMAS values be a sensible piece of suggested additional work?

I suggest the implications of the paper need more discussion. As it appears the errors you find are only as low as Laxon et al 2013 if you apply a bias correction to your method, it doesn't seem clear whether the method is an improvement or not. Does your method imply higher or lower Arctic sea ice volume than reported in Laxon et al? If the errors are similar what scope is there for improved estimate of sea ice volume using a combination of both methods (presumably you would want to give more weight to your method for areas that are a long way from a lead)?



Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #135 on: January 27, 2014, 05:49:43 PM »

If Laxon et al get difference of <0.1m then couldn't an alternative explanation be that as ice gets thinner while snow remains similar (possibly increases slightly) the density that should be used changes over time such that 925kgm-3 is a sensible value for 2009, 915 sensible for 2011, while in 2013 a more sophisticated 917 / 882 split for FYI / MYI is sensible.

Snow is a big problem. It was already discussed in this thread. I speculate it is also the main reason why there has been no Cryosat-2 data reported yet for the Antarctic (where snow is often thicker than the ice).

I am still left feeling I don't know if the method is an improvement or not.

It seems that Cryosat2 volume is likely to be fairly similar to what has been previously reported (not 1m thinner).

Even if no better data on thickness and volume come out of it, the advantages of getting the freeboard right are huge:
- better constraints on snow;
- better constraints on ice densities;

These have important impacts on ice forming and melting and as such will improve the modelling. In other words helps us to understand better what is going on.

Think of it, density depends mostly on the salt contents (brine inclusion). That influences the melting point, the mechanical properties but also the porosity of ice: forming and draining of melt pools.

Further Cryosat-2 data is seen as a vindication of PIOMAS. That may not be as strong as it looked if this paper has it right.
 
These guys with their new method have already identified a slight misalignment of the satellite (nadir is not exactly nadir). Even looks like this is worthy stuff.

Would a comment on the paper something like the following be sensible?
Commenting is open until 21st of March, there is some time.
I think from science pov, these guys have covered the issues. Yet a clarification that the big 1.5m
improvement does not translate into similar Cryosat's new estimate seems in order (to avoid misunderstandings). 

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #136 on: January 27, 2014, 06:22:07 PM »
Yes probably want to scratch most of that suggested comment because I wasn't thinking straight and almost certainly I still don't understand enough about it.

Yes even if the method isn't better, it is good that the work has been done: It may not be better for one type of task but it may be better for another. It is just that the title of the paper is 'an improved ..." and without comparing to the less than 0.1 m in Laxon et al, do we know whether it is improved or not?

Further Cryosat-2 data is seen as a vindication of PIOMAS. That may not be as strong as it looked if this paper has it right.

With the errors looking to be of similar magnitude, AFAICS the vindication might be slightly worse or slightly better. If you see reason to think it will be worse rather than better then please explain.

the advantages of getting the freeboard right are huge

Sure, but do we think the freeboard is more correct than Laxon et al?

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #137 on: January 27, 2014, 06:59:46 PM »

Further Cryosat-2 data is seen as a vindication of PIOMAS. That may not be as strong as it looked if this paper has it right.

With the errors looking to be of similar magnitude, AFAICS the vindication might be slightly worse or slightly better. If you see reason to think it will be worse rather than better then please explain.
If I had know that the Cryosat-2 data includes a bias of about 1.5m, then I would have been less impressed. That is all I wanted to say.
the advantages of getting the freeboard right are huge

Sure, but do we think the freeboard is more correct than Laxon et al?
It is what this paper is claiming  (with the assumption that IceBridge is correct).

ChrisReynolds

  • ASIF Upper Class
  • Posts: 1711
    • View Profile
    • Dosbat
Re: Piomas vs. CryoSat
« Reply #138 on: January 27, 2014, 08:43:53 PM »
I've not read much on IceBridge, but when I was looking at the raw data with a view to intercomparison with PIOMAS I was stunned at the uncertainties stated with each reading. IIRC they were often as large as the stated thickness.

Jim Hunt

  • ASIF Governor
  • Posts: 2970
    • View Profile
    • The Arctic sea ice Great White Con
Re: Piomas vs. CryoSat
« Reply #139 on: April 30, 2014, 08:31:04 PM »
A new paper has just been published in The Cryosphere, entitled "Uncertainties in Arctic sea ice thickness and volume: new estimates and implications for trends"

http://www.the-cryosphere.net/8/705/2014/tc-8-705-2014.html

16 pages of interesting stuff, including discussion about the uncertainties in ice density and snow depth.

The ultimate conclusion:

Our results still reveal a decline in sea ice volume between the ICESat (2003–2008) and CryoSat-2 (2010–2012) periods, but less dramatic than reported in previous studies. However, final quantitative conclusions about a change in sea ice volume are hard to make, considering the large uncertainties and unresolved biases found in our study.

Reality is merely an illusion, albeit a very persistent one - Albert Einstein

ChasingIce

  • ASIF Lurker
  • Posts: 82
    • View Profile
Re: Piomas vs. CryoSat
« Reply #140 on: May 01, 2014, 12:52:29 AM »
So decline over the last 33 years is probably more than the 75% to 80% reduction in minimum volume since 1979 (16.855 down to 3.261 K Km^3).
Right Chris, that would be an 80.7% decline Dr. Schweiger refers to. But he goes on to say that the newly published data makes that 80% decline look too conservative. I still don't see that.
Anyone?

Yes, you go from 1979 maximum to 2012 minimum to come up with the ~80%

ktonine

  • ASIF Citizen
  • Posts: 324
    • View Profile
Re: Piomas vs. CryoSat
« Reply #141 on: May 01, 2014, 05:39:29 AM »
Yes, you go from 1979 maximum to 2012 minimum to come up with the ~80%

1979 MIN 16.855   2012 MIN 3.673   % Change 78.21
1979 MAX 33.035   2012 MIN 3.673   % Change 88.88

It's minimum to minimum that they're comparing.  It would be rather meaningless to compare maximum to minimum.



crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #142 on: May 01, 2014, 03:43:02 PM »
A new paper has just been published in The Cryosphere, entitled "Uncertainties in Arctic sea ice thickness and volume: new estimates and implications for trends"

http://www.the-cryosphere.net/8/705/2014/tc-8-705-2014.html

16 pages of interesting stuff, including discussion about the uncertainties in ice density and snow depth.

The ultimate conclusion:

Our results still reveal a decline in sea ice volume between the ICESat (2003–2008) and CryoSat-2 (2010–2012) periods, but less dramatic than reported in previous studies. However, final quantitative conclusions about a change in sea ice volume are hard to make, considering the large uncertainties and unresolved biases found in our study.



In case anyone has difficulty getting pas thalf way in the abstract where it says

The mean total sea ice volume is 10120±1280 km3 in October/November and 13250±1860 km3 in February/March for the time period 2005–2007.


There is a restricted area of study shown by fig 1: it is cut off at Bering Strait, Severnaya Zemlya to Franz Josef to Svalbard to NE Greenland. Also Canadian Archipelago, Baffin Bay... are excluded.

10120 is near PIOMAS values (2005: 10.209 12.892 2006: 9.838 12.303 2007: 7.135 10.472 average 10.475). After deducting remaining ice in Canadian Archipelago and Greenland sea, IceSat is probably a little higher than PIOMAS.

Feb March 2007 per PIOMAS of 20.863 23.031 and higher in 2005 and 2006 obviously cannot be compared with 13250 without deducting lots of volume from PIOMAS figures for the excluded areas.

.

The difference between 10120 and 13250 seems small, but given other comparisons of ICESat to PIOMAS, perhaps PIOMAS difference is even smaller for this limited region.

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #143 on: April 17, 2015, 03:36:10 PM »
This February, Cryosat saw average sea-ice floe thicknesses of just over 1.7m, giving a volume across the Arctic of nearly 24,000 cubic km. Back in the winter of 2013, following strong melting during the previous summer, floe thicknesses averaged 1.5m and the volume fell below 21,000 cu km.

http://www.bbc.co.uk/news/science-environment-32348291

PIOMAS has

2015  32  20.348
2015  60  22.524
avg 21.502

2013 avg Feb 19.376
so Cryosat continues to have more volume in winter.

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #144 on: April 17, 2015, 03:48:36 PM »
And to mark the spacecraft's fifth birthday in orbit, the team is switching on a new, near-real-time service to aid science and maritime activities.


Sea Ice thickness products from CryoSat NRT (near real time) released. These include 2, 14, and 28-day Arctic maps and thickness timeseries from Oct-2010 until the latest available Cryosat NRT products.

Monthly thickness and volume maps from the archive (2010-2015) will be released in the next few months.


http://www.cpom.ucl.ac.uk/csopr/index.html
http://www.cpom.ucl.ac.uk/csopr/seaice.html


Siffy

  • ASIF Citizen
  • Posts: 178
    • View Profile
Re: Piomas vs. CryoSat
« Reply #145 on: April 17, 2015, 04:11:54 PM »
And to mark the spacecraft's fifth birthday in orbit, the team is switching on a new, near-real-time service to aid science and maritime activities.


Sea Ice thickness products from CryoSat NRT (near real time) released. These include 2, 14, and 28-day Arctic maps and thickness timeseries from Oct-2010 until the latest available Cryosat NRT products.

Monthly thickness and volume maps from the archive (2010-2015) will be released in the next few months.


http://www.cpom.ucl.ac.uk/csopr/index.html
http://www.cpom.ucl.ac.uk/csopr/seaice.html




Odd how the ice in the Barrents is a metre thick even on the ice edge, you'd expect or at least I expected some kind of tapering off of ice thickness.

crandles

  • ASIF Upper Class
  • Posts: 1594
    • View Profile
Re: Piomas vs. CryoSat
« Reply #146 on: April 17, 2015, 05:48:21 PM »

Odd how the ice in the Barrents is a metre thick even on the ice edge, you'd expect or at least I expected some kind of tapering off of ice thickness.

I am suspecting that some areas shown as white are too thin to reliably measure as well as possibly also being no measurements or no ice. Even so, there is a lot of green and little blue in Southern Baffin Bay as well.

Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 3272
    • View Profile
    • Arctic Sea Ice Blog
Re: Piomas vs. CryoSat
« Reply #147 on: April 17, 2015, 06:25:41 PM »


Compared to SMOS:

Il faut cultiver notre jardin

Wipneus

  • ASIF Governor
  • Posts: 3147
    • View Profile
    • Arctische Pinguin
Re: Piomas vs. CryoSat
« Reply #148 on: April 17, 2015, 07:16:21 PM »
Just posting this here, not completely on-topic but I might add Cryosat later.

This is AMSR2 volume, compared with PIOMAS, calculated from the ice thickness maps in the ADS Sea Ice monitor. Inverting the colorscale for thickness and calculating the area of each pixel (for a 10 km polar stereo graphic projection)  volume can be calculated.

I am ignoring:
-pole hole
-fraction of melting ice
-difference between PIOMAS domain and AMSR2 domain (I am using all the ice in the image).

There is a warning that the AMSR2 thickness is not reliable during the melting season, but at least we can say that the results are not all that crazy.




ghoti

  • ASIF Citizen
  • Posts: 445
    • View Profile
Re: Piomas vs. CryoSat
« Reply #149 on: April 17, 2015, 07:18:27 PM »

Odd how the ice in the Barrents is a metre thick even on the ice edge, you'd expect or at least I expected some kind of tapering off of ice thickness.

If you go to the map page and click on a spot you get a more zoomed in map and a chart of the data for the specific point you clicked. When you do this it looks like you get a much larger range of thicknesses. I guess the overview map is "averaging" the colours and thicknesses. The 1 meter thick ice along the edge of the Barents is shown as being between 0.5m and 1.5m.