Support the Arctic Sea Ice Forum and Blog

Author Topic: SMOS  (Read 13526 times)

Sterks

  • ASIF Middle Class
  • Posts: 537
  • Member # 1000
    • View Profile
  • Liked: 76
  • Likes Given: 46
Re: SMOS
« Reply #100 on: June 10, 2019, 01:51:07 PM »
Bumping this live plot,  one click less away. Thank you Steven.
In summer SMOS is sensitive to surface melting.


https://www.dropbox.com/s/fl2xs6aeop3ioen/SMOS_beige_pixels.png
« Last Edit: June 15, 2019, 01:02:33 PM by Sterks »

Steven

  • ASIF Middle Class
  • Posts: 514
    • View Profile
  • Liked: 98
  • Likes Given: 15
Re: SMOS
« Reply #101 on: June 18, 2019, 11:57:45 PM »
The number of beige pixels for 2019 has been slightly below 2012 in the last 3 days.  But it may also be interesting to have a look at the other colors in the SMOS images.  Note that 2019 currently has relatively many of the purple/violet/red/orange-ish pixels, especially when compared to 2012.  To quantify this, I run a pixel-counting algorithm for several color segments of the SMOS color legend, for the date 17 June, see the bottom chart below (blue pixels are ignored, and I made some heuristic corrections for data gaps for the years 2010 and 2011).






Rich

  • ASIF Citizen
  • Posts: 427
    • View Profile
  • Liked: 71
  • Likes Given: 38
Re: SMOS
« Reply #102 on: June 19, 2019, 02:43:16 AM »
The number of beige pixels for 2019 has been slightly below 2012 in the last 3 days.  But it may also be interesting to have a look at the other colors in the SMOS images.  Note that 2019 currently has relatively many of the purple/violet/red/orange-ish pixels, especially when compared to 2012.  To quantify this, I run a pixel-counting algorithm for several color segments of the SMOS color legend, for the date 17 June, see the bottom chart below (blue pixels are ignored, and I made some heuristic corrections for data gaps for the years 2010 and 2011).






This is extremely cool. Have you considered running a simulated volume with this?

Simply multiply the pixel count by my implied thickness and area. For the beige, make an assumption of average thickness (perhaps 75-100 cm).

You're making a good point here that all of pixel colors matter.... it's not just beige v. other.

oren

  • ASIF Governor
  • Posts: 3918
    • View Profile
  • Liked: 539
  • Likes Given: 1113
Re: SMOS
« Reply #103 on: June 19, 2019, 02:59:26 AM »
Steven thanks for your analysis of SMOS chart (and for the automated dropbox link for the "beige index"). I agree an aggregate SMOS index taking all the colors with some kind of weighting could be quite useful in summer. Just don't call it volume, as it doesn't represent thickness at this stage, but mostly surface wetness.

Rich

  • ASIF Citizen
  • Posts: 427
    • View Profile
  • Liked: 71
  • Likes Given: 38
Re: SMOS
« Reply #104 on: June 19, 2019, 03:41:20 AM »
Just don't call it volume, as it doesn't represent thickness at this stage, but mostly surface wetness.

I respectfully disagree with this point Oren.

The sensors are still picking up enough information that allow it to assign a thickness to the ice which conforms to a common sense understanding of where we expect the thickness to be.

If it was simply measuring a binary state of wet v. dry, the map would only have two colors.

The presence of water on the surface might impair the precision of the thickness data, but there is still something functional going on there which allows the sensor to assign a thickness output resembling reality.

It's a tall order to ask, but if someone wanted to reconcile the SMOS pixel counts with past PIOMAS data, I'm guessing they could find a good fit.
« Last Edit: June 19, 2019, 04:34:02 AM by Rich »

Tor Bejnar

  • ASIF Governor
  • Posts: 2776
    • View Profile
  • Liked: 189
  • Likes Given: 115
Re: SMOS
« Reply #105 on: June 19, 2019, 05:15:58 AM »
I imagine (= guess - based on what I've been hearing for years) when the sensor responds with "purple", it is sensing an area that has more dry spots than when it responds with "red", and has 100% nothing to do with ice thickness (except for where it is sensing actually dry thin ice).
Arctic ice is healthy for children and other living things.

oren

  • ASIF Governor
  • Posts: 3918
    • View Profile
  • Liked: 539
  • Likes Given: 1113
Re: SMOS
« Reply #106 on: June 19, 2019, 05:36:43 AM »
Check out this post by A-Team in the Test Space thread for some more information.
https://forum.arctic-sea-ice.net/index.php/topic,2558.msg206386.html#msg206386
Quote
Smos and Smos-Smap can't help but still report ice thinness but that is overlain -- and overwhelmed -- by other factors contributing dielectric. Note the color sequence in the palette is maintained south to north, with the periphery 'thinner'. This does not agree well with melt pond distribution (blueish in WorldView) because a number of other factors contribute to the overall artifact picture, such as ice surface roughness and floe vs open water distribution relative to instrument ground resolution.

Both are passive satellite instruments that do not use altimeters. In winter, they measure the change in stokes parameters as emitted blackbody radiation from the upper cms of ocean water passes through the anisotropic ice above. Like so many hand-me-down Arctic tools, Smos and Smap have been re-purposed from their originalmission design which required a near-polar orbit and so provided accidental coverage of the Arctic Ocean.

Smos and Smap in summer have been wrongly written off in journals. There is still information there if you 'let go' of ice thinness; inter-year comparisons must indicate something (but what?). A glance at the non-beige colors above the CAA (where the ice is entirely >0.5m) shows an interesting correlation with a dramatic shift in Ascat brightness. So it is an interpretive matter of what the colors are telling us.

The gif investigates an overlay of SMOS and high-resolution WV visible on a clear day. Again, it is not clear what visible features on the ice surface correlate with Smos and Smap map colors.

Rich

  • ASIF Citizen
  • Posts: 427
    • View Profile
  • Liked: 71
  • Likes Given: 38
Re: SMOS
« Reply #107 on: June 19, 2019, 05:58:28 AM »
I imagine (= guess - based on what I've been hearing for years) when the sensor responds with "purple", it is sensing an area that has more dry spots than when it responds with "red", and has 100% nothing to do with ice thickness (except for where it is sensing actually dry thin ice).

Studying the map closely reveals that the story you imagine is probably wrong.

The pixels are arranged in a continuum according to thickness.

They run from thinnest to thickest as follows

dark blue
light blue
light green
dark green
yellow
orange
red
purple
beige

The pattern on the map reveals no deviation in the continuum. You don't see any purple areas adjacent to any green areas. Yellow always has green on one side and orange / red on the other. This is exactly what one  would expect to see with a body of ice that is thickest in the center and melting from the shallow sections toward the thick sections.
.
It is unrealistic to expect that surface moisture would appear along a continuum like that. .

This appears to be an ASIF legend that has gotten traction over the years. I'm challenging the legend.

binntho

  • ASIF Citizen
  • Posts: 459
    • View Profile
  • Liked: 121
  • Likes Given: 46
Re: SMOS
« Reply #108 on: June 19, 2019, 07:01:15 AM »
This is exactly what one  would expect to see with a body of ice that is thickest in the center and melting from the shallow sections toward the thick sections.

Sounds reasonable. But the thickness distribution that we "know" from products like Piomas does not match the color distribution of the SMOS graph.

It seems to me that the more reasonable explanation for the colour gradiation is that hot air flows from one area to another, and cools down as it reaches the coldest parts (in the center). Surface melts then mostly follows the temperature of the hot air.

This can be seen happening very clearly north of CAA and Greenland where we know that the thickest ice is.

oren

  • ASIF Governor
  • Posts: 3918
    • View Profile
  • Liked: 539
  • Likes Given: 1113
Re: SMOS
« Reply #109 on: June 19, 2019, 07:49:27 AM »
I recommend to read
https://seaice.uni-bremen.de/thin-ice-thickness/
Quote
SMOS
The thickness of thin sea ice (SIT) is daily retrieved from observations of the L-band microwave sensor SMOS (Soil Moisture and Ocean Salinity). Horizontal and vertical polarized brightness temperatures in the incidence angle range of 40° to 50° are averaged. The ice thickness is then inferred from the polarization difference and the intensity using an empirical method (Huntemann et al., 2014).

Thin sea ice occurs during the freezing season. In the melting season, the thickness of sea ice is highly variable and the emission properties in the microwave change due to the wetness of the surface and occurrence of melt ponds in the Arctic. Therefore, thickness data are calculated only during the freezing season, that is from October to April in the Arctic and from March to September in the Antarctic. During the melting season, the procedure does not yield meaningful results.

As the resolution of SMOS at the used incidence angle range is about 40 km, only larger regions of thin ice will be retrieved correctly. The rim of thin ice shown in many cases not necessarily indicates thin ice, but can also be caused by the smearing effect (convolution) of the low resolution.

Each day of ice thickness data product are calculated twice to ensure that all swath files were available in the archived product. First processing is done directly on the next day with only about 7 hours delay. At this time it can happen that not all swath files are available and another processing of the same day is initiated 23 hours later. In more than 50% of the time the first processing does not include all swath but usually provides sufficient coverage for Arctic and Antarctic regions.

This service has been developed in the framework of the EU project SIDARUS. After completion of the SIDARUS project end 2013, the service is continued on a best effort base in the context of the Polar View and of the Arctic Regional Ocean Observing System (Arctic ROOS).

Other useful links which I haven't read in full yet:
https://icdc.cen.uni-hamburg.de/1/daten/cryosphere/l3c-smos-sit.html
Uni Hamburg SMOS page.

http://seaice.de/Kaleschke_2016.pdf
SMOS sea ice product: Operational application and validation in the
Barents Sea marginal ice zone

https://www.the-cryosphere.net/13/675/2019/tc-13-675-2019.html
Combined SMAP–SMOS thin sea ice thickness retrieval


Rich

  • ASIF Citizen
  • Posts: 427
    • View Profile
  • Liked: 71
  • Likes Given: 38
Re: SMOS
« Reply #110 on: June 19, 2019, 08:46:02 AM »
This is exactly what one  would expect to see with a body of ice that is thickest in the center and melting from the shallow sections toward the thick sections.

Sounds reasonable. But the thickness distribution that we "know" from products like Piomas does not match the color distribution of the SMOS graph.

It seems to me that the more reasonable explanation for the colour gradiation is that hot air flows from one area to another, and cools down as it reaches the coldest parts (in the center). Surface melts then mostly follows the temperature of the hot air.

This can be seen happening very clearly north of CAA and Greenland where we know that the thickest ice is.

If you want to prove your point, it might be better to out up a current PIOMAS chart next to the current SMOS chart and identify areas which are different.

Comparing May 2018 PIOMAS to current SMOS isn't very scientific.


Rich

  • ASIF Citizen
  • Posts: 427
    • View Profile
  • Liked: 71
  • Likes Given: 38
Re: SMOS
« Reply #111 on: June 19, 2019, 09:02:52 AM »
Thank you Oren for providing the quotes from A-Team and the links.

The interpretations of those is vague.

I have acknowledged that the surface moisture impairs the thickness measurement which may reduce the precision to a point where the folks in Bremen may not consider it meaningful.

That doesn't mean that it still isn't producing an output which is representative of thickness.

Neven

  • Administrator
  • ASIF Royalty
  • *****
  • Posts: 6955
    • View Profile
    • Arctic Sea Ice Blog
  • Liked: 510
  • Likes Given: 355
Re: SMOS
« Reply #112 on: June 19, 2019, 10:26:11 AM »
If you want to prove your point, it might be better to out up a current PIOMAS chart next to the current SMOS chart and identify areas which are different.

Comparing May 2018 PIOMAS to current SMOS isn't very scientific.

Coincidentally, Wipneus just posted a mid-month PIOMAS update.
Il faut comparer, comparer, comparer, et cultiver notre jardin

Tor Bejnar

  • ASIF Governor
  • Posts: 2776
    • View Profile
  • Liked: 189
  • Likes Given: 115
Re: SMOS
« Reply #113 on: June 19, 2019, 06:38:47 PM »
I imagine (= guess - based on what I've been hearing for years) when the sensor responds with "purple", it is sensing an area that has more dry spots than when it responds with "red", and has 100% nothing to do with ice thickness (except for where it is sensing actually dry thin ice).

Studying the map closely reveals that the story you imagine is probably wrong.

The pixels are arranged in a continuum according to thickness.

The pattern on the map reveals no deviation in the continuum. You don't see any purple areas adjacent to any green areas. Yellow always has green on one side and orange / red on the other. This is exactly what one  would expect to see with a body of ice that is thickest in the center and melting from the shallow sections toward the thick sections.
.
It is unrealistic to expect that surface moisture would appear along a continuum like that. .

This appears to be an ASIF legend that has gotten traction over the years. I'm challenging the legend.
I think I am probably approximately correct, still.  From SMOS (via Oren)
Quote
In the melting season, the thickness of sea ice is highly variable and the emission properties in the microwave change due to the wetness of the surface and occurrence of melt ponds in the Arctic.
My experience of drawing topographic maps from surveying data tells me that a map with a 40 km grid with points of presumed thickness (or wetness) across it can have a topographic-like map drawn over it, creating continuous lines of demarcation between areas of equal thickness (wetness).

Old fashioned reason tells me that the juxtaposition of the thickest ice on the Arctic Ocean (residing perennially in the area north of the CAA) with a zone of SMOS-determined thin ice (presuming non-beige color means variations on thin) is not tenable.  Given that the publishers of the SMOS data declare their data is not useful for thickness determination during the melting season, I can conclude that the colored zones do not describe any aspect of thinness when the surface has degrees of wetness.

This does not mean I am right.  But you are probably wrong.  The ASIF legend lives a little longer!  :)
« Last Edit: June 21, 2019, 04:28:39 PM by Tor Bejnar »
Arctic ice is healthy for children and other living things.

lurkalot

  • ASIF Lurker
  • Posts: 26
    • View Profile
  • Liked: 0
  • Likes Given: 0
Re: SMOS
« Reply #114 on: June 21, 2019, 01:51:58 PM »
A large proportion of Arctic ice is no more than 2m thick, we are told. This would have a freeboard of around 25cm, which ought to be reachable by any self-respecting wave in the vicinity. Is SMOS capable of distinguishing between ice that is wet from melting, from waves or indeed from rain?

oren

  • ASIF Governor
  • Posts: 3918
    • View Profile
  • Liked: 539
  • Likes Given: 1113
Re: SMOS
« Reply #115 on: June 21, 2019, 02:31:38 PM »
As floes can easily be hundreds of meters across, and some reach tens of kilometers across, the effect of wave-induced wetness should be negligible.

Tor Bejnar

  • ASIF Governor
  • Posts: 2776
    • View Profile
  • Liked: 189
  • Likes Given: 115
Re: SMOS
« Reply #116 on: June 21, 2019, 04:51:28 PM »
I know little about SMOS, but their sensors do detect "salinity".  So conceptually, it would register surface melt differently from a salt water wash.  However, SMOS reports on a 35 or 40 km grid (I've read both; I presume it is one or the other, unless it is latitude dependent), so the sensors-to-data-output-system is doing a great deal of averaging.  As open water is very wet, as soon as SMOS looks at floes that are 'small' (and not perfectly packed together), it starts averaging (convoluting) thickness with wetness and the thickness output becomes seriously/totally irrelevant (per SMOS literature).  As Oren suggested, a salt water wash will occur mostly along the edges of floes, so the salinity SMOS detects from open water might be very slightly exacerbated by salty water on a floe.  I conclude with Oren, "negligible", and add "irrelevant after review."
Arctic ice is healthy for children and other living things.

Steven

  • ASIF Middle Class
  • Posts: 514
    • View Profile
  • Liked: 98
  • Likes Given: 15
Re: SMOS
« Reply #117 on: June 21, 2019, 09:19:27 PM »
an aggregate SMOS index taking all the colors with some kind of weighting could be quite useful in summer

Below is a first version.  In this graph, I weighted the pixels according to their numerical value on the color legend of the SMOS maps.  So the non-beige pixels are weighted by a factor between 0 (for dark blue pixels) and 0.5 (for light purple pixels).  For the beige pixels I used a weighting factor 0.6 for now.  That is somewhat arbitrary, but I'm not sure if there is a better choice.

I excluded 2010 and 2011 from this graph for now, since I haven't automated the correction of data gaps for those 2 years.  (For the beige pixel graph, I did some manual corrections for that, but that would be more complicated for the weighted version).  Perhaps I'll include those 2 years later if I have some time.


prokaryotes

  • ASIF Citizen
  • Posts: 191
    • View Profile
    • ClimateState.com
  • Liked: 14
  • Likes Given: 18
Re: SMOS
« Reply #118 on: June 21, 2019, 11:03:38 PM »
Just curious, but can this pickup on flaw polynyas?

Quote
Areas of flaw polynyas in the ESAS increased dramatically (by up to five times) during the last decades, and now exceed the total area of Siberian wetlands.
Link
CLIMATE STATE WEBSITE | YOUTUBE | USCREEN

oren

  • ASIF Governor
  • Posts: 3918
    • View Profile
  • Liked: 539
  • Likes Given: 1113
Re: SMOS
« Reply #119 on: June 22, 2019, 12:07:57 AM »
Thank you very much Steven.
I think the beige pixels should carry a larger weight, I would put it at around 0.75. Beyond that, the other factors make sense and the graph is useful as a complement to the beige-only graph.
Is there any way for you to generate a permanent link for this one as well?

Tor Bejnar

  • ASIF Governor
  • Posts: 2776
    • View Profile
  • Liked: 189
  • Likes Given: 115
Re: SMOS
« Reply #120 on: June 22, 2019, 12:24:40 AM »
I don't recall ever seeing the term "flaw polynya" before, and it took some searching to find a definition, finally (from here):
Quote
flaw polynyas (band-like ice-free areas), which form
simultaneously with land-fast ice in November. Flaw polynyas
reach tens of kilometres in width and migrate out of fast ice
hundreds of kilometres northward (Smolyanitsky et al., 2003),
and here
Quote
A polynya is defined as any nonlinear-shaped area of open water and/or sea ice cover < 30 cm thick enclosed by a much thicker ice cover (WMO 1970). It can be restricted on one side by a coast, terrned shore polynyas, or bounded by fast ice, termed flaw polynyas.
Arctic ice is healthy for children and other living things.

Steven

  • ASIF Middle Class
  • Posts: 514
    • View Profile
  • Liked: 98
  • Likes Given: 15
Re: SMOS
« Reply #121 on: June 22, 2019, 12:22:30 PM »
I think the beige pixels should carry a larger weight, I would put it at around 0.75. Beyond that, the other factors make sense and the graph is useful as a complement to the beige-only graph.
Is there any way for you to generate a permanent link for this one as well?

The link is here:
https://www.dropbox.com/s/97j1gbr7zdostz8/smos_weighted.png?dl=0

I guess the weighting for beige pixels would depend on the time of year.  A large contiguous group of beige pixels would probably have higher numerical values in the middle of the group than at the edges.  Anyway there isn't much beige left at this time of year so they get less important for the calculation.

P.S.  To illustrate the problems with 2010 and 2011, see e.g. this image for 2010.  The pole hole for those 2 years is larger than for the years 2012-2019.  Moreover 2010 and 2011 have some irregular data gaps at other places (e.g. Greenland Sea) which vary from day to day.

Steven

  • ASIF Middle Class
  • Posts: 514
    • View Profile
  • Liked: 98
  • Likes Given: 15
Re: SMOS
« Reply #122 on: June 30, 2019, 02:47:38 PM »
I excluded 2010 and 2011 from this graph for now, since I haven't automated the correction of data gaps for those 2 years

Update: I made some improvements to the pixel counting algorithm and included 2010 and 2011 in the graph.  Note that those two years were very low at this time of year.  That is consistent with some other datasets that are also showing strong surface melting in 2010/2011 at the end of June, see e.g. Worldview bands 7-2-1 or NSIDC sea ice area, where 2010 was also the lowest on record in the past few days.


https://www.dropbox.com/s/97j1gbr7zdostz8/smos_weighted.png