Support the Arctic Sea Ice Forum and Blog

Author Topic: Home brew AMSR2 extent & area calculation  (Read 1573217 times)

A-Team

  • Young ice
  • Posts: 2739
    • View Profile
  • Liked: 685
  • Likes Given: 34
Re: Home brew AMSR2 extent & area calculation
« Reply #3550 on: September 29, 2020, 09:17:35 PM »
Forums focus so much on measuring ice area that it's easy to forget sea ice loss is open sea gain. Open water has drastically lower albedo than all the variants of ice (Perovich 2011: snow, rain-on-snow, melt ponds, age etc) so in terms of oceanic heat gain from solar insolation, this gain is highly significant to the extent it matches the insolation season.

The flip side is less solar reflected back out to space, a previous service provided by Arctic Ocean  ice that moderated climate. (Old-timers recall when the Barents, Bering, Greenland and Baffin seas had late season reflective ice too, not just the inner basin.)

This year, open water took up 55% of the surface area of the Arctic Ocean basin by mid-September (based on pixel counts of 0% sea ice concentration on AMSR2_UHH adjusted for conformal projection). That percentage would be higher classifying pixels in the 0-20% range as open water but not by much (~2%). The one million sq km version of BOE has ~85% open water at end season. BOEs can only be understood in their whole-season context, below.

Historically, the match has been very poor so decay of the 'planetary refrigerator' reflectance out to space was somewhat a non-issue. There was not that much open sea area and it formed late in the melt season long after sunshine had peaked. Under incipient BOE however, the lag and mismatch will be dissipating.

The double graph below shows open sea areas on days spaced over the 2020 melt season overlain on the (fixed) surface solar radiation budget (neglecting clouds) restricted to the Arctic Ocean basin. (These years, there's a surprising amount of open water even before the summer solstice. Peak open water in mid-Sept is still quite distant from peak insolation and lingering persistence into October would have no effect on total solar heat seasonal adsorption.)

However solar insolation at 75ºN continues a long ways into the modern melt season at the lower latitudes that matter -- those with the most and earliest open water. Below, the weekly area of open sea can be weighted by its week of solar input on out to the right (ie relevant area under the curve) to see the overall affect which is to disproportionately lift up the left end of the open water graph.

For example, a parcel of open water in back in late May may be small in area but it will be taking in some 16 weeks of insolation at 90+% efficiency including the peak solstice and so exceed a much larger parcel only opening up in mid-August. Sunlight converted to heat will be redistributed in various ways but once converted that incident sunlight cannot be reflected back out to space the way it once was (from higher albedo ice).

Pistone 2019 and Guarino 2020 fast-forward the quantitation (clouds included) to a full-on ever-earlier BOE with catastrophic impacts on climate change amelioration efforts.

Reposting another version of the albedo graphical calculator that has components for ice at various stages in albedo evolution and also for permafrost land. These use 1º latitude bands restricted to surface type; in the limit, band width could go to zero for an integral. The satellite data comes in polar stereographic EPSG 3413 (ofter without netCDF) which only has equal area pixels at 70º.

However pixel counts can be converted to accurate areas by a small latitude-dependent adjustment reaching only 6% at the pole. Thus the first terms of the taylor series expansion of the trig function suffice given the errors elsewhere. For a proper treatment of grid choices, see:

 A Comparison Study of Three Polar Grids
TC Chen et al
https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1087&context=ge_at_pubs

The most commonly used reanalyses generated by global data assimilation systems of various operational centers [eg ERA-40) are archived on the rectangular latitude–longitude (EL) grid with a 2.5° latitude  2.5° longitude resolution. Because the radius of latitude circles is proportional to the cosine of the latitude, the circumference of latitude circles reduces toward the Poles. For example, from 47° to 70°N, the circumference of a latitude circle is almost halved.

Meteorological data are often presented using an EL grid on a polar stereographic projection such as EPSG 3413. The analysis quality of some meteorological fields on this grid, such as wind vectors, may be degraded by the convergence of analysis points near the Poles. The closer analysis points allow smaller scales to be resolved, but there is no corresponding increase in constraining observations, so that the added smaller scales may not be weather/climate signals.
« Last Edit: September 29, 2020, 10:20:21 PM by A-Team »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3551 on: October 01, 2020, 01:59:18 PM »
daily ImageJ pixel count on v103, sep4-30. Might be interesting to run it on Beaufort tail only

A-Team

  • Young ice
  • Posts: 2739
    • View Profile
  • Liked: 685
  • Likes Given: 34
Re: Home brew AMSR2 extent & area calculation
« Reply #3552 on: October 01, 2020, 04:55:55 PM »
Nice combo! Might think about cropping off lower Greenland, looks like there is room for 2x the inset resolution. That consistent spike at 50% suggests a round-off bias. Or, if you are taking pixel counts off the grayscale, that conversion could also do it if there is a palette collision on fairly uncommon concentration classes.

Found some more open water papers. The first distinguishes between 'ice albedo effect' (much discussed on forums) and 'ice-ocean albedo effect' (my current interest) which focuses on dominant lower albedo of open water and the non-radiation of sunlight back out to space.

Cumulative solar adsorption by a small patch of open water on May 20th in the Chukchi really starts to add up after fourteen weeks (September), as do ice-free Bering Kara and Barents seas which also affect the Arctic basin, compared to a big patch opening up in mid-August at 80ºN which barely gets a couple of weeks of post peak solstice insolation. (Multiply area under the surface insolation curve by sq km of open water to get 20:1 effect.)

Evidence for ice-ocean albedo feedback in the Arctic Ocean shifting to a seasonal ice zone
H Kashiwase, K Ohshima, S Nihashi & H Eicken
Scientific Reports v7 8170 (2017)
https://www.nature.com/articles/s41598-017-08467-z

"Ice-albedo feedback due to the albedo contrast between water and ice is a major factor in seasonal sea ice retreat, and has received increasing attention with the Arctic Ocean shifting to a seasonal ice cover. However, quantitative evaluation of such feedbacks is still insufficient.

"Here we provide quantitative evidence that heat input through the open water fraction is the primary driver of seasonal and inter-annual variations in Arctic sea ice retreat. Divergent ice motion in early melt season triggers large-scale feedback which subsequently amplifies summer sea ice anomalies; divergence has doubled since 2000 due to a more mobile ice cover.

"Until recently, the Arctic Ocean has been characterized by a thick multiyear ice cover that persisted throughout the summer, with melt confined to its upper surface. In the seasonal ice zone, presence of an open water fraction with a much lower albedo results in high solar radiation absorption by the upper ocean which in turn serves as the dominant heat source for sea ice lateral and bottom melt. Since the seasonal ice zone is dominated by thin and undeformed first-year ice, the melting of sea ice immediately increases the fraction of open water in the ice-covered area and thus ice-ocean albedo feedback drives up absorption of solar energy in the upper ocean.

"Here we show the dominance of heat input through the open water fraction on sea ice loss We selected the Pacific Arctic Sector (fan-shaped area in figure) as the main study area. This region experienced the largest reductions in summer ice extent and volume anywhere in the Arctic Ocean beginning in the 2000s. Inter-annual variation of ice retreat in this region explains about 86% of the variance over the entire Arctic Ocean (p < 0.001).

Dominance of heat input through the open water fraction: "For the ice-covered area defined by ice concentrations >15%, we have analyzed the daily heat budget separately for the water and ice surfaces from 1979 to 2014. During the summer season, net heat flux at the water surface is much larger than that at the ice surface because shortwave radiation is the dominant component of heat budget in the analysis area…

"The fraction of multiyear ice based on ice age data has decreased from 49 to 31%. This reduction affects sea ice dynamics, in particular through decreases in ice mechanical strength and internal ice interaction forces, and increases in ice deformation rates

"Other factors such as changes in atmospheric circulation patterns, influence of cloud cover, long wave radiative forcing due to anthropogenic CO2 emission, melt pond distribution in the early summer season, release of the solar heat stored in a near-surface layer of the ocean, and increases in the heat inflow through Bering Strait may also contribute to drastic ice reductions. However, we note that these factors are intrinsically linked to divergence in the ice pack, because increased heat input from any source may enhance sea ice mobility.

https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2015JD023712

Yet another reason not to use DMI 80: "The European ERA‐Interim reanalysis data and NCEP‐CFSv2 analysis data are used to investigate the surface energy budget and atmospheric conditions in 2012 and 2013. The ERA‐Interim reanalysis data are the latest global atmospheric reanalysis produced by the ECMWF is available from 1989 onward The ERA‐Interim reanalysis routines provide major improvements compared with ERA‐40 such as better vertical consistency of the air temperature in the Arctic region and an improved representation of the hydrological cycle.

"The open water fraction is typically small (often less than 10%), but the fact that open water absorbs 2 to 3 times as much broadband solar radiation as bare ice means that these small areas can contribute significantly to the overall uptake of solar radiation in the region. Changes that increase the amount of melt, such as an earlier melt onset, will lead to thinner ice that can more easily allow the dynamic formation of leads or be completely melted through.

"The conditions in 2012 exhibited a longer and more continuous period of ice and snowmelt, with earlier melt onset and later freeze‐up than in 2013, resulting in more ice melt in 2012 than in 2013 [see Perovich  2014a].

"The earlier melt onset in 2012 likely preconditioned the system to allow a longer melt season with lower albedo, resulting in much more solar heat input to the ice‐ocean system in 2012 than in 2013. This additional deposited solar energy would melt the surface, thin the ice, and warm the upper ocean, resulting in more melting and longer melt period. This enhances the positive ice‐albedo feedback."
« Last Edit: October 01, 2020, 09:07:40 PM by A-Team »

gerontocrat

  • First-year ice
  • Posts: 9665
    • View Profile
  • Liked: 3812
  • Likes Given: 30
Re: Home brew AMSR2 extent & area calculation
« Reply #3553 on: October 02, 2020, 03:21:41 PM »
Found some more open water papers. The first distinguishes between 'ice albedo effect' (much discussed on forums) and 'ice-ocean albedo effect' (my current interest) which focuses on dominant lower albedo of open water and the non-radiation of sunlight back out to space.
I attach some examples of open water graphs. You can see the gradual developemnt of earlier open water over the years both through less winter ice and earlier melt (the purple line of May-June-July open water).

I also chose the seas from the Barents to the ESS that seem to demonstrate the eastward progression of Atlantification
"Para a Causa do Povo a Luta Continua!"
"And that's all I'm going to say about that". Forrest Gump
"Damn, I wanted to see what happened next" (Epitaph)

A-Team

  • Young ice
  • Posts: 2739
    • View Profile
  • Liked: 685
  • Likes Given: 34
Re: Home brew AMSR2 extent & area calculation
« Reply #3554 on: October 02, 2020, 10:44:47 PM »
Quote
You can see the gradual developemnt of earlier open water over the years both through less winter ice and earlier melt (the purple line of May-June-July open water).
Interesting! Trends toward earlier rain-on-snow, earlier melt ponds and earlier open water will have consequences. (Melt ponds 'don't count' here as their lower but still high albedo is largely determined by their cupping ice -- Perovich 2007.)

The date of first appearance and the latitude involved are quite important in determining the cumulative effect over the insolation season on loss of reflection of incoming solar back out to space (respectively adsorption of that heat in the upper water column, which might be upper upper given an algal bloom).

Using the graphs in #3550 and #3552, the data below pushes a parcel of water from its origin date forward through the end of melt season, adding up all the energy input as the insolation season progresses (discrete integral, area (t) under the time-forward surface insolation curve for 80ºN).

Note 75ºN would make a better choice for mean reference latitude; open water is assumed to stay open water for the season rather than be covered by re-freeze, ice drift or dispersed.

While the open water area is very low during late May, it has many weeks to take in insolation thus making it a larger contributor than say a larger area that opens in early August. For example, the chart below calculates a factor of 4.48x. So 100,000 sq km on the earlier date amounts to 448,000 sq km on the later.

A mid-September BOE would have almost no effect on solar radiation reflected back into space despite it being very important to many other considerations.

The bars in the blue graph in #3550 (areas of open water by date) should thus be adjusted in height according to the multiplier column in the table below to get the insolation season match adjustment (offset red bar graph).
« Last Edit: October 03, 2020, 01:32:33 PM by A-Team »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3555 on: October 06, 2020, 01:39:26 PM »
Daily pixel count of the Beaufort Sea area, sep4-oct5

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3556 on: October 07, 2020, 12:12:59 PM »
A quick look in more detail at refreeze in the Beaufort/southern CAB. The v103 presentation shows some nice detail suggesting a weather front passing over the area.
v103, sep30-oct6

comparison of UHH and v103, oct4 (click for correct resolution)

polarview S1A of the same area, oct6 (click twice for full res)
« Last Edit: October 07, 2020, 04:28:10 PM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3557 on: October 11, 2020, 08:23:00 PM »
No data for 2012 but 2020 is about to test the boundary set by 2018 for slow refreeze in the CAB.

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3558 on: October 15, 2020, 12:42:21 AM »
2020 again setting a new boundary for CAB refreeze (wipneus' CAB)
updated below
« Last Edit: October 15, 2020, 11:06:23 PM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3559 on: October 15, 2020, 11:06:53 PM »
CAB extent update
« Last Edit: October 17, 2020, 12:54:47 AM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3560 on: October 17, 2020, 12:55:53 AM »
updated below
« Last Edit: October 18, 2020, 12:16:47 AM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3561 on: October 18, 2020, 12:17:30 AM »
CAB extent update.
Atlantic side, oct7-17
updated below
« Last Edit: October 18, 2020, 09:35:50 PM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3562 on: October 18, 2020, 09:36:43 PM »
CAB extent update.
Atlantic side, oct8-18
added updated pixel count
« Last Edit: October 18, 2020, 11:23:02 PM by uniquorn »

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3563 on: October 24, 2020, 12:04:42 AM »
atlantic side oct22-23
wipneus' CAB extent

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3564 on: October 26, 2020, 09:49:55 PM »
atlantic side oct25-26
CAB extent

sja45uk

  • New ice
  • Posts: 21
    • View Profile
  • Liked: 3
  • Likes Given: 17
Re: Home brew AMSR2 extent & area calculation
« Reply #3565 on: October 29, 2020, 10:18:57 AM »
Interesting! Trends toward earlier rain-on-snow, earlier melt ponds and earlier open water will have consequences. (Melt ponds 'don't count' here as their lower but still high albedo is largely determined by their cupping ice -- Perovich 2007.)
I struggled to understand 'don't count', and worry that the SHERPA data (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2000JC000449) may not reflect the current situation, as it seems to be almost 20 years old.

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3566 on: October 30, 2020, 12:16:15 AM »
amsr2 awi v103, oct28-29, a small move south into the Laptev, the sea ice that advanced west and south of SZ has mostly melted, Kara coastal ice hoping to take its place. Atl front retreating north of FJL, some drift towards N Svalbard
« Last Edit: October 30, 2020, 01:01:34 AM by uniquorn »

A-Team

  • Young ice
  • Posts: 2739
    • View Profile
  • Liked: 685
  • Likes Given: 34
Re: Home brew AMSR2 extent & area calculation
« Reply #3567 on: October 30, 2020, 03:42:42 PM »
Quote
what is meant above by 'melt ponds don't count; is SHEBA data 20 yrs old, still applicable?
The Surface Heat Budget of the Arctic Ocean (SHEBA) involved an icebreaker freezing itself into the icepack for over a year, from 02 Oct 1997 to 11 Oct 1998. The ice station drifted from its initial position at 75° N, 144° W to 80°N, 162°W, ie 555 km north but 462 km towards the ESS rather than the Fram.

The ice was much thicker then and the study did not have the extreme drift and floe disintegration problems that beset this year's Mosaic expedition being frozen outside the TransPolar Drift in the southeastern Laptev and the earlier effort off Svalbard called N-ICE2015.

Ice physics has not changed over the history of the universe but the earth's climate certainly has. So certain things studied by SHEBA like cloud cover, snow, ice thickness, ice age, timing of melt pond onset only provide an important baseline of change relative to changing conditions today whereas straight physics like radiative flux and melt pond albedo are the same (for a given ice class).

Then and now, sunlight on a melt pond is largely reflected back the ice remaining on the sides and bottom of the melt pond. Those vary with sun angle, geometry of the melt pond, its drainage status but not so much with ice thickness or ice age. Seawater in the pond does not notably scatter or absorb insolation in the visible where it peaks (algal blooms do not have time to establish).

Measured albedo of melt ponds is around 0.65 then and now, similar to its ambient surface ice whereas true open water is far less reflective, taking in 90% of the sunlight. It follows, since the Arctic basin exceeded 55% open water as percent of area, that to some zeroth order approximation, melt ponds can be thrown in with the ice (not counted) with the heat gain story emphasizing open water.

Radiative fluxes: longwave and shortwave
Heat flux: turbulent fluxes of latent and sensible heat
Cloud height, thickness and other properties
Energy exchange in the boundary layers of the atmosphere and ocean
Snow depth and ice thickness
Ocean salinity, temperature and currents

https://www.nsf.gov/pubs/2003/nsf03048/nsf03048_3.pdf
https://web.archive.org/web/20110726132916/http://www.crrel.usace.army.mil/sid/perovich/DKPpdf/SHEBABAMS.pdf
https://www.eol.ucar.edu/content/sheba-isff-flux-pam-project-report
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2001JC001079

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3568 on: November 05, 2020, 09:27:38 PM »
atlantic side wind driven retreat, as expected, nov4-5
CAB extent nov4. 2020 still setting a new boundary.

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3569 on: November 14, 2020, 10:36:43 PM »
2020 now testing the boundary set by 2016 for wipneus' CAB

uniquorn

  • Young ice
  • Posts: 2834
    • View Profile
  • Liked: 1293
  • Likes Given: 258
Re: Home brew AMSR2 extent & area calculation
« Reply #3570 on: November 23, 2020, 11:45:01 PM »
last chart for now. After setting new boundaries since the end of September, the 2020 CAB rejoins its forebears