Support the Arctic Sea Ice Forum and Blog

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Topics - Tealight

Pages: [1]
Arctic sea ice / Ice Melt AWP
« on: July 17, 2020, 04:00:23 AM »
A few people follow my NRT AWP site and use it as a predictor for future melt. The model though only calculates how much energy went into a grid cell and thus is only good to predict re-freeze. It does not give a good indication about the amount of ice that has melted. Ice-free regions can appear because the ice has melted, or the ice was transported away by winds and currents. If it gets compacted on the other side of the Arctic, it is harder to melt. A smaller area taken up by the ice means less sunlight is available to melt it. This phenomenon is already known in the compaction ratio (area/extent). During the summer, a low ratio will result in more melt.

The new Ice-melt AWP model better visualises this phenomenon by only accumulating AWP if the sea ice concentration (SIC) is above 25%. In case of an ice-free region driven by wind, the new model will remain at zero while the regular AWP model will accumulate huge amounts due to the low ocean albedo. In the Eastern Greenland Sea, the sea ice is continuously replenished and melting throughout the summer. On average this region melts the most ice.

Initially I intended to run this model just as described, but while the maps looked good the values for graphs were not helpful for determining strong melting seasons. Due to its size the central Arctic had by far the most accumulated AWP even though there is still little melt. Instead of an upwards trend in melting over the decades there was a significant downwards trend. This was not a model error. In the past there was simply more ice present to absorb more sunlight. Additionally, less greenhouse gases meant more of the absorbed energy was lost to space instead of melting ice.

To address these two issues, I decided to test the model with a “heat loss to space” component from my Sea Ice Forecast Model for the Sea Ice Prediction Network. The heat loss is 4-6 MJ/m2/day depending on a mean temperature and CO2 level. This level of heat loss is just a third of the measured outgoing longwave radiation for the Arctic.  Its purpose was to improve the forecast model, which is only a local energy model without any heat transfer from lower latitudes. Since the Ice Melt AWP model does not include any heat transfer either it seemed appropriate to use the same level of intensity.

The heat loss adjustment per CO2 level is necessary to avoid over prediction of ice loss in the 1980s and under prediction of ice loss in the 2000s and 2010s. In the model the late 2010s at around 410ppm CO2 have an 11% reduced heat loss compared to the early 1980s at 340ppm.
With these model additions the central Arctic now has the lowest ice melt AWP of all regions and the melting energy has a clear upwards trend over the decades. All years with the lowest sea ice extent/area (2007,2012,2016,2019) are also the ones with the highest Ice Melt values. Apart from comparison to other years I would not put any other use case into the presented values. Especially relating negative values with freezing conditions. It is still not even close to a volume model. For sea ice volume we already have PIOMAS, Cryosat2 and my AMSR2 Snow & Ice Volume.

Maps & Graphs:

NRT charts will follow soon

Greenland and Arctic Circle / Bedrock overlay poll
« on: April 14, 2019, 03:40:48 AM »
I've often seen users asking about a glaciers bedrock elevation to estimate future retreat. Then some veteran users post a bedrock map, but it's still difficult to compare it to current satellite images. Therefore I want to build a bedrock overlay webpage for all major glaciers. However I'm not sure which of the two presentation styles is better.

You can vote for both options if you have no major preference.

Horizontal overlay

Full image Opacity

Permafrost / Snow Cover changes on regional scale
« on: January 07, 2019, 01:25:37 AM »
I finished calculating regional snow extent data and will post my analysis here. The main snow cover thread doesn't quite fit for this detailed long term analysis. At the moment all data is still in one long list, but after formatting we can graph things like snow extent for region x in month y. I attached a map showing all regions and an example for Greenlands snow extent.

Eventually regional graphs should also get daily updates on my main snow cover webpage

Data Download (csv & formatted ExcelSheet)

Arctic sea ice / AMSR2 Snow & Ice Volume/Thickness
« on: March 04, 2018, 12:53:45 PM »
A few people already know the ADS/JAXA sea ice thickness and melt concentration product. It is derived from various AMSR2 scanning frequencies.

The melt concentration makes it hard to estimate the sea ice volume, because you can't just read out all individual cell values and calculate the sum. To overcome this issue I developed a melt algorithm which estimates thinning based on the melt concentration percentage and the number of days it occurred. The algorithm also estimates freezing of open water from the melt concentration.

My calculated Sea Ice Volume & Thickness:

Anyone interested in the details can look at my short documentation.

Due to the JavaScript download-site from ADS I can't set up daily updates, but I'm able to download a whole month of sea ice thickness products and get rid of unwanted files and decompress the right ones for processing. With these scripts I can update it every month with a few minutes work.

The AMSR2 thickness isn't as accurate as PIOMAS for actual sea ice thickness. It's definitely affected by melt pond refreezing in August, which almost cancels out further volume losses and thickness increase is rather slow in October-November. As indicated with 2017 it counts snow cover as additional sea ice volume like CryoSat did.

If ones aware of these artefacts it's still a very useful tool to judge current sea ice conditions. Refreezing melt ponds mean a quick end to the melting season and additional snow has to melt in spring before the sea ice melts.

Further it has a higher spatial resolution (10km) than PIOMAS and a higher temporal resolution for the entire Arctic than CryoSat.

Attached is a comparison to PIOMAS

ADS Raw Data visualization:

Greenland and Arctic Circle / Glacier Lake 66N, 50W / West Greenland
« on: September 01, 2017, 01:08:59 PM »
Near the icecaps in West Greenland is a glacier lake without any visible surface outflow.

This lake has one of the biggest icebergs of any lake in the world. The biggest I have measured calved in 2016 and was 1km long, 200m wide and 120m in height. The height is estimated from an image in 2017 after the iceberg turned over. Other icebergs in the lake are similar in size, several reaching 300-600m in length. The whole front of the glacier retreated over 100m from August 2016 until August 2017.

The lake itself is 15 km W-E, 10.5 km N-S and at least 120m deep (otherwise the icebergs can't turn over). According to Google Earth and Greenland Bedrock Map the lake is at 800m above sea level. Maybe it has an outflow beneath a glacier, but I can't say under which one, the bedrock data isn't accurate enough. It should be a good place to study non-ocean terminating glaciers. If the ice sheet retreats further inland it might expose more lakes, which don't have warm ocean water to melt the glaciers from beneath and can't transport icebergs away.

On Google Earth/Maps this lake doesn't even exist and just shows as white snow unless you zoom really close. I'm not sure if it has a name, but feel free to reference it or any of the glaciers terminating in it.

The actual center of the lake is 66N, 55.15W, but in a few years it will melt the ice sheet and reach 50W.

Antarctica / Melt water in Antarctica
« on: December 23, 2016, 03:07:16 PM »
Most of Antarctica is far too cold for surface melting in summer, but with global warming more and more regions go periodically above the freezing point. Regions which already experience surface melt will have longer melting periods.

Its time to keep track of these early changes and document if the melt water can form melt lakes like on Greenland or if it flows towards the coast. For the start I attach four images from East Antarctica. Three are upstream of the Amery ice shelf and one is at Cape Ann.

Arctic sea ice / Albedo-Warming Potential
« on: September 27, 2016, 09:29:22 AM »
The Albedo-Warming Potential is an attempt to quantify the additional warming from a lower ice cover at the poles. At the moment these calculations don't include cloud cover, therefore it is called "Warming Potential" and not actual warming. However, over six-month weather tends to average out and warm areas correlate well with low ice extend in September. The basis of all calculations is a self-developed global surface radiation model and NSIDC Sea Ice Concentration data.
It is in essence a much better version of my “Quantifying albedo effect / Rating daily area values” topic. In order to present the results better I created my own website and only use links on the forum. If I update my calculations all changes will be applied to this first post too and not scattered across several posts. 

Link to my website CryosphereComputing:

former website

The following images are all cumulative results from the last day of the melt season. The end of the astronomical summer to be precise. For daily Animations click on the link below the year. All daily values are also available on my website.

All anomalies are calculated against the 2007-2016 sea ice concentration average.
Red indicates lower albedo and above average warming.
Blue indicates higher albedo and below average warming.
One extra day of peak insulation on open ocean is about 20 MJ/m2.
One extra month of peak insulation on open ocean is about 600 MJ/m2.

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Daily Animation

Pan Arctic Graphs:

Regional graphs:

Global surface radiation model details:
The model calculates the incoming solar energy per day per m2 for all latitudes between 40N and 90N (0.2 degree steps). Considered are solar zenith angles, the atmospheric reduction (Air mass), Projection effect and water albedo for every 15min interval. If a grid cell has 55% ice concentration, then it is treated as a water area 55% the size of the grid cell.

NSIDC Sea Ice Concentration details:
   Ice concentration average: 2007-2016
   Pole hole ice concentration is calculated from a 2-pixel wide ring around the hole
   Lake ice is ignored to reduce noise
   Pixel Area corrected
   warming potential for each individual pixel (max. 0.44-pixels off from pixel center)

Cavalieri, D., C. Parkinson, P. Gloersen, and H. J. Zwally. 1996, updated yearly. Sea Ice Concentrations from Nimbus-7 SMMR and DMSP SSM/I-SSMIS Passive Microwave Data, Version 1. [indicate subset used]. Boulder, Colorado USA: NASA DAAC at the National Snow and Ice Data Center.

Glaciers / Barnes Ice Cap / Penney Ice Cap
« on: August 26, 2016, 11:59:19 PM »
On the 8th of August most of the snow on Barnes Ice Cap has melted apart from the very top. This reveals an interesting pattern of white parallel lines flowing down the ice cap. Most of these are dried out meltwater streams. Only a few are light blue and have meltwater flowing down.

The pattern reminds me of a skeleton from a prehistoric animal buried in sand or rock.

Click to see full resolution

Developers Corner / Arctic sea ice forecast model
« on: June 05, 2016, 08:12:47 PM »
This thread will be used to discuss and improve my arctic sea ice loss forecast model, which is based on my albedo calculations.

Albedo thread:,1543.msg79289.html#new

Question : what is that "outgoing kWh" variable and how did you determine it ?
Also, which period did you analyze to determine "the best results so far" ?

So far I only compared my model to recent years, 2007 and later.
The outgoing kWh is based on my heat loss to space calculations. (10 degrees Celsius black body radiation with 35% back radiation) The actual values don't matter so much as the time when this sub-part becomes relevant. It had to match the June cliff losses from end of May to July.

I did not find any significant change in fit by just assuming that thickness does not change.
I think its very important to prevent the model from shooting towards zero area. If for example the area losses are overestimated for a week, then the average thickness will increase and reduce future area losses, because the ice is thicker.  At the moment this function isn't implemented because all losses are based on historic area data and not predicted ones. Maybe thats a big part of the rather large errors for 2008 and 2011.

I'm confused about your 0.000001 factor for snow cover compared to 0.75 for extent-area.
In my experience the snow cover factor should be about 1/4 of the extent-area factor.
Or am I misinterpreting the units on each variable ?
Ha, I made a mistake. All my ice data is in million km2, but my snow area is in km2. So all my land energy calculations were too high by 6 orders of magnitude, which is exactly the number of zeros in the factor.

Edit1: I changed the model to use average thickness calculated with my own area values instead of historic area. This reduced the error by around 50%.

Edit2: By using calculated values for (extent - area) instead of historic ones the error reduces even further.

new errors for each year
year   2007   2008   2009   2010   2011   2012   2013   2014   2015
diff   0.35   0.74   0.46   -0.13   -0.27   0.29   -0.33   0.32   0.05
%   12%   25%   13%   -4%   -9%   13%   -9%   9%   2%

I wanted to investigate if the arctic albedo is the main factor during a melting season or if warmth from lower latitudes and weather is a bigger factor. On the forum we always talk about preconditioning from melt ponds and ice free areas leading to self-reinforcing melt.

Another motivation was to give each daily area value a weight. A low value in June should be more significant than a low value in September, since most sunlight is already gone by then.

To estimate the albedo I used Cryosphere Today area data, which is prone to show low concentrations in low albedo areas. Normally this is bad to estimate the true sea ice area, but for my calculations it gives a good albedo value. The first image shows an example from 2nd July 2012.

Currently the model is only able to calculate with one latitude for the entire arctic and doesn’t give a perfect representation. However the error isn’t huge since the sea ice area value is most important. A gridded concentration dataset would improve accuracy, but might not be possible in Excel because it requires millions of calculations steps. If I develop a second version, it will consider gridded concentration.

The model calculates the following every half hour for each day: solar zenith angles, the atmospheric reduction (Air mass), Projection effect and albedo for water and snow surfaces.
The final value (kWh/m2) is then multiplied with the water or ice area respectively. See Block diagram for an overview of the calculations.

For simplicity the model assumes clear sky all the time and no energy loss to space. It is not a weather model, just a surface albedo rating.

A review of snow and ice albedo and the development of a new physically based broadband albedo parameterization, Alex S. Gardner and Martin J. Sharp -

Pages: [1]