Support the Arctic Sea Ice Forum and Blog

Author Topic: The 2017/2018 freezing season  (Read 26318 times)

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #200 on: October 15, 2017, 03:22:53 PM »
Pavel notes on #187: Cryosat has resumed. don't see any 3m or thicker ice and there's very little of thicker than 2.5m ice
This is a difficult data set to work with; taking a quick look at netCDF file offered on the CPOM web page (1st image below), the data was there but not in a form (Geo2D) that would allow redrawing the ice thickness map at a larger scale without the lat lon overlays.

Panoply was able to draw out the thickness observations (which seem to be head-to-tail abutted swaths over a month of orbits) and the standard deviations (2nd image). It would be possible to export the thickness numbers to excel to plot the thickness distribution (but what to do about negative ice thicknesses?). Other forums provide very knowledgable posts on Cryosat data and how to compare it to model data such as Piomas.

As the satellite completes a full cycle of its near-polar orbits, the swaths overlap in some places but don't quite come together elsewhere. I looked at several methods of filling in the gaps in ice thickness that would utilize nearby measured thicknesses before settling on D Tschumperlé's graphical algorithm in the Repair section of online G'mic, an amazing French site that allows visitors to conduct a full range of graphical manipulations over the web.

The infill came out rather nice, though it's hard to say how it would compare to kriging or other numeric estimations, much less to the real situation on the ice (which we will never know as this time frame has passed on by).

http://www.cpom.ucl.ac.uk/csopr/seaice.html
https://gmicol.greyc.fr

Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 4068
    • View Profile
    • Arctic Sea Ice Blog
Re: The 2017/2018 freezing season
« Reply #201 on: October 15, 2017, 10:22:42 PM »
Some threshold got re-set way too high on attachment security after the hack scare, hopefully our admin can dial that back a bit. These gifs have had zero contact with any Adobe product and have nothing whatsoever to do with "gif89a format.aip" of Adobe Illustrator.

I don't know if anything got re-set, but here are the attachment settings I can adjust as admin:
Il faut cultiver notre jardin

Pavel

  • ASIF Citizen
  • Posts: 121
    • View Profile
Re: The 2017/2018 freezing season
« Reply #202 on: October 15, 2017, 11:56:23 PM »
The Arctic atmosphere tries to cool but it fails. According to the weather forecasts no significant coldness will come or even things may get warmer

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #203 on: October 16, 2017, 05:58:11 PM »
According to the weather forecasts no significant coldness will come or even things may get warmer
Here is the ESRL forecast out to Oct 24th for 80ºN. Forty 2m air temperature maps are provided at 6hr intervals and the average determined. The identical result is displayed in a variety of color tables with a scale that runs from -22º to +6º C. These variations illustrate how interpretations can be helped or hindered by presentation choices.

 Warmer air appears to be intruding well into the interior from the North Atlantic though the CAA remains cold. However it is not warm enough to melt any snow on ice. This time of year, snow retards bottom ice formation by insulating the top ice from air.

The second animation shows these temperatures for the Arctic Ocean as a whole over the same time frame. This shows air flow well but it is not easy to get a sense of the time-averaged temperature from it. The averaged whole ocean temperature has a red line indicating the southern boundary of sea water above its freezing temperature of -1.8ºC

Technical note: Panoply was run in linear grayscale mode on REB.2017-10-15.nc. The 40 frames are then averaged to a single grayscale in Gimp. All extraneous pixels are removed, leaving only the image plus its palette as 256 grays. Lookup tables in ImageJ are applied, those that seem informative are saved as .png, reloaded as an ImageJ stack, and saved out as a gif. Gimp has a bad bug in gifs that causes it to seek a global color table whereas gif89 allows each frame to have its own color table. The cluts used here are gray, glow, redHot, ICA3, physics, royal, rainbow, rire, cool, and inverted glasbey with the addition of G'mic contouring in some instances.
« Last Edit: October 18, 2017, 11:43:17 PM by A-Team »

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #204 on: October 18, 2017, 09:46:40 AM »
It's that time of year again, when thousands of abstracts for AGU meeting become available. Hardly anyone discloses results, posters aren't available for poster sessions, talks won't be videoed, their powerpoints won't be archived, and already-published articles won't be linked.

Still, AGU17 does allow a look ahead to the coming year of journal articles. A name search can show what a particular scientist has been up to; for example Neven asked upforum what J Stroeve is doing, she is on three of the abstracts.

https://agu.confex.com/agu/fm17/meetingapp.cgi/SearchResults/0

Snow on Arctic sea ice is an active topic. Like ice thickness and clouds, it is very difficult to characterize basin-wide, in part because depth alone doesn't capture its insulating properties in freeze season: it's blown into windrows, it may be dunked in sea water on a floe with negative freeboard, or be rained upon and refreeze. Still, it looks like some better products than what we have now may be in the pipeline.

C23E-08: Merging observations and reanalysis data to improve estimates of snow depth on Arctic sea ice
NT Kurtz et al

Snow is an important controlling factor in the heat and radiation balance of the Arctic sea ice pack. Knowledge of snow on sea ice is also required for retrievals of sea ice thickness from airborne and spaceborne altimeters, and is presently the largest source of uncertainty in the conversion of freeboard to sea ice thickness from these altimetry data.

Multiple sources of observational snow depth data exist such as those from the Operation IceBridge (OIB) snow radar, passive microwave satellites, and ice mass balance buoys. However, these observational data sources are limited in spatial and/or temporal extent, which makes their usage impractical when used for basin-wide sea ice thickness retrievals in a standalone fashion.

We show how the use of snow depth observations from the OIB snow radar can be used as a primary means to improve basin-scale snow depth results from a simple snow model forced by reanalyses and satellite-derived ice drift estimates. We also show how different observational data sets impact the snow depth estimates, and how best to incorporate data sets of differing temporal and spatial scales to provide snow thickness estimates of consistent quality over the entire sea ice growth season. Particular focus is given to the new 2017 OIB data set which included new flights into the eastern Arctic sector where interesting differences were seen between the first year and multiyear ice areas.

C32B-02: Snow accumulation on Arctic sea ice: is it a matter of how much or when?
M Webster  et al

Snow on sea ice plays an important, yet sometimes opposing role in sea ice mass balance depending on the season. In autumn and winter, snow reduces the heat exchange from the ocean to the atmosphere, reducing sea ice growth. In spring and summer, snow shields sea ice from solar radiation, delaying sea ice surface melt. Changes in snow depth and distribution in any season therefore directly affect the mass balance of Arctic sea ice.

In the western Arctic, a decreasing trend in spring snow depth distribution has been observed and attributed to the combined effect of peak snowfall rates in autumn and the coincident delay in sea ice freeze-up. Here, we present an in-depth analysis on the relationship between snow accumulation and the timing of sea ice freeze-up across all Arctic regions.

A newly developed two-layer snow model is forced with eight reanalysis precipitation products to: (1) identify the seasonal distribution of snowfall accumulation for different regions, (2) highlight which regions are most sensitive to the timing of sea ice freeze-up with regard to snow accumulation, and (3) show, if precipitation were to increase, which regions would be most susceptible to thicker snow covers. We also utilize a comprehensive sensitivity study to better understand the factors most important in controlling winter/spring snow depths, and to explore what could happen to snow depth on sea ice in a warming Arctic climate.

C33C-1215: Rainy Days in the New Arctic: A Comprehensive Look at Precipitation from 8 Reanalysis
L Boisvert  et al

Precipitation in the Arctic plays an important role in the fresh water budget, and is the primary control of snow accumulation on sea ice. However, Arctic precipitation from reanalysis is highly uncertain due to differences in the atmospheric physics and use of data assimilation and sea ice concentrations across the different products. More specifically, yearly cumulative precipitation in some regions can vary by 100-150 mm across reanalyses. This creates problems for those modeling snow depth on sea ice, specifically for use in deriving sea ice thickness from satellite altimetry.

In recent years, this new Arctic has become warmer and wetter, and evaporation from the ice-free ocean has been increasing, which leads to the question: is more precipitation falling and is more of this precipitation rain? This could pose a big problem for model and remote sensing applications and studies those modeling snow accumulation because rain events will can melt the existing snow pack, reduce surface albedo, and modify the ocean-to-atmosphere heat flux via snow densification.

In this work we compare precipitation (both snow and rain) from 8 different reanalysis: MERRA, MERRA2, NCEP-R1, NCEP-R2, ERA-Interim, ERA-5, ASR and JRA-55. We examine the annual, seasonal, and regional differences and compare with buoy data to assess discrepancies between products during observed snowfall and rainfall events. Magnitudes and frequencies of these precipitation events are evaluated, as well as the “residual drizzle” between reanalyzes. Lastly, we will look at whether the frequency and magnitude of “rainy days” in the Arctic have been changing over recent decades.

C21B-1122: Synoptic weather conditions, clouds, and sea ice in the Beaufort and Chukchi Seasonal Ice Zone
Z Liu et al
http://journals.ametsoc.org/doi/abs/10.1175/JCLI-D-16-0887.1

The connections between synoptic conditions and clouds and sea ice over the Beaufort and Chukchi Seasonal Ice Zone are examined. Four synoptic states with distinct thermodynamic and dynamic spatial and vertical signatures are identified using a k-means classification algorithm and the ERA-Interim reanalysis data from 1979 to 2014.

The combined CloudSat and Calipso cloud observations suggest control of clouds by synoptic states. Warm continental air advection is associated with the fewest low-level clouds, cold air advection under low pressure generates the most low-level clouds. Low-level cloud fractions are related to lower-tropospheric stability and both are regulated by synoptic conditions. Observed cloud vertical and spatial variability is reproduced well in ERA-Interim, but winter low-level cloud fraction is overestimated.

Sea ice melt onset is related to synoptic conditions. Melt onsets occur more frequently and earlier with warm air advection states. The warm continental air advection state with the highest temperature is the most favorable for melt onsets even though fewer low-level clouds are associated with this state. The other warm advection state is cloudier but colder.

In the Beaufort and Chukchi Seasonal Ice Zone, the much higher temperature and total column water of the warm continental air advection state compensate the smaller cloud longwave radiative fluxes due to the smaller low-level cloud fraction. In addition, the higher shortwave radiative fluxes and turbulent fluxes to the surface are also favorable for sea ice melt onset.

C21G-1186: There goes the sea ice: following Arctic sea ice parcels and their properties.
MA Tschudi et al
http://www.mdpi.com/2306-5729/2/3/25

Arctic sea ice distribution has changed considerably over the last couple of decades. Sea ice extent record minimums have been observed in recent years, the distribution of ice age now heavily favors younger ice, and sea ice is likely thinning. This new state of the Arctic sea ice cover has several impacts, including effects on marine life, feedback on the warming of the ocean and atmosphere, and on the future evolution of the ice pack.

The shift in the state of the ice cover, from a pack dominated by older ice, to the current state of a pack with mostly young ice, impacts specific properties of the ice pack, and consequently the pack’s response to the changing Arctic climate. For example, younger ice typically contains more numerous melt ponds during the melt season, resulting in a lower albedo. First-year ice is typically thinner and more fragile than multi-year ice, making it more susceptible to dynamic and thermodynamic forcing.

To investigate the response of the ice pack to climate forcing during summertime melt, we have developed a database that tracks individual Arctic sea ice parcels along with associated properties as these parcels advect during the summer. Our database tracks parcels in the Beaufort Sea, from 1985 – present, along with variables such as ice surface temperature, albedo, ice concentration, and convergence.

We are using this database to deduce how these thousands of tracked parcels fare during summer melt, i.e. what fraction of the parcels advect through the Beaufort, and what fraction melts out? The tracked variables describe the thermodynamic and dynamic forcing on these parcels during their journey. The attached image (it’s not) shows the ice surface temperature of all parcels (right) that advected through the Beaufort Sea region (left) in 2014.

C33C-1210: Towards development of an operational snow-on-sea-ice product
GE Liston et al

While changes in the spatial extent of sea ice have been routinely monitored since the 1970s, less is known about how the thickness of the ice cover has changed. While estimates of ice thickness across the Arctic Ocean have become available over the past 20 years based on data from ERS-1/2, Envisat, ICESat, CryoSat-2 satellites and Operation IceBridge aircraft campaigns, the variety of these different measurement approaches, sensor technologies and spatial coverage present formidable challenges. Key among these is that measurement techniques do not measure ice thickness directly – retrievals also require snow depth and density.

Towards that end, a sophisticated snow accumulation model is tested in a Lagrangian framework to map daily snow depths across the Arctic sea ice cover using atmospheric reanalysis data as input. Accuracy of the snow accumulation is assessed through comparison with Operation IceBridge data and ice mass balance buoys (IMBs). Impacts on ice thickness retrievals are further discussed.
« Last Edit: October 18, 2017, 09:24:01 PM by A-Team »

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #205 on: October 18, 2017, 12:15:22 PM »
Here are those same three RASM-ESRL precipitation forecasts at 24 hour intervals out to Oct 24th. A moderate amount of rain-on-snow is foreseen for a small area north of Svalbard. Snow depth is moderate, at most 0.25m, and quite uneven in providing thermal insulation after wind-blown drifting is considered (2nd image),
« Last Edit: October 18, 2017, 12:43:57 PM by A-Team »

Neven

  • Administrator
  • ASIF Governor
  • *****
  • Posts: 4068
    • View Profile
    • Arctic Sea Ice Blog
Re: The 2017/2018 freezing season
« Reply #206 on: October 18, 2017, 02:31:44 PM »
Snow on Arctic sea ice is an active topic. Like ice thickness and clouds, it is very difficult to characterize basin-wide, in part because depth alone doesn't capture its insulating properties in freeze season: it's blown into windrows, it may be dunked in sea water on a floe with negative freeboard, or be rained upon and refreeze. Still, it looks like some better products than what we have now may be in the pipeline.

Thanks for those abstracts, A-Team. Very interesting stuff. Snow on ice is one of those things I'd always known about, but my interest in it really got kindled during this past melting season.

A couple of days ago I also received this interesting message in my mailbox:

Dear colleagues and sea ice friends,

POLAR2018 is a *unique**joint event* organized by the Scientific
Committee on Antarctic Research SCAR and the International Arctic
Science Committee IASC, which will take place in Davos, Switzerland,
from 15 - 26 June 2018 with the open science conference from 19 - 23
June; see http://www.polar2018.org for general information.

Following up with our first invitation on September 27 we would like to
encourage you to submit your presentation to the conference session
entitled "*The role of snow on sea ice for sea-ice parameter retrieval
and variability*".

We invite studies dealing with in situ observations, with retrieval from
satellite observations, modeling and combinations thereof for snow
parameters on sea ice. We also invite studies on methods for quantifying
the influence of (unknown) snow properties on the satellite retrieval of
sea-ice parameters, on reducing the noise, improving the accuracy of
retrieved sea-ice parameters due to snow properties, and related studies.

Conveners of this session are: Stefan Kern, Burcu Ozsoy, Georg Heygster and Leif T. Pedersen

Please find details about the program as well as deadlines here:
http://www.polar2018.org/program.html
Il faut cultiver notre jardin

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #207 on: October 18, 2017, 09:12:41 PM »
unique event which will take place in Davos, Switzerland
I am skipping both Bilderberg and Davos this year in favor of a staycation. ;) 

The abstracts below speak to common themes on the forum; it's hard to say which will emerge as game-changers vs incremental improvements vs never-to-be-seen-agains.

C21G-1188: Estimation of Melt Ponds over Arctic Sea Ice using MODIS Surface Reflectance Data
Y Ding et al

Melt ponds over Arctic sea ice is one of the main factors affecting variability of surface albedo, increasing absorption of solar radiation and further melting of snow and ice. In recent years, a large number of melt ponds have been observed during the melt season in Arctic. Moreover, some studies have suggested that late spring to mid summer melt ponds information promises to improve the prediction skill of seasonal Arctic sea ice minimum.

In the study, we extract the melt pond fraction over Arctic sea ice since 2000 using three bands MODIS weekly surface reflectance data by considering the difference of spectral reflectance in ponds, ice and open water. The preliminary comparison shows our derived Arctic-wide melt ponds are in good agreement with that derived by the University of Hamburg, especially at the pond distribution. We analyze seasonal evolution, inter-annual variability and trend of the melt ponds, as well as the changes of onset and re-freezing.

The melt pond fraction shows an asymmetrical growth and decay pattern. The observed melt ponds fraction is almost 25% in early May and increases rapidly in June and July with a high fraction of more than 40% in the east of Greenland and Beaufort Sea. A significant increasing trend in the melt pond fraction is observed for the period of 2000-2017.


C21G-1179: A Novel Approach To Retrieve Arctic Sea Ice Thickness For Prediction And Analysis
L Brucker et al

In spite of October-November Arctic-sea-ice-volume loss exceeding 7000 km3 in the decade following ICESat launch, most global ocean reanalysis systems are not able to reproduce such a drastic decline.

Knowledge of the sea ice properties and its thickness distribution is critical to our understanding of polar ocean processes and the role of the polar regions in the Earth's climate system. Existing large-scale sea ice thickness datasets are derived from freeboard observations made by different satellite altimeters (radar and lidar). These datasets are significantly different due to the remote sensing technique and spacecraft orbit, and they are limited in time. These differences increase the difficulty of using such data for sea ice initialization and assimilation, and increase the challenge for studying sea ice processes and interactions with the ocean and atmosphere.

For the first time, we were able to reproduce the Arctic sea ice thickness field at 10 km resolution with success for fall, winter, and spring (April/May depending on melt conditions) from passive microwave data. Our results reveal the same patterns of thickness distribution in the Arctic basin and peripheral seas as CryoSat-2, and the majority of the retrievals are within 0.5 m of CryoSat-2. The range of CryoSat-2 ice thickness is correctly retrieved, including in the upper range (3-5 m). The amplitude is well reproduced too, as the distribution of differences is centered on 0 m (no bias).

Some underestimations are visible between islands of the Canadian Archipelago, but due to the size of the field of view our confidence will always be lower in this region where there is land contamination. An initial comparison of the AMSR2 ice thickness with IceBridge airborne products in different sectors (Beaufort sea, central Arctic) demonstrates the quality of the retrievals.

We will also quantify the prediction and nowcast gain obtained from assimilating these new retrievals. We carried-out the integration of 36 members of coupled NASA Goddard Earth Observing System Model, version 5 (GEOS-5) to enable the implementation of an Ensemble Kalman Smoother (EnKS) over the period September 2012 - January 2013. Assimilating our retrievals improves the nowcast of ice volume, the forecast and the retrospective forecast.


C11D-06: Regional Arctic sea-ice prediction: A direct comparison of potential versus operational seasonal forecast skill
M Bushuk et al

Seasonal predictions of Arctic sea ice on regional spatial scales are a pressing need for a broad group of stakeholders, however, most forecast skill assessments to date have focused on pan-Arctic sea-ice extent (SIE). In this work, we present a direct comparison of potential and operational seasonal prediction skill for regional Arctic SIE. This assessment is based on two complementary suites of seasonal prediction ensemble experiments performed with a global coupled climate model.

First, we assess the operational prediction skill for de-trended regional SIE using a suite of retrospective initialized seasonal forecasts spanning 1980-2017. These retrospective forecasts are found to skillfully predict regional winter SIE at lead times of 3-11 months and regional summer SIE at lead times of 1-4 months, owing partially to subsurface ocean temperature and sea-ice thickness initial conditions, respectively. Second, we present a suite of perfect model predictability experiments with start dates spanning the calendar year, which are used to quantity the potential regional prediction skill of this system.

These perfect model experiments reveal that regional Arctic SIE is potentially predictable at lead times beyond 12 months in many regions, substantially longer than the current operational skill of this system. Both the retrospective forecasts and perfect model experiments display a spring prediction skill barrier for regional summer SIE forecasts, indicating a fundamental predictability limit for summer regional predictions. The skill gap identified in this work indicates a promising potential for future improvements in regional SIE predictions.


C21G-1190: Assessing surface radiative fluxes and developing surface turbulent heat fluxes over Arctic sea ice
M Song et al

In this study, we have developed a new satellite-based surface heat and moisture flux data set over the ice-covered ocean in the Arctic using a recently developed flux algorithm based on the theory of maximum entropy production (MEP model). First, the accuracy and uncertainty associated with surface radiative fluxes and temperature for three available satellite products are evaluated against the assembled in-situ data.

The three satellite products are the Surface Radiation Budget project (SRB), the International Satellite Cloud Climatology Project (ISCCP), and the Extended AVHRR Polar Pathfinder version-2 (APP-x).

Our comparisons suggest that 1) in terms of the overall bias, root mean square error, and correlation, the net surface radiative flux of ISCCP is closer to in-situ observations than that of SRB and APP-x; 2) in terms of the bias by local times, it is not very clear which satellite product is superior to others; and 3) in terms of inter-annual variability of the bias, the net surface radiative flux of ISCCP is more accurate than that of SRB and APP-x. Based on the above comparison, we use the ISCCP surface radiative fluxes as input values for the MEP model to calculate surface turbulent heat fluxes over Arctic sea ice.


C21G-1184: Improving Arctic sea ice edge forecasts by assimilating high resolution VIIRS sea ice concentration data into the U.S. Navy’s ice forecast system
OM Smedstad et al

This study presents the improvement in ice edge error within the U.S. Navy’s operational sea ice forecast system gained by assimilating the high horizontal resolution visible/infrared satellite-derived VIIRS ice concentration products. A series of hindcast studies are performed for the period of 1 January – 31 December 2016 using Global Ocean Forecast System (GOFS 3.1), a 1/12° HYbrid Coordinate Ocean Model (HYCOM) that is two-way coupled to the Community Ice CodE (CICE) in a daily update cycle with the Navy Coupled Ocean Data Assimilation (NCODA).

Comparisons using the VIIRS ice concentration products (< 1km resolution) show lower ice edge location errors than the current system, which assimilates near real-time passive microwave data from the Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave/Imager (SSMIS) and the Advanced Microwave Scanning Radiometer (AMSR2) ice concentration products (25 and 12.5km resolution, respectively).

The daily ice edge locations from the model simulations are compared against independent observed ice edge locations. Results from the Pan-Arctic and regional areas along with seasonal time scales will be presented. A previous study using the Arctic Cap Nowcast/Forecast System (ACNFS), a 1/12° coupled HYCOM/CICE/NCODA for the Northern Hemisphere only, has shown that by assimilating the VIIRS (along with SSMIS and AMSR2) ice concentration products reduced the ice edge location errors by 25% in the pan-Arctic region for the same year-long time period.
« Last Edit: October 18, 2017, 09:19:16 PM by A-Team »

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #208 on: October 18, 2017, 11:31:33 PM »
Speaking of snow depth, here is another ESRL product, snowdepthchange.gif from their web page or archival REB_plots. It is somewhat peculiar in that D0 is not provided, only the D5 forecast. Thus the animation is of these day fives -- 15 Sep to 22 Oct -- rather than the presumably more accurate initial states.

Still, it gives an idea how rapidly snow depth changes from day to day as well as the expected prevailing wind. The final frame averages these out, even though the palette is not really designed to support this.

This time of year, when thermal insulation not solar insolation is the issue for the rate of bottom ice growth induced by frigid surface air, the relevant property of snow is its conductivity.

Is it still, as often assumed, a uniform basin-wide porous medium with a large immobilized air component (like a foam pad) after being blown around for weeks, possibly getting dunked, rained on, and soaked with sea spray? If so, is the current ankle-deep mean snowpack enough to seriously inhibit bottom growth, relative to not-so-cold prevailing mean air temperatures?

That's hard to say directly with no buoys, no ships, and no one out there but satellites can measure bulk properties. The scale though is not commensurate with that of snow features, though Sentinel-1 comes fairly close.

http://www.inscc.utah.edu/~campbell/snowdynamics/reading/Pomeroy.pdf
http://acwc.sdp.sirsi.net/client/search/asset/1005644;jsessionid=CE14DA1FFAEF3D6FD98ABAD517B04B81.enterprise-15000
http://www.sciencedirect.com/science/article/pii/S0165232X1100187X
http://arc.lib.montana.edu/snow-science/objects/issw-1994-176-184.pdf
« Last Edit: October 18, 2017, 11:42:09 PM by A-Team »

Peter Ellis

  • ASIF Middle Class
  • Posts: 549
    • View Profile
Re: The 2017/2018 freezing season
« Reply #209 on: October 19, 2017, 10:58:44 AM »
Speaking of snow depth, here is another ESRL product, snowdepthchange.gif from their web page or archival REB_plots. It is somewhat peculiar in that D0 is not provided, only the D5 forecast.

Given that it's a depth change, then surely it has to be estimated over a given period?  At D0, there is no change from D0...

Images for the absolute values of snow/ice thickness and area are in one tab, images for the 5-day changes are in another.
https://www.esrl.noaa.gov/psd/forecasts/seaice/


Edit to add: Moreover, given that the "5 day change" values from two successive days' forecasts will necessarily include four out of the same 5 days, then I think rapid changes from day to day may be more indicative of model variability than anything to do with actual weather.

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #210 on: October 19, 2017, 06:37:53 PM »
Right. ESRL is primarily interested in making forecasts whereas we are primarily interested in archival initial state time series because short-term predictions have such a limited shelf life. The idea here was to finish scrolling through all their precipitation products to see which are worth scripting (Panoply --> ImageMagick --> cloud --> forum) into hindcasts + today + forecast time series.

The ESRL web site presents this one well enough, though too large to display well here. It might be of heads-up interest should more moisture-laden storm sweeps north from the Caribbean again this fall. However ice thickening takes place on a much slower time scale, so daily comings and goings of the insulating blanket of snow are of less interest than mean snowpack.

While it's hard to see the thermal relevance of blowing ankle-deep snow to bottom ice formation rates, maybe it will be knee-deep by late winter and suppress early melt pond formation through reflectance.

There being little purpose in simply replicating daily changes in NOAA's web site, the question becomes where we can 'add value'. Among the many opportunities explored in previous posts (eg SMOS-ESRL thin ice hybrids), are combined time series across the three ESRL (and other) archives.

These can be seasonal: the forecast below combines an open water property with a sea ice measure, namely temperature. That product diminishes in utility along with residual exposed water later in the fall. Salinity is another option; it mixes SMOS bulk ice salinity with that of ESRL open water. That too is seasonal since UH SMOS availability is melt-limited.

Note the Chukchi north of the Bering Strait is still far too warm for ice to form. That stayed open to mid-December last year. The map also shows a pronounced intrusion of warm surface water in the Yermak Plateau area north of Svalbard. Spurious open water is shown around CAA islands which are very difficult to get at accurately with gridded data (UH AMSR3 3.125 km is a better option there).

Technical note: these are easy to make since 'not sea ice' on the sea ice layer provides a pixel-perfect cut-out allowing any open water characteristic to show through. As long as the data sources are both available to Panoply as netCDF files, Gimp will receive the maps in perfect co-registration with compatible and operable palette legends. This readily scales to times series via tile 'n' slice.
« Last Edit: October 20, 2017, 10:02:39 AM by A-Team »

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #211 on: October 19, 2017, 10:54:01 PM »
Here are 27 days of sea water salinity from RASM-ESRL for October. As noticed before, each ten day forecast series begins at hour 24 rather than hour 00, the initial state. (Some even start at hour 48, skipping the first two days.) The odd boundary on the Svalbard side apparently results from a lack of data (or maybe it's off-scale on the high side).

There's ample room in a netCDF file for an explanation of the satellite (or oceanographic) source of the data but there is none. It's not clear what salinity under the ice pack means in terms of depth. The salinity range is also mistakenly set, showing large negative salinities.

Indeed, the whole file system of this project is seriously mis-configured. File names for a given product are all the same; they're supposed to be inseparably concatenated with their date. The daily RASM-ESRL archive is presented as nine separate files but these in effect just represent an animatable time sequence. They could have been folded into a single file with each time an animation frame (and there's a simple command line for doing just that).

This project reminds me of an autonomous 18-wheeler driving without incident from NY to LA but continuing on, only to plunge off the Santa Monica pier. That is, is anyone really driving this project, who is using it without reporting the flat tires, and how long can it run on fumes without  interventional refueling?

Whatever, it's interesting to watch salinity evolve along the Alaskan and East Siberian coasts. Salinity lowers the freezing point of sea water somewhat but here it is not determinative because though the remaining open water is fresher, its temperature (and that of the air above) are warmer.

SMOS provides bulk ice salinity of the ice pack surface which is more or less directly observable from its dielectric. As sea ice ages, it extrudes its brine which lowers its perceived bulk salinity here.
« Last Edit: October 20, 2017, 07:11:11 PM by A-Team »

Michael

  • ASIF Lurker
  • Posts: 41
    • View Profile
Re: The 2017/2018 freezing season
« Reply #212 on: October 20, 2017, 11:08:31 AM »
Here are 27 days of sea water salinity from RASM-ESRL for October. As noticed before, each ten day forecast series begins at hour 24 rather than hour 00, the initial state. (Some even start at hour 48, skipping the first two days.) The odd boundary on the Svalbard side apparently results from a lack of data.

The "t024" / "t48" indicate "hours since analysis" (Tau) rather than a time period.

variables:
   double tau ;
      tau:long_name = "Tau" ;
      tau:units = "hours since analysis" ;
   double time(time) ;
      time:long_name = "Valid Time" ;
      time:units = "hour since 2000-01-01 00_00_00" ;
   double time_bounds(time, d2) ;
      time_bounds:long_name = "boundaries for time-averaging interval" ;
      time_bounds:units = "days since 0000-01-01 00:00:00" ;
data:

 tau = 24 ;
 time = 155856 ;
 time_bounds =
  736488.75, 736489 ;

"time" in the RASM-ESRL files is calendar hours, "time" in the REB files is model days ("All years have exactly 365 days").

These data in the RASM-ESRL files are the same as the fourth set of data in the REB files but converted from float to short, in the process removing the Nans.

A-Team

  • ASIF Upper Class
  • Posts: 1879
    • View Profile
Re: The 2017/2018 freezing season
« Reply #213 on: October 20, 2017, 04:51:05 PM »
Thanks. You are talking about ice and snow thicknesses?

Salinity, compressive strength, sea water temperature, the three melts, and two precips are not to be found in REB. None of these are attributed within RASM_ESRL. Their model might be able to derive some from others but salinity, water temperature and so on must be external inputs. From where though, there might be something better out there that could be stubbed in.

The REB files overlap do considerably in name. And they do provide start-stop ranges. However they provide 40 animation frames, the first of which I've been taking as t00 whereas RASM_ESRL provide only 9.

So I'm not sure what you mean by same as every 4th bit of data is the same. That would only use up 4x9 = 36 (sometimes 4x8 = 36) of the 40, suggesting the initial (or final?) state is missing in RASM_ESRL. Or rather, the latter uses intervals, n times has n-1 intervals but what does this mean in tangible terms for observational validation or animation frames, very little.

It seems better just to use REB whenever possible since they didn't see the merge app as applicable to RASM_ESRL intervals. But REB doesn't have the data to generate all the forecast animations that RASM_ESRL can. No way am I going to interpolate four 6hr frames out of one 24 hour to complete the file set in REB.

NaNs, float etc seem to be non-issues suppressed by Panoply and have no impact on visualizations or grepped csv coming out of ncdump.

It appears that not nearly enough information is provided in RASM_ESRL and REB.nc together to draw all the REB plots. That's unfortunate, those files might have been provided so users could correct the many inept products provided in REB plots, make omitted ones, compare to other observational sources, run an alternative model, or compare to competitive products like ECMWF. 

NOAA states this project is experimental. Fair enough but in its 3rd year, it's time to pull things together, maybe lay on some documentation and make the five minute fixes. It's true though that they didn't need to provide a public archive at all, much less the most thorough one around providing comprehensive Arctic forecasts. Expired forecasts have such limited interest that the real value may lie in archival initial states (or their reanalysis), which need attending to before letting this go on as unattended robo-ware.

Going around the web to the netCDF data sources we commonly use for forum graphics, I see a tremendous range in quality from zero (take this map and shove it), outdated (defective variable treatment disabling Geo2D), inadequately commented files, okay, and fantastic. In the instances where I know the authors, there's been a perfect correlation of open sourcing effort with the quality of their journal publications.

Data is not open source accessible in my view if it can't be viewed and manipulated without purchasing proprietary software, working in terminal mode, or emailing a deceased author. Site users and journal readers should have the capacity in most instances to reproduce major graphics.

I see a goodly number of totally incompetent graphical products, both in archives and after peer review. That is the real purpose of posting netCDF files -- the next person who comes along might have the skills to fix the graphic, re-project or re-palette it, delete over-writing  layers, test it for accuracy, or combine it in novel ways with other data sources. There is no purpose to climate science if it is not communicated.

In every collaborative project I've worked on, everyone including myself had moved on and lost all interest long before the draft worked its way through the publication process. We all knew what was in the data, making derivative charts from it was considered a total bore, the only thing worse being a remake six months later. Here again it's in everyone's interest to have a proper archive.

Other scientific communities with even bigger data sets, such as genomics, laid down the law fifteen years ago (GenBank) and enforce it via conditions in the grant (both govt and foundation). There was a lot of initial resistance to sharing, people acted like they somehow 'owned' the data even though the public had bought and paid for every scrap of it with the understanding they could see it.

Nobody has to share: just click the 'Don't Accept' option on the grant application, do the work at home, pay for it out of your non-salaried savings, and post on your facebook page to sidestep journal data requirements. If you don't go that route, then an adequate open archive is, scientifically speaking, obligatory. And it's especially important in the case of climate change.
« Last Edit: October 20, 2017, 07:22:49 PM by A-Team »

Michael

  • ASIF Lurker
  • Posts: 41
    • View Profile
Re: The 2017/2018 freezing season
« Reply #214 on: October 21, 2017, 10:46:50 AM »
So I'm not sure what you mean by same as every 4th bit of data is the same. That would only use up 4x9 = 36 (sometimes 4x8 = 36) of the 40, suggesting the initial (or final?) state is missing in RASM_ESRL. Or rather, the latter uses intervals, n times has n-1 intervals but what does this mean in tangible terms for observational validation or animation frames, very little.

Apologies, this is very much off topic.

My explanation. was very poor.  If the first file in the RASM-ESRL archive is labeled t048 it simply means that the ensemble hasn't been run and and the previous day 2 has been carried forward as day 1 etc.

To get the relationship between the RASM-ESRL and REB files, I would suggest extracting the relevant data from all the RASM-ESRL files in an archive and comparing it with the same data from the corresponding REB file.

ncdump -v tau,time,time_bounds "RASM-ESRL_2017-10-dd-00_t0hh.nc" > RASM-ESRL201710ddhh.txt
ncdump -v time,time_bounds "REB.2017-10-dd.nc" > REB20171010time.txt

To convert RASM-ESRL "time" to REB "time" :  time =  (time + 17519880) / 24

The actual source for the data in the individual .nc files can be found near the bottom of the history section of the header.
The syntax is ncrcat -v [variable list] [source files] [ouput files]
and ncks -v [variable list] [source file] [ouput files]