iacs-cryo / delineation-wg Goto Github PK
View Code? Open in Web Editor NEWIce sheet and glacier boundaries and basins
Home Page: https://cryosphericsciences.org/activities/delineation-of-glaciers-ice-sheets-and-ice-sheet-basins/
Ice sheet and glacier boundaries and basins
Home Page: https://cryosphericsciences.org/activities/delineation-of-glaciers-ice-sheets-and-ice-sheet-basins/
Doc: https://www.glims.org/rgi_user_guide
The input data comes from optical imagery, mostly Landsat. No other mask is used, unless the outlines authors did.
In Greenland, RGI uses Raster et al (2012). From the paper its not clear what / if they used a mask for the ice sheet proper to help their connectivity level design.
In Antarctica, RGI uses Bliss et al., (2013), which itself relies on the Antarctic Digital Database quite heavily. RGI7 corrected quite some geometries for poor georeferencing.
ADD Consortium (2000) Antarctic Digital Database, Version 3.0, database, manual and bibliography. Scientific Committee on Antarctic Research, Cambridge
Bliss, A., Hock, R., and Cogley, J.G. A new inventory of mountain glaciers and ice caps for the Antarctic periphery. Annals of Glaciology, 54(63):191–199, jul 2013. https://doi.org/10.3189/2013AoG63A377.
Rastner, P., Bolch, T., Mölg, N., Machguth, H., Le Bris, R., and Paul, F.: The first complete inventory of the local glaciers and ice caps on Greenland, The Cryosphere, 6, 1483–1495, https://doi.org/10.5194/tc-6-1483-2012, 2012.
These two papers (2012 and 2013, mind!) decided to do so. Raster suggests a definition, Bliss also contains arguments as to why certain outlines were chosen to be "peripheral" or not.
It would be based on a discussion happening here, and together with a clear definition of "main".
If its just about removing outlines, RGI could be amended very quickly. The redefinition of a new standard has, however, quite important implications for downstream products. It's also very likely some sort of mapping would be needed.
N.A.
N.A.
Vector
N.A.
WGS84
The two original inventories have been slightly adapted for RGI. Importantly, connectivity level 2 is filtered out in Greenland. The main rationale for picking the masks in Bliss and Rastner remains valid.
What upstream mask product did you use? Citterio & Ahlstrøm (2013)
Reason for using this mask
citet:mankoff_2020_liquid used the PROMICE citet:citterio_2013 product as the base mask because that was estimated to be the “truest” representation of the glaciated area of Greenland, including small peripheral glaciers.
The effort to change masks in a future release is trivial - probably a few days of effort for testing and documenting the changes. The downstream effects should be negligible, as this is a dynamic product that updates every few years - not just extended time series, but using the latest BedMachine for example, which changes location and number of outlets. Therefore, downstream users are cautioned to use a programmatic interface and not expect stability between product versions.
N/A
Type: vector
Resolution if raster:
Vector digitized to 100 m because that's the resolution of the ArcticDEM product used for sub-aerial off-ice surface routing. That resolution appeared to correctly route all terrestrial streams when comparing with streams visible in Google Earth satellite imagery.
Projection information:
+proj=stere +lat_0=90 +lat_ts=70 +lon_0
-45 +x0=0 +y0=0 +datum=WGS84 +units=m +nodefs +type=crs= PROJCRS["WGS 84 / NSIDC Sea Ice Polar Stereographic North",
BASEGEOGCRS["WGS 84",
ENSEMBLE["World Geodetic System 1984 ensemble",
MEMBER["World Geodetic System 1984 (Transit)"],
MEMBER["World Geodetic System 1984 (G730)"],
MEMBER["World Geodetic System 1984 (G873)"],
MEMBER["World Geodetic System 1984 (G1150)"],
MEMBER["World Geodetic System 1984 (G1674)"],
MEMBER["World Geodetic System 1984 (G1762)"],
MEMBER["World Geodetic System 1984 (G2139)"],
ELLIPSOID["WGS 84",6378137,298.257223563,
LENGTHUNIT["metre",1]],
ENSEMBLEACCURACY[2.0]],
PRIMEM["Greenwich",0,
ANGLEUNIT["degree",0.0174532925199433]],
ID["EPSG",4326]],
CONVERSION["US NSIDC Sea Ice polar stereographic north",
METHOD["Polar Stereographic (variant B)",
ID["EPSG",9829]],
PARAMETER["Latitude of standard parallel",70,
ANGLEUNIT["degree",0.0174532925199433],
ID["EPSG",8832]],
PARAMETER["Longitude of origin",-45,
ANGLEUNIT["degree",0.0174532925199433],
ID["EPSG",8833]],
PARAMETER["False easting",0,
LENGTHUNIT["metre",1],
ID["EPSG",8806]],
PARAMETER["False northing",0,
LENGTHUNIT["metre",1],
ID["EPSG",8807]]],
CS[Cartesian,2],
AXIS["easting (X)",south,
MERIDIAN[45,
ANGLEUNIT["degree",0.0174532925199433]],
ORDER[1],
LENGTHUNIT["metre",1]],
AXIS["northing (Y)",south,
MERIDIAN[135,
ANGLEUNIT["degree",0.0174532925199433]],
ORDER[2],
LENGTHUNIT["metre",1]],
USAGE[
SCOPE["Polar research."],
AREA["Northern hemisphere - north of 60°N onshore and offshore, including Arctic."],
BBOX[60,-180,90,180]],
ID["EPSG",3413]]
What upstream mask product did you use? MAR, RACMO, and HIRHAM
Reason for using this mask
Although the bask mask was Citterio & Ahlstrøm (2013), the bulk of the inputs were from the HIRHAM, MAR, and RACMO RCMs, and significant work was spent aligning these three, and adjusting to match ‘reality’ as defined by citet:citterio_2013: cropping and scaling down RCM pixels that were only partially glaciated, or extrapolating RCM values into neighboring cells that were not glaciated in the RCM but were glaciated in citet:citterio_2013.
Effort required by you if upstream product changed (to a community standard)
A few days to re-validate workflow.
Type: Rasters
Resolution if raster: 1 km
Projection information: EPSG:3413
What upstream mask product did you use? HIRHAM
Projection information
Working with HIRHAM took a lot of extra effort because it's native format is 'rotated pole' with Greenland near the equator. If it is reprojected to EPSG:3413, significant aliasing occurs. Therefore, it was easier to reproject the vectorized hydrologic basins to the HIRHAM domain. This work is fully documented in citet:mankoff_2020 (both the publication and the online digital workbooks that contain everything that went into the work), but that work is ~5 years old, and the specific reasons for the decisions around HIRHAM are not well remembered at this point.
Output is a vector product of stream discharge locations. There is no mask. Metadata provides the nearest internal ice basin for each stream outlet for a variety of products (e.g. Zwally basin, Mouginot basin and sector, Rignot region, etc.)
This work also assumed subglacial routing, which required the use of the BedMachine bed and a surface DEM to estimate ice thickness. Because the ArcticDEM surface DEM was different than both citet:citterio_2013 and BedMachine, hydraulic “jumps” occurred at the ice edge when streams transitioned from subglacial to subaerial. We could not use the BedMachine surface because we need a DEM at higher than 250 m resolution when off-ice for sub-aerial stream routing.
Suggestions from community survey:
What upstream mask product did you use? GIMP mask v1 https://nsidc.org/data/nsidc-0714/versions/1
Reason for using this mask. This mask was the most complete, accurate, highest resolution available at the time when BedMachine was released, and coincided temporally with other datasets.
Effort required by you if upstream product changed (to a community standard). I had to make some minor corrections (e.g. remove supraglacial lakes)
Type: raster
Resolution if raster: 150 m
Projection information: EPSG 3413
What upstream mask product did you use?
The ice sheet mask is set at init, typically inherited from either the SMB forcing product (MAR, RACMO, ...) or from a geometric input dataset (BedMachine).
The ice sheet mask is often first held constant during spinup, but typically released for a projection.
Reason for using this mask
Matching the SMB product is practical and allows to constrain the ice sheet model to the 'observed' mask.
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why?
Masking out the connected periphery can be awkward dynamically. Removing disconnected ice caps and glaciers is more obvious. The reason to do that would be that the ice sheet model is not very good simulating those smaller glaciers.
Effort required by you if upstream product changed (to a community standard)
No problem if ahead of time. But initialisation and spinup are the most expensive. So better to change once.
If you used internal basins, which product and why?
We used internal basins diagnostically, following ISMIP6. They used both IMBIE basin sets (Rignot and Zwally).
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why
So far only simulated entire ice sheets Greenland and Antarctica.
Type: raster
Resolution if raster: 32x32km, 16x16km, 8x8km, 4x4km, 2x2km, 1x1km (GrIS)
Projection information: EPSG 3031, EPSG 3413
Year or years of input product(s): Same as observational datasets, typically early 2000
Internal ice sheet mask is the result of a dynamic ice flow simulation. Unless a fixed mask is imposed, the ice sheet produces it's own mask.
All masks have boundaries. How, and why did you define yours?
Operating on a computationally feasible domain, or trying to match the diagnostic grid (ISMIP6/7).
For GrIS, often masking out Ellesmere Island, Iceland and other land outside of Greenland.
If your product includes distinction between main and peripheral ice sheet, how did you define this? If possible, cite source, equation, pseudo-code, or text description)
CISM has an option to detect disconnected ice and remove it. This is a search algorithm from a known internal divide location. Not applied by default.
CISM can apply an imposed and changing ice mask at runtime. This comes with the ISMIP6 forcing implementation.
Can remove overlapping glaciers in post-processing, based on RGI.
Effort required if your internal product changed (to a community standard)
No effort.
If you used internal basins, how did you define them, or which product and why?
We used internal basins diagnostically, following ISMIP6. They used both IMBIE basin sets (Rignot and Zwally).
Geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier)
So far only simulated entire ice sheets Greenland and Antarctica.
Type: raster
Resolution if raster: adjusted to upstream
Projection information: EPSG 3031, EPSG 3413
Ice sheet model diagnostic output.
Typically interpolated from the native model grid to match a target diagnostic grid (ISMIP6/7, PROTECT) at a given resolution.
Even if the native grid is regular and at the same resolution, it could be offset in x/y to the diagnostic grid
As per https://essd.copernicus.org/preprints/essd-2022-355/ BedMap3 is in progress, and includes,
The Bedmap3 gridded products [...] a newly updated grounding line [...] and updated ice extent
One thing to consider is the balance between a new project defining new ice extent
and grounding line
default values, and existing product(s).
We'll need to balance between here is our best estimate
and here is what is best for the community
.
What upstream mask product did you use? Basin definition from Jeremie Mouginot to define grounded vs floating ice (IceBoundaries_Y2014-2016_Antarctica.tif
may not be published but is similar to IMBIE's basins), and Rock mask from Antarctic Digital Database (Rock_outcrop_medium_res_polygon.tif
)
Reason for using this mask. This mask was the most complete, accurate, highest resolution available at the time when BedMachine was released, and coincided temporally with other datasets.
Effort required by you if upstream product changed (to a community standard). I had to make some minor corrections (grounding line further retreated, etc)
Type: raster
Resolution if raster: 500 m
Projection information: EPSG 3031
[Please enter text below, and fill in sentences where prompted. Remove text like this in brackets. Skip any questions that are not relevant for your product. Optionally select appropriate "labels" from the side-bar (e.g. greenland, antarctica, peripheral)]
[Use the sections below to describe the upstream input mask for your work (if any), masks you generate internally (if any), and downstream output masks you release (if any), reasons for their usage, and cost of changing.]
[For each upstream mask product you used, copy and repeat this section. Or remove if no upstream mask]
What upstream mask product did you use?
HIRHAM5 has two resolutions of 0.11degree and 0.44 degree , with an ice mask derived from data created by the United States Geological Survey (USGS) Earth Resources Obser- vation and Science (EROS) Center and consisting of Ad- vanced Very High Resolution Radiometer (AVHRR) data at a 1 km resolution collected from 1992 to 1993 (Eidenshink and Faudeen, 1994)
Reason for using this mask
Historical reasons I think, it has never been updated
Projection information: equatorial rotated pole
Year or years of input product(s): Constant over time
The ice mask from HIRHAM5
The HIRHAM5 ice mask includes ice shelves. All kind of processing on grounded ice or basins are done after the model has been run.
What upstream mask product did you use?
Reason for using this mask
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why?
Effort required by you if upstream product changed (to a community standard)
If you used internal basins, which product and why?
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why
Type: vector, raster
Resolution if raster:
Projection information:
Year or years of input product(s):
Here's an animation of the final product:
These data products were used to generate our monthly ice masks:
Here's how many line-kilometers of terminus position data we used for each glacier, in each month:
The description below is copied and pasted from the Methods section of the published paper:
We use 237,556 manually derived and AI-derived glacier terminus picks from 1972 to 2022, obtained from the sources described below. We focus our analysis primarily on the years since 1985, during which time 236,328 terminus picks were acquired. Although data coverage is generally poor before 1985, we include all available observations to help constrain the state of the ice sheet at the beginning of our analysis period. Extended Data Fig. 6 shows the temporal distribution of the acquisition times of all terminus picks, which were selected from the following datasets:
AutoTerm: we use 153,281 terminus positions from the AutoTerm dataset55,56, including 153,250 positions acquired since 1985. AutoTerm provides data from several optical and radar satellite sensors, spanning nearly four decades and includes winter data in recent years. Through visual inspection, we found that AutoTerm performs particularly well at some of the 295 glaciers it covers; however, data quality clearly suffers at other glaciers. Terminus-position accuracy is often dependent on satellite sensor and corresponds reasonably well with error estimates that are provided with the AutoTerm data. For our purposes, we inspected all AutoTerm picks visually to manually determine separate error thresholds for each of the 295 glaciers, such that we eliminate all data corresponding to error values that are associated with obvious outliers or asynchronous behaviour. For this reason, we use only 153,281 of the 278,239 terminus positions available in the full AutoTerm dataset.
MEaSUREs weekly to monthly: we use the 21,990 weekly to monthly terminus positions18,57 collected using the Sentinel-1 synthetic aperture radar since January 2015. Although the full dataset contains 23,676 terminus positions, we only use positions whose quality flag is 0.
MEaSUREs Annual v2: we use 3,437 terminus positions from the MEaSUREs v2 dataset58, including 2,987 picks acquired since 1985. We only use the highest-confidence data, with quality flags 0 or 2. Quality flags 1 and 3 correspond to uncertain picks or Landsat-7 SLC-off images and are not used in our study. We also eliminate redundant data by discarding any positions obtained from the same images used in MEaSUREs weekly to monthly data.
CALFIN: we use 19,835 terminus positions from the CALFIN dataset59,60, including 19,665 picks acquired since 1985. In this subset, we have discarded any CALFIN picks in which MEaSUREs Annual v2 manual picks are available for the same satellite image.
TermPicks: we use 39,013 terminus positions from the TermPicks dataset61,62, including 38,436 picks acquired since 1985. We discard any TermPicks data in which MEaSUREs Annual v2 manual picks are available for the same satellite image.
We note that the AutoTerm dataset includes a large amount of Landsat imagery that is also included in the MEaSUREs Annual v2, CALFIN and TermPicks datasets, meaning that there is some redundancy and probably some discrepancies between the various methods of terminus-position picking. We find that AutoTerm provides the most comprehensive record overall, but the width of fjord walls tend to be defined more narrowly in AutoTerm than in other datasets, meaning that the full widths of glaciers are sometimes not captured in AutoTerm. Also, in some cases, the bounding boxes of the AutoTerm picks seem to cut off the full extents of calving-front migration.
Here are the datasets we used:
Observations of terminus positions were combined from the five datasets listed in the section above. We find that human and AI-derived terminus positions both contain errors and often disagree with each other or even themselves. To reduce "flicker" (as ice can sometimes magically appear or disappear if a terminus position pick is incorrect) we constrain the advance rate of the ice masks from one month to the next using observed velocities.
Masks are from Mouginot 2019. The text below is copied and pasted from the methods section of our paper:
We use 260 named Glacier catchments for the GrIS78. To account for terminus activity that may have occurred beyond the extents of the predefined glacier catchments, we extrapolate each catchment downstream following flowlines from our velocity grid. Each catchment is then dilated by 5 km to fill any gaps between extrapolated flowlines and fjord walls or neighbouring catchments. Our extrapolated catchment delineations are shown in Extended Data Fig. 11.
Geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier)
Type: raster
Resolution if raster: 120 m
Projection information: EPSG 3413
What upstream mask product did you use? Ice mask created from the International Geosphere-Biosphere Programme (IGBP) data in combination with 1 km AVHRR data from the period of 1992 to 1993.
Reason for using this mask: This is the default for the MetUM
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why? N/A
Effort required by you if upstream product changed (to a community standard) Would need to also modify the orography fields used by the MetUM.
If you used internal basins, which product and why? N/A
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why N/A
Type: vector, raster
Resolution if raster:
Projection information: [Example: WKT string, proj4 string, EPSG code, or other]
Year or years of input product(s):
Greenland Ice Mapping Project (GIMP) Ice and Ocean Mask
Howat, I.M., A. Negrete, B.E. Smith, 2014, The Greenland Ice Mapping Project (GIMP) land classification and surface elevation datasets, The Cryosphere, 8, 1509-1518, doi:10.5194/tc-8-1509-2014
NOTE that this is the original data product and it comes packaged with the BedMachine dataset for Greenland. I use it via BedMachine.
This mask is provided as part of the BedMachine dataset and is consistent with the bed and surface geometry in BedMachine, required for model initialization.
When initializing the model further back in time (e.g., in the 1980s or during the Little Ice Age), the mask has to be modified to extend ice further into outlet glacier fjords.
If the mask provides similar delineations (ice, ocean, land, etc.), and has a similar resolution, it would not be much effort to switch to the new mask.
I typically use this product for internal basins: https://earth.gsfc.nasa.gov/cryo/data/polar-altimetry/antarctic-and-greenland-drainage-systems
Zwally, H. Jay, Mario B. Giovinetto, Matthew A. Beckley, and Jack L. Saba, 2012, Antarctic and Greenland Drainage Systems, GSFC Cryospheric Sciences Laboratory
Sometimes subset to either (1) drainage basin using above or (2) individual glacier using catchment derived from ice sheet surface velocity observations
Type: raster
Resolution if raster: 150 m (BedMachine provides GIMP mask at this resolution)
Projection information: EPSG3413
Year or years of input product(s): 1999-2002
Remove this section if no downstream output mask from your product
The model can simulate changes in the extent of the ice sheet (e.g., due to outlet glacier retreat). The model uses an unstructured mesh and a levelset method to track the ice extent. The value of the levelset is output at every mesh node at every model timestep and the ice extent can be directly calculated from the levelset.
[For each upstream mask product you used, copy and repeat this section. Or remove if no upstream mask]
PISM does not use upstream masks. Internally, a mask is computed to distinguish land ice, ocean, land, and floating ice. This mask is computed at each time step. For accurate accounting, PISM reports scalar and map-plane field such as ice mass, surface mass balance and grounding line flux. This allows the user to post process PISM simulations any way they want using their preferred masks.
Reason for using this mask
This is best way to ensure conservation of mass within the model.
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why?
N/A
Effort required by you if upstream product changed (to a community standard)
N/A
If you used internal basins, which product and why?
N/A
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why
N/A
Type: vector, raster
N/A
Resolution if raster:
N/A
Projection information: [Example: WKT string, proj4 string, EPSG code, or other]
N/A
Year or years of input product(s):
N/A
[Remove this section if no internal mask for your product]
All masks have boundaries. How, and why did you define yours?
The modeling domain is chosen by the user.
If your product includes distinction between main and peripheral ice sheet, how did you define this? If possible, cite source, equation, pseudo-code, or text description)
PISM does not distinguish between main ice sheet and periphery.
Effort required if your internal product changed (to a community standard)
N/A
If you used internal basins, how did you define them, or which product and why?
N/A
Geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier)
Type: vector, raster
Resolution if raster:
Projection information: [Example: WKT string, proj4 string, EPSG code, or other]
Remove this section if no downstream output mask from your product
What upstream mask product did you use?
BedMachine Antarctica for regions covered by ice, ice thickness, surface elevation and bedrock elevation: https://doi.org/10.1038/s41561-019-0510-8
Reason for using this mask
Comprehensive and consistent dataset of ice geometry
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why?
N/A
Effort required by you if upstream product changed (to a community standard)
Changes in geometry require redoing all the model inialization process (data assimilation, etc.). We do this regularly so effort is limited if dataset includes comprehensive and consistent geometry.
If you used internal basins, which product and why?
Internal basins are not used for the simulations but afterwards to compute regional metrics
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why
We use all Antarctic regions covered by ice reported in BedMachine Antarctica
Type: raster
Resolution if raster: BedMachine Antarctica is 500 m, model resolution varies between 500 m around grounding lines and in shear margins to 20 km inland
Projection information: EPSG 3031
Year or years of input product(s): 2007-2300
Ice sheet models typically include all glaciated areas on the continent, connected ice shelves and peripheral glaciers that may or may not interact with the main ice sheet.
Papers and products:
RGI version 6.
RGI Consortium, 2017. Randolph Glacier Inventory - A Dataset of Global Glacier Outlines, Version 6. Boulder, Colorado USA. NSIDC: National Snow and Ice Data Center. doi: https://doi.org/10.7265/4m1f-gd79.
Reference mask.
We removed connectivity level 2 from the RGI - it was deemed "too close" to the ice sheet proper. It's also "common practice".
Relatively straightforward for future versions. Effort will be to explain large changes to scientific community, IPCC, public, etc.
Often the output is aggregated by RGI region, in netcdf files.
If you're interested in joining,
Watch
for this repository on the main page[For each upstream mask product you used, copy and repeat this section. Or remove if no upstream mask]
What upstream mask product did you use? RGI6, Mouginot, J., & Rignot, E. (2019).
Reason for using this mask: RGI is only thing available for mountain glaciers, Mouginot and Rignot seemed like best option and most widely used
If you modified the upstream mask, for example, to define and split ‘main’ vs. ‘peripheral’, how and why? I had to do lots of modifications, RGI clumps some glaciers together with multiple termini, so I needed to split this. Same with the Greenland Ice Sheet.
Effort required by you if upstream product changed (to a community standard): Would likely still need to modify to deal with multiple termini considered as one glacier, but this project on frontal ablation is not operational.
If you used internal basins, which product and why? RGI6, Mouginot, J., & Rignot, E. (2019) - RGI is only thing available for mountain glaciers, Mouginot and Rignot seemed like best option and most widely used
If you subset to a geographical region (Greenland, Antarctica, sub-region, peripheral, individual glacier), how and why: No subset
Type: vector
Projection information: EPSG 3995
Year or years of input product(s): 2000-2020
Publications:
https://www.nature.com/articles/s41467-022-33231-x
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023GL104095
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021GL096501
What upstream mask product did you use? Greenland (GIMP), Antarctica (Bedmachine 2), Arctic (Ice mask ESA CCI Land Cover User Tool (v.3.10)). By default, it is the Ice mask ESA CCI Land Cover User Tool (v.3.10)) which is used as it provides information also over Alps, Alaska, Patagonia, Himalaya, ... where MAR is also used.
Reason for using this mask: the MAR pre-processing tool is able to use these masks. Having a same data set on a common grid for both hemisphere will be the best
Effort required by you if upstream product changed (to a community standard): having a Netcdf file with lat/lon of each pixel is the best for me and I can easily adapt the MAR pre-processing tool.
Type: raster
Resolution if raster: 1km
Projection information: EPSG 3031 for Antarctica and EPSG 3413 for Greenland/Arctic but the mask does not need to be on these grids.
Year or years of input product(s): the last one
We envision the bulk of the dialog occurring under the discussions section of this website, and actionable items distilled from the discussions will then be worked on under the issues section of this website.
Discussions is where we will work as a community to decide what to address and how to address it.
Issues is where individuals or smaller groups or teams will work on specific actionable items, possibly linked to specific data features or code.
The purpose of this issue (#12) is simply to be pinned at the top of the Issues list and to help people target new items in the best location.
After we have started the discussion within the steering committee, we will write to CRYOLIST (probably January 2024) to call for more contributors.
Here's an animation of the final product:
Radarsat Antarctic Mapping Project (RAMP) Antarctic Mapping Mission-1 coastline used after_coast_continuous.shp
file for late 1997 and the RAMP Modified Antarctic Mapping Mission cst2000line.shp
, which corresponds to late 2000.
MODIS Mosaic of Antarctica coastlines We use coastlines provided by the MODIS Mosaic of Antarctica (MOA) project for the years 2004, 2009 and 2014.
Circum-Antarctic landfast sea ice extent, 2000-2018 Annually from March 2000 to March 2017, we interpolate ice-sheet areas in a 1-km-resolution optical- and thermal-band Moderate Resolution Imaging Spectroradiometer (MODIS)-based fast-ice dataset to our 240-m-resolution grid.
Sentinel 1a Annually from March 2015 to March 2021, we digitize coastlines using the same methods that were used to generate the annual MODIS-based fast-ice dataset, but we apply the methods to Sentinel 1a radar image mosaics from http://seaice.dk/ that were collected concurrently with the MODIS data each March, thus providing two independent datasets obtained from different types of sensor.
All coastlines were masked to the 240-m-resolution ITS_LIVE Antarctic mosaic grid. We immediately found that the uncorrected masks from each dataset are not directly intercomparable, as different research groups sometimes differ in how they designate ice types, different sensors vary in their ability to differentiate shelf ice from sea ice, and certain islands are included in some coastline datasets but neglected in others. To ensure that any changes we see from one coastline mapping to the next reflect true changes in coastline rather than changes in sensors or methodology, we create a composite dataset by adjusting each contributing dataset as follows. After each adjustment described in the following, any potential new holes in any ice mask are filled to preserve a continuous ice sheet. (1) Remove the attached iceberg D15 from all datasets, because D15 calved from West Ice Shelf in 1992, then abutted the coast for decades after62 and appears in some coastline products but not others. (2) Remove any islands that are not present in all datasets. (3) Remove from the MODIS-based data any ice that is present in all 18 MODIS-based mappings but does not appear in any other datasets. Add to all MODIS-based data any ice that is present in all other datasets but does not appear in any of the MODIS-based data. (4) Remove from the Sentinel 1a-based data any ice that is present in all seven years of Sentinel 1a mappings but does not appear in any other datasets. Add to all Sentinel 1a data any ice that is present in all other datasets but does not appear in any of the Sentinel 1a data. (5) Remove from the MOA-based data any ice that is present in all three MOA-based mappings but does not appear in any other datasets. Add to all MOA data any ice that is present in all other datasets but does not appear in any of the MOA-based data. (6) Add to all RAMP-based data any ice that is not present in either of the RAMP mappings but is present in all other datasets. (7) Using the years 2015, 2016, 2017 and 2021 when the same techniques were applied to MODIS and Sentinel 1a data collected concurrently, we remove from both datasets any ice that does not appear at least once in each dataset. (8) Add to all datasets any ice that is present within all three MOA coastlines, but is not present in any of the three concurrent MODIS-based mappings. (9) Use projected x and y velocity components vx and vy to calculate grid-cell centre displacements that should occur, in increments of six months, up to six years for the entire grid. For this step we multiply vx and vy by 1.1, which allows some tolerance for variations in velocity, which we describe below. (10) It is evident by inspection that the 1997 RAMP coastline was digitized at higher resolution and with greater care than the 2000 RAMP coastline. Therefore, we put some faith in the 1997 coastline as a more reliable reference. Using the expected displacements from the velocity fields to tell us the maximum amount of coastline growth that could possibly occur from 1997 to 2000, including the 10% velocity tolerance, any ice that is present in the 2000 mask but is missing from the 1997 mask, and could not have advected to the new location in just three years, is removed from the 2000 mask. (11) Adjust the MOA2004 mask following the same logic described in the previous step, but tie MOA2004 to RAMP2000 and MOA2009. Any pixels in MOA2004 that could have advected from ice that is present in RAMP2000 and will advect to a location that will appear as ice in MOA2009, then must be ice in MOA2004. Similarly, any ice that is present in MOA2004, but could not have advected there from RAMP2000 and will not advect to an ice location in MOA2009, cannot be ice in MOA2004. Adjust MOA2004 accordingly. (12) Follow the previous step to tie MOA2009 to the adjusted MOA2004 and MOA2014. (13) Tie MOA2014 to MOA2009 and any ice that is present in both the Sentinel 1a- and MODIS-based mappings for 2021. (14) The adjusted RAMP, MOA and combined 2021 mappings now serve as anchors to tie the MODIS- and Sentinel 1a-based mappings. Following the same method described above, tie each annual MODIS- and Sentinel 1a-based mapping to the closest past and future anchor mappings.
The resulting 24 ice masks achieve higher resolution than the underlying MODIS or Sentinel 1a mosaics because we exploit the offset of the MODIS and Sentinel 1a mosaic grids, and because we use known velocity to interpolate coastline migration between coarse-resolution grid postings to create our 240-m grid.
We note that some islands do not appear in every constituent dataset, so through the methods described above, we have constrained them to a constant area, or a nearly constant area in some cases where the algorithm introduces or otherwise allows a small amount of noise. As a result, if any bias exists in our overall estimates of ice-sheet-area change, we suspect it would be towards underestimation of area reduction, because most islands are small, susceptible to changes in their environment and located around the Antarctic Peninsula, where major reductions in ice-shelf area are known to have occurred.
We have the most confidence in our ice-shelf-area time series that show the largest amplitudes of change, so we recommend considering the range of values that are presented in Supplementary Information and Supplementary Table 1 when interpreting the time series of smaller ice shelves or ice shelves that are adjacent to any islands that may not appear in all contributing datasets.
To quantify changes in the area and mass of each ice shelf, we create a gridded mask of 181 ice-shelf names based on the MEaSUREs Antarctic Boundaries for IPY 2007–2009 from Satellite Radar, Version 258. We dilate the boundaries of each ice shelf by 100 km in all directions, then use constant extrapolation along flowlines following the procedure described above for velocity and thickness. The result shown in Extended Data Fig. 3 is a set of masks that cover areas much larger than any observed ice-shelf extents, but are certain to fully capture changes at the ice front while providing extra tolerance in the grounding zone, which is beneficial when working with multiple datasets that may have been created with different grounding-line masks. The ice-shelf mask in Extended Data Fig. 3 shows the fully dilated ice-shelf areas; although, to be clear, we limit analysis of the area and mass of each ice shelf to pixels where ice is observed and the BedMachine mask indicates ice shelf or ocean.
Type: raster
Resolution if raster: 240 m
Projection information: EPSG 3031
Raster data available here for now. The dataset is currently in review at NSIDC. We've submitted a revised version of the NetCDF which will be posted to NSIDC soon. Only the metadata have been changed in the revisions, and the ice mask data itself will not change from the original version here.
A vectorized version of the data is available here.
Ken and I just had a call to get things started while he is off to Antarctica.
We agree to use discussions to start brainstorming, and find people / assign responsibilities.
I'll continue to work on formalizing this process next week. I'd like to have a few discussion threads on the forum and a few instructions about how / what to contribute online before we start to advertise more broadly.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.