[Liwg-core] Fwd: ISMIP6 grid for Greenland

William Lipscomb lipscomb at ucar.edu
Thu Sep 21 13:38:00 MDT 2017

Hi all,

I'm forwarding a couple of emails from Heiko Goelzer. Earlier this month
(see his email of Sept. 1), he began a discussion about the use of standard
grids for displaying Greenland ice sheet results for ISMIP6/CMIP and other
community efforts. The discussion is summarized in his most recent email

There seems to be a consensus to switch to the EPSG_3413 projection (as
already done for CESM/CISM), with broad agreement on other issues.

I was wondering if any of you would like to add to the discussion.  In
particular, Heiko proposed a paleo-Greenland grid that is extended relative
to the present-day grid.  Does that grid look like it would suit your
purposes?  I've attached copies of his proposed present-day and paleo grid

Note that these grids would be for reporting diagnostics, not necessarily
for running the ice sheet model. But if we can run CISM using the same
projections and resolutions, it makes our lives a bit easier.


Bill L.

---------- Forwarded message ----------
From: Heiko Goelzer <h.goelzer at uu.nl>
Date: Thu, Sep 21, 2017 at 7:03 AM
Subject: Re: ISMIP6 grid for Greenland
To: "sophie.nowicki at nasa.gov" <sophie.nowicki at nasa.gov>, "
Tamsin.Edwards at open.ac.uk" <Tamsin.Edwards at open.ac.uk>, "
abeouchi at aori.u-tokyo.ac.jp" <abeouchi at aori.u-tokyo.ac.jp>, "
aaschwanden at alaska.edu" <aaschwanden at alaska.edu>, "calov at pik-potsdam.de" <
calov at pik-potsdam.de>, "olivier.gagliardini at univ-grenoble-alpes.fr" <
olivier.gagliardini at univ-grenoble-alpes.fr>, "
fabien.gillet-chaulet at univ-grenoble-alpes.fr" <
fabien.gillet-chaulet at univ-grenoble-alpes.fr>, "Nicholas.Golledge at vuw.ac.nz"
<Nicholas.Golledge at vuw.ac.nz>, "greve at lowtem.hokudai.ac.jp" <
greve at lowtem.hokudai.ac.jp>, "Angelika.Humbert at awi.de" <
Angelika.Humbert at awi.de>, "phuybrec at vub.ac.be" <phuybrec at vub.ac.be>, "
kennedyjh at ornl.gov" <kennedyjh at ornl.gov>, "eric.larour at jpl.nasa.gov" <
eric.larour at jpl.nasa.gov>, William Lipscomb <lipscomb at ucar.edu>, "
sebastien.leclech at lsce.ipsl.fr" <sebastien.leclech at lsce.ipsl.fr>, "
V.Lee at bristol.ac.uk" <V.Lee at bristol.ac.uk>, "fpattyn at ulb.ac.be" <
fpattyn at ulb.ac.be>, "A.J.Payne at bristol.ac.uk" <A.J.Payne at bristol.ac.uk>, "
cr at dmi.dk" <cr at dmi.dk>, "martin.rueckamp at awi.de" <martin.rueckamp at awi.de>, "
isaitofuyuki at jamstec.go.jp" <isaitofuyuki at jamstec.go.jp>, "
nicole-jeanne.schlegel at jpl.nasa.gov" <nicole-jeanne.schlegel at jpl.nasa.gov>,
"Helene.Seroussi at jpl.nasa.gov" <Helene.Seroussi at jpl.nasa.gov>, "
Sainan.Sun at ulb.ac.be" <Sainan.Sun at ulb.ac.be>, "Wal, R.S.W. van de
(roderik)" <R.S.W.vandeWal at uu.nl>, "florian.ziemen at mpimet.mpg.de" <
florian.ziemen at mpimet.mpg.de>, "Price, Stephen F" <sprice at lanl.gov>, Rob
Deconto <deconto at geo.umass.edu>, Dave Pollard <dxp21 at psu.edu>
Cc: "mathieu.morlighem at uci.edu" <mathieu.morlighem at uci.edu>, "
ismip6-leads at climate-cryosphere.org" <ismip6-leads at climate-cryosphere.org>

Dear all,

thanks for your comments and suggestions about the grid. I am attempting a
summary to reflect the current state of the discussion, add some
considerations and possibly wrap up.

- EPSG_3413: so far general agreement. In my opinion that's the most
important aspect of the new grid question.

- Higher grid resolution than 1 km: As a reminder, the main purpose of the
grid is as a diagnostic grid to register Greenland ice sheet results for
ISMIP6/CMIP and other community efforts. It is already foreseen that
results can be submitted at different (but specific) resolutions and as
mentioned by Ralf, higher resolutions are always possible with the grid as
proposed. So, we are good for now and for the future. In my experience as
an analyst, there is a limit to what resolution is still feasible to
manage. No doubt detail is important for the physics and we should
encourage high resolution, but we have to provide results (also) at a
resolution that can be easily analysed. For me, 1 km was already beyond
that point, hence asking for output at 5 km for initMIP. A future MIP on
our favourite outlet glacier could indeed define a subset at very high
resolution if needed and a paleo MIP could ask for every model to run at,
say 20 km.

- Grid registration: Questions arose about the grid ranges and false
easting/northing for the grid. Mathieu has suggested to use
node-registration, and to only specify the coordinates of the cell-centers
in x and y coordinates (not the ranges) to avoid confusion about grid
limits. In other words, we will specify the x and y points where model
output should be defined. In our 1 km example, the x/y-coordinates of the
lower left cell *center* are (-720000m,-3450000m) with nx=1681 and ny=2881
cells in x and y-direction at full km positions. (xmin =  -720 km, xmax =
+960 km, ymin = -3450 km, ymax = -570 km, thanks Ralf).
Background reading for this discussion point is here

- Reference resolution: the questions here boil down to whether to try to
match the resolution/spacing of a particular data set (BedMachine in this
case). This is an important point, so some background. ISMIP6 has decided
early on not to promote any particular data set. Our decision to move to
EPSG_3413 has certainly been encouraged by existence of BedMachine and its
use by many modellers, but decisive is that EPSG_3413 is standardised and
that *many* data sets (ESA, NASA, ...) are released for this projection.
Data sets exist at various different resolutions already (30m, 90m, 150m,
1000m) and providers of new data sets will typically chose an ideal
resolution to maximise detail in their products. This will equally be the
case for future generations of BedMachine. I believe with a reference
resolution of 1000 m (which can be further subdivided if needed: 100, 125,
200, 250, 500) we create a good chance to match future releases (as integer
subdivisions). We also can and should of course try to convince data set
providers (like Mathieu) to meet an integer subdivision if possible. I can
see that 1200 m offers additional subdivision by 3 and 6 that for 1km give
somewhat uneven resolution numbers (that maybe nobody would use), but not
sure it is worth it, considering resulting numbers for the lower

- Paleo grid: No return on this one so far. I would suggest to postpone the
question when the need arises. This could be any time soon, as we are
discussing the ISMIP6 Last Interglacial experiments for Greenland next week
at the PMIP conference. However, an extension of the proposed grid to a
larger domain seems unproblematic.

Is there something I've overlooked? Mathieu, do you want to add something?

Best greetings


On 01/09/2017 16:23, Heiko Goelzer wrote:

> Dear initMIP-Greenland participants and interested parties
> this email is about defining a new grid for Greenland ice sheet model
> results within ISMIP6. We would be interested in your opinion. Please read
> on for more details (apologies for the long email).
> The initMIP-Greenland experiments have been operated (and also performed
> by some fixed-grid models) on the Bamber grid. In the course of the project
> it has been pointed out that the projection used by the Bamber grid is not
> standardised (important e.g. for smooth QGIS analysis and other processing)
> and that many observational data sets are becoming available on a different
> projection (epsg_3413). Some modellers have already started to move to this
> projection. This is why we (ISMIP6 SSC) have agreed to find a new grid on
> the epsg_3413 projection for ISMIP6 on Greenland.
> The goal of defining a new grid now is directed towards using it for other
> ISMIP6 experiments of the GrIS (most importantly the future projections).
> The most important bit is changing the projection to epsg_3413, because
> this will avoid re-projection in pre- and post-processing, which produces
> confusion and mistakes.
> The new grid will become the standard diagnostic grid to report and
> archive results within ISMIP6 and in the CMIP archives. Similarity to the
> grid used for existing and upcoming observational data sets is therefore
> important, because we anticipate the ice sheet results to be used by a
> broader audience and analysts that may want to compare to observations.
> Multiple resolutions (see next point) is also useful here, because it will
> allow storing ice sheet results adopted to the grid resolution of the
> native grids.
> Another aspect is that the grid will be used by many fixed-grid modellers
> as native grid. This has been the case with the Bamber grid, which has
> become the 'quasi-standard' for that purpose for many years, mainly because
> of the use in both ice2sea and SeaRISE. It is therefore essential that the
> new grid is practical from a fixed grid perspective. The main point is that
> the grid allows for different and multiple grid resolutions in the range of
> what modellers use now and in the future.
> If we start with a highest resolution of e.g. 1 km, the grid should
> accommodate sub-setting to lower resolutions by full integer division. This
> is best achieved with an uneven grid size n, where m=n-1 can be divided by
> many commonly used grid resolutions r in km.
> An example is n=2161, where m=2160 divides by r=[1 2 4 5 8 10 15 16 20 30
> 32 40]
> The grid size for those grid resolutions r is then
> n_r=(n-1)/r+1;
> Based on this logic, we would first suggest the following present-day
> grid, which covers all of Greenland and extends a bit outside of the
> present coast-line. The margins are relatively tight around, so it is not
> directly suitable for paleo-modelling. So this would be for present-day
> diagnostic output only (see next point).
> The problem in defining a larger grid is really where to stop. If we try
> to accommodate glacial grounded ice extend, why not accommodate glacial
> floating ice extent, why not a bit more? Instead, we can propose a natural
> extension of the grid with additional margin, see below. Given that ISMIP6
> is focussed on future projections and that additional grid size means
> increasing storage and processing costs, we believe this may be the way to
> go.
> The paleo grid extension below is the next that fulfils all the multiple
> resolution criteria, but may be too large from a practical point of view.
> Here I am particularly interested in input from people running (glacial)
> paleo setups (in terms of size and what resolutions should be covered). It
> is clear that it would be easy to extract the PD diagnostic grid
> information from any model, even if it has been run with a larger native
> grid. It would therefore also be possible to only define the present-day
> grid and leave the rest for individual needs of modellers/projects.
> I attach figures that show Bamber bedrock elevation and mask projected for
> the selected grids to get an idea of the sizes. Below are the associated
> parameters, generally operating with pixel centers not corners!
> We would be interested in any comments on the proposed grid and if/how a
>  paleo extension should be defined in your opinion.
> It would also be good to get an estimate (i.e. time scale) of how
> difficult it would be for you to move to the new grid with your
> pre-/post-processing routines and/or fixed grid models and what could be
> done to facilitate that move from our side.
> We'll try a 'reply all' discussion style until the first complaint to move
> to a more controlled response format. OK? If you feel an important voice is
> missing in the conversation, please let us know, too.
> Best greetings
>      Heiko and Mathieu
> # Small PD grid
> proj_info.earthradius=6378137.0;
> proj_info.eccentricity=0.081819190842621;
> proj_info.falseeasting=720000;
> proj_info.falsenorthing=3450000;
> proj_info.standard_parallel=70.;
> proj_info.longitude_rot=315.;
> %1km resolution
> grid(1).dx=1000.;
> grid(1).dy=1000.;
> grid(1).nx_centers=1681;
> grid(1).ny_centers=2881;
> %10 km
> grid(1).dx=10000.;
> grid(1).dy=10000.;
> grid(1).nx_centers=169;
> grid(1).ny_centers=289;
> # Bigger paleo grid
> proj_info.earthradius=6378137.0;
> proj_info.eccentricity=0.081819190842621;
> proj_info.falseeasting=960000;
> proj_info.falsenorthing=3690000;
> proj_info.standard_parallel=70.;
> proj_info.longitude_rot=315.;
> %%1km resolution
> grid(1).dx=1000.;
> grid(1).dy=1000.;
> grid(1).nx_centers=2161;
> grid(1).ny_centers=3361;
> %10km resolution
> grid(1).dx=10000.;
> grid(1).dy=10000.;
> grid(1).nx_centers=217;
> grid(1).ny_centers=337;
Heiko Goelzer
PostDoctoral Researcher
IMAU, Utrecht University
Princetonplein 5, 3584 CC Utrecht
PO box 80000, 3508 TA Utrecht
The Netherlands

William Lipscomb
Climate & Global Dynamics
National Center for Atmospheric Research
1850 Table Mesa Drive
Boulder, CO 80305
(505) 699-8016
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cgd.ucar.edu/pipermail/liwg-core/attachments/20170921/f454eae2/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PDGrid_mask.png
Type: image/png
Size: 49228 bytes
Desc: not available
URL: <http://mailman.cgd.ucar.edu/pipermail/liwg-core/attachments/20170921/f454eae2/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PaleoGrid_mask.png
Type: image/png
Size: 39533 bytes
Desc: not available
URL: <http://mailman.cgd.ucar.edu/pipermail/liwg-core/attachments/20170921/f454eae2/attachment-0003.png>

More information about the Liwg-core mailing list