[Liwg-core] target: ablation areas

Lenaerts, J.T.M. (Jan) j.lenaerts at uu.nl
Thu Jan 26 12:20:31 MST 2017


Hi all,

Based on the flood of emails, I can summarize the following:

- the climate over Greenland looks very good, except for the north - northeast - east region, where sea ice is extensive and ocean, tundra, and ice sheet too cold.

- melt rates are lower than in RACMO2, most prominently in the cold regions (see above).

- the firn physics developed by Leo (10 m w.e. Hmax, density changes) are based on empirical and/or theoretical evidence, are tested in detail, and included in the CESM2 code, so we should try to prevent any retractions of these. This wil not be in time for the code freeze and not acceptable for the community.

- any of the sensitivity tests Leo performed so far with 126/129 have relatively little effect on the melt and SMB, including lowering ice albedo, LW downscaling, snow fraction bug, etc.. Although these runs are short, they show that the model is stubborn for changes (which is bad), but also robust (which is good!). I agree with Bill L. about the strategy to start ‘cold runs’. Leo has done some of these ‘cold starts’ already to investigate the sensitivity to initial and maximum snow thickness, so that should be no problem.


That said, I have two personal comments:

1. I think we are approaching the end of the testing phase, and we should report to the community - mostly the CAM and CICE groups - that we are concerned about the high Arctic cold bias in CESM. Possibly sea ice albedo could be tuned. Miren, feel free to contact Dave Bailey and please keep us in the loop. Leo, could you group your most important findings in a powerpoint and send it over to J-F and us, so he is aware that you (and the rest of us) are working hard on this?!

2. The Last Resort: one possible additional test could be to lower snow albedos, and test the sensitivity to the grain size of refrozen snow (i.e. increase it to 2000 microns), and/or include a fixed amount of soot in the Greenland snow. Using RACMO2 and CESM, we have shown before that these changes (potentially strongly) impact snow melting, and none of the two are well constrained by observations at all (so there is some parameter space allowed). However, I am not sure if such changes would be acceptable for the land modellers and the community as a whole.


Thanks!

Jan



On 26 Jan 2017, at 17:59, Bill Sacks <sacks at ucar.edu<mailto:sacks at ucar.edu>> wrote:

Hi Bill & others,

Cold start = 100 mm SWE.

I'm not sure exactly how to reset the snow pack, but I think it would "just" be a matter of temporarily deleting the restartvar calls for various snow variables. The tricky part of this is that (as I discovered recently) if you reset some snow variables but not others, you can get a balance error at the start of therun, so this might take some trial & error.

We should be good on our allocation: I just moved over a bunch of core-hours from cheyenne to yellowstone. On dev (if I recall correctly) we have barely used any of our allocation yet.

Bill S

On Jan 26, 2017, at 9:28 AM, Lipscomb, William Henry <lipscomb at lanl.gov<mailto:lipscomb at lanl.gov>> wrote:


One clarification:  When I say “cold start”, I mean specifically CLM glacier columns, not other CESM components.  This might mean simply taking a CLM restart file and overwriting the initial snow thickness and temperature with the default cold-start values.  Bill S. or others, do you have thoughts on how best to do this?

Bill L.


On Jan 26, 2017, at 9:18 AM, Lipscomb, William Henry <lipscomb at lanl.gov<mailto:lipscomb at lanl.gov>> wrote:

Hi all,

Thanks to Leo for starting the latest experiments.  I’m unfamiliar with how ground heat flux is computed, so I’m not sure how much an effect the snow conductivity could be having on the melt.  I’d be in favor of running one or two tests with modified snow density (within reasonable physical bounds) just to have a sense of how much difference it makes.

Following up on earlier threads:  I talked with Miren this morning, and wanted to suggest an experimental protocol for testing possible improvements to melting.

My understanding is that some of the recent runs were spun off from simulations that start from a thick 10-m snowpack in much of Greenland.  This means it could take a long time to expose all the areas that would have ice melting (negative QICE) in equilibrium.  With a run of ~10 years, it's hard to be sure how far we are from our target SMB.

An alternative would be to begin coupled simulations from cold start with a relatively low initial value of Hsno for Greenland glacier columns.  (I’ve forgotten what the default is—is it 0.1 m, or 0.5 m, or something else?)  Then we run the model for a decade or so.  That won’t be long enough to develop snowcapping in the entire accumulation zone, so total accumulation would be underestimated.  But what we most care about for now is having more extensive ablation regions.  I’m guessing that a few simulation years would be enough to melt through a thin initial snow pack and start melting ice wherever it’s going to be melting in the long term.

I’m also thinking that much of the tundra is too cold, and snow is still going to accumulate where we wish it would melt.  But this approach would give us more certainty.

Leo and Jan, does that make sense to you?  If so, we could start with a control experiment using our current preferred parameter choices, and then run sensitivity experiments (e.g., revised snow density, artificially warmer surface air temperatures, modified sea ice parameters) and compare.

And a question for Bill S.:  How are we doing with our allocation, and how many coupled years can we reasonably plan to run in the near future?

Thanks,

Bill L.

On Jan 26, 2017, at 8:49 AM, Kampenhout, L. van (Leo) <L.vanKampenhout at uu.nl<mailto:L.vanKampenhout at uu.nl>> wrote:

Hi all,

I’ve started two new experiments B6 and B7 where I perturb the surface temperature, as per Bill Lipscomb’s suggestion. That is, I only apply the perturbation over downscaled columns, so grid cells without any glaciers are not affected.
https://docs.google.com/document/d/1l6SU0PPba8eOuB8fiI3LYVDdyDMrwZkQf2lyh7Ztaqw/edit?usp=sharing

The runs haven’t finished their first year yet, but the first year already reveals a significant response:

B3 total JJA melt = 182.2 Gt (average over 6 years)
B6 total JJA melt = 179.8 Gt
B7 total JJA melt = 261.7 Gt

B6 melt exceeds B3 melt in June and July, but suffers from a cold August. I’ll send an update to these numbers when the runs have reached at least 6 years of runtime (probably tomorrow). Arguably 6 years is not long enough for the model to adjust, but it hopefully gives an insight into the sensitivity.

In the meantime, I’ve been looking at some regional differences, see this Powerpoint<https://drive.google.com/file/d/0BwZ8r8pyf3aUUnhxZFNvS2w1aUk/view?usp=sharing>. It demonstrates that the regional differences are large, with the West least biased, and the North & East most.




Regarding revisiting snow density: if snow density would inhibit melt, we would see a large ground heat fluxes (GHF), or not? Although GHF is regionally a little on the high side, it does not explain the melt biases. Indeed, the GHF could even be a little high due to the changes that we made in this run, which make the soil a little warmer w.r.t. standard 129 (which has the fractional snow cover bug). Moreover, in my head lower surface densities correspond to higher albedo, which could be a bad thing. However, I’m willing to give this a try if there is a clear wish to do so. Personally, I believe the 129 is still a little too cold, so when it would warm we start to see a much improved SMB simulation. It’s not only the ice sheet that is cold, also the tundra in the North. But then, I could be wrong.

Cheers,
Leo



On 26 Jan 2017, at 14:53, Lipscomb, William Henry <lipscomb at lanl.gov<mailto:lipscomb at lanl.gov>> wrote:

Hi Miren,

I agree that if we had sufficient snow melt to expose bare ice in (at least most of) these ablation areas, we’d have a much improved SMB simulation.  Without more snow melt, it’s hard to see how we get there.

The snow physics is much improved from CESM1 (thanks to Leo and Jan), and I’d be very reluctant to backtrack.  My sense is that the snow albedo has been pretty well validated, and any tuning we did could have undesirable effects outside ice sheets, so that’s not a good option.  We do now have a fixed treatment of fractional snow (thanks again to Leo), which should give us weighting similar to RACMO.  I like Leo’s idea of revisiting the snow density (looking at thermal effects that could inhibit melt).

With respect to sub-grid simulation, our potential tuning knobs include elevation-based downscaling of temperature, radiation and precip.  Currently we downscale temperature and LW, but not SW or precip. (To be precise, we downscale the phase of precip via the temperature ramp, but not the amount.)
* I’m not aware of a simple, well validated downscaling scheme for either SW or precip—is there such a scheme that we’ve overlooked?
* Leo has been experimenting with LW, and my understanding is that turning off the downscaling makes the melt smaller rather than bigger.
* Has anyone looked at sensitivity to the prescribed temperature lapse rate?  I’d be surprised if it made a big difference, but you never know.

My guess is that climate forcing is the major issue.  So I think it’s a good idea to run the column experiments you mention and to contact Dave B.

Since many of the recent runs have been short, I’m not quite clear on how much the initial snow thickness matters and how long it takes the model to equilibrate.  So I also agree it would be good to look at Hinit.  Could you please remind me what value of Hinit we’ve been using for Greenland?

Thanks,

Bill L.

On Jan 26, 2017, at 2:52 AM, Miren Vizcaino <M.Vizcaino at tudelft.nl<mailto:M.Vizcaino at tudelft.nl>> wrote:

Morning, -

Let’s plan on immediate steps for the LIWG towards CESM2.0

Target: having sufficient snow melt in (all) the observed ablation areas as to expose bare ice.

Motivation: ablation areas expose bare ice at some point during the melt season (this is confirmed by remote sensing - I checked with Stef - and regional modeling - this is how RACMO operates. He told me there is no super-imposed ice in the model, so ice albedo comes from getting rid of the snow. There is some weighting in RACMO between bare ice albedos - around 0.4, that come from MODIS- and snow albedos, when the snow thickness is low)

Where we are now: difficulties to get the ablation areas in the north and northwest, and too narrow SW ablation area

Please let me know asap if you disagree with target.

There are three ingredients here:

1- snow physics
2- climate forcing
3- sub-grid simulation

1-

It could be worthwhile to revisit some of the changes that I made to fresh snow density as perhaps this prevents early melt and associated albedo changes.  But I doubt whether this solves the bias completely.

Leo, would you like to discuss how to do some testing here?

2-

I'd like to think that CESM suffers from a cold bias in the North and East of about 2 degrees:
https://www.dropbox.com/s/wqu2n4f6ad5lemb/maps_JJA_b.e20.B1850.f09_g16.pi_control.all.129_liwg_Hmax1m_0019-0021.pdf?dl=0


Raymond is going to look to the sensitivity of snow melt to the forcing in N Greenland with the column model

I’ll contact Dave Bailey

I vaguely remember from CESM1.0 that there was N cold bias and permanent snow pack over the tundra, - same problem?-, but ablation areas where ok. Perhaps the bias in snow melt from wrong densities explains this.

3-  Hinit, Hmax : here we have a bunch of runs, none very conclusive, because of bugs, multiple things changed, too short runs (e.g., B5 is only 3-years-long).

Hinit seems to matter, and should be 0.1-0.5 m w.e. both for ice-sheet and tundra - For Hmax, let’s address it at a later stage

Thanks, Miren


_______________________________________________
Liwg-core mailing list
Liwg-core at cgd.ucar.edu<mailto:Liwg-core at cgd.ucar.edu>
http://mailman.cgd.ucar.edu/mailman/listinfo/liwg-core

---
William Lipscomb
Los Alamos National Laboratory
Group T-3, MS B216
Los Alamos, NM 87545
505-667-0395




_______________________________________________
Liwg-core mailing list
Liwg-core at cgd.ucar.edu<mailto:Liwg-core at cgd.ucar.edu>
http://mailman.cgd.ucar.edu/mailman/listinfo/liwg-core


---
William Lipscomb
Los Alamos National Laboratory
Group T-3, MS B216
Los Alamos, NM 87545
505-667-0395





---
William Lipscomb
Los Alamos National Laboratory
Group T-3, MS B216
Los Alamos, NM 87545
505-667-0395




_______________________________________________
Liwg-core mailing list
Liwg-core at cgd.ucar.edu<mailto:Liwg-core at cgd.ucar.edu>
http://mailman.cgd.ucar.edu/mailman/listinfo/liwg-core

_______________________________________________
Liwg-core mailing list
Liwg-core at cgd.ucar.edu<mailto:Liwg-core at cgd.ucar.edu>
http://mailman.cgd.ucar.edu/mailman/listinfo/liwg-core

- - - - - - - - - - - - - - - - - - - - - - - - - - -
Jan Lenaerts
IMAU, Utrecht University
@lenaertsjan<https://twitter.com/lenaertsjan> || IMAU website<http://www.staff.science.uu.nl/~lenae101/> || CU<http://www.colorado.edu/lab/icesheetclimate/> website<http://www.colorado.edu/lab/icesheetclimate/>





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cgd.ucar.edu/pipermail/liwg-core/attachments/20170126/dc684844/attachment-0001.html>


More information about the Liwg-core mailing list