Model implementation

[gavinny]

This is the discussion on how climate models can/should implement the input of freshwater fluxes into the ocean.

[gavinny]

Consensus thoughts on the model implementation:

Different approaches needed for models that fix discharge at PI levels (Type-1) vs. models that maintain ice sheet mass balance (Type-2). Consequences are possibly different for global mass balance and ocean freshwater sources.

Commonalities:
Separation of spatial and time-varying parts.
Regionalisation of fluxes to large drainage basins. Maybe 5 to 7 based on existing practice.
Spatial domain of iceberg melt and depth melt profiles can be derived from iceberg resolving models (Merino et al, 2016, Moon, Enderlin et al)
Depth profiles can be provided for the basal runoff/melt terms taken from fjord/calving front modeling.
Pre-industrial controls stay the same (ie mass balance is assumed).
Annual resolution is sufficient (need to fill 1850-2021, with extensions to 2023 desirable). Link to CMIP7 protocol.
Simplifications can be to only use spatial map for icebergs (not depth), amalgamate regions per hemisphere, smooth timeseries. Groups can also pick and choose regions or to use iceberg flux or not depending on model capabilities.

Suggestion that climatological fluxes by treated the same way as the anomalous flux - which may involve some new coding for existing models. This is not essential, but makes dealing with possibly negative FW anomalies easier (since the total discharge is never going to be negative).

In historical period: Estimates of mass imbalance from multiple sources (see Bamber, Davison notes), including floating ice shelf variations and trends (more important in Antarctica). Possible use of ensemble mean calibrated output from Coulon et al (1950 onwards) as estimate of forced trend?

Scenarios from 2022 - for the time being, derived from existing ISMIP6 or other single model efforts (Coulon et al, Edwards group etc., UKESM?). Reconciliation with historical obs will be needed. Secondary priority. Note that anomalous flux in scenarios can be of both signs (i.e. net gain of mass in Antarctica for instance).

Implementation:

Type 1 approach:

These models will have imbalances in the historical runs due to changes in the SMB on the ice sheets - for instance an increase in precipitation due to rising water vapor and convergence over Antarctica. Adding the anomalous flux to the PIControl discharge will provide a more accurate discharge into the Southern Ocean, but the implicit mass balance on the ice sheets will be different to observed (or simulated by the ISM), and so the overall ocean salinity/sea level will not necessarily match observations.

Type 2 approach:

These models assume mass balance of the ice sheets, via a change in iceberg/runoff flux (however implemented) on either instantaneous or decadal scale.
Since there is an assumed balance, addition of the Anomalous FW terms will make the implicit mass balance of the ice sheets match observations and consequently the ocean salinity/sea level change will match expectations. However, the total discharge into the oceans may be different from observations depending on the accuracy of the resolved net integrated SMB.

Assessment of the errors associated with each approach should be included in any paper.

[nicojourdain]

For type-2 models that remain type-2 throughout projections, we could discuss what to do with the freshwater anomaly that compensates future precipitation anomalies over the ice sheet. If the aim is really to close a global budget, then the freshwater anomaly could be spread globally, not under the form of a physical glacial freshwater flux around the ice sheets (which has strong impacts on sea ice). Then we can prescribe historical freshwater fluxes on top of this in the way we discussed at this meeting, without adding too unrealistic freshwater near ice sheets.

[swartn]

Thanks @nicojourdain - I think this makes sense to me conceptually - sorry I think I misunderstood that you proposed to spread the whole flux, not just the anomaly part. From a practical standpoint of implementation across modelling centres, it would be interesting to see. From my perspective, some things I would wonder about would be, is there a potential to introduce a drift, and for which experiments to apply the constant background/anomaly spreading - only those with specified anomalous FW forcing, or for every experiment even if no additional FW forcing anomaly is applied?

[gavinny]

I think it might be useful to think about what the biases in the ESM might be due to as we consider how to work around them. For instance, excessive precipitation on the AIS could be due to too diffusive water vapor transport up on to the plateau. In the real world, this vapor rains out before it gets there. So a fix that puts the excess precip (back) in the oceans around the continent might actually produce a FW budget in the Southern Ocean that was more realistic, not less. If WV transports at 60S (say) are realistic, but it's just where the precip ends up that's the problem, this might all be finr.

[QianLi-Ocean]

Is there any feedback between the Anomalous FW and precipitation around the ice sheets? After adding a large amount of Anomalous FW, due to atmospheric cooling, the P-E around the ice sheets will decrease in future projections.

[gavinny]

in our runs it was small ~30 Gt/yr and not highly significant.

[QianLi-Ocean]

What amount of Anomalous FW does this correspond to? About 5%?

[gavinny]

This could be informed by an analysis of how large these fluxes actually are - it may well be that the net discharge with this approach is still within the uncertainties of the observed discharge - in which case it's probably not too bad.

[therealpaulholland]

Very naive question that might be off topic - sorry if so. Until the models have dynamic ice sheets implemented, do we as a group prefer type 1 or type 2? And does the introduction of a FW flux field change our preference?

I think that if the models don't have a FW flux field imposed then maybe we might prefer type 2, in that at least the ice sheet is interacting with the rest of the climate and there will be an increasing discharge in forced runs.

But as soon as there is a FW flux field imposed I think maybe we might prefer type 1. If there is a trend of increasing snowfall onto Antarctica and that causes a sea level drop/salinification in a model, well that makes physical sense, right, as part of the simulation? It might not match observations, but we'd like to know that because it will cause us to think carefully about the projection from that model.

Is the above reasonable do we think? If we did have the chance to influence what the modelling centres do on this topic, actually the switch from type 1 to type 2 or vice versa might be an important outcome. If there is a majority agreement on some of this we should probably say so.

[therealpaulholland]

@swartn nice points as ever, but I am feeling slow this morning after a pub trip and I don't follow your third paragraph above. Are you saying that in models with ice shelf melting and iceberg input, you are necessarily inputting FW over a small/constrained area, and so large future changes in a type 2 approach could lead to large local FW fluxes and model instability?

IF that is the argument THEN
There is an implication for our present endeavour. If we advocate a FW flux spatial field that is highly localised on the regions of real anomalous FW input, we will be causing problems for the type 2 models. So then there would be a tension between those of us advocating for a very localised FW input, implying a type 1 approach, and those happy with a wider input, permitting a type 2 approach.
ELSE
please clarify
END

[gavinny]

I think Neil's point is that if you have an ice shelf cavity that calculates basal melting, but no way of reducing the size of that cavity in a strong melt scenario, you could end up with an unbounded source of FW into the ocean.

[swartn]

Yes, what @gavinny said.

[therealpaulholland]

Ah, I think I see what you are getting at. For example, if the Ross Ice Shelf cavity floods with warm water, the melt rate goes up massively. In a coupled ISM the ice shelf will thin, and hence melting will drop because the thinner ice shelf is not sitting in the warm water. In a static ice shelf the melting will stay elevated. In the UKESM that stabilising feedback due to ice thinning is only just kicking in at 2100 in SSP5-8.5 (Figure 7b in https://doi.org/10.5194/tc-16-4053-2022).

This is not true in general though. E.g. there is good evidence that in the Amundsen Sea, the ice thinning and retreat leads to an increase in melting.
https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023GL103088
https://tc.copernicus.org/preprints/tc-2023-77/
https://doi.org/10.5194/egusphere-2023-1587
So then a static non-retreating ice shelf cavity would actually have a lower FW flux than would arise from a coupled ISM.

[swartn]

@therealpaulholland, yes I'm thinking of that longer timescale-e.g. ISMIP6 results, which consistently shows melt increases under ssp585, but plateau's towards 2100. The reason given is that all the ice shelves have disappeared. I think Helene's talk showed it here. Sounds roughly consistent with what you are seeing in UKESM.

Interesting papers- thanks. Maybe my presumption is incorrect. I guess one would need to run a model like UKESM coupled and then with fixed cavities in a type 2 configuration and compare the resulting melt rates to test these ideas.

[dringeis]

Hi all,

Following all the discussions above, I made some graphics to wrap my head around the

• type-1 and type-2 models
• the implementation of these historical and scenario experiments
• and the conservation of freshwater mass in the models' component interactions.

I thought it might be interesting to share this to make things clearer about what is going on in the different models and what we want to do for the experiment. Please let me know if you have some comments, if you think that I made a mistake or mixed things up.

I think the historical experiment is quite straightforward, However, where I am not too sure, it is on the implementation for the scenario case (CMIP6 forcings, ...), see my question marks.

Here is document with the four cases:

1. Current situation
2. Historical experiment
3. Scenario experiment
4. and finally coupled models.

FW-workshop.pdf

[swartn]

Thanks @dringeis, diagrams are always helpful!

I have a question about the equation setup for type 2 models. Currently you have 1+2+3=0. However, as I see it, 2 and 3 are not different fluxes, they are one and the same flux (unless I don't understand what you are drawing). For mass balance (i.e. type 2), what we enforce in models like canesm is that 1=3 (or equivalently 1=2). Am I missing something? (perhaps if 2 is meant to be an explicit calving/basal melt flux, in which case, 1+2+3=0, then the name for 2 should be changed to reflect this)

[dringeis]

Thanks Neil!

The way I see it, I think 2 and 3 are different fluxes:

• 2 is a physical surface runoff flux. Snow/firn is melting from radiation on the surface, and percolates through the snow/firn layers until it reaches the bottom of it and makes it way to the ocean as liquid FW flux through a river-like mecanism. The snow/firn model is of limited thickness, 3-15m, depending on the number of layers and personal model choices.
• 3 is a flux approximation designed to emulate ice sheet/shelves dynamics under the assumption that the ice sheet is in steady state and the accumulation/ablation on the surface is almost directly represented in the basal melt and iceberg calving fluxes. Maybe implicit is not the good term here...
  • For more details: The top of the ice sheet is kept at a constant atmospheric height, the snow/firn model on top of it. The mass or thickness of the snow/firn model is kept somewhat constant and as we have accumulation of snow on the ice sheet high parts and ablation on the lower parts. These are added and subtracted respectively and accumulated and then send to the ocean as FW fluxes without or with a delay to represent some kind of ice sheet response time (I think we used 10 years in modelE). Note that the flux cannot be negative, no "iceberg sucking" allowed! 😄 There is also the question of where this flux is added in the ocean... These details are surely very model-dependent.

This all a big approximation which is getting less and less true with climate change.

[swartn]

@dringeis - ok thanks for clarifying what you mean. I agree the explicit surface melt, and implicit accumulation/calving=FW flux can be separated. What led to my confusion, is that for us, and I understand quite a few other models, both of these fluxes are practically added into river runoff. Diagrammatically it might be useful to clarify.

[dringeis]

Oh ok, I see, interesting!

hmm I can try something, but I am a bit afraid that it would make the diagrams a bit too cluttered, because this is too model-dependent and I cannot represent all model flavors. This is the reason why I did not try to include the geographical distribution of the FW in the ocean to focus on the FW fluxes balances.

[xylar]

I think it would be best if different centers make their own versions of the figures, rather than making one that covers everyone.