Is adding freshwater the conceptually correct thing to do?

[gavinny]

Neil?

[Recalculate]

From Andrew Meijers - don't make your professional github name based on your teenage gamer tag...The question as I understood it in the session was whether or not the cryosphere freshwater forcing should be included in the historical runs (and presumably piControl?), or kept in separate MIPs? As a simple observational oceanographer, I feel it would be very valuable if this could be done. The historical Southern Ocean (SO) is so poorly captured in the CMIP models, notably sea ice, temperature anomalies and vertical exchange, and at least some of these do appear to improve in the presence of added fw fluxes (eg Gavin's recent paper). If we are leaving these fluxes out, then we are either getting the 'right' SO for the wrong reasons, or settling for a bad SO...which given its role in global heat/carbon uptake could have significant impacts. Conceptually therefore it would seem to be the right thing to do; if we can.

I realise that adding fw makes things like getting a decent deep cell even harder, so perhaps the realistic answer is 'we want to but there are so many things to fix if we do, eg deep convection, dense water cascading etc, that it cannot realistically be done in time for CMIP7? I'd be interested to know modeller's thoughts.

[jlbamber]

@gavinny. Sorry misread your comment first time. What would be driving barystatic pre 1850? NH glaciers and ice caps (GIC), which are the largest volume compared to the total, were still growing in early 19th C and roughly reached their maximum extent between 1850-1900 depending on region. Could have internal variability driving AIS change but that's not well constrained.

For the Common Era, the rate fluctuated around ~ 0.5 mm/yr in RSL. The reconstruction by Bob is RSL not ASL, which is extremely difficult to do pre satellite. The GIA correction of 0.3 mm/yr is for ASL not RSL. Here's the GrIS reconstructed TMB for 1840-present. Obviously, pre 1950 it's not amazingly reliable but still useful IMHO

[jlbamber]

So, making a small contribution to SLR from 1840-1900 of ~0.05 mm/yr => 24 Gt/yr which is less than 1 mSv...

[jlbamber]

@gavinny in this paper we calc something we call the balance temperature, required for ASL to flucuate due to internal variability only and you will see that the value is pretty close to pre-industrial (if you discount the recent papr on sponges). Grinsted, A., J. Bamber, R. Bingham, S. Buzzard, I. Nias, K. Ng, and J. Weeks (2022), The Transient Sea Level response to external forcing in CMIP6 models, Earth's Future, n/a(n/a), e2022EF002696, doi:https://doi.org/10.1029/2022EF002696.

[bobkopp]

We addressed this in the Zoom chat, but for documentation purposes:

The centennial variability in Kopp et al 2016 dampened out in later reconstruction, c.f.

• Walker, J. S., Kopp, R. E., Little, C. M., & Horton, B. P. (2022). Timing of emergence of modern rates of sea-level rise by 1863. Nature communications, 13(1), 966.

• Walker, J. S., Kopp, R. E., Shaw, T. A., Cahill, N., Khan, N. S., Barber, D. C., ... & Horton, B. P. (2021). Common Era sea-level budgets along the US Atlantic coast. Nature Communications, 12(1), 1841.

Key points from Walker et al. (2022):

• “Over the pre-industrial Common Era, the global component of reconstructed RSL exhibits fluctuating rates between −0.3 ± 0.2 and 0.2 ± 0.3 mm/yr from 0 to 1700 CE." [These are 60-year average rates]
• 0-1700 CE average rate of -0.08 ± 0.05 mm/yr (2s)

[jlbamber]

@xylar OK, useful to know. Assumption that FW anamoly for NH land ice is ~0 for 1850-1950 is not great. The flux is not huge and the uncertainties I quite large but post LIA both GrIS and Arctic GIC were losing mass, with lots of inter-annual varability of course. Magnitude is probably 2nd or 3rd order for the ocean but still there is a monotonic component. See, e.g. Kjeldsen, K. K., et al. (2015), Spatial and temporal distribution of mass loss from the Greenland Ice Sheet since AD 1900, Nature, 528(7582), 396-+, doi:10.1038/nature16183.

[swartn]

[Reformulated for clarity/brevity]

Are we ready to apply specified anomalous freshwater forcing in core policy relevant CMIP experiments, like historical, as opposed to testing these ideas and their many uncertainties more comprehensively in dedicated MIPs/experiments (e.g. NAhosMIP / SOFIAMIP / etc)?

All CMIP6 (and likely many CMIP7) models lack processes representing dynamic ice-ocean interaction, and we all agree the missing anomalous freshwater input is an important feedback on climate change. Adding anomalous freshwater as an additional "forcing" is arguably a good way to simply represent this missing interaction, given the large expense and difficulty in coupling an Ice Sheet Model. Indeed, this is the approach we have taken in the SOFIA initiative and propose to extend into a CMIP7 community MIP.

However, I feel there are some possible complications with applying such forcing more generally in CMIP simulations outside of a dedicated MIP, which is focused on science not policy, and has clear goal and caveats. Some of these are:

• Climate change varies widely across models. In coupled ESM-ISMs, the icesheet derived anomalous freshwater input will vary in strength according to the model's climate change. In models without coupled ISMs, specifying a freshwater input timeseries (from observations or offline ISMs), will lead to some inconsistencies. For example, models with high and low climate sensitivity will experience the same anomalous meltwater input, and models with either warming or cooling CDW at the Antarctic shelf will experience the same freshwater input. Do we know the impacts of these inconsistencies, and how to interpret them? Would we need to, and could we explain this to the broader community using policy relevant experiments? (This arises in a sense because anomalous meltwater input is a feedback to climate change, as opposed to a purely external forcing).

• The anomalous meltwater timeseries is I suspect far more uncertain than other core external forcings (as reflected in many of the observational talks). However, for practical application in DECK/historical, likely only a single timeseries could be specified (due to expense). Is our estimate input good enough to apply in policy relevant experiments? Maybe CMIP7 will treat historical uncertainty differently (Paul hinted at this), but quantifying this forcing uncertainty maybe easier/more appropriate in a dedicated MIP (could be in addition to doing it in historical).

• If applying anomalous freshwater as a forcing in CMIP7 historical, how are other experiments handled? CMIP6 had >100 different experiments, and CMIP7 will likely have more (most in community MIPs, but CMIP7 core will still contain dozens). It is not tractable to generate anomalous freshwater timeseries for all of these experiments. So how to decide which experiments to apply this additional forcing in and which to exclude? How does this affect interpretation, since many experiments refer back to DECK as a reference? For instance, in Detection and Attribution MIP, would an additional "meltwater-only" experiment be required to disentangle the effects? Would it be inconsistent having meltwater in the ALL forcing (historical) experiment, but not in GHG-only, or AER-only?

• Given the diversity of models, with varying representations of the relevant processes, is it plausible to design a consistent way to apply anomalous freshwater forcing, and on the timeline of CMIP7, is it possible for modelling centres to implement a new type of forcing correctly, and differentially across core policy relevant experiments? In SOFIA, we overcome this challenge by specifying a very simple design (surface uniform input), and then having higher tier experiments to test the sensitivity to how/where we input the freshwater. However, for CMIP7 core, how to design something "realistic" to apply across diverse models is not trivial, particularly when many modelling centres are actively working to explicitly include at least some of the relevant processes directly.

Whether in CMIP7 DECK or in a dedicated MIP, building greater community consensus on the best forcing timeseries/uncertainty and implementation into models is highly valuable, and this workshop is much appreciated. Thanks!

[durack1]

I guess it remains to be seen whether FW would be accepted formally into CMIP core runs, but either way integration would be good.

My two cents. There is a (ridiculously, I think this word is warranted) ambitious timeline for AR7 (which is yet to be finalized), and consequently CMIP7 that will be putting pressure everywhere (forcing, modeling groups, infrastructure, etc). I advocate that this community/activity works on a timeline that is science (not IPCC or CMIP-phase) driven, work out the kinks and then engage with modelling groups when there is a tested framework to implement.

We've been hoping a "continuous DECK" to be realized since CMIP6, with evolving forcings (somewhat continuously, "operational" is the aspiration), which means that a 2024 historical (and definitely a CMIP6 historical) and 2026 historical simulation, etc may be quite different, as the forcings (and science driving this) have evolved.

[jlbamber]

@durack1 No single answer to first q. as it depends on the time period. For satellite era it will be as good as you need. Prior to that regional resolution degrades rapidly. For AIS, pre 1979, really we can only say at the whole ice sheet scale but then the anomaly is small (probably) for early 20th C. For glaciers, we just discussed that briefly and I guess we'd generate a mass anomaly that would map to the scale of the hydrology or land surface scheme grids but these are the kind of detail that are important to fix early on.

[durack1]

@jlbamber is your SLB figure (5 comments up) with all the goodies 1900 to ~2020 published somewhere?? This is a very, very nice figure that I would like to educate myself all about!

[ben-davison]

@durack1 it's from Frederikse et al. (2020): https://www.nature.com/articles/s41586-020-2591-3

[jlbamber]

@durack1 So the paper by Frederikse is pretty important as it "solves" the long standing Munk sea level enigma re. the early 20th accel in sea level and the sources. It's also a really nice analysis of uncertainties in 20th C SLB. There's been a lot of work on the SLB during the altimetry era and this paper is pretty comprehensive https://essd.copernicus.org/articles/10/1551/2018/essd-10-1551-2018.html

[mankoff]

Historical balance for Greenland is addressed by Box and Colgan (2013), and that is then somewhat improved and incorporated into Mankoff et al. 2021 (http://doi.org/10.5194/essd-13-5001-2021). For the record, I don't love the pre-1980 data and there are large uncertainties.

⚠️ This product only provides net SMB and discharge, which is not sufficient to provide total freshwater. For example, a year with net 0 SMB could have high snowfall and high meltwater runoff, or low SMB inputs (snowfall) and low SMB outputs (runoff). For a discussion of the difference between mass balance and freshwater, see #6 .

[therealpaulholland]

So, if I understand it, currently most models assume steady state ice, implying that the historical and future surface mass balance trends over the ice sheets lead to an anomalous freshwater flux into the ocean. Can anyone point me towards a study that looks at the numbers for these, in the CMIP models? Maybe they are of the correct magnitude?

[Recalculate]

(Andrew Meijers here) @swartn I have a student, Jon Rosser, looking at exactly the driver of oscillations in CanESM5. In the piControl the period is closer to 100 years. Associated with this is a mode of heat distribution in the SO south of 40, oscillating around 60S, driven by the deep cell and the presence/absence of coordinated dense water formation in the Weddell/Ross seas. The 100 year timescale (and presumably the synchronicity of both gyres) appears to be associated with the transit time of dense water from the formation sites to the ACC, and the subsequent return flow of slightly lighter CDW. It seems to be something like dense water production in the southern gyres -> increased AABW export/southern intensification/ACC acceleration -> increase in gyre size and southward GM eddy flux, entraining relatively warmer waters into gyres -> warmer return flow shutting down dense water formation -> cessation of warmer inflow -> convection starts again. Jon has noted similar behaviour in a couple of other CMIP6 models, as well as a different mechanism for oscillations driven in the GFDL model that is more of a deep heat accumulation -> tipping point -> wild convection/ice loss/heat loss -> re stratification.

We've not looked at the precipitation associated with these runs, but there is a significant surface warming/sea ice loss associated with the cycle, so I'd not be terribly surprised if that impacted evap/precip regionally. Interestingly surface fluxes over the gyres seems to respond to the oscillation, rather than drive it, and there isn't any meaningful signal in the freshwater fluxes/winds/atmospheric modes that might hit at an external driver. Likewise if you look at hovemollers of the ocean heat content, the SO signal propagates north, but only weakly once you look past 40S, so it doesn't really appear to be linked to wider AMOC variability either. It seems to be very much a SO internal oscillation.

He has looked at forced runs of both the CanESM5 and HadGEM3 (which has somewhat similar behaviour after a point in its run), and interestingly noted that under most forms of increased climate forcing the oscillations are very quickly damped out, presumably by increased stratification. However, in runs with 1/2 co2 forcing, the oscillations reappear, but with a 50 year cycle! We're still wondering about what that is, and any suggestions would be welcome. In any event, it seems that these models exist in something of a 'goldilocks zone' where it is just right for sustained oscillations of this nature, but moving the mean state much seems to kill them (or change their frequency).

[swartn]

@Recalculate -yes I have spoken to Jon a few times and always interested to hear his progress - I agree that cycle is predominantly 100 y. We also see in our SOFIA runs, with only FW forcing, that the oscillations completely disappear. The part about shifting to a 50yr cycle under 1/2 CO2 is interesting.

[swartn]

I wonder if an expanded analysis like this might fit into the workshop paper, or perhaps merit a paper. I.e. examining the evolution and drivers of Antarctic "runoff" in CMIP6, as a basis for understanding how prescribed FW fluxes would differ from the status quo.

[torgemartin]

Just knowing the spread in runoff magnitudes from Antarctica for CMIP6 models would be an insightful basis for the discussion here, I guess. But in principle one would also need to have information on how precipitation (liquid and solid) is treated individually and routed to the coast, how the ice sheet is represented (thin snow layer, near-infinite snow pile, ...), is there any time delay included (seasonal effects, e.g. precip timing, instant/delayed runoff, delayed iceberg "calving", ...), to actually know what these models are doing and see the potential wealth of individual solutions.

[swartn]

Agreed @torgemartin . The survey @gavinny did of modelling centres provides a lot of insight into what was done in this respect.

[mankoff]

@WilliamColgan (not part of this workshop so far?) may have thoughts on historic Greenlandic mass balance (and mass balance partitioning).

[WilliamColgan]

Thanks for the H/T Ken.

Hi Folks -- Yes, I did once spend a lot of time looking into the influence of assumed pre-satellite era GrIS balance flux on our understanding of contemporary GrIS mass loss partition (i.e. during the satellite period). Long story short, previous studies have assumed a variety of balance fluxes from the GrIS. Generally, the lower balance flux you assume, the higher the contemporary mass loss you attribute to ice dynamics.

So, in terms of this thread, I agree with @jlbamber -- I would encourage folks to assume a negative mass balance baseline for the GrIS during the 1850-1950 period, largely based on Kjeldsen2015. Imposing an equilibrium assumption during the historical period will otherwise bias both the iceberg calving and SMB contributions differenced from the historical flux.

I attach some relevant figures here, and can share the triple-rejected, and now quite dead, manuscript if anyone wants.

Abstract:
Accurately partitioning past- and present-day ice-sheet mass loss into meltwater runoff and iceberg discharge fractions is critical for accurately projecting future mass loss partition. The sea-level rise contribution of iceberg discharge (or meltwater runoff) is the departure of this flux from a theoretical balance flux. Here, we develop a novel framework to estimate this balance flux, which avoids the common assumption of recent ice-sheet equilibrium. Our balance flux of 422±57 Gt/yr resides within the ~130 Gt/yr range of previous balance flux estimates. We difference this balance flux from an ensemble of iceberg discharge estimates to infer that iceberg discharge was responsible for 52±24 % (or 5.4±0.4 mmSLE) of ice-sheet mass loss over the 2000-2018 period. We find that ice dynamic mass loss from Southeast Greenland is disproportionately low in comparison to that from Northwest Greenland. We speculate that longer term ice dynamic thickening may be buffering the recent ice-dynamic sea-level rise contribution from South Greenland.