environment · family: the answer is in a field you’ve never heard of
theicemodelandtheclimatemodelwon'tmerge
Climate models and ice sheet models cannot talk to each other, so sea level projections remain unreliable
Problem statement
Projecting future sea level rise requires coupling climate models (which simulate atmosphere, ocean circulation, and heat transport) with ice sheet models (which simulate ice flow, calving, and mass loss from Greenland and Antarctica). A 2012 National Research Council report describes sea level projection as an Earth system problem involving ice-sheet models "that have not traditionally been included in climate modeling efforts," and says most ice-sheet models use only minimal climate data, without interactive or physically based coupling with climate models. The two kinds of model also work at different scales: the report puts the atmospheric grids of state-of-the-science climate models at about 100 km, and says ice sheet models need about 5 km resolution near the ice-sheet margin and possibly finer where ice is in contact with the ocean; Aschwanden et al. found that Greenland outlet glacier behavior is well captured at grid resolutions of less than 2 km. Coupling them means simulating melt rates and the fluxes of energy and mass at the ice-ocean interface, which the report says requires models of ocean circulation at the ice front and in the cavities beneath ice shelves, and schemes that conserve energy and mass. IPCC AR6 reported the "continued absence of active ice-shelf and ice-sheet coupling in the CMIP6 model suite," and it has only low confidence in the likely Antarctic contribution to sea level by 2300 and in the ice-sheet instability processes that could push sea level well above its likely range by 2100.
Why this matters
How much the sea will rise is one of the questions the 2018 decadal survey for Earth observation sets for the coming decade: "How much will sea level rise, globally and regionally, over the next decade and beyond, and what will be the role of ice sheets and ocean heat storage?" IPCC AR6 projected a likely range of 0.63-1.01 m of global mean sea level rise by 2100 (relative to 1995-2014) under the SSP5-8.5 high-emissions scenario, counting only processes that can be projected with at least medium confidence. Including ice-sheet processes in which there is low confidence widens that to 0.6-1.6 m (17th-83rd percentile), with 5th-95th percentile projections extending to 0.5-2.3 m. In AR6's table of contributions (Table 9.9), Antarctica has the widest likely range of any component. The chapter says the deep uncertainty in these ice-sheet processes arises from "limited process understanding, limited availability of evaluation data, uncertainties in their external forcing and high sensitivity to uncertain boundary conditions and parameters." The 2012 National Research Council report states that "Interactive two-way coupling is required for simulation of decadal- to century-scale sea-level rise"; AR6 later reported that the CMIP6 climate models still had no active ice-sheet coupling.
What’s been tried and why it hasn’t worked
"Offline" coupling — running stand-alone ice sheet models driven by climate model output — is the approach behind the ice sheet projections AR6 assessed: in the ISMIP6 project, ice sheet models were forced with output from CMIP5 and CMIP6 climate models, with ocean temperature and salinity extrapolated under the ice shelves and basal melt computed as a quadratic function of ocean thermal forcing. Because the CMIP6 climate models have no active ice-shelf or ice-sheet coupling, AR6 notes, they ignore the feedback of basal meltwater and calving on ocean circulation. Coupled attempts exist — AR6 cites CESM2-CISM2, an Earth system model coupled to an ice sheet model for Greenland, which includes ice-sheet-atmosphere interactions and routes ice-sheet surface meltwater to the ocean — but the chapter repeats an earlier IPCC finding that remaining deficiencies in the coupling between climate and ice sheet models, such as low spatial resolution, limit how well the feedbacks between them are represented. AR6 also has low confidence in the ability of present models, especially at coarse resolution, to project the processes that bring warm deep water into the cavities under ice shelves. Marine ice cliff instability (MICI) — the proposed failure of ice cliffs after ice shelves disintegrate — is, together with marine ice sheet instability, described as having the potential to strongly increase Antarctic mass loss on century to multi-century timescales, but AR6 reports that "Only one ice-sheet model represents MICI" and repeats an earlier assessment that "the validity of MICI remains unproven."
What would unlock progress
Global variable-resolution models on unstructured grids, which refine the mesh locally without the boundaries of a nested regional model — the 2012 National Research Council report says these have become feasible, but that developing physics parameterizations that work well across the variable resolution is a significant challenge. Better physical understanding of calving, basal melting at the ice-ocean interface, and grounding line dynamics, which the same report says need to be better understood and included in ice sheet models; the 2018 decadal survey says high-resolution (100 m) measurements are required to gain insight into ice fracturing, including calving cliff failure. Emulators of ice sheet models — AR6 already used an emulator of the ISMIP6 simulations to produce projections across emissions scenarios, and it notes the increasing use of statistical emulation to sample the parameter space of ice sheet models more fully. Observations from the ice-ocean interface: the 2018 decadal survey calls for seafloor bathymetry in fjords and beneath ice shelves and a novel network of ocean measurements at the ice-sheet periphery, and lists seal data, autonomous underwater vehicles, and gliders among the ways to observe ocean subsurface temperature.
Entry points for student teams
A student team could implement and benchmark a simple ice-ocean coupling framework using a 1D melt parameterization (e.g., quadratic melt-rate dependency on ocean thermal forcing) within an open-source ice sheet model (ISSM, Elmer/Ice, or PISM), testing how different coupling frequencies and melt parameterizations affect modeled retreat rates for a simplified glacier geometry. Alternatively, a team could train a neural network emulator of an ice sheet model's response to varying ocean forcing scenarios, assessing whether ML surrogates can reproduce multi-century ice dynamics at reduced computational cost. Relevant disciplines: computational fluid dynamics, glaciology, physical oceanography, machine learning, numerical methods.
Genes
Structural cousins: stuck the same way, in other fields
Sources
"Thriving on Our Changing Planet: A Decadal Strategy for Earth Observation from Space," NASEM, 2018. accessed 2026-02-16. Also: "A National Strategy for Advancing Climate Modeling," National Research Council, 2012, IPCC AR6 WG1 Chapter 9 (Ocean, Cryosphere and Sea Level Change), Aschwanden et al., Science Advances 2019. go to source 1 ↗ go to source 2 ↗ go to source 3 ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
- The `failure:disciplinary-silo` tag is the primary failure mode: ice-sheet models "have not traditionally been included in climate modeling efforts" (National Research Council, 2012), and most were built without interactive coupling with climate models.
- The `failure:unrepresentative-data` tag reflects that ice sheet models are evaluated against present-day observations, while AR6 notes the difficulty of evaluating simulations of processes that are not currently observed, rare, or only indirectly deduced — in particular the ice-shelf disintegrations and cliff failures that would drive the proposed marine ice cliff instability.
- This brief complements the existing environment-ice-sheet-collapse-timeline brief, which focuses on the scientific uncertainty in collapse timing. This brief focuses on the specific engineering/computational barrier — model coupling — that prevents narrowing that uncertainty.
- Cross-domain connection: shares the multi-model coupling challenge with manufacturing-multiscale-materials-modeling-gap (different physical models at different scales that don't interoperate) and the disciplinary-silo challenge with environment-critical-zone-process-prediction (hydrology, geomorphology, and biogeochemistry as separate modeling communities).
- IPCC AR6 Table 9.9 shows that the Antarctic contribution is the component of projected sea level rise by 2100 with the widest likely range.
Reconciliation 2026-10-01: Fact repair after reading the whole write-up against all four sources it names, each fetched on 2026-10-01. Read as full text: the 2018 decadal survey, through the National Academies' online reader (https://www.nationalacademies.org/read/24938, 22 chapter pages; the PDF needs a login); the 2012 report, the same way (https://www.nationalacademies.org/read/13430, 24 chapter pages); IPCC AR6 Working Group I Chapter 9 (https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_Chapter09.pdf); and Aschwanden et al. 2019 (Europe PMC, PMC6584365). The two reports and the chapter are long; the full text of each was searched for every claim and the passages on ice sheets, sea level and coupling were read. (1) "0.28-1.01 m of global mean sea level rise by 2100 under a high-emissions scenario" joined two scenarios: in the chapter, 0.28 m is the low end for the lowest scenario (SSP1-1.9) and the likely range for SSP5-8.5 is 0.63 to 1.01 m. The text now gives 0.63-1.01 m. (2) The quotation "a rise approaching 2 m by 2100 cannot be ruled out" is not in Chapter 9. The chapter says low-confidence processes could lead to 0.6 to 1.6 m under SSP5-8.5, with 5th to 95th percentile projections extending to 0.5 to 2.3 m; the text now says that, and the "factor-of-7 spread" built on the old numbers was removed. (3) Removed because none of the four sources contains them: 680 million people in low-lying coastal zones; about $1 trillion of coastal infrastructure at risk per meter of rise; 30-minute climate model timesteps and annual-to-decadal ice sheet timesteps; the grounding line as "the most poorly observed" region; E3SM and UKESM as coupled examples; numerical artifacts where grids do not align; coupled runs too expensive for ensembles; MPAS-Ocean; borehole access under ice shelves; and the sentence on coastal adaptation planning. (4) The text said AR6 named the lack of coupled ice-ocean models as "the primary reason" for low confidence in the Antarctic contribution beyond 2100, and that the spread exists "primarily because" climate models cannot predict ocean heat delivery. The chapter does not give a primary reason. It reports the absence of ice-sheet coupling in the CMIP6 models, and it attributes the deep uncertainty to limited process understanding, limited evaluation data, uncertain forcing and sensitivity to uncertain parameters; the text now quotes both. (5) Model scales now follow the sources: climate model grids of about 100 km and ice sheet models needing about 5 km or finer (2012 report), and outlet glaciers well captured below 2 km (Aschwanden et al.), in place of "50-100 km" and "1-5 km." (6) The statement that the two model families were developed independently for decades by separate communities (and "30+ years" in the notes above) now follows the 2012 report, which says ice-sheet models have not traditionally been included in climate modeling efforts. (7) The marine ice cliff sentence said the mechanism could add "meters" on century timescales and that "no coupled model has resolved it." The chapter says the instabilities have the potential to strongly increase Antarctic mass loss on century to multi-century timescales and that only one ice-sheet model represents the mechanism; the text now says that. (8) The paragraph on what would unlock progress now ties each item to what the reports say: variable-resolution unstructured grids (2012 report, which does not name a specific ocean model), the processes that need to be better understood, the emulator AR6 used, and the ocean measurements the 2018 survey calls for. (9) In the notes above, "no modern analog" and the statement that the range in Table 9.9 is "directly attributable to the coupling problem" are not in the sources; those two notes now say what the chapter and the table show (Antarctica has the widest likely range of any component). (10) On the Source line, the 2012 report's author was given as NASEM; the record gives the National Research Council, and the Source line now says so. The chapter title was matched to the IPCC's ("Ocean, Cryosphere and Sea Level Change"). The links that were read for the 2012 report and the chapter were added to the Source line beside their names. Confirmed in the sources and left as written: the titles and years of all four sources; that climate model output is used to drive stand-alone ice sheet models (ISMIP6); that the missing coupling leaves out the effect of meltwater on ocean circulation; the CESM2 coupled example; and the quadratic dependence of melt on ocean thermal forcing used in the first entry point. Could not be checked in these sources: that ISSM and Elmer/Ice are open-source (the Aschwanden paper gives a public code repository for PISM only). No title changed. Genome tags untouched. A second reader went over these repairs on the same day, fetching the same four sources independently (the two reports through the National Academies online reader, the chapter PDF, and the Europe PMC full text); all ten repairs stand as described. One sentence was tightened: the emulator item in the unlock paragraph had said emulators could provide fast approximations of ice sheet response without running a full ice sheet model, which none of the sources says; it now says what the chapter says, that AR6 used an emulator of the ISMIP6 simulations and notes the increasing use of statistical emulation to sample the parameter space of ice sheet models more fully.