As the ice sheet loses mass at an ever faster rate, scientists are increasingly worried that part of these huge frozen reservoirs is preparing to start irreversibly retreating [Cornford et al., 2015; DeConto et al., 2021]. In order to adapt to the subsequent changes in the coastline, the authorities responsible for coastal planning and climate mitigation work need operational forecasts of sea level rise. However, recent studies using climate and ice sheet models have increasingly drawn very different conclusions about the future rate of sea level rise and even the sensitivity of the ice sheet to future warming [DeConto et al., 2021; Edwards et al. People, 2021].
Paying attention to the uncertainty of long-term sea level rise model predictions is a trap we must avoid.
In the face of huge uncertainty, how can climate scientists help decision makers navigate vague or conflicting information to formulate practical coping strategies? One solution that might provide the required clarity is to shift our focus from what we don’t know to what we know.
There are huge differences between model predictions of long-term sea level rise, which has prompted the scientific community to call on scientists to work hard to reduce uncertainty. However, focusing on uncertainty is a trap we must avoid. Instead, we should pay attention to the adaptive decisions we can already make on the basis of the current model, and communicate and build confidence in the long-term decision-making model.
The emphasis on uncertainty is wrong for two main reasons. First, more and more studies have shown that providing uncertainty estimates to decision makers actually reduces the availability of climate predictions [Lemos and Rood, 2010]. This is partly because it is not always clear how to best incorporate uncertainty into planning. Are we planning for the most probable sea level rise forecast knowing that the protection measures we have taken may be insufficient, or are we planning for the most extreme sea level forecast at an additional cost? The planning process is complex, and the uncertainty of global sea level forecasts is just one of many factors that must be considered by decision makers. For example, when people cannot leave their homes due to air quality issues or cannot drink tap water due to pollution, it does not seem urgent to invest in sea level protection that they will not experience for 70 years. In addition, future planning and infrastructure decisions must directly face unfair practices that have long placed disadvantaged and marginalized people at a disadvantage.
Planning for short-term sea level rise does not mean ignoring the ghost of sea level rise further away.
Second, although the model provides a vague picture of the extent of sea level rise by the end of this century, it has a much clearer estimate of what will happen in the next few decades. This clarity is important because the most urgent adaptation decisions that communities are now facing—related to addressing climate vulnerabilities and historical inequalities—mainly reflect needs on a ten-year, not a hundred-year time scale. Therefore, instead of emphasizing remote goals that are elusive and constantly changing, communities need help to successfully adapt to near-term climate risks.
Planning for short-term sea level rise does not mean ignoring the specter of further sea level rise. Long-term climate and sea level predictions are still needed. For example, adaptive decisions, such as where to place infrastructure designed to last more than a century (for example, new sewers) require information about long-term and short-term changes, and require significant direct costs.
However, taking full adaptation measures based on unclear long-term forecasts is like planning a dinner a few years in advance: thinking ahead is good, but it may be too early to buy groceries. In addition, sea level rise does not suddenly inundate coastlines like a tsunami (although this seems to be the case when sea level rise and storm surge conspire to inundate communities). The rate of sea level rise, even at the extremely high end, is measured in centimeters per year. Given the reality that sea levels will rise in the short term, today’s plan can focus on the expected changes in the next year or two, and then adjust as more obscure long-term changes become the focus.
Climate and ice sheet model predictions are getting farther and farther in the future—reflecting uncertainty—because the physical processes and conditions that we have not observed before occur in a climate that is very different from the climate we experience in modern times. However, the inherent problem with this divergence is not the magnitude of uncertainty, but the resulting lack of confidence in the model that has the necessary skills to express the underlying physics that leads to changes, especially rapid changes.
A common method of estimating uncertainty used by climate and ice sheet modelers is to examine the spread of sea level rise predictions related to a set of different ice sheet models driven by the same input climate forcing. Each model simulates the same system, but its construction and initialization are slightly different. This method is a bit like viewing different answers to a group of students' exam questions. Students may use different methods to answer questions, resulting in a series of answers, although it is hoped that most of them will get answers that are close to the correct answer.
But what happens if the question asked involves untaught material? Well, based on the physical principles learned by the students (and the model), the students (and the model) can still find the correct answer. In the case of climate and ice sheet models, some of these principles, such as conservation of mass and conservation of momentum, have been established and are always applicable. But others just make simple working assumptions, called parameterization.
It is believed that under certain conditions, the instability of ocean ice cliffs may cause the loss of ice and the domino effect of rapid sea level rise to get out of control. However, this process has not been observed in nature.
Parametric attempts to use simpler representations to represent complex processes that rely on adjustable values (parameters) to define the system and how the system evolves. However, many parameters can take a series of poorly constrained values, leading to a wider spread of potential model results. For example, part of the uncertainty regarding the expected fate of the Antarctic ice sheet involves a recent controversial assumption about the unstable process of ocean ice cliffs, which suggests that ice cliffs formed where glaciers flow into the ocean may become structurally unstable. . Height becomes too high [Bassis et al., 2021; Bassis and Walker, 2012; De Cantor and Pollard, 2016]. It is believed that under certain conditions, this instability may cause the domino effect of ice loss and rapid sea level rise to get out of control. However, this process has not been observed in nature, and current models either do not include ice cliff collapse at all, or rely on empirical parameterization based on modern Greenland glaciers [DeConto and Pollard, 2016].
Another way to estimate uncertainty is to explore the range of simulation model results related to different parameters or parameterizations. The challenge here is that the parameterization is usually adjusted to represent the physical processes observed in modern times. As climate change continues to expose ice sheets to conditions beyond our modern observation range, existing parameterizations may no longer truly represent the expected process. Similarly, if processes that remain to be observed (such as the instability of ocean ice cliffs) become important, the model's estimate of uncertainty may no longer represent reality. It is essential that including more processes in the model—especially those for which we have limited observations—in order to improve the accuracy of the model may increase the uncertainty of long-term sea level rise forecasts, at least when these processes are better Understand this before.
So how do we know when a model is physically complex enough that we can rely on its predictions for the future under completely different conditions than today? Answering this question boils down to two related concepts: model confidence and model skills.
Confidence level reflects an assessment (qualitative or quantitative) of whether we believe the physics and assumptions supporting the model are fundamentally correct. In addition to being correct, the model assumptions must be sufficiently complete so that the model can still produce accurate results even if the model is pushed beyond the calibrated conditions or regime. For example, before we can confidently predict the role of ocean ice cliff instability in future sea level rise, we must be able to reliably predict the time and speed of ice breaking and collapse. Therefore, building confidence in the models requires using them to make—and then test—predictions. Model skills measure how accurately the model predicts past changes. Higher model skills will bring greater confidence, but improving model skills is not easy.
Our modern observational record of ice sheet changes is relatively short, dating back to the beginning of the satellite era in the 1970s, and ice sheet models do not have a long record in predicting rapid changes. In 2002, the Larsen B ice shelf on the Antarctic Peninsula disintegrated in less than 6 weeks, an unprecedented and unpredictable rate [Banwell et al., 2013]. When this happens, the flow of the tributary glaciers that feed it accelerates, providing clear evidence that the ice shelf has been supporting the ground ice behind it [Berthier et al., 2012; Scambos et al., 2004] And prove that the ice shelf plays a key role in regulating the discharge of the ice sheet. However, during the collapse, ice sheet modelers are still debating whether large-scale instability will occur, and the potential of this rapid process is not considered in the model [Hindmarsh and Le Meur, 2001].
Building confidence in the model will require the synthesis of broader observations (beyond just past sea level), which allows us to test the model’s ability to represent key processes in many different situations.
The model’s ability to reconstruct past sea levels continues to make significant progress, but this ability itself provides little guidance on whether the model fundamentally correctly represents the physical process. Building confidence in the models — and demonstrating that they have the skills required to accurately represent the rapid changes in the ice sheet — requires the synthesis of a wider range of observations (beyond sea level), allowing us to test the model’s ability to represent key processes in many ways. System.
For example, an ever-growing catalog of observed glacier behavior in Greenland can be used to test models [Catania et al., 2020]. Continued changes in Antarctica, such as the weakening of the Thwaites Ice Shelf and the retreat of the floating part of the Songdo Glacier, may also provide an opportunity to test whether current models can represent a large amount of ice retreat or collapse. In addition to short modern observation periods, paleo records showing changes on longer time scales can provide additional clues about past ice sheet instability and responses to extensive climate forcing. Modern or ancient data sets are not enough by themselves, but piecing them together provides a richer and broader set of conditions for testing models and identifying their inappropriate behavior.
The way to test models and increase confidence in their sea level predictions is not to adjust them to reproduce certain observations. On the contrary-although this sounds contradictory-find examples where the model cannot reproduce the observations. Identifying model failures is the key to improvement because it highlights processes that are incorrectly represented or completely absent in the model. Correcting these deficiencies will lead to a slow and steady move towards models—whether based on machine learning or physics—that accurately integrate more basic physics that affect climate, ice, and sea level. This method of finding and repairing faults is necessary to build confidence that the model will produce realistic predictions when faced with conditions that are radically different from today.
Forecasts of sea level rise that extend to the end of this century and the next may be uncertain. But this uncertainty is not a bad thing for science or adaptation planning. The divergence between the current model predictions is actually a good sign, because it means that scientists are exploring different parameterizations, process representations, and assumptions. Some of them may eventually be abandoned, but others will develop and be widely adopted because of their improved predictive capabilities.
Models are highly skilled in predicting sea level changes on ten-year time scales, and we have already made operational predictions on these time scales. We should emphasize this fact in discussions with community members, stakeholders and decision makers.
Models are highly skilled in predicting sea level changes on ten-year time scales, and we have already made operational predictions on these time scales. We should emphasize this fact in discussions with community members, stakeholders, and decision makers so that they can advance important adaptation and mitigation plans. These adaptation decisions need to be initiated immediately, while scientists continue to work on model improvements.
In the short term, as we explore a wider range of processes and conditions, making these improvements may increase the uncertainty of future sea level rise forecasts. But the increase in uncertainty will be accompanied by an increase in confidence that the model does not omit key physics. Compared to worrying about everything we still don’t know, this increased confidence is more useful in formulating long-term adaptation strategies.
Banwell, AF, DR MacAyeal, and OV Sergienko (2013), The rupture of the Larsen B ice shelf caused by the drainage of a chain reaction from an ice lake, Geophys. Res. Lett., 40(22), 5,872–5,876, https://doi.org/10.1002/2013GL057694.
Bassis, JN, and CC Walker (2012), calculating the upper and lower limits of the stability of disintegrating glaciers from the yield strength envelope of ice, Proc. R. Soc. A, 468(2140), 913-931, https://doi.org/10.1098/rspa.2011.0422.
Bass, JN, etc. (2021), Transition to ocean ice cliff instability controlled by ice thickness gradient and velocity, Science, 372(6548), 1,342-1,344, https://doi.org/10.1126/science.abf6271.
Berthier, E., TA Scambos and CA Shuman (2012), The mass loss of Larsen B tributary glacier (Antarctic Peninsula) has been increasing since 2002, Journal of Geophysics. Reservoir Wright, 39, L13501, https://doi.org/10.1029/2012GL051755.
Catania, Georgia et al. (2020), The future evolution of the Greenland Ocean Terminating Export Glacier, J. Geophys. Reservoir Earth Surfing, 125(2), e2018JF004873, https://doi.org/10.1029/2018JF004873.
Comfortford, SL etc. (2015), Century-scale simulation of the response of the western Antarctic ice sheet to climate warming, Cryosphere, 9(4), 1,579–1,600, https://doi.org/10.5194/tc-9-1579-2015.
DeConto, RM and D. Pollard (2016), Antarctica’s contribution to past and future sea level rise, Nature, 531(7596), 591-597, https://doi.org/10.1038/nature17145.
Decanto, RM, etc. (2021), Paris Climate Agreement and Future Sea Level Rise in Antarctica, Nature, 593(7857), 83-89, https://doi.org/10.1038/s41586-021-03427-0.
Edwards, TL, etc. (2021), Estimated contribution of land ice to sea level rise in the 21st century, Nature, 593(7857), 74-82, https://doi.org/10.1038/s41586-021-03302-y.
Hindmarsh, RCA and E. Le Meur (2001), Dynamic processes involved in ocean ice sheet retreat, J. Glaciol., 47(157), 271–282, https://doi.org/10.3189/172756501781832269.
Lemos, MC and RB Rood (2010), Climate prediction and its impact on policy and practice, Wiley Interdiscip. Pastor climate. Change, 1(5), 670–682, https://doi.org/10.1002/wcc.71.
Scambos, TA etc. (2004), Glacier acceleration and thinning after the collapse of the Larsen Bay B ice shelf in Antarctica, Geophysics. Reservoir Lett., 31, L18402, https://doi.org/10.1029/2004GL020670.
Jeremy Bassis (email@example.com), Department of Climate and Space Sciences, University of Michigan, Ann Arbor
The most shared geophysical research bulletin "America's Increasing Tornadoes", authored by James B. Elsner and others.
Highly cited space weather "2016 International Reference Ionosphere: From Ionospheric Climate to Real-Time Weather Forecast", author: D. Bilitza et al.
Popular article Geophysical Research Letters "P-wave and S-wave velocities of metamorphic sediments excavated from the Alaska subduction zone: effects on in-situ conditions of large thrust faults", author: Peter K. Miller et al.