A Climate Science Toolkit for High Impact‐Low Likelihood Climate Risks

An important component of the risks from climate change arises from outcomes that are very unlikely, but whose impacts if they were to occur would be extremely severe. Examples include levels of surface warming, or changes in the water cycle, that are at the extreme of plausible ranges, or crossing of a climate system “tipping point” such as ice sheet or ocean circulation instability. If such changes were to occur their impacts on infrastructure or ecosystems may exceed existing plans for adaptation. The traditional approach of ensemble climate change projections is not well suited to managing these High Impact‐Low Likelihood (HILL) risks, where the objective is to “prepare for the worst” rather than to “plan for what's likely.” In this paper we draw together a number of ideas from recent literature, to classify four types of HILL climate outcome and to propose the development of a practical “toolkit” of physical climate information that can be used in future to inform HILL risk management. The toolkit consists of several elements that would need to be developed for each plausible HILL climate outcome, then deployed individually to develop targeted HILL risk management approaches for individual sectors. We argue that development of the HILL toolkit should be an important focus for physical climate research over the coming decade, and that the time is right for a focused assessment of HILL risks by the Intergovernmental Panel on Climate Change in its 7th Assessment Cycle.

doubt, and societies are increasingly focusing on what actions are needed to address the risks, as well as opportunities, that climate change presents. This implies a new set of questions for climate science.
As the Covid pandemic has reminded us, the greatest risks often come from hazards that are not the most likely. In risk assessment and planning in many fields, attention is paid not only to likely outcomes but also to what would have the highest impact (what is the worst that could happen), even if its likelihood is low or uncertain. Societies need to know what the worst outcomes of climate change could be, to inform action to limit climate change to avoid such outcomes, and to build resilience if they are not avoided (Sutton, 2019).
We have identified several categories of high impact-low likelihood (HILL) climate hazard ( Figure 1): • Weather events that go beyond the established study of likely changes in extreme weather types. This includes record-shattering extremes (Fischer et al., 2021), compound events due to coincidence of several factors (Zhang et al., 2022;Zscheischler et al., 2018) and rapid shifts between opposite extremes (e.g., drought/flood) • Levels or rates of global climate change (and hence regional changes) that are above the likely ranges assessed by IPCC (e.g., because the climate sensitivity of the real world, or the response of the hydrological cycle to a given warming, turns out to be at the upper end of plausible ranges (IPCC, 2021)) • Crossing large scale tipping points in the climate system, for example, instability of ice sheets, major shifts in atmosphere/ocean circulation systems, or loss of major ecosystems (Armstrong McKay et al., 2022). • Climatic consequences of unexpected human actions, possibly by specific sectors of society (e.g., major increases in greenhouse gas emissions, or attempted geoengineering, by individual groupings) Working Group I of the IPCC recently noted the importance of such HILL outcomes in a risk-based approach to climate change assessment (IPCC, 2021), but while the Policymakers' Summary of Working Group II takes such outcomes into account in its "Reasons for Concern," they receive little explicit discussion (IPCC, 2022). This reflects the relatively low level of research focus on HILL outcomes in physical climate and impacts science, and is a disconnect between climate science and society's needs to inform responses to climate risk. The World Climate Research Program's Lighthouse Activity on "Safe Landing Climates" has identified "High-Risk Climate Events" as one of its five science themes, and the need to improve knowledge of HILL outcomes is also recognized in the Lighthouse Activities "Explaining and Predicting Earth System Change" and "My Climate Risk" (https://www.wcrp-climate.org/lha-overview).
In this paper we propose a new research agenda for the coming decade, to respond to this need by developing a suite of climate information needed to inform societal decisions on responses to HILL risks.

From Projection to Risk Management
In many fields, risk is assessed through a likelihood-impact matrix ( Figure 2). The high impact-high likelihood outcomes clearly require most attention. But differing approaches are needed for the low-to-mid impact-high likelihood, and high impact-low likelihood (HILL) outcomes.
Reducing emissions to reduce the overall rate of climate change produces widespread benefits by moving nearly all hazards toward the lower left of the diagram. But plausible pathways to net zero emissions still result in a residual commitment to climate change. For these unavoidable hazards, adaptation and resilience-building across sectors is a key response. The traditional climate science approach, focusing on projections, is well designed to inform adaptation to the high likelihood side of the risk matrix ("plan for what's likely"). But building resilience to the HILL side of the matrix could require extremely high levels of investment, which may never be used and which may have undesirable side effects (e.g., building a high sea defense which reduces a community's attractiveness for tourism). While some "no regrets" actions may be available, investment in resilience to HILL risks Figure 1. Schematic illustrating the four types of high impact climate hazard discussed in this paper. Clockwise from top: compound or unprecedented weather extremes (image shows a forest fire in California, USA in July 2021, during the heat wave that produced unprecedented temperature extremes in the region); levels of climate change (e.g., warming, water cycle changes) that are above the assessed likely ranges; crossing tipping points/thresholds in the physical climate system such as rapid ice sheet collapse; and unexpected human actions such as a rapid increase in emissions from a particular sector.
("preparing for the worst") may be best deferred until the need becomes clear. This approach needs a different type of climate information.
In some cases (e.g., compound extremes), approaches using large climate model ensembles may be useful to assess the likelihood of the hazard at different warming levels (Fischer et al., 2021;Thompson et al., 2017;Zhang et al., 2022;Zscheischler et al., 2018). In other cases (e.g., some climate tipping points) uncertainty may be so deep that robust quantitative estimates of likelihood are impossible. Nevertheless, decision making can still take account of such risks (Desai & Hulme, 2011). Methods that support decision making under deep uncertainty, including robustness analyses or adaptive policy pathways that retain flexibility to respond as new information emerges, are becoming more widely used in the adaptation and policy communities (Marchau et al., 2019). However, despite some examples of successful application in coastal planning (e.g., Ranger et al., 2013), physical climate science has so far paid relatively little attention to the climate information that is needed for such approaches. A new research agenda is needed for physical climate science that enables societies to develop a risk-based approach to decision making, that includes the HILL quadrant of the risk matrix.  (Sharpe, 2019). Without such interactions we risk developing knowledge that is attractive in its specialist field but fails to provide information that enables decisions.

Informing Management of High Impact-Low Likelihood Climate Risks
However, a purely vulnerability-focused approach cannot drive the necessary climate science insights. The climate outcomes we are considering here would have impacts across multiple sectors (e.g., health, ecosystems, agriculture, built infrastructure, energy systems, finance). Furthermore, specific sectors may be vulnerable to some low-likelihood climate outcomes but not others. For example, a coastal planner in western Europe would be concerned about tipping points in both the West Antarctic Ice Sheet (WAIS) and the Atlantic Meridional For each quadrant, the implications for the broad response areas of mitigation (reducing the drivers of climate change such as greenhouse gas emissions) and adaptation and resilience building (adapting societal systems to the climate changes that remain after mitigation) are shown in gray. The types of climate science information needed to support those responses are shown in blue. Mitigation and adaptation/resilience responses tend to move outcomes in the directions shown by the dashed arrows. For many of the HILL outcomes discussed in this paper, mitigation action moves that outcome to a lower likelihood. But for high climate sensitivity, mitigation action reduces the impact.
Overturning Circulation (AMOC), as both these events would result in accelerated regional sea level rise (Bouttes et al., 2014). However an inland farmer in western Europe might be relatively unaffected by WAIS tipping, but still highly vulnerable to AMOC tipping due its cooling and drying impacts (Ritchie et al., 2020). Hence it is not possible to produce a single "HILL climate scenario" to inform all sectors. Instead, we introduce the idea of a "HILL Climate Toolkit": a package of climate information to be developed for each plausible HILL climate outcome, and used as input to the process of developing tailored risk management approaches for individual sectors.

A HILL Climate Risk Toolkit
The elements of the "HILL Climate Toolkit" would need to be developed separately for each HILL climate outcome. For each identified HILL outcome H (e.g., a specific tipping point): • Storylines of dangerous climate system properties, pathways and events. What are the properties or pathways of the climate system that could lead to outcome H? This may involve properties of the climate system itself (e.g., a particular cloud-climate feedback turns out to be strong (Sherwood et al., 2020), or a specific combination of weather events occurs (Sillman et al., 2021)); or it may be a response to specific human actions (e.g., fast vs. slow paths to net zero, geoengineering). These storylines inform mitigation action by better defining "safe operating pathways" of the climate system, and they enable the development of Impact Storylines and Early Warning Indicators. • Storylines of impacts and impact thresholds. An increased focus on HILL outcomes is needed in impacts modeling, which has historically concentrated on the most likely range of climate drivers. To build resilience it is necessary to understand, for each outcome H, its potential impacts across multiple sectors, in isolation and in combination with possible changes in other climate elements. Such information is essential to underpin regional and sector-specific risk management. • Early warning indicators. Where likelihood cannot be estimated, are there indicators to detect whether H is becoming more likely over time (e.g., Boers, 2021)? Or could improving knowledge of a specific climate process lead to better understanding of the likelihood of H? Would such warnings give time to avoid H through mitigation action, or would it be committed/"baked-in," leaving adaptation or forced transformation as the only options (Jackson & Wood, 2018;Ritchie et al., 2021)? How much warning time would there be to adapt? (Jackson & Wood, 2017)? • Monitoring and attribution. How do we build and maintain operational systems of observation and modeling to flag these early warning indicators and to interpret unfolding changes? Can early warning indicators based on dynamical systems ideas (e.g., Boers, 2021) or on simplified process-based models (e.g., Alkhayuon et al., 2019) offer useful detection and warning times in the context of real-world climate noise?
These elements will provide a baseline of climate information for each hazard H, that can be built into tailored climate services for decision makers to develop sector-and locally specific approaches to managing HILL climate risks. Climate scientists will need to work closely with these sectors to co-design and refine the toolkit to meet application needs. The needs of decision makers and scientists in low-income countries, where vulnerabilities and long-term impacts may be greatest, will be particularly important. Some key challenges will include: • Thresholds. Identify physical, biological and socioeconomic thresholds or limits to adaptation (IPCC, 2022), and assess whether these thresholds may be crossed in the storylines above. A specific example would be crossing temperature and humidity thresholds that are beyond the limits of human tolerance (Andrews et al., 2018) • Responses. What feasible responses could reduce dangerous impacts under these climate pathways (adaptation, resilience building, mitigation), while minimizing risks of damaging side-effects? When would transformative, rather than incremental measures be needed?

Conclusions
Climate scientists need to broaden their thinking from quantifying the most likely climate changes, to considering as well what plausible changes could cause the greatest impact. We propose a "HILL Climate Toolkit", a core set of climate information based around storylines, early warning and monitoring, that can be used by decision makers to develop actions to manage HILL climate risks in their specific sectors.
Some elements of the Toolkit will require long term research programmes, and some may prove to be unattainable for particular climate outcomes. Nonetheless the approach is progressive in that each element, as it is added to the toolbox, enhances the overall ability to build resilience to HILL hazards. However the Toolkit on its own will not be enough, as the pathway to use the tools to inform sector-specific decision making will need to be developed through close interaction between climate scientists and decision makers.
Physical climate and impacts science are only just starting to consider these tools. The IPCC recently developed "low likelihood, high impact" climate storylines for high levels of warming and global sea level rise (IPCC, 2021), and a few studies have evaluated impacts at high levels of warming (e.g., Arnell et al., 2019). Similar assessments are needed for other global-and regional-scale climate hazards, and multiple impact sectors. Some international and national research programmes are now recognizing the need for improved information on HILL outcomes (e.g., https://cinea.ec.europa.eu/programmes/horizon-europe/climate-action-horizon-europe_en, https://www. ukclimateresilience.org/themes/climate-resilience/, Stocker et al., 2022), leading to the prospect of real scientific progress over the coming years.
Progress must be underpinned by improved understanding and modeling. As we see increasing numbers of extreme climate events, we need to use these to challenge climate models. Basing assessments entirely on ensembles of "best-estimate" models may systematically underplay high-impact "tails" (Valdes, 2011), while some key processes and feedbacks (e.g., ice sheets) may be missing from many Earth System Models. This suggests that a model hierarchy approach, going beyond the traditional design of climate model intercomparison projects (https://www.wcrp-climate.org/wgcm-cmip), may be needed to explore the full range of possibilities.
Balanced communication on the science of HILL climate hazards will be a particular challenge. Science needs to inform society about the full range of risks and responses, without either inducing feelings of helplessness or fearing accusations of "scaremongering." Such communication needs to be supported by a balanced and comprehensive assessment of current knowledge and research practices, such as can be provided by the IPCC.
As the focus of climate change policy moves from defining the problem to implementing solutions, the need for reliable scientific information on HILL outcomes is becoming ever greater. With prospects of real scientific progress over the coming years in the areas we have outlined, we believe the time is right for IPCC to place a particular focus on High Impact-Low Likelihood Events, and the associated risks, consequences and responses, in its 7th assessment cycle.

Data Availability Statement
There was no actual data collected or used for writing this commentary.