The sectors and systems subject to climate-related risks do not exist in isolation; they interact with one another and with other sectors and systems. For example, agricultural systems require water for irrigation, which is supplied from lakes, rivers, dams, and reservoirs. Forest management influences the runoff that makes its way into these water systems. Electricity systems use water for hydroelectric power as well as for cooling thermoelectric power plants. Many urban transportation systems rely on electricity to power subways and buses. Meanwhile, medical services, and public health more broadly, are enabled by transportation, water, electricity, and communications (Ch. 11: Urban, KM 3). To most effectively assess the risks associated with climate-related stressors such as floods, droughts, or heat waves, the interactions among these systems must be considered in addition to the effects of these stressors on individual systems.⁸

In addition, climate-related stressors are not the only influences to which natural, built, and social systems are exposed. For example, population movements and demographic changes, economic growth, and changes in industrial activity can all influence systems exposed to climate-related stressors as well as systems that interact with them (see, for example, Box 17.3). Such factors can have powerful impacts on these systems or alter their vulnerability to climate-related stressors. For example, rapid population growth in the coastal United States over the past half-century has significantly increased society’s exposure to extreme weather events like hurricanes.⁹ These demographic trends may have a greater impact on future hurricane damages than sea level rise or changes in storm intensity.¹⁰

A long history of research on complex systems (e.g., Simon 2000¹¹), spanning disciplines from meteorology¹² to ecology¹³ to paleontology¹⁴ to computer science, ¹⁵ has shown that systems that depend on one another are subject to new and often complex behaviors that do not emerge when these systems are considered in isolation. These behaviors, in turn, raise the prospect of unanticipated, and potentially catastrophic, risks.¹⁶ For example, failures can cascade from one system to another; that is, failures in one system can lead to increased risks or failures in other systems. Such cascades have been observed with Hurricane Harvey (see Box 17.1), the Northeast blackout (see Box 17.5),¹⁷ and erosion and permafrost thaw in Alaska (Ch. 26: Alaska, KM 3), where failures in physical infrastructure systems had downstream consequences for human health and safety. Tightly connected supply chains can quickly transmit impacts from events such as floods, droughts, heat waves, and tropical cyclones in one region or part of the world to systems in another (see Ch. 16: International, KM 1). For example, the spike in food prices in 2010–2011 was driven in part by drought-related declines in production of basic grains in Australia and Eastern Europe, which provided a short-term income increase to U.S. farmers of those commodities (see Ch. 16: International, KM 1).¹⁸

Similarly, changes in one part of a system may alter the thresholds and tipping points in other parts of the system (see Kopp et al. 2017¹⁹). For example, the overuse and depletion of groundwater removes a backstop in times of drought (see Box 17.3). Forest wildfires can affect water and air quality and render soil impermeable, altering both health and flood risks (see Box 17.4). Interactions among systems can also buffer systems from shocks and introduce a measure of system stability or recovery potential that might not have otherwise existed (see Box 17.5). For example, social networks, which are increasingly reflected in social media enabled by communications infrastructure, can have an important influence on the resilience of communities to natural hazards. Compound events, such as simultaneous temperature extremes and drought, can produce greater economic costs than events considered separately.¹⁹ The complexity of the interactions that exist among these various systems limits the ability to predict the consequences of climate-related stressors with confidence. This poses important challenges for risk assessment as well as the management of those risks.

The number and complexity of possible interactions among systems affected by climate expand the scope of climate change risk assessment. Recent assessments have acknowledged the importance of this expanded perspective. For example, Chapter 10 of the Third National Climate Assessment (NCA3) ⁴⁴ highlighted interactions among energy, water, and land systems that people and economies depend on. Other recent climate change impact assessments (e.g., Oppenheimer et al. 2014, Houser et al. 2015, Rosenzweig et al. 2017^45,46,47) have highlighted risks emerging from interactions among different sectors, geographic regions, and stressors.

A key factor in assessing risk in this context is that it is hard to quantify all the ways in which climate-related stressors might lead to severe or widespread consequences for natural, built, and social systems. A multisector perspective can help identify critical risks ahead of time, but uncertainties will always remain regarding exactly how consequences will materialize in the future. In some cases, interactions are well known. For example, yearly fluctuations in river flows affect hydropower generation, in turn shaping energy costs and profits and reliance on other energy sources (see Box 17.3). Yet even in these cases, it is often difficult to quantify all relevant processes and interactions. Sometimes, interactions are only obvious in retrospect, such as those associated with many past hurricanes (see Box 17.1) or the Northeast blackout (see Box 17.5), with impacts cascading through different parts of the built environment and affecting human health, well-being, and livelihoods. In still other cases, all the relevant interactions are simply not fully understood, for example in the context of the linkages between wildfires, pine bark beetles, and forest management (see Box 17.4).

Therefore, effectively assessing multisector risks requires different tools and approaches than would be applied to understand a single sector by itself. For example, as land management, infrastructure, and climate all change through time, statistical analysis of extreme weather events based on the past becomes less accurate in predicting future outcomes (Ch. 28: Adaptation, KM 2).⁴⁸ Approaches can be applied to integrate diverse evidence, combining quantitative and qualitative results and drawing from the natural and social sciences and other forms of analysis.^49,50,51 As one example, models and numerical estimates can be complemented by methods quantifying expert judgment in order to consider uncertainties not well represented by the models. For instance, models and expert judgment have been used together to inform understanding of future sea level rise.⁵² Scenarios can also be used to explore preparedness across possible futures, including extreme outcomes that have been rare in historical experience but may be particularly consequential in the future.^50,53,54,55 Such scenarios in assessment can inform understanding of different decisions and choices for managing climate risks, responding to uncertainties about the future by starting with goals and priorities people have.

In interacting systems, management responses within one system influence how other systems respond. Within water basins, for example, upstream management decisions can constrain downstream water-dependent management decisions that affect agriculture, transportation, domestic and commercial use, and environmental protection. Failure to anticipate such interdependencies can lead to missed opportunities for managing the risks of climate change; they can also lead to management responses that increase risks to other parts of the system. For example, the use of groundwater in California as an agricultural backstop in the recent drought may alter California’s resilience to future droughts (see Box 17.3).

In practice, managers of agricultural, natural resource, or infrastructure systems do manage at least some degree of system interdependencies. For example, electrical utilities account for the management of water resources to provide power plant cooling capacity, manage fuel supply chains through transportation networks,⁶⁴ and manage demand from customers. This requires utilities to acquire appropriate operational permits, licenses, and contracts relevant to other systems and to incorporate the characteristics of those systems in strategic planning and risk management. At the same time, water utilities are users of electricity, particularly those that rely on desalination, which is very energy intensive. Hence, efforts to enhance the resilience of water supply systems to drought can have important consequences for the energy sector and electricity costs.⁶⁵ Such indirect risks can be challenging to manage, particularly when system managers have no operational control or jurisdiction over the interacting system. When drought reduced barge traffic on the Mississippi in 2013, farmers had limited options other than seeking more expensive transportation options or incurring delays in getting their products to market.^66,67 More holistic management approaches therefore hold the potential for anticipating these risks and developing effective strategies and practices for risk reduction.

Despite the challenge of managing system interactions, there are opportunities to learn from experience to guide future risk management decisions (Ch. 28: Adaptation, KM 3). The financial sector has invested significantly in understanding and managing systemic risks—including those associated with climate change and climate policy.⁶⁸ Mechanisms include risk assessment, financial disclosures, contingency planning, and the development of regulations and industry standards that recognize system interdependencies. Another example is that of the Department of Defense (DoD), which integrates consideration for the implications of climate change and variability for food, water, energy, human migration, supply chains, conflict, and disasters into decision-making and operations around the world.⁶⁹ In so doing, the DoD focuses on enhancing preparedness, building partnerships with other public and private organizations, and including climate change in existing planning processes.^69,70 These strategies are relevant to any organization attempting to enhance its resilience to climate change.

A multisector perspective recognizes that systems have multiple points of vulnerability, that risk can propagate between systems, and that anticipating threats and disruptions requires situational awareness within and between systems.^71,72 Translating the growing awareness of such complexities into the design of policies and practices that effectively address climate change risks, however, requires rethinking how systems are managed in order to identify opportunities for risk reduction or greater efficiency. For example, risk can be reduced by building excess capacity, flexibility, and redundancy into systems.⁷³ Reserve margins for electricity grids, multi-objective reservoir management, grain storage, multimodal transportation networks, and redundant communications are all mechanisms that provide flexibility for coping with a broad range of risks. Resilience can also be enhanced by planning for system recovery in the event of diverse types of disruptions. For example, restoring power in the wake of a natural disaster is critical for restoring other services such as transportation, water, health, and communications (see Box 17.5). Nevertheless, the costs of designing, constructing, operating, and monitoring resilient, interacting systems that are robust to multiple sources of stress can be significant. Hence, consideration of the costs and benefits of resilience over the entire life cycle of the system may be necessary to make the business case.

Although it is clear that climate-related and non-climate stressors impact multiple natural, built, and social systems simultaneously, thereby altering societal risks, the tools available for predicting these dynamics lag those that predict the dynamics of individual systems. There are many existing modeling efforts that explore complex natural systems, including climate models and numerical weather forecasting models. Although these models are applied to scenarios driven by social and policy decisions, the models themselves rarely incorporate the feedback relationships to social systems and policy contexts.^85,86,87,88 Engineering and resource management models that explicitly incorporate societal economic decisions and represent built systems at a very high resolution have not traditionally been linked to climate projections. Some integrated human–earth system models are explicitly designed to identify system linkages but frequently lack key features or sufficient resolution of the inherent dynamics of the natural environment.^89,90 These and other intersectoral models are also used to create scenarios of how combined natural–human systems might evolve (for example, the Shared Socioeconomic Pathways [SSPs]) (see Scenario Products section of App. 3).⁵³ Such scenarios can be useful for exploring the range of possible outcomes of larger trends, but the results should not be considered predictive. There is a large gap, therefore, in the multisector and multiscale tools and frameworks that are available to describe how different human systems interact with one another and with the earth system and how those interactions affect the total system response to the many stressors they are subject to, including climate-related stressors.⁹¹ However, increasing recognition of this gap has given rise to a number of innovative research projects that seek to directly link climate scenarios or earth system models to high-resolution models of built infrastructure and human systems (e.g., Allen et al. 2016; Voisin et al. 2016; Ke et al. 2016; Zhou et al. 2017, 2018; Tidwell et al. 2016^{92,93,94,95,96,97}).

The responses of interacting systems to both climate-related and non-climate stressors exhibit deep uncertainty, especially when interactions with societal decisions are included. Deep uncertainty arises when there is a lack of clarity about the appropriate models to apply, the relative probability of various scenarios, and the desirability of alternative outcomes.⁹⁸ Risk management decisions must therefore be made in the face of these uncertainties (see Key Message 2). An important research challenge is therefore advancing scientific methods and tools that can be applied in climate research, risk assessment, and risk management for complex, interdependent systems under deep uncertainty.⁹⁹