THE END OF STATIONARITY–THE FUTURE IS DIFFERENT

Jay Gulledge and I argued previously that we are facing what scientists call the end to “stationarity” (Busby et al. Reference Busby, Gulledge, Smith and White2012, 7–8). The current geologic age, the Holocene—which began 12,000 years ago—has been characterized by “stationarity,” or relatively stable global climatic patterns. That relative stability allowed human civilizations to evolve from nomadic lives to settled agriculture to more contemporary forms of urban living—made possible by relatively predictable environmental conditions and human ingenuity that generated sufficient food and energy to support those city dwellers (Milly Reference Milly2008; Milly et al. Reference Milly, Betancourt, Falkenmark, Hirsch, Kundzewicz, Lettenmaier, Stouffer, Dettinger and Krysanova2015). Scholars who examine the near-term historical record—which is most similar to the present in terms of social systems—will apply their insights to a world that will differ quite dramatically in terms of the magnitude, frequency, and geography of the physical effects of climate change.

The climate security field and the social sciences, however, are skeptical about scholarship that makes broad claims about future security risks. Gleditsch (Reference Gleditsch1998, 394) trenchantly wrote: “‘There will be water wars in the future’ is no more a testable statement than the proverbial ‘The end of the world is at hand,’ unless terms such as ‘the future’ and ‘at hand’ are clearly specified.”

Given these concerns, the climate security field, as Werrell and Femia (Reference Werrell and Femia2019) noted, “favors forensic analysis—case studies or information on past events, rather than future scenarios that social science methods cannot reliably test.” Social scientists generally look to the past to understand the implications for the future. Many use historic evidence of droughts, temperature change, rainfall volatility, and other physical phenomena as proxies for future climate change to assess whether they historically contributed to negative security outcomes such as armed conflict. In keeping with normal scientific practice, scholars (myself included) have examined how states responded in the recent past to climate-related extreme weather (Busby Reference Busby2022). These extreme weather events are consistent with what we expect from anthropogenic climate change, but they may not be attributable to climate change.

These proxies for climate change mostly correspond to short-run shocks or deviations in rainfall or temperature. However, this emphasis on the short run is different from long-run changes that we associate with climate change (Theisen Reference Theisen2017; von Uexkull and Buhaug Reference von Uexkull and Buhaug2021). These changes will unfold over decades and shape the baseline conditions for agriculture, economic development, and even the habitability of societies. We have only begun to experience these effects. Rather than a one-off storm or even a multi-month drought, we are discussing changes in long-term means for rainfall and temperature as well as the tails of observed extreme weather (Intergovernmental Panel on Climate Change 2001;Wagner and Weitzman Reference Wagner and Weitzman2015).

Future climate change may deviate in terms of temperature and rainfall from the normal bounds of what people have been accustomed to for thousands of years. The geography of places affected by weather extremes also may be significantly different from past patterns. Parts of the world may become uninhabitable or barely habitable without major adaptation efforts (Im, Pal, and Eltahir Reference Im, Pal and Elfatih2017; Wallace-Wells Reference Wallace-Wells2019; Xu et al. Reference Xu, Kohler, Lenton, Svenning and Scheffer2020).

DEEP HISTORICAL ANALYSIS

What, then, are potential responses by scholars to this conundrum? Some have mined the deep historical record to see how human societies hundreds or thousands of years ago dealt with such cataclysmic changes in climatic conditions. Diamond (Reference Diamond2004) studied the way that preindustrial societies such as the Greenland Norse, populations on Easter Island, the Anasazi, and the Maya all faced collapse from environmental damage. In their quantitative meta-analysis of the influence of climate on human conflict, Hsiang, Burke, and Miguel (Reference Hsiang, Burke and Miguel2013) included numerous examples from the distant past and the impact of climate changes on various civilizations, including the collapse of the Maya, the Angkor kingdom in modern-day Cambodia, and the Akkadian empire in ancient Mesopotamia—as well as changes in dynastic transitions in China, abandonment of the Lake Titicaca region in modern-day Peru, and political instability in Europe (Hsiang, Burke, and Miguel Reference Hsiang, Burke and Miguel2013).

These historical episodes are important and interesting in their own right. However, differences in adaptive capacity, technological options, and trade and humanitarian networks, as well as changes in social systems, limit their transferability to contemporary situations, barring an explicit effort to surface these differences. While horrendous outcomes are still possible in the wake of extreme weather events and even long-run changes in climatic conditions—particularly in poorly governed places on the planet—it is not clear which lessons we should draw from the premodern era. There remains a role for using historical analogies to elucidate possible future security impacts, but scholars must be more careful and critical in their case selection when they surface analogies from the distant past (Tubi et al. Reference Tubi, Mordechai, Feitelson, Kay and Tamir2022).

SCENARIOS AND PROJECTIONS

Another alternative is scenario analysis. There are different types, some of which are more accepted in social science.Footnote ² Scenarios often have been used by the business community and policy makers to envision possible futures and surface the assumptions that might lead to different outcomes (Task Force on Climate-Related Disclosures ND). These scenarios can be either qualitative, with different descriptions of possible future worlds, or quantitative.

For qualitative scenarios, several future worlds are depicted, often ranging from a worst case, in which countries or actors do not work together, to those involving a more collaborative outcome. In between are mixed scenarios in which some factors favor good outcomes while others do not. For example, the National Intelligence Council (2021) published the 2040 Global Trends report with five scenarios along these lines with climate change as a common stressor. An early scenario-driven report in the climate security space was published by Schwartz and Randall (Reference Schwartz and Randall2004), purportedly commissioned by the US Department of Defense. They depicted a fanciful scenario of abrupt climate change that led to global cooling that prompted outmigration from the United States to Mexico. In a 2020 report, the Center for Climate and Security used a near-term warming scenario of 1 to 2 degrees C for the period up to 2050 and a medium-term warming scenario of 2 to 4 degrees C for 2050–2100. Security practitioners then assessed the impacts on security in different regions (Guy Reference Guy2020). Although qualitative scenarios are popular in the grey literature and the policy/business world, they may have limited academic credibility. They are not designed to be preferred end states and neither are they intended to be falsifiable or even true. Rather, the question is whether they are useful for prodding decision makers from their inertia to consider alternative futures that could prove costly for their organization’s interests (Wilkinson and Kupers Reference Wilkinson and Kupers2013).

Other scenarios are more formal and quantitative. The Intergovernmental Panel on Climate Change (IPCC) uses different types of scenarios. In their 2014 fifth assessment report, it used so-called Representative Concentration Pathways (RCPs) to model different emissions trajectories for their future climate impacts. In their 2021 sixth assessment report, the IPCC used so-called Shared Socio-Economic Pathways (SSPs), which assume different levels of ambition on climate mitigation and adaptation and which can be used alongside Integrated Assessment Models to assess future climate impacts.Footnote ³

These and similar scenarios have been used for climate-security–related work by social scientists. Most use an historical period to “train” the data to identify patterns and associations and then project forward based on a trajectory of climate change, emissions, economic growth, and/or other parameters. For example, a 2009 study by economists on climate change and civil wars in Africa projected the number of likely future civil wars and battle deaths by 2030 based on extrapolations of past conflict mortality and ensembles of climate model projections of future temperature and rainfall (Burke et al. Reference Burke, Miguel, Satyanath, Dykema and Lobell2009). In another study, Hegre et al. (Reference Hegre, Buhaug, Calvin, Nordkvelle, Waldhoff and Gilmore2016) set aside connections between climate and conflict to model future conflict incidence based on the IPCC’s SSP scenarios of future economic growth. Hoch et al. (Reference Hoch, de Bruin, Buhaug, von Uexkull, van Beek and Wanders2021) projected future conflict risks in Africa, using a combination of RCP emissions pathways and SSP socioeconomic pathways, training their data on 1990–2015 using machine-learning techniques and then projecting forward to the 2041–2050 period.Footnote ⁴

A related approach by economists combined assessment of future mortality risks from a changing climate that accounted for adaptation capability through projections of future income (Carleton et al. Reference Carleton, Jina, Delgado, Greenstone, Houser, Hsiang and Hultgren2022). They used available mortality data from 40 countries to estimate relationships between temperature and mortality, extrapolating to countries that lacked historic mortality data. For some studies, one problem—the previous study included—is the reliance on worst-case scenarios of climate change. For example, RCP 8.5, a frequently used worst-case scenario, may overstate the emissions trajectory that humanity currently is on that actually is more midrange like the RCP 4.5 scenario (Hausfather Reference Hausfather2019).

Although these approaches are promising avenues for research, one limitation is that they depend on assumptions of stationarity—that is, stable relations between cause and effect across periods. In their provocatively titled paper, Bowlsby et al. (Reference Bowlsby, Chenoweth, Hendrix and Moyer2020) noted that drivers of conflict in one period may not hold for different periods. For example, they found that drivers of political instability explained the 1995–2004 period well but did not explain instability associated with the subsequent Arab Spring. As the climate changes, this problem will become more acute (Bowlsby et al. Reference Bowlsby, Chenoweth, Hendrix and Moyer2020). As Hoch et al. (Reference Hoch, de Bruin, Buhaug, von Uexkull, van Beek and Wanders2021, 10) noted, “Relations will most likely not remain stable over time; especially when climate change impacts worsen.” For example, using historic data, the causal contribution of climate factors to conflict often is small compared to other factors, such as governance. As climate change worsens, that relationship likely will change (Hoch et al. Reference Hoch, de Bruin, Buhaug, von Uexkull, van Beek and Wanders2021; Mach et al. Reference Mach, Kraan, Adger, Buhaug, Burke, Fearon and Field2019). Thus, if patterns and drivers are not robust across time, out-of-sample validation and efforts to predict future trends may be suspect, particularly as the climate changes.

Although experimentation with forecasting is worthwhile, available data may lead to an elusive quest for prediction that may struggle to accurately forecast future incidence of security problems. Given human agency and unforeseen causal factors, scholars should have some modesty about their models’ forecasting precision.

EXPERT JUDGMENT

Some scenarios use expert judgment to assess regional climate security risks over time. Expert judgment has more established support in the academic literature.Footnote ⁵ For example, Brooks, Adger, and Kelly (Reference Brooks, Adger and Kelly2005) used expert judgment to validate which indicators to include in a composite index of climate vulnerability and which weights to assign to them. Busby et al. (Reference Busby, Smith, Krishnan, Wight and Vallejo-Gutierrez2018) used a similar expert elicitation to conduct sensitivity tests for their composite index of likely South Asia and Southeast Asia climate security hot spots. Both projects were intended to identify relative vulnerabilities to future climate change.

Similarly, Mach et al. (Reference Mach, Kraan, Adger, Buhaug, Burke, Fearon and Field2019) assessed the role of climate relative to other factors in the incidence of conflict. This approach asked a small number of experts to rank, weigh, and assess the uncertainty of key drivers of internal conflict. The experts suggested that low economic development and low state capability were the most important drivers of conflict with climate among the lowest of the 15 indicators. However, the respondents also indicated wide uncertainty about climate if magnitudes of climate change exceed previous experiences, given how climate trajectories are bound up with future socioeconomic trajectories.

Finally, another approach asked regional experts to write about what climate change would mean for the security of particular countries. In an edited volume by Moran (Reference Moran2011), area experts assessed the potential significance of climate change out to 2030 for 42 countries and regions.Footnote ⁶ Similar expert assessments may be idiosyncratic with varying levels of quality and depth of scientific understanding. They also may miss patterns across cases that are only visible through comparisons. However, scholars steeped in the history and geography of particular places may understand the implications of physical changes better than generalists.