Main  主页

Widespread attention is being given to the growing risk from wildfires as one of the signature threats of climate change1. The empirical basis for this concern is considerable, with a convergence of scholarship revealing upward trends in wildland fire severity and size2,3,4, shorter durations between re-burning due to accelerated interactions between droughts and wildfire activity5, longer fire weather seasons6 and an increase in the global burnable area7,8. Recent extreme incidents, such as the 2023 Canadian and Hawaiian wildfires, are clarion calls signalling growing wildfire risk and associated demands for human adaptation. Research suggests that escalation in wildfire emissions and associated negative impacts to air quality will continue9. In other words, there is growing scientific consensus globally of the need for human and institutional adaptation to a changing climate, both now and into the future. And yet, knowing adaptation is needed and understanding what adaptation requires are two different things.
作为气候变化的标志性威胁 1 之一，野火带来的风险越来越大，这一点正受到广泛关注。这种担忧有大量的经验依据，研究结果表明，野地火灾的严重程度和规模呈上升趋势 2,3,4 ，干旱和野火活动之间的相互作用加快，导致重新燃烧的持续时间缩短 5 ，火灾天气季节延长 6 ，全球可燃烧面积增加 7,8 。最近发生的极端事件，如 2023 年的加拿大和夏威夷野火，发出了野火风险和人类适应相关需求日益增长的号角。研究表明，野火排放的升级和对空气质量的相关负面影响将继续 9 。换句话说，全球范围内越来越多的科学界一致认为，无论是现在还是将来，人类和机构都需要适应不断变化的气候。然而，知道需要适应和理解适应的要求是两码事。

The growing threat of wildfire to human populations is driven by a complex array of factors rooted in a changing climate, historical land management practices and the expansion of human settlements into the wildland2,10,11,12. While the understanding of the interactions between climate, land use and wildfire has advanced, and scientists are increasingly aware of humanity’s role in contributing to climate change1,13, the ability to improve fire management outcomes remains elusive14. A key impediment lies in the limited attention given to the conceptual linkage between environmental changes and human institutions15. Wildfire management, while rooted in biophysical phenomena, is fundamentally a human activity constrained and enabled by governing institutions—both long-standing organizations and the system of rules and logics that govern their behaviour16. Successful adaptation requires understanding how environmental changes translate into new demands on these institutions17. Despite this, the state of knowledge concerning the institutional dimensions of wildfire remains nascent. Little is known about where institutional adaptation pressures are most acute or about the forces that reinforce stability and inertia in systems governing wildland fire management. While wildfires are increasingly described as more complex, far less is understood about how changes in wildfire complexity systematically alter the demands on responding institutions at national and regional levels. This study examines trends from 1999 to 2020 in the characteristics of complex wildfire incidents in the USA, focusing on incident characteristics associated with institutional complexity. Because institutions differ across contexts, consideration of environment and institutional linkages must be understood within national contexts, which vary in their institutional wildfire response18.
野火对人类的威胁与日俱增，这是由一系列复杂的因素造成的，这些因素植根于不断变化的气候、历史上的土地管理方法以及人类住区向荒野的扩张 2,10,11,12 。虽然人们对气候、土地利用和野火之间相互作用的认识有所提高，科学家们也越来越意识到人类在气候变化中所扮演的角色 1,13 ，但改善火灾管理结果的能力仍然遥不可及 14 。一个关键的障碍在于人们对环境变化与人类制度之间的概念联系关注有限 15 。野火管理虽然植根于生物物理现象，但从根本上说，它是一种人类活动，受到管理机构--包括长期存在的组织以及管理其行为的规则和逻辑系统--的制约，并由其促成 16 。成功的适应需要了解环境变化如何转化为对这些机构的新要求 17 。尽管如此，有关野火的制度层面的知识仍处于起步阶段。人们对哪些方面的制度适应压力最大，以及哪些力量加强了野火管理制度的稳定性和惰性知之甚少。虽然野火被描述得越来越复杂，但人们对野火复杂性的变化如何系统地改变国家和地区层面的应对机构的需求却知之甚少。本研究探讨了 1999 年至 2020 年美国复杂野火事件特征的变化趋势，重点关注与机构复杂性相关的事件特征。 由于不同国家的体制各不相同，因此在考虑环境与体制之间的联系时，必须结合各国的国情，因为各国的野火应对体制也不尽相同 18 。

Institutional complexity is an important concept for advancing climate research because it conceptually links the biophysical environment and associated task environment. In this context, the task environment refers to the nature and scope of demands placed on institutions with authority and responsibility to manage some aspects of the biophysical environment. Changes in the biophysical environment can create variation in the task environment for these institutions. However, case studies reveal that the nature of this coupling is not necessarily linear nor obvious19. Agencies and organizations evolve over time to reflect the demands of their task environments20. The task environment of an agency is not exclusively a function of their physical environment but is also linked to demands in their policy and their political, economic, cultural and social environment21,22. The logics that guide these organizations are rooted in past experience23. Institutional logic refers to the tacit and explicit rules that govern thought and action within an organization and are tied to normative expectations of what ‘right’ looks like within larger institutional fields. Institutions are, by definition and design, stable and enduring16. Their very nature makes them resistant to change and oriented towards stability and incrementalism. New logics can take years, sometimes decades, to become institutionalized. In general, the older, larger and more complex an institution is, the more challenging and time intensive adaptation can be24.
机构复杂性是推进气候研究的一个重要概念，因为它从概念上将生物物理环境与相关任务环境联系起来。在这里，任务环境指的是对有权力和责任管理生物物理环境某些方面的机构提出的要求的性质和范围。生物物理环境的变化会使这些机构的任务环境发生变化。然而，案例研究表明，这种耦合的性质不一定是线性的，也不一定是显而易见的 19 。机构和组织会随着时间的推移而演变，以反映其任务环境的需求 20 。机构的任务环境不仅是其物理环境的函数，而且还与其政策需求以及政治、经济、文化和社会环境相关 21,22 。指导这些组织的逻辑植根于过去的经验 23 。制度逻辑指的是支配一个组织内部思想和行动的默示和明示规则，这些规则与更大的制度领域中对 "正确 "的规范性预期息息相关。根据定义和设计，制度是稳定和持久的 16 。其本质决定了它们抵制变革，并以稳定和渐进为导向。新的逻辑可能需要数年，有时甚至数十年才能制度化。一般来说，机构越老、越大、越复杂，适应工作就越具有挑战性，需要的时间就越长 24 。

Land management and fire management are both old, highly bureaucratized and deeply professionalized institutions within the USA. Effective institutional adaptation is understood as readiness to meet the demands of the new task environment25. However, during times of environmental change, institutions—and their governing policies—may struggle to re-align logics to reflect the demands of the new task environment, particularly if those changes are not well understood19,26.

In the USA, wildfire response is managed under the incident command system by incident management teams (IMTs) who are called in by affected land jurisdictions to assume operational command of fire suppression activities27. Several different factors can influence the institutional complexity of wildfire and their implications may vary on the basis of country context28. For this study, we focus on change over time in five key features of wildfire incidents in the USA: incident frequency, jurisdictional complexity, incident size, incident days as preparedness level (PL) 4 or 5 and the scale of threat to human populations. Detailed justifications for considering these features as dimensions of institutional complexity can be found in Methods.

Here we document and investigate changes in the institutional complexity of wildfire incidents in the USA from 1999 to 2020 to address the following research questions: How have these features of wildfire institutional complexity changed over the past 20 years? What aspects of institutional complexity have seen the greatest change? And do these changes vary by region? We focus on the wildfire incidents that require either a type 1 or a type 2 IMT, which are reserved to manage the most extreme wildfire incidents among the five types (types 1 to 5). Our interest in type 1 and 2 incidents is twofold. First, consequences of climate change are often observed in the extremes rather than the averages3. Second, national resilience in the context of wildfire should be the capability to prepare for and mitigate harm from extreme events instead of the average incident. Understanding regional variation patterns is critical because, in most cases in the USA, the initial phase of wildfire response is heavily dependent on local and regional capacity.

Prevalence of complex wildfires

Wildfire incidents in the USA are classified on the basis of the qualifications of the IMT deployed for the incident. IMT classifications range from types 5 to 1 based on team size and qualifications. Type 1 and 2 IMTs are the largest, most elite teams and are reserved to manage incidents of the greatest complexity and national consequence (www.nifc.gov/resources/firefighters/incident-management-teams, accessed 15 June 2023). First, we consider change over time in the total number of incidents involving type 1 or 2 IMTs from 1999 to 2020 (Fig. 1a). The number of incidents is an aspect of institutional complexity as the national cache of qualified personnel to staff type 1 and 2 IMTs is decreasing29, leading to the potential situation of more deployments of fewer IMTs. While type 1 and 2 incidents are consistently concentrated in the western USA, all US regions have experienced type 1 and 2 wildfires (Fig. 1a).

Fig. 1: Spatiotemporal trends in the number of type 1 and 2 wildfire incidents and change in incident counts by region from 1999 to 2020.
figure 1

a, Locations of type 1 and 2 wildfires coloured by year. The bar chart shows annual frequency of type 1 and 2 wildfires. b, Degree of change in annual type 1 and 2 incident frequency 1999–2020 by region based on linear slopes normalized within region. Note: Hawaii is grouped by NWCG as part of the Northern California region.

Full size image

Nationally, on average, the linear trend associated with the prevalence of type 1 and 2 incidents has decreased slightly in number since 1999, by 2%; this is due to very active fire years in 2006 and 2008 (Fig. 1a). The key exception to this is the Pacific Northwest (Fig. 1b), which has seen, on average, a 2% per year linear increase in type 1 and 2 incidents since 1999. This suggests that increased regional pressure on type 1 and 2 resources is possible, but these pressures vary greatly from year to year and region to region.

Jurisdictional complexity

Second, we investigate trends in the changing jurisdictional complexity of complex wildfire incidents. Authority for managing wildfire in the USA is fragmented across a mosaic of local, state, federal and tribal jurisdictions. Jurisdictional complexity reflects the number of discrete land jurisdictions affected by, and engaged on, a wildfire incident as well as the extent to which affected jurisdictions span several levels of government28. As jurisdictional complexity increases, the governance complexity of the incident intensifies as each jurisdiction may have differing priorities and no single jurisdiction has singular authority over the incident. Evidence suggests that the average jurisdictional complexity of wildfires has increased slightly (Fig. 2). Thirty per cent of the most jurisdictionally complex incidents during the study period occurred since 2016. However, as shown in Fig. 2, the trends differ from east to west. For example, in Northern California, the average wildfire in 2020 was 24% more jurisdictionally complex when compared with the average wildfire in 1999, whereas the southern region showed a 13% linear decrease in the average type 1 or 2 jurisdictional complexity. Standardization is based on the cumulative average fire size within the Geographic Area Coordinating Center (GACC). Slopes represent average percentage change in size within GACC by year. It is noteworthy that the Theil–Sen estimator used in calculating these trends subdues the impact of outlier incidents. For example, even though the region overall showed a decrease in complexity, two of the three most jurisdictionally complex incidents in the dataset occurred in the southern region (Supplementary Information).

Fig. 2: Trends in jurisdictional complexity of complex wildfires.
figure 2

a, Change in jurisdictional complexity 1999–2020 by region based on linear slopes normalized within region. Points reflect the locations of the top-5% most jurisdictionally complex wildfires coloured by time period (1999–2003, 2004–2009, 2010–2015 and 2016–2020). b, Theil–Sen linear trend (black) 90th percentile bootstrapped confidence bounds (dotted) of the annual national average for jurisdictional complexity of type 1 and 2 wildfires. The linear fit had a slope of 0.01, median absolute error of 0.28 and 90% slope confidence bounds of [0.05, −0.02]. c, Annual frequency of the top-5% most jurisdictionally complex wildfires, coloured by time period.

Full size image

Size of complex wildfires

Larger wildfires require larger fire organizations with more fire-fighters and equipment spread out across a larger area, which increases the management complexity of the incident. Results confirm that nationally, on average, type 1 and 2 wildfire incidents are increasing in size annually by 2.13 km2 (Fig. 3). This means an increasingly greater distance of fire perimeter must be managed and coordinated. Nearly one-third of the largest incidents in the study period have occurred since 2016. All regions but three have experienced an upward trend in fire size. California and the southwest region showed the strongest relative change, where the average incident size within each of these regions in 2020 is approximately 90% larger than it was in 1999.

Fig. 3: Trends in the average size of type 1 and 2 wildfire incidents.
figure 3

a, Change in average size of incidents 1999–2020 by region based on linear slopes normalized within region. Points reflect the locations of the top-5% largest wildfires coloured by time period (1999–2003, 2004–2009, 2010–2015 and 2016–2020). b, Theil–Sen linear trend (black) and 90th percentile bootstrapped confidence bounds (dotted) of the national annual average for size of type 1 and 2 wildfires. The linear fit had a slope of 2.35, median absolute error of 25.72 and 90th percentile slope confidence bounds of [4.98, 0.72]. c, Annual frequency of the top-5% largest wildfires, coloured by time period.

Full size image

Incident days at PL 4 or 5

The national PL is an index in the USA managed by the National Multi-Agency Coordinating Group (NMAC) based on national resource availability (www.nifc.gov/sites/default/files/2020-09/National_Preparedness_Levels.pdf). This index ranges from 1 to 5. Higher PL levels indicate that many or most fire response resources have been deployed and there is a scarcity of resources that can be allocated to any given incident. When a complex incident occurs during times when the nation is at PL 4 or 5, incident commanders may not have access to the fire-fighters and equipment they need and will need to calibrate their strategy on the basis of what can be accomplished with the resources that are available. Nationally, analysis suggests a slight increase in the average number of incident days at PL 4 or 5 per year over the past 22 years, although substantial annual variation was observed (Fig. 4). Of the incidents with the greatest number of incidents days at PL 4 or 5, 30% have occurred since 2016. Regional analysis reveals that this trend is driven by changes in the western USA. In these regions, wildfires are increasingly managed under conditions of resource shortages. For example, the average complex wildfire incident in Northern California in 2020 has seen, on average, a 75% increase in days at PL 4 or 5 since 1999. Interestingly, while the southern region has increasingly larger fires, the average type 1 or 2 incident is not experiencing more days at PL 4 or 5. This may suggest that regions with peak wildfire activity occurring off-cycle from the peaks in the west may benefit from increased availability of national resources.

Fig. 4: Trends in the number of incident days at national PL 4 or 5.
figure 4

a, Change in average number of incident days at PL 4 or 5 from 1999 to 2020 by region based on linear slopes normalized within region. Points reflect the locations of the top-5% of incidents with the most days at PL 4 or 5, coloured by time period (1999–2003, 2004–2009, 2010–2015 and 2016–2020). b, Theil–Sen linear trend (black) and 90th percentile bootstrapped confidence bounds (dotted) for incident days at PL 4 or 5. The linear fit had a slope of 0.13, median absolute error of 4.46 and 90th percentile slope confidence bounds of [0.75, −0.28]. c, Annual frequency of the top-5% most institutionally complex wildfires in terms of number of days at PL 4 or 5, coloured by time period.

Full size image

Structures threatened, damaged or destroyed

The findings indicate that complex wildfire incidents are becoming more threatening over time. On average, nationally, a complex wildfire in the USA is threatening, damaging or destroying approximately 38 more structures every year (Fig. 5). Of those wildfires representing the top-5% in terms of the number of structures threatened, damaged or destroyed, nearly half the most threatening wildfires between 1999 and 2020 have occurred since 2016. Findings further indicate that wildfires are increasingly threatening, damaging and destroying more structures in just about every region (Fig. 5). However, the southwest has seen the greatest relative change over time, with the average incident in 2020 being 220% more threatening or destructive than the average incident in 1999. In California, the linear average type 1 or 2 incident is 112% more threatening or destructive. This trend is driven by spikes in threatened, damaged or destroyed structures from key incidents between 2015 and 2020. A previous study29 cautioned that incident data on structures threatened, damaged and destroyed may represent the information available at the time and should be viewed as estimates.

Fig. 5: Change in average number of structures threatened, damaged or destroyed.
figure 5

a, Change in average number of structures threatened, damaged or destroyed by region based on linear slopes normalized within region. Points reflect the locations of the top-5% most complex wildfires in terms of the number of structures threatened, damaged or destroyed coloured by time period (1999–2003, 2004–2009, 2010–2015 and 2016–2020). b, Theil–Sen linear trend (black) and 90th percentile bootstrapped confidence bounds (dotted) of the national annual average of structures threatened. The linear fit had a slope of 47.66, median absolute error of 97.21 and 90th percentile slope confidence bounds of [99.13, 29.30]. c, Annual frequency of the top-5% most institutionally complex wildfires in terms of the number of structures threatened, damaged or destroyed, coloured by time period.

Full size image

Discussion

Changes in fire regimes, driven by changes in climate, invasive species, land use and land management, are affecting the nature of wildfire and its consequences for human settlements2,15,30. Effective adaptation will require, in part, an understanding of how these changes translate to changes in the task environment for the institutions that are responsible for mitigating the consequences of wildfire to communities. This study documents changes in key areas of institutional complexity for complex wildfire incidents in the USA. Findings indicate a national picture of an increasingly complex task environment for organizations and agencies responsible for managing these incidents; one that is likely to require the management and support of larger and more complex fire organizations and strain historical wildfire governance systems with a growing portfolio of competing jurisdictional priorities and interests that must be reconciled31. While this portfolio of potentially competing interests is expanding for IMTs, the west is experiencing more incident days when there are limited national resources to respond. Finally, the scale of threat to human settlements for the average type 1 and 2 wildfire incident has increased over the past 22 years. In addition to being a significant concern to local communities, this escalation of risk further exacerbates the complexity of incident response as it requires tighter coordination between wildfire management and emergency management operations, which are governed by different actors operating at different scales27.

These findings contribute to the understanding of the nature of the demands for institutional adaptation, raising important empirical and practical questions about the extent to which, and ways in which, existing institutions at local, state, tribal and federal levels are adapting to these new demands, providing an empirical basis to consider potential trade-offs in adaptive action. One example of this is a recent policy development in the USA to create complex IMTs32. These teams combine the previous two highest qualified categories of IMTs (type 1 and 2) into a single level of qualification and allow for greater ad hoc integration of type 1 or 2 individuals from other teams into key team-member roles. The intent of this institutional adaptation is, in part, to increase the number and availability of IMTs that can be deployed to complex incidents formerly requiring type 1 IMTs32. This adaptation aims to address critical resource availability challenges in staffing type 1 teams33 while the number of incident days at PL 4 and 5 increase in some regions. What is yet unknown is whether this adaptation may be associated with trade-offs with other areas of growing institutional complexity, as team members with potentially fewer qualifications face increasingly complex wildfire governance challenges both across jurisdictions and in the coordination with emergency management operations. Monitoring these changes and investigating how different aspects of institutional complexity interact to affect incident outcomes over time is an important direction for future research.

Clarifying the geographies where the pressures for adaptation are most acute is also critical to advancing a programme of study about the institutional adaptation of wildfire. Findings reveal regional variation not just in the magnitude of change but, importantly, in the direction of change. Implications of the risk profile by region are different based on directionality of change. Regions that have limited recent experience with complex wildfires may be particularly vulnerable to extreme events, such as the case of the 2016 Gatlinburg and 2023 Lahaina wildfires34. Hawaii is administratively grouped with Northern California GACC. Regional analysis shown here may mask within-GACC variation. On the other end of the spectrum, the magnitude and scope of increasing complexity in the west has been comparatively greater relative to other regions. The institutions in these regions have had to adapt in more substantial ways to managing larger and more jurisdictionally complex fire organizations, requiring more complex coordination with emergency operations to protect human settlements, and under more frequent conditions of resource scarcity. Future research should look to these regions as potential front runners of innovation in the management of this new task environment. The institutional innovations and adaptations that are proving effective in managing in this more complex task environment may hold valuable lessons for the rest of the nation as it prepares for the coming decades. Future research should also consider how national changes may have variable impacts across regions in consideration of these divergent task environments.

Wildfire has been recognized as a growing global challenge, linked to changes in climate35. However, little is known about how changes in biophysical characteristics of wildfire may have differing consequences in different societal and national contexts. The present study can speak only to changes in institutional complexity within the US context; however, this study raises important questions for future research about how these trends may vary across countries and globally. Research suggests that trends in the biophysical characteristics of fire are variable across country context28,36,37. Wildfire climatology is likely to be associated with different institutional demands for adaptation across different country contexts. Further, countries vary in their institutional systems for managing wildfire38. For example, it is possible that similar changes in biophysical environments could have very different consequences based on different institutional characteristics of wildfire response across countries.

Findings must take into account potential data limitations. Under extreme fire years, raw counts of team assignments may mask or under-represent institutional complexity as there are finite numbers of type 1 and 2 teams available to be assigned. As indicated by our analysis of changes in incident days at national PL 4 or 5, western regions are experiencing, on average, a growing number of incident days when national response resources are scarce due to the level of fire activity. This may necessitate IMTs assigned to manage multiple fires as part of the same incident (referred to as complexing; 29). It should also be noted that data carry across the pandemic of 2020, which had an impact on the availability of wildfire personnel39.

Last, findings from this study illustrate the need to give greater empirical attention to the coupling between changes in the biophysical aspects of climate change and needs for institutional adaptation in response to corresponding changes in the task environment. Our study examines large-scale trends aggregated by incident by year. However, temporal variation in institutional complexity is also present within incidents over time and different patterns, such as the number of days when structures are threatened, may also signal increases in institutional complexity. Further, institutional complexities associated with changes in fire funding and policy, insurance, an ageing workforce and Covid 19 offer fertile, policy-relevant directions for future research. Normal accidents theory has long argued that, as systems become more complex and interdependent, risks for failures also increase40. Despite this, the current state of scientific knowledge concerning the coupling of environmental change and requisite institutional adaptation is comparatively nascent. Successful adaptation to a changing climate will require an integrated understanding of coupled change in the biophysical environment to the associated task environment of governing institutions.

Methods

This study examined trends over time associated with five dimensions of institutional complexity associated with wildfire incidents. All analysis was conducted on publicly available datasets. All data and code are available41. Justification of each dimension of institutional complexity as well as a description of the analytic approach are outlined below.

Five dimensions of wildfire incident institutional complexity

Institutional complexity was conceptualized as incident characteristics that would affect the task environment for responsible agencies and their IMTs. We analysed five wildfire incident characteristics associated with increased institutional complexity: (1) prevalence of type 1 and 2 incidents, (2) jurisdictional complexity, (3) incident size, (4) incident days at PL 4 or 5 and (5) number of structures threatened, damaged or destroyed.

First, the frequency of incidents is an important feature of institutional complexity because the extent to which demand for the most elite and qualified incident response resources is increasing will place greater burden and strain on national resources. Further, in the USA, incident response personnel are predominantly reservists who work full-time elsewhere, often in key land or fire management leadership positions. More frequent deployment of these individuals from their primary posts increases institutional complexity of incidents because it will add further strain on both these agencies and the team members, who may struggle to balance these competing demands.

Second, trends in the jurisdictional complexity of complex incidents are critical for understanding changing demands in wildfire governance. IMTs work under the authority of one or more jurisdictions who have requested their assistance. When managing an incident wholly contained within the bounds of a single jurisdiction, such as a National Forest, there is a clear principal–agent relationship between the National Forest and the IMT. However, if a fire crosses jurisdictional boundaries, the objectives of multiple jurisdictions must now be taken into consideration and balanced into a unified response strategy. Previous scholarship has demonstrated that different land jurisdictions have different missions and often perceive competing interests in the strategic management of wildfire42,43. For example, federal agencies such as the US Forest Service and the Bureau of Land Management have a multiple-use mandate from Congress requiring them to balance land interests associated with conservation, recreation and natural resource use. Other jurisdictions have more narrow mandates, such as a commercial timber manager whose mandate is to protect the financial interests of their shareholders. The more jurisdictions threatened or affected by a wildfire, the more institutionally complex a wildfire is to manage in terms of reconciling a more diverse array of interests and objectives31,43.

Third, we consider changes over time in the size of incidents. Most often, incidents with more acreage are more institutionally complex, not just because they may involve more jurisdictions, but also because they generally require a larger fire organization. Larger fire organizations spread out over a larger fire perimeter result in more complex planning and logistics as well as potentially larger and more complex spans of managerial control.

Fourth, we consider trends in the number of incident days managed at national PL 4 or 5. The national PL is an index in the USA managed by the NMAC on the basis of national resource availability (www.nifc.gov/sites/default/files/2020-09/National_Preparedness_Levels.pdf). The national PL reflects the cache of IMTs, fire crews and other wildfire resources that can be called upon to respond to a wildfire incident. This index ranges from 1 to 5. At PL 1, fire activity is minimal and there are ample wildfire response resources to deploy on any incident. As the PL index increases, there is more wildfire activity and resources are increasingly committed to incidents. At PL 5, at least 80% of the US resources for managing wildfire are already committed. This means that if new incidents occur or existing incidents intensify, resources may not be available or may need to be pulled from other active incidents. In other words, at higher PL levels, IMTs may not have the resources available to protect everything; trade-offs may be necessary. Consideration of PL is an important aspect of institutional complexity because it speaks to the realities of resource constraints when managing wildfire. If more complex wildfire incidents are concentrated during the same time period, this will have greater consequences than if these incidents are spread out across a longer timeframe.

Last, we consider trends over time in the number of structures threatened, damaged or destroyed on type 1 or 2 incidents. This metric speaks to both the human cost of a wildfire and to changes in the institutional complexity of wildfire, as fire management activities must become more tightly coordinated with emergency management. Wildfire is unique in that incident responders do not attempt to direct tornados or hurricanes away from populated areas but fire-fighters do attempt to intervene to keep wildfire away from populated areas. When wildfire threatens human settlements, emergency response and structure protection operations become activated in the same geographic area as wildfire operations. There is increasing pressure on public information for coordinated and timely communication44. Emergency management efforts are managed by an entirely separate set of organizations and agencies, yet these activities are highly interdependent with fire operations. The institutional complexity of the incident increases as the number of different organizations and agencies responsible for these emergency support functions become engaged.

Database

The dataset used in this study reflects the population of type 1 and 2 wildfire incidents from 1999 to 2020 as identified in Federal 209 incident reports41. We identified relevant wildfire incidents using the all-hazards dataset of ref. 29, which compiles Federal 209 data into incident-level summaries. Federal 209 reports are daily reports required of all type 1 and 2 incidents and are submitted by the assigned IMT. The unit of analysis is a discrete incident. We included incidents if they contained at least one 209 record as a type 1 or 2 incident. Consistent with ref. 29, wildfire complexes were included as a single unique incident rather than separated by their component fire perimeters. This sampling strategy resulted in a population-level dataset of 3,974 type 1 or 2 wildfires. However, the specific number of incidents used to analyse each variable varies on the basis of data availability in the all-hazards dataset and subsequent processing.

Analysis approach

For each variable, we evaluated trends nationally, regionally and through time by integrating nine national datasets, including the all-hazards dataset29, the monitoring trends in burn severity fire footprints spatial data45, the National Interagency Fire Center (NIFC) PL dataset and six multiscale jurisdictional boundary datasets summarized above (for detailed description by variable, see Supplementary Information).

All analyses were performed using R statistical software (v.4.2.1)46. We used the START_YEAR field in the all-hazards dataset of ref. 29 to assign each wildfire to a year (1999–2020). For each variable, values for all incidents were averaged annually and trend lines fit to these averaged values. We used the Theil–Sen estimator47 with 90th percentile bootstrapped confidence intervals from the openAir R package (v.2.18)48 to estimate linear trends through time for each variable. Analysis code is available49.

We relied on the GACC boundaries to delineate the regional analyses. GACCs were selected because they are a central administrative entity for coordinating incident support. GACC delineations are governed by the NIFC and the National Coordination Center. Currently, there are ten GACC regions: (1) Northwest, (2) Northern California, (3) Southern California, (4) Northern Rockies, (5) Great Basin, (6) Southwest, (7) Rocky Mountain, (8) Eastern, (9) Southern and (10) Alaska. GACC boundaries changed in 2001, so we assigned fires that occurred between 1999 and 2001 to their present GACCs for consistency through time. See https://gacc.nifc.gov/ for more information.

GACC boundaries were identified using a GACC shapefile (National GACC boundaries, 2023, see ‘Data availability’). We used the point of origin (latitude and longitude) to assign wildfires to individual GACCs. Eighty-four incidents did not have a point of origin to assign to a region. The total number of incidents assigned to the 10 regions is 3,890 instead of 3,974.