Model fires, not ignitions: Capturing the human dimension of global fire regimes

Fire regimes are intrinsically shaped by humans, but current global fire models do not reflect the diverse objectives humans have for managing fire. With new data sources and collaboration across disciplines, an improved understanding of human influences on fire regimes is increasingly possible and essential to represent. Fire regimes are intrinsically shaped by humans, but current global fire models do not reflect the diverse objectives humans have for managing fire. With new data sources and collaboration across disciplines, an improved understanding of human influences on fire regimes is increasingly possible and essential to represent. The future of fire on Earth will depend on the interaction between climate change and multiple aspects of wider socio-economic change. Fire regimes have been shaped through millennia of human fire use, multi-decadal fire suppression, and ongoing land use change and intensification.1Bowman D.M.J.S. Balch J. Artaxo P. Bond W.J. Cochrane M.A. D'antonio C.M. DeFries R. Johnston F.H. Keeley J.E. Krawchuk M.A. et al.The human dimension of fire regimes on Earth.J. Biogeogr. 2011; 38: 2223-2236https://doi.org/10.1111/j.1365-2699.2011.02595.xCrossref PubMed Scopus (825) Google Scholar Today, humans continue to live with fire across the world, often in ecosystems that have co-evolved with fire,2Coughlan M. Magi B. Derr K. A global analysis of hunter-gatherers, broadcast fire use, and lightning-fire-prone landscapes.Fire. 2018; 1: 41https://doi.org/10.3390/fire1030041Crossref Scopus (18) Google Scholar and in which fire use can provide a vital means of sustaining and enabling human livelihoods.3Smith C. Perkins O. Mistry J. Global decline in subsistence-oriented and smallholder fire use.Nat. Sustain. 2022; 5: 542-551https://doi.org/10.1038/s41893-022-00867-yCrossref Scopus (11) Google Scholar Understanding how human and ecosystem interaction combine to produce observed patterns of fire, and may produce changes in future fire regimes, is vital to informing data-driven and effective adaptive management responses to fires. However, this view of global fire regimes, as the product of active and passive human participation in underlying biophysical processes, is not reflected in current state-of-the-art global fire models. Rather, humans are usually represented as an ignition source by a single global function.4Rabin S.S. Melton J.R. Lasslop G. Bachelet D. Forrest M. Hantson S. Kaplan J.O. Li F. Mangeon S. Ward D.S. et al.The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions.Geosci. Model Dev. (GMD). 2017; 10: 1175-1197https://doi.org/10.5194/gmd-10-1175-2017Crossref Scopus (153) Google Scholar This disregards the fact that human fire use varies markedly around the world. Humans often actively set and manage fires for distinct reasons and with the intention of burning specific areas (often small), intensities (usually low), and locations (e.g., their livestock pastures). Here, we argue that this mis-conceptualization of human fire interactions prevents robust assessment of the drivers of present and future fire regimes and detailed quantitative assessment of the effectiveness of possible climate change adaptation options. We then present a range of new data sources that can enable improved representation of human influences on fire regimes in global models, before critically evaluating recent attempts toward this goal. Global fire models arose out of a need to represent fire as an essential earth system process in dynamic global vegetation models (DGVMs) and Earth system models (ESMs), influencing carbon emissions and vegetation distribution.4Rabin S.S. Melton J.R. Lasslop G. Bachelet D. Forrest M. Hantson S. Kaplan J.O. Li F. Mangeon S. Ward D.S. et al.The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions.Geosci. Model Dev. (GMD). 2017; 10: 1175-1197https://doi.org/10.5194/gmd-10-1175-2017Crossref Scopus (153) Google Scholar,5Hantson S. Arneth A. Harrison S.P. Kelley D.I. Prentice I.C. Rabin S.S. Archibald S. Mouillot F. Arnold S.R. Artaxo P. et al.The status and challenge of global fire modelling.Biogeosciences. 2016; 13: 3359-3375https://doi.org/10.5194/bg-13-3359-2016Crossref Scopus (254) Google Scholar DGVMs and ESMs are typically process-based numerical simulation models, used, for example, within the Intergovernmental Panel on Climate Change (IPCC) framework to simulate the large-scale changes in physical climate and land cover that have occurred historically and when exploring future climate scenarios. This has, understandably, led to fire being represented as a purely biophysical process: a sub-grid-scale phenomenon that can be simulated as a function of gridbox-mean biophysical variables, using either empirical or process-based relationships that are assumed to hold for any time period from the pre-industrial through the end of this century or beyond. Increasingly, these global models are also being used to explore climate impacts and adaptive responses and to attribute observed changes in fire regimes over recent decades.6Burton C. Lampe S. Kelley D. Thiery W. Hantson S. Christidis N. Gudmundsson L. Forrest M. Burke E. Chang J. et al.Global burned area increasingly explained by climate change.Research Square. 2023; (Preprint at)https://doi.org/10.21203/rs.3.rs-3168150/v1Crossref Google Scholar At the same time, advances in satellite remote sensing of fires have improved our ability to detect the substantial cumulative contribution of small fires to global burned area and emissions. The fifth and latest version of the Global Fire Emissions Database (GFED5), after incorporating calibration data from high-resolution Sentinel-2 and Landsat images, now estimates global burnt area to be 61% higher than the previous GFED4s product and 93% higher than the Moderate Resolution Imaging Spectroradiometer (MODIS) MCD64A1 burnt-area product.7Chen Y. Hall J. van Wees D. Andela N. Hantson S. Giglio L. van der Werf G.R. Morton D.C. Randerson J.T. Multi-decadal trends and variability in burned area from the fifth version of the Global Fire Emissions Database (GFED5).Earth Syst. Sci. Data. 2023; 15: 5227-5259https://doi.org/10.5194/essd-15-5227-2023Crossref Scopus (12) Google Scholar These previously underestimated small fires are frequently human-set (e.g., for agricultural waste removal), and so if we wish to represent the total contribution of fire in the earth system, and correctly understand the drivers of changing fire regimes, our models will need to model the human dimension of fire more explicitly. Systematic and quantitative evidence now abounds that humans actively apply fire to the landscape, and frequently control its spread, in support of economic activity and human well-being.3Smith C. Perkins O. Mistry J. Global decline in subsistence-oriented and smallholder fire use.Nat. Sustain. 2022; 5: 542-551https://doi.org/10.1038/s41893-022-00867-yCrossref Scopus (11) Google Scholar At the most basic level, there are pronounced quantitative differences in fire regimes generated by human fire use in croplands and where fire is "broadcast" onto the landscape—where fire is used in pastures, forests, and savannas for livestock farming, hunting, gathering, and fire regime management (Figure 1). Based on the Database of Anthropogenic Fire Impacts (DAFI),8Millington J.D.A. Perkins O. Smith C. Human fire use and management: a global database of anthropogenic fire impacts for modelling.Fire. 2022; 5: 87https://doi.org/10.3390/fire5040087Crossref Scopus (4) Google Scholar human fires in croplands have median and mean sizes of 0.5 hectares (ha) and 20.5 ha, respectively, compared to 7.3 ha and 142.0 ha for broadcast fires. This cropland-broadcast distinction is far from the only example of humans intentionally shaping fire regimes. The reintroduction of Aboriginal burning in Australia led to a distinct shift in the fire regime—from one dominated by few, large, and intense fires under fire exclusion to numerous, small, and low-intensity fires under Aboriginal management.9Bowman D.M.J.S. Walsh A. Prior L.D. Landscape analysis of Aboriginal fire management in Central Arnhem Land, north Australia.J. Biogeogr. 2004; 31: 207-223https://doi.org/10.1046/j.0305-0270.2003.00997.xCrossref Scopus (99) Google Scholar Notably, there was less overall burned area after the reintroduction of Aboriginal fire, and the smaller flame heights were less likely to damage tree canopies. Similarly, fire regimes in the Brazilian Amazon follow a predictable trajectory under land use intensification: from initial deforestation-driven fires, through rotational burning under extensive livestock farming, through fire exclusion under intensive soybean farming.10Ribeiro A.F.S. Santos L. Randerson J.T. Uribe M.R. Alencar A.A.C. Macedo M.N. Morton D.C. Zscheischler J. Silvestrini R.A. Rattis L. et al.The time since land-use transition drives changes in fire activity in the Amazon-Cerrado region.Commun. Earth Environ. 2024; 5: 96https://doi.org/10.1038/s43247-024-01248-3Crossref Scopus (1) Google Scholar Simply put, different kinds of human fire use and management generate categorically different types of fire events, with predictable differences in the resulting fire frequency, fire size, and vegetation types burned. However, across the current generation of global fire models, humans are almost always conceptualized simply as an ignition source, with human ignitions typically represented as a function of population density.4Rabin S.S. Melton J.R. Lasslop G. Bachelet D. Forrest M. Hantson S. Kaplan J.O. Li F. Mangeon S. Ward D.S. et al.The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions.Geosci. Model Dev. (GMD). 2017; 10: 1175-1197https://doi.org/10.5194/gmd-10-1175-2017Crossref Scopus (153) Google Scholar Following an "ignition," the spread, duration, and subsequent size of that fire are parameterized as a physical process dependent on the vegetation and weather, sometimes also with a probability of fire suppression (active firefighting) calculated again as a simple function of population density or gross domestic product. This conceptualization of humans as a numerical parameter, a predictor with a universal relationship that modifies a physical process of fire occurrence, disregards the deliberate, variegated, and intentional way that humans interact with fire. It supposes that the rate of human fire ignition is analogous to the way that lightning strikes create natural ignitions, and the only preference humans express related to fire is to remove it from the landscape. But in reality, humans have objectives behind their use of fire and desire the landscape to be a certain way. For example, while much human fire use results in small and low-intensity fires, in rangeland environments people may set fires late in the dry season to ensure they are intense enough to remove woody shrubs that are unpalatable to livestock.11Bloesch U. Fire as a Tool in the Management of a Savanna/Dry Forest Reserve in Madagascar.Appl. Veg. Sci. 1999; 2: 117-124https://doi.org/10.2307/1478888Crossref Scopus (52) Google Scholar Humans are active participants in both the inputs and outputs of fire regimes: humans are not like lightning. One consequence of the current conceptualization of humans in global models is that there is little capacity to systematically evaluate climate change adaptation options. For example, the absence of representation of both prescribed fire (fire used to reduce fuel loads and manage wildfire risk) and Indigenous/cultural burning makes it challenging to assess the role of fuel reduction methods. While vegetation modelers increasingly recognize the importance of landscape fragmentation, unlike cropland conversion or road building, prescribed fire is often applied to intentionally fragment fuels.12Moreira F. Ascoli D. Safford H. Adams M.A. Moreno J.M. Pereira J.M.C. Catry F.X. Armesto J. Bond W. González M.E. et al.Wildfire management in Mediterranean-type regions: paradigm change needed.Environ. Res. Lett. 2020; 15011001https://doi.org/10.1088/1748-9326/ab541eCrossref Scopus (280) Google Scholar Assuming that prescribed fire behaves on the landscape in a similar way to an uncontrolled wildfire is unlikely to provide a robust measure of its efficacy for fuel management. Despite the complexities, real progress has been made toward representing human-fire interactions in global scale models. Perhaps most critically, several global datasets of human fire use and management now exist, which can be used to support improved numerical modeling (Table 1). These data span multiple sources: human observations, field measurements, remote sensing, institutional records, and expert elicitation. Such data provide the possibility of creating "bottom-up" parameterizations of the conditions under which humans use fire, including, for example, the bio-climatic conditions under which prescribed fire use is possible. This, in turn, can inform assessment of how the conditions regulating human fire use may alter under climate change.Table 1Global datasets on human fire use and managementSourceData typeReporting unitKey characteristicsStatusLivelihood Fire Database (LIFE)literature case study dataarea of underlying case studiesfocus on qualitative understanding of human fire use & governancepublished (Smith et al.3Smith C. Perkins O. Mistry J. Global decline in subsistence-oriented and smallholder fire use.Nat. Sustain. 2022; 5: 542-551https://doi.org/10.1038/s41893-022-00867-yCrossref Scopus (11) Google Scholar)Database of Anthropogenic Fire Impacts (DAFI)literature case study dataarea of underlying case studiesquantitative information on all forms of human fire use & governancepublished (Millington et al.8Millington J.D.A. Perkins O. Smith C. Human fire use and management: a global database of anthropogenic fire impacts for modelling.Fire. 2022; 5: 87https://doi.org/10.3390/fire5040087Crossref Scopus (4) Google Scholar)Global Fire Use Survey (GFUS)expert elicitationnational-ecoregion polygonswide-ranging assessments of fire regime, human drivers, active policy & its effectivenessin review (Smith et al.13Smith C. Mistry J. Perkins O. Global Fire Use Survey data and metadata.Dataset. 2024; (Figshare [Internet])https://doi.org/10.17637/rh.23659737.v2https://royalholloway.figshare.com/articles/dataset/Global_Fire_Use_Survey_data_and_metadata/23659737/2Crossref Google Scholar)GlobalRxinstitutional recordsgeolocated fire perimetershighly detailed documents on prescribed firein preparation (Hsu et al.14Hsu A. Thurgood J. Smith A. Anderson L. Clarke H. Doerr S. Fernandes P. Kolden C. Santín C. Strydom T. et al.GlobalRx: A global assemblage of regional prescribed fire records for use in assessments of climate change impacts.EGU General Assembly. 2024; 2024 (EGU24–12529)https://doi.org/10.5194/egusphere-egu24-12529Crossref Google Scholar)GloCABremote sensing0.25° gridspatio-temporally consistent detection of cropland firepublished (Hall et al.15Hall J.V. Argueta F. Zubkova M. Chen Y. Randerson J.T. Giglio L. GloCAB: global cropland burned area from mid-2002 to 2020.Earth Syst. Sci. Data. 2024; 16: 867-885https://doi.org/10.5194/essd-16-867-2024Crossref Scopus (4) Google Scholar)LIFE, DAFI, GFUS, and GloCAB are all freely available online; see references for detail. Open table in a new tab LIFE, DAFI, GFUS, and GloCAB are all freely available online; see references for detail. Limitations of such data include a need for further studies in regions such as Central Asia and Southern Russia,8Millington J.D.A. Perkins O. Smith C. Human fire use and management: a global database of anthropogenic fire impacts for modelling.Fire. 2022; 5: 87https://doi.org/10.3390/fire5040087Crossref Scopus (4) Google Scholar where data reported in the literature are sparse and GFED5 shows a large increase in burned area from GFED4. Just as crucially, given ongoing uncertainty about how much of the global burned area is captured by remote sensing, field data that link qualitative research about local fire use practices with quantitative outcomes in fire regimes are invaluable to modeling but still in comparatively short supply. For example, as noted by Millington et al.8Millington J.D.A. Perkins O. Smith C. Human fire use and management: a global database of anthropogenic fire impacts for modelling.Fire. 2022; 5: 87https://doi.org/10.3390/fire5040087Crossref Scopus (4) Google Scholar in the DAFI, only one in two (52%) documented instances of hunter-gatherer fire use reported any quantitative characteristics of the resulting fires. The Wildfire Human Agency Model (WHAM!)16Perkins O. Kasoar M. Voulgarakis A. Smith C. Mistry J. Millington J.D.A. A global behavioural model of human fire use and management: WHAM! v1.0.Geosci. Model Dev. (GMD). 2024; 17: 3993-4016https://doi.org/10.5194/gmd-17-3993-2024Crossref Scopus (0) Google Scholar is one example of recent progress, in which the drivers of human fire use and management have been represented with a global-scale agent-based model. Based on the DAFI, this model represents human fire use in its underlying land use context and seeks to capture broad patterns in what vegetation people do (and do not) burn, how much they burn, and how this varies around the world. Importantly, such a representation could allow vegetation models to distinguish between fires that are free-spreading according to their biophysical drivers and those whose spread is being controlled by human management. However, there remain limitations in this work. First, imposing a Western conceptualization of rationality on communities with diverse cultural and environmental understandings may miss important details surrounding how humans make decisions around fire use. Laris17Laris P. Burning the seasonal mosaic: preventative burning strategies in the wooded savanna of southern Mali.Hum. Ecol. 2002; 30: 155-186https://doi.org/10.1023/A:1015685529180Crossref Scopus (194) Google Scholar argues, for example, that many fires serve multiple purposes and that humans use fire to create an intended landscape outcome rather than a particular narrow economic purpose. To some degree, current global fire models take this narrow view of human decision-making to an extreme, by representing human agency in fire suppression but not in active fire use.4Rabin S.S. Melton J.R. Lasslop G. Bachelet D. Forrest M. Hantson S. Kaplan J.O. Li F. Mangeon S. Ward D.S. et al.The Fire Modeling Intercomparison Project (FireMIP), phase 1: experimental and analytical protocols with detailed model descriptions.Geosci. Model Dev. (GMD). 2017; 10: 1175-1197https://doi.org/10.5194/gmd-10-1175-2017Crossref Scopus (153) Google Scholar It is arguable that this reproduces a global North perspective, in which intensive fire suppression combined with unfamiliarity with fire among the general population creates regimes dominated by accidental fires whose threat is addressed primarily through firefighting.18Fischer A.P. Spies T.A. Steelman T.A. Moseley C. Johnson B.R. Bailey J.D. Ager A.A. Bourgeron P. Charnley S. Collins B.M. et al.Wildfire risk as a socioecological pathology.Front. Ecol. Environ. 2016; 14: 276-284https://doi.org/10.1002/fee.1283Crossref Scopus (159) Google Scholar Similarly, the barrier between managed and unmanaged (controlled and uncontrolled) human fire use is often blurry. For example, in Indigenous patch-burning practices, fire may be applied to certain vegetation types at cooler or wetter times of year in order to create a landscape of fragmented patches that serve to prevent large, late-dry-season wildfires.17Laris P. Burning the seasonal mosaic: preventative burning strategies in the wooded savanna of southern Mali.Hum. Ecol. 2002; 30: 155-186https://doi.org/10.1023/A:1015685529180Crossref Scopus (194) Google Scholar However, each individual fire may not be controlled in the way a forestry service may control it, but rather is controlled through an overall system of community fire practice. Alternatively, in some more fuel-constrained grasslands, communities may choose to burn hot and intense fires late in the dry season to, for example, remove woody vegetation and prevent the spread of tsetse flies.19Lamprey R.H. Reid R.S. Expansion of human settlement in Kenya's Maasai Mara: what future for pastoralism and wildlife?.J. Biogeogr. 2004; 31: 997-1032https://doi.org/10.1111/j.1365-2699.2004.01062.xCrossref Scopus (185) Google Scholar Such fires are intentional, but it is not clear how far they can be considered controlled. There also remain practical and conceptual challenges to better representing human fire use in DGVMs and ESMs, even if we had the data to inform a perfect prediction of fire use choices. One challenge is compatibility with underlying model structures: to make numerical simulation of the earth system tractable, global models divide the globe into gridboxes with typically coarse spatial resolutions of tens or hundreds of kilometers. Implicit in this approach is the assumption that the aggregate effect of unresolved sub-grid processes can be represented as a gridbox-average value. But for certain types of human fire use, such as prescribed fire to create fuel breaks and reduce fuel loads, the sub-grid spatial structure is an essential aspect of this fire use. Representing prescribed burning as a non-spatial fraction of a coarse model gridbox neglects the fact that this fire was applied with an intentional placement to fragment landscape vegetation and reduce subsequent fire risk. Parameterizing the seasonal timing of human fire use is another unresolved challenge. Vegetation and climate models typically operate on daily or sub-daily timesteps and assume that the variables of interest have well-defined rates of change that can be numerically integrated from one timestep to the next. The seasonality of unmanaged fire, for instance, is emergent from the time evolution of fuel load and meteorological fire risk, both of which are continuous physical quantities. By contrast, human fire use may only take place sporadically at particular times of year, centered, for instance, around discrete economic activities such as crop harvest.3Smith C. Perkins O. Mistry J. Global decline in subsistence-oriented and smallholder fire use.Nat. Sustain. 2022; 5: 542-551https://doi.org/10.1038/s41893-022-00867-yCrossref Scopus (11) Google Scholar,8Millington J.D.A. Perkins O. Smith C. Human fire use and management: a global database of anthropogenic fire impacts for modelling.Fire. 2022; 5: 87https://doi.org/10.3390/fire5040087Crossref Scopus (4) Google Scholar Whether this seasonality can be inferred from physical properties of the climate system, at least for certain categories of fire use, is the subject of ongoing research. Lastly, DGVMs and ESMs are routinely employed to study the historical periods from pre-industrial to present day, where they have contributed enormously to our understanding and attribution of historical climate change and its impacts.20Eyring V. Gillett N.P. Achuta Rao K.M. Barimalala R. Barreiro Parrillo M. Bellouin N. Cassou C. Durack P.J. Kosaka Y. McGregor S. et al.Human Influence on the Climate System.in: Masson-Delmotte V. Zhai P. Pirani A. Connors S.L. Péan C. Berger S. Caud N. Chen Y. Goldfarb L. Gomis M.I. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, 2021: 423-552https://doi.org/10.1017/9781009157896.005Google Scholar When modeling physical processes, we expect our models to be grounded in universal physical laws, or at least empirically derived relationships that can be assumed to remain constant over space and time. Not so for human-fire relationships: human behaviors and responses to the same input can vary widely both between different time periods and across different locations, based, for instance, on past experience, cultural and societal knowledge, education, policy choices and governance, and myriad other factors. Even if we had complete data for present-day fire use practices, there is a very limited basis for assuming that relationships and drivers of fire use/management today will have remained constant through history—an issue that may require us to re-appraise the epistemological status we give our models and which questions they can meaningfully be applied to. As climate change adaptation increasingly becomes an immediate rather than future-focused question, credible consideration and representation of the human dimensions of such questions in process-based biophysical models will become an increasingly important research challenge. Researchers working across disciplinary boundaries, leveraging the anthropological insight of social studies with the generalizability of process-based models and the global perspective of remote sensing, will be critical to addressing this challenge. The authors are grateful for funding from the Leverhulme Trust under grant RC-2018-023. Conceptualization: M.K. and O.P.; writing – original draft: M.K. and O.P.; writing – review and editing: all authors; visualization: O.P. The authors declare no competing interests.