Global decline in subsistence-oriented and smallholder fire use

Controlled fire use for hunting, gathering, smallholder agriculture and pastoralism shapes ecologies and enhances livelihoods worldwide. Yet, at global scale, we know little about how these practices influence human well-being, ecologies and wildfire risk. As a basis for global syntheses, we collated information from the literature about fire practices in 587 case study locations spanning the globe. Here, we assess the coverage and completeness of these data. Limited quantitative data, particularly, present a challenge for improved modelling of anthropogenic influences on fire regimes. We also analyse global trends in fire practices from these studies, finding evidence that subsistence-oriented fire practices have declined in recent decades, while market-oriented fire practices have increased. Implications of these changes can include reduced biodiversity in fire-dependent ecosystems, increased wildfire risk, reduced household income and loss of cultural identity. The case studies point to important drivers of changing fire practices, especially economic pressures and state governance. Controlled use of fire for subsistence and smallholder livelihoods has undoubtedly shaped ecosystems but we have limited research on the practices and extent. This analysis of nearly 600 case study locations finds fire use is changing in ways that could pose risks to smallholder livelihoods as well as wildlife and biodiversity.

W orldwide, controlled fire use is culturally and economically significant across human societies for subsistence-oriented or smallholder livelihoods, including for hunting and gathering, pastoralism and agriculture 1 . These fire practices often draw on deep, place-based knowledge of the feedbacks between fire and its biophysical and social environment 2 . Beyond immediate livelihood benefits that structure the application of anthropogenic fire to landscapes, such fires have co-benefits. People tend to make small, frequent fires that produce more heterogenous landscapes, supporting greater biodiversity and fragmenting fuels, which reduces wildfire risk 3,4 . Simultaneously, especially in changing climates and environments or where local governance institutions and fire knowledge are being lost, anthropogenic fire can potentially lead to uncontrolled wildfires, posing risks to human health and well-being. Generally, such wildfire events dominate public and policy discourse around fire 5 rather than the social and ecological relevance of controlled fire use.
Recent global studies demonstrate that annual burned area has decreased in recent decades 6 , while extreme wildfire events are becoming more frequent and intense [6][7][8] . Drivers of these trends include fragmentation of fuel landscapes with land use change, accumulation of flammable vegetation after suppression of human fire use and lengthening fire weather windows due to climate change 9,10 . In this context we need to better understand how smallholder and subsistence livelihood-oriented fire use is currently changing and the implications for fire ecologies, human well-being and wildfire events 11 . It would be virtually impossible to use available remotely sensed burned area or active fire datasets to assess global trends in these fire practices because their effects are not easily disentangled from those of lightning fires and other forms of anthropogenic fire use and suppression. The global remotely sensed datasets for which there are time-series data available are also at too coarse a resolution to capture many forms of smallholder and subsistence fire use 12 . Most global fire models represent all human fire use and suppression as a function of variables such as population density or gross domestic product (GDP), failing to account for different ways in which people use fire or how this varies across environmental and social contexts 13 . Similar variables are used to make inferences about the human signal in global fire remote sensing studies 6,14 .
Since the 1990s, a growing number of case studies have examined local fire knowledge and practices but with limited synthesis of this literature at the global scale. Existing literature reviews have drawn on few sources, have often been biased geographically towards North America and Australia and have examined narrow research questions 2,4,[15][16][17][18] . To build an empirical basis for studying livelihood-oriented fire practices at the global scale, we conducted a systematic review of 592 published and unpublished sources to create an open-source database, the livelihood fire database (LIFE) 19 , describing 1,708 contemporary subsistence-oriented and smallholder fire use and mitigation practices in 587 case study locations. LIFE contains data on both social and biophysical aspects of fire practices. This resource can support, for example, future analyses covering the spatiotemporal patterns of fires set for diverse livelihood purposes, different forms of fire governance and drivers of change in fire practices. Here, we assess the coverage and completeness of the LIFE database, present a typology of fire-use purposes developed from the database and assess contemporary trends in these fire practices.

Results and discussion
The LIFE database demonstrates that fire use and mitigation practices have remained important within subsistence-oriented and smallholder livelihoods worldwide over the past 30 years. Yet, subsistence-oriented fire practices have declined at the global scale, while market-oriented fire practices have increased. The following sections speak to these findings in turn and discuss their implications for human well-being, ecologies and wildfire risk. Finally, we reflect on gaps in our knowledge of these fire practices, as highlighted by LIFE, and implications for future research and fire policies.
Fire practices contribute to diverse livelihoods worldwide. LIFE contains information about 1,708 fire practices, of which 92% were actively practiced at the time of the case study research and 8% had Global decline in subsistence-oriented and smallholder fire use Cathy Smith 1,2 ✉ , Ol Perkins 1,3 and Jayalaxshmi Mistry 1,2 Controlled fire use for hunting, gathering, smallholder agriculture and pastoralism shapes ecologies and enhances livelihoods worldwide. Yet, at global scale, we know little about how these practices influence human well-being, ecologies and wildfire risk. As a basis for global syntheses, we collated information from the literature about fire practices in 587 case study locations spanning the globe. Here, we assess the coverage and completeness of these data. Limited quantitative data, particularly, present a challenge for improved modelling of anthropogenic influences on fire regimes. We also analyse global trends in fire practices from these studies, finding evidence that subsistence-oriented fire practices have declined in recent decades, while market-oriented fire practices have increased. Implications of these changes can include reduced biodiversity in fire-dependent ecosystems, increased wildfire risk, reduced household income and loss of cultural identity. The case studies point to important drivers of changing fire practices, especially economic pressures and state governance.
been lost but were practiced within the living memory of research participants. A total 95% of practices in LIFE are examples of fire use. For 23% of these fire-use practices, we have information about associated fire control measures. Of practices in LIFE, 1% are examples of 'opportunistic' use of fires or burned areas created by lightning or other people. The remaining 3% of practices are general fire mitigation measures taken independently of fire use, including fire prevention and suppression practices. Our focus here is largely on fire use but there is need for future research addressing fire control and mitigation specifically.
These practices were recorded for case study locations spanning 84 countries across all continents except Antarctica, with most locations in Brazil, Indonesia, Australia, Mexico and India (Fig.  1). The United States, Canada, Australia and China have the highest number of practices that were lost at the time of the case study research. Research biases shape the distribution of case studies to some extent. The high number of cases in Australia and the United States, for example, is probably more reflective of a bias towards research in these countries than it is of widespread contemporary subsistence-oriented and smallholder fire practices. Conversely, the low number of cases in eastern Europe and northern, western and central Asia probably reflects lack of research rather than lack of fire practices. Despite research biases, the high concentration of cases in the equatorial tropics probably reflects a high prevalence of subsistence-oriented and smallholder fire practices in these regions.
The fire practices in LIFE contribute to livelihoods on a spectrum between subsistence oriented and market oriented. A total 43% of fire practices in LIFE are subsistence oriented (associated with a product produced, gathered or hunted mainly for subsistence, where surpluses may be marketed), 15% are market oriented (associated with a product produced, gathered or hunted mainly for sale), 2% are practices that are themselves marketed as a service and 20% are neither associated with markets nor subsistence.
Using LIFE, we developed a hierarchical fire-use purpose classification system including eight higher-tier categories: agriculture, pastoralism, hunting and fishing, gathering, charcoal and firewood production, movement, human health and well-being and social signals. Each of these higher-tier categories is associated with several of 29 lower-tier categories (Table 1). This scheme aligns broadly with, but is more detailed than, those developed in previous studies 4,16,20 . Our lower-tier categories include both direct reasons for fire use, such as driving game for hunting or clearing vegetation for swidden agriculture, as well as co-benefits, such as reducing fuel loads or maintaining cultural identity. It is likely that many anthropogenic fires, especially those set in tropical savannas, are associated with multiple purposes, including proximate reasons and co-benefits 21 . We listed multiple purposes against a fire practice only where a source explicitly stated that multiple reasons were associated with the same ignitions (17% of cases). It is probable that a higher proportion of anthropogenic fires are associated with multiple purposes. Most sources analysed did not distinguish between direct reasons for fire use and co-benefits, so we could not make this distinction in LIFE. This is a limitation for future analyses in that it is the more proximate reasons that are more likely to sustain fire use and consciously shape how fire is applied to the landscape (see comment by R. Bliege Bird and D. W. Bird in ref. 22 ).
Decreasing trends in fire practices. We recorded trends in the proportion of the population in the study area practising the practice, the overall proportion of the study area affected by the practice annually and the return period for the practice in each affected landscape patch. Practices associated with two-thirds of our 29  fire-use purposes were significantly more likely to show decreasing trends in at least one of the variables in a two-tailed binomial test (Table 2). Except for pre-and post-harvest crop residue burning, for those fire purpose types for which we had insignificant results in the binomial tests, we had very small sample sizes (n < 15). Declining trends in fire use have a historical precedent. From the fifteenth century, European colonialism, development of capitalist economies and intensification of agriculture directly and indirectly suppressed fire use associated with subsistence and smallholder livelihoods worldwide [23][24][25] . The LIFE database sheds light on contemporary drivers of changes in fire use. Subsistence-oriented fire-use practices in LIFE are significantly more likely to show decreasing trends, while market-oriented fire-use practices are more likely to show increasing trends in all three variables (Fig. 2). This suggests that economic pressures are a global driver of changing fire use.
Case studies in LIFE illustrate some of the mechanisms by which economic pressures are driving decreasing subsistence-oriented fire use. Sometimes, commercial land uses limit the land available to populations with subsistence-oriented livelihoods. In the Brazilian Cerrado, for instance, expansion of industrial-scale capitalized agriculture is driving Indigenous and traditional populations into smaller territories, limiting fire use associated with hunting and gathering 26 . Elsewhere, market-oriented activities are replacing subsistence-oriented activities within livelihoods, reducing subsistence-oriented fire use. For example, in Guinea-Bissau, structural adjustment policies in the late 1980s created new export markets for cashew, driving smallholders to clear land for cashew orchards in following decades 27 . Adoption of the cashew cash crop has driven declines in swidden and savanna fire use because farmers have less time and land available for more traditional land uses and they are wary of the risks of wildfire damage to their cash   crops. Deepening capitalist relations also see the replacement of subsistence activities with wage labour and purchased goods. For example, in Australia, many Aboriginal people now live and work in large settlements 'off-country' , decreasing their use of fire for hunting and gathering 28 . Notably, the substitution of market-oriented production or wage labour for subsistence activities is often only partial. For Maya people in southern Belize, for instance, the insecurity and seasonality of labour and cash crop markets mean that subsistence-oriented swidden agriculture remains important within diversified livelihoods 29 . Sometimes, diversification of livelihoods to include more market-oriented activities can allow more customary, subsistence-oriented practices to persist in the face of pressures such as land constraints 30 . State governance can be another direct or indirect driver of decreasing fire use. Over recent centuries, a narrative that fire use was incompatible with 'rational' land use was used as justification for policies aiming to eradicate anthropogenic fire in many countries, especially in European colonies [23][24][25] . Although many nation states have since adopted 'prescribed fire' (the controlled application of fire for specific land management objectives), this is usually solely deemed the prerogative of specialized agencies, while livelihood-oriented fire use is heavily regulated. For instance, in the Borana rangelands of Ethiopia, pastoral fire use has largely ceased after enforcement of a general ban on fire use 31 . Fire regulations may be specific to, or only see enforcement in, protected areas such as the Biligiri Rangaswamy Temple Tiger Reserve in India, where Soliga people have stopped customary fire use since it was made illegal 32 . Regulations may also impose restrictive conditionality upon fire use. In Laos, laws that severely limit the amount of land allocated to each household and restrict fallow periods to 5 years have made it difficult to maintain swidden agriculture 33 . Many countries have strong anti-fire regulations but it is important to note that these are not always enforced: it is often easy for fire users to remain anonymous; states may lack the resources for strong enforcement; and those state officials responsible for enforcement may collude with fire users 24,34,35 . Besides regulation, economic governance mechanisms are driving declines in fire use. In recent decades one such mechanism has been payments for ecosystem services (PES) schemes, where these have made payments conditional upon eliminating fire use. For instance, in the páramos of Ecuador, communities participating in the Socioparamo PES scheme are paid to cease burning and grazing in certain areas 36 . Elsewhere, such as in Mexico, agricultural subsidies from the government have only been available for permanent agriculture rather than swidden and grants have supported the adoption of agrochemicals and mechanical tillage, which replace fire use 37 .
Case studies in LIFE highlight potential implications of declines in or loss of subsistence-oriented fire use. Declining fire use can severely undermine the livelihoods of rural people. For example, for Borana pastoralists in Ethiopia, loss of fire use in rangelands has led to bush encroachment and forage scarcity 31 and prohibition of fire use has reduced household income from agriculture and forest produce for Soliga people in India 32 . Sometimes, as for some Pemón communities in Venezuela, declining fire use can lead to a loss of cultural identity and conflict between older generations who retain fire knowledge and younger generations who are losing this knowledge 38 . Notably, its association with cultural identity may also contribute to the persistence of fire use in the face of countervailing drivers, as with the swidden agriculture of Tagbanua people in the Philippines 39 .
Loss of human fire can reduce the biodiversity of fire-dependent ecosystems. In the Biligiri Rangaswamy Temple Tiger Reserve in India, loss of Soliga fire use has increased vegetation density to the point of suppressing the growth and recruitment of many tree and plant species 32 . Meanwhile, in Australia, hill kangaroos are more abundant in regions with fine-grained vegetation mosaics resulting from Martu hunting fires, than in regions dominated by lightning fire 40 . Loss of smallholder and subsistence-oriented fire use can also lead to increased wildfire risk where it leads to more homogenous fuel landscapes. Case studies in LIFE suggest that fire use associated with reducing fuel loads at the landscape scale or creating firebreaks is declining. In the Brazilian Cerrado, Indigenous reserves where Xavante fire use for hunting and gathering is retained suffer significantly less wildfires than reserves where fire use has been lost or has declined 26 . When fire use reduces in importance or is made illegal, local governance institutions and knowledge that ensure that fire use is controlled can also be lost, increasing the likelihood of anthropogenic wildfires. For example, anti-fire legislation in Ghana is leading to the loss of traditional village institutions that organized collective early dry season burning to protect trees of spiritual and economic importance 41 . In Belize, Maya farmers with less time for swidden agriculture due to engagement in wage labour, cash cropping and pasture development are more likely to practice agricultural burns alone than in shared labour groups and less likely to construct firebreaks before burning or remain onsite until a fire is extinguished 29 .
Increasing trends in fire practices. Economic pressures can also drive increases in fire use: while subsistence-oriented fire use is declining, market-oriented fire-use practices are increasing. New markets for certain products are driving increases in fire use associated with their production. For example, in Indonesia, a high-value market for turtle meat emerged in the 1990s, driving increases in fire use to hunt turtles 42 . New technologies or infrastructural development may also drive increasing fire use. On the Trans-Gangetic Plains, 'green revolution' technologies have enabled smallholders to adopt double cropping of rice-wheat, with the use of combine harvesters 43 . Under this cropping system there is limited time between harvest of rice and planting of wheat and a greater biomass of crop residue left in-field. This means that farmers are now more likely to burn crop residues than collect them for other uses.
Among specific fire-use purposes, only fire use to clear land for permanent agriculture and protest fires (arson) show increasing trends in all three variables (proportion of population involved, proportion of landscape affected and return period of fire use in affected landscape patches) ( Table 2). The increasing trend in protest fires is often linked to disputes over land, especially where commercial enterprises or protected areas restrict subsistence or smallholder land use. Its anonymity makes fire a well-documented 'weapon of the weak' 24,44 . Arson is commonly used by smallholder farmers in Indonesia, for example, to assert their rights to land where large landholders have obtained land concessions from government 45 .
Transitions to cash cropping in permanent agricultural systems tend to be driven by economic pressures, as with adoption of cashew as a cash crop by farmers in Guinea-Bissau 27 . In many cases, government policies also support transitions to permanent agriculture, like in Mexico, where it is promoted through agricultural subsidies 37 .
Where economic pressures and policies support market-oriented permanent agriculture this can also be an indirect driver of increasing fire use associated with crop residue burning. This has been the case with smallholder rice-wheat cropping on the Indo-Gangetic Plains 43 . Crop residue burning does not, however, show a significant increasing trend across all case studies. This may relate to recent Government regulations on crop residue burning in many countries. In China, for example, crop residue burning has been banned since 1997 46 .
While burning to clear land for swidden agriculture is significantly more likely to be decreasing in terms of the proportion of population practising it or the proportion of the landscape affected, it is more likely to be increasing in return period in affected landscape patches. This accords with observed global trends of declining area and declining fallow periods in swidden agriculture 47 . Shorter rotations and corresponding increases in fire return period are often linked to declining land availability per household. This can be related to expansion of commercial land uses or protected areas, or to population growth. It may also be that government policies or incentive programmes supporting permanent agriculture drive shortening fallow periods. In Brazil, participants in the Bolsa Floresta Program receive payments for 'zero deforestation' , where 'deforestation' refers to opening fields in mature forest areas 48 . Clearing plots in secondary forests under a certain age of regrowth is permitted, incentivizing farmers to practice short-fallow swidden.
Generally, where economic pressures are driving increasing fire use, this fire use is less likely to be sensitive to environmental cues, which may increase wildfire risk and lead to negative ecological impacts. For example, in Indonesia, repeated burning to open areas for market-oriented fishing is transforming peatland forests into open floodplains 42 . Positive feedbacks between fire, removal of tree cover and loss of peat are increasing the susceptibility of these areas to wildfires.

Discussion
In each source, we looked for information about 37 variables for each fire practice. Most sources only provided information for some of the variables of interest, so the coverage of LIFE is incomplete (Fig. 3). Most case studies, for example, do not provide quantitative data regarding the fire return periods or burned area of anthropogenic fires; such data are often sensitive and difficult to collect and most researchers studying fire use in the field use qualitative research methods. This presents limitations for use of the database to improve global fire models. Geographically, there are also data gaps, such as in eastern Europe and northern, western and central Asia (Fig. 1). There are also far more studies discussing fire use than fire control and mitigation practices. These are areas that future case study research might address. Future research might also compare the sources in the LIFE database with research and historical texts pre-dating 1995, to look at longer term trends in fire practices.
Few sources we reviewed included fire users as authors or used participatory research methodologies and those that did were commonly for cases in the Global North. Given that these fire practices have often been misrepresented as irrational and environmentally destructive, future case study research should aim to facilitate cultural affirmation for fire users. This might be done by including Indigenous language quotes in publications 49 , facilitating deliberation within and between communities 50 or including fire users as authors and editors of publications 51 . Centring the voices of fire users would better demonstrate the ways in which fire use contributes to cultural identity and how Indigenous and local communities' cosmologies shape anthropogenic fire.
Despite the data gaps, there is potential to combine other forms of fire data with the quantitative and qualitative data from LIFE and future case studies, to improve our understanding of the implications of changes to smallholder and subsistence livelihood-oriented fire practices. Local or regional studies provide a precedent here. For example, in Mexico, social research on contemporary fire use has been used to help interpret the human fire signal in a dendrochronological record 52 and, in Indonesia, local case study data have been combined with fire remote sensing data to study how local fire-use practices shape the regional fire regime 45 . So far, case study data on fire use have not been combined with these other types of data at the global scale, nor have they been used to inform global fire models 13 . Our work to collate case study data at the global scale opens future possibilities to analyse how fire regimes and ecologies are shaped by fire use directed towards different livelihood outcomes.
Since the 1990s, there have been some efforts to design fire management programmes based on traditional fire knowledge, in recognition of the ways in which subsistence-oriented fire use has shaped ecosystems or contributed to wildfire risk reduction. Perhaps the most advanced example of this is in Australia, where Aboriginal fire practices have informed the management of protected areas and Aboriginal rangers are now being funded to conduct fire management via government-accredited methods for counting carbon credits 53 . Importantly, such programmes often assume that standardized fire management can substitute for the contingent, livelihood-oriented practices of fire users but this does not necessarily have the same ecological outcomes, nor lead to meaningful recognition of local fire users 24,54,55 . Future programmes and policies must make space for controlled fire use as it is already practised within local livelihoods, as is currently being done in Venezuela 50 . This will only be possible if the drivers of livelihoods changes are explicitly considered. This includes the social and environmental conditions under which economic pressures or state governance drive loss of fire use or less controlled fire use and the cultural and economic factors that drive the persistence of fire use in some places. The case studies in LIFE will be an important resource for further systematic analyses of these drivers.

Methods
A systematic literature review was used to find sources for inclusion in the LIFE database. Below we outline our search methodology, database structure and methods for the analysis of trends in fire practices presented in this article.
Data sources. Published articles and books and 'grey literature' in the form of dissertations and unpublished reports, including non-English language sources, were included in LIFE. We were unable to include sources where we could not access them either in online repositories, by contacting the authors or in libraries at the time of the research. The following criteria were used for inclusion of a source in LIFE: (1) is written or published during or after the year 1995, with relevant data collection carried out during or after the year 1990; (2) has a subnational scale of analysis; (3) is based on original empirical research (sources based on secondary data or review articles were not included); (4) discusses fire use or mitigation practices by smallholders or households with non-agricultural subsistence-oriented livelihoods, as practiced at the time of the research or practised within the living memory of research participants. We defined 'smallholders' as households for which agriculture is the principal source of livelihood and which rely predominantly on family labour 56 . Such households usually produce at least a proportion of their crops and/or livestock for subsistence but may also produce a substantial amount for the market. We did not consider farm size in our definition, given that the amount of land required to make agriculture a viable livelihood varies significantly between countries and different governments vary significantly in their legal definitions of the farm area considered a 'smallholding' 57 .
The systematic literature review was initially conducted using the Clarivate Analytics Web of Science Core Collection. We applied a search string to the titles, keywords and abstracts of all sources in the collection (Supplementary Methods). We grouped sources where they relate to research by the same author(s) in the same case study location(s) over a continuous period. Where a source or group of sources provided data separately for case studies in multiple locations, these were recorded as separate cases. Case studies in LIFE vary greatly in study area, from studies focused on single settlements to regional studies. For all cases, a single latitude and longitude point is provided to indicate the approximate centre of the study area and each is coded into one of four categories to indicate the broad scale of the study area. When determining what constituted separate fire practices for inclusion in LIFE, we used the criteria by which the author(s) of the source(s) distinguished between practices. In most cases, authors distinguished fire practices on the basis of their purposes (for example, 'burning to clear land for swidden agriculture' or 'burning to drive game for hunting'). Sometimes authors made it clear that multiple fire-use purposes related to the same ignitions on the landscape, in which case these were listed together as a single practice. In some cases, authors distinguished practices on the basis of the seasons or locations in which they take place (for example, 'fires set in the early dry season').
Coding of practices as subsistence oriented or market oriented. We recorded the orientation of a fire practice vis-à-vis subsistence and markets, where a source provided this information and where applicable (some practices, such as burning to protect settlements from wildfires, are not associated with subsistence or markets). Our analysis is not intended to suggest that most livelihoods fall neatly into either category. Households often combine subsistence-and market-oriented activities and the relative importance of these activities may vary year upon year. Yet, burning strategies may differ in association with subsistence-oriented versus market-oriented activities. We coded a practice as subsistence-oriented if it was associated with growing, hunting or gathering products for household consumption, with possible marketing of surpluses. Where a practice was associated with growing, hunting or gathering products predominantly for sale, we coded it as market oriented. Sometimes burning strategies do not differ between subsistence-and market-oriented activities. Where the same fire practice was associated with both subsistence-and market-oriented activities, we coded it as both. An example of this might be if staple and cash crops were grown in the same fields prepared using fire or if households hunting only for subsistence used fire in similar ways to households in the same village hunting mostly for local markets. We only applied both codes to the same practice in this way where a source described no differences in the practice whether it was subsistence-or market-oriented.
Analysis of trends in fire practices. For each fire practice, we recorded whether it was increasing or decreasing in each of three variables: (1) the proportion of the population in the study area practising the practice; (2) the overall proportion of the study area affected by the practice annually; (3) the return period for the practice in each patch of the landscape. Many sources did not provide quantitative data for these variables, so, where possible, we made judgements based on the qualitative information available. For example, if a source noted that expansion of intensive cropland in the study area had taken place at the expense of savanna areas in which subsistence hunting took place, we would record fire use for hunting as taking place over a decreasing proportion of the study area. Or, for example, if a source reported that younger generations were no longer learning to use fire for hunting, we would report fire use for hunting as taking place among a decreasing proportion of the population. There was sufficient information to record the trend in at least one of these three variables for 40% of the fire practices in LIFE (Fig. 3). We also noted where practices had been lost but were remembered by research participants or where practices were being (re)introduced on the basis of traditional knowledge.
We then ran a two-tailed exact binomial test for each of the three variables for all subsistence-oriented practices, all market-oriented practices and all fire-use practices categorized under each of our 29 lower-tier fire-use purpose types (Table 1). Practices coded as both subsistence-and market-oriented were counted twice in the analysis. Here, we tested whether the proportion of practices in each of these categories with a decreasing trend, or that had been lost, differed from what we would expect by chance alone (0.5). We treated results as significantly different from the null hypothesis with a P value of <0.05. In running this test, we assumed that the case studies in LIFE were an unbiased sample of all smallholder and subsistence-oriented fire practices, that is, that the authors of the studies in LIFE were not more likely to focus on practices that were increasing rather than decreasing or vice versa. We believe that this assumption holds, given that LIFE draws on studies from a wide range of academic disciplines and with a diversity of primary research topics (many studies discussed fire practices tangentially to other phenomena).

Data availability
The full LIFE database and accompanying metadata are available on figshare at https://doi.org/10.17637/rh.c.5469993.

Code availability
The R computer code supporting the analysis presented in this study and a file containing the subset of the LIFE database analysed using the code are available on figshare at https://doi.org/10.17637/rh.c.5469993.