Timber is a globally important natural resource used for construction materials, paper and energy. In 2015, the forestry sector contributed over US$1.5 trillion to national economies¹, and in 2020 at least one-third of all forest globally was being used for timber production². Timber demand is expected to almost triple by 2050³, via greater human demand and population growth, increasing urbanization, and as net-zero climate targets promote the replacement of carbon-intensive building materials, such as concrete and steel, with wood⁴. Ensuring that timber supply can meet future demand is therefore a key challenge in the twenty-first century.

Wildfires are a natural ecological process in many ecosystems. However, forest wildfires and associated forest loss have been increasing throughout the twenty-first century^,,7, with >110 million hectares (Mha) of forest lost to wildfires between 2001 and 2019⁷. Fire season length and fire extent are expected to increase significantly by 2100 because of climate change^8,9, placing forests under increasing threat of high-severity burning. Given the long-term nature of timber production, typically on 40–100+ year cutting cycles^10,11, future crops of timber trees will face a very different climate as they mature towards harvest.

We currently lack a robust understanding of how wildfires have influenced global timber production. High-severity, stand-replacing wildfires represent a large threat to timber stocks within forests managed for wood production¹², and the increasing frequency of such wildfires in the twenty-first century is a serious concern. Understanding where fire-induced losses of timber-producing forests occur and at what rate is therefore of critical importance in initiating efforts to ensure that timber production, and associated trade and financial investment, can be maintained in a more climatically hostile future. Here we combine spatial data detailing the global extent of forestry practices^6,13 with annual layers of forest loss due to wildfire⁷ to conduct a global assessment of the threat of wildfire to timber production. We answer the following key questions: (1) how much timber-producing forest is being lost to stand-replacing wildfires globally, and where do these losses occur? and (2) what are the temporal trends in annual burned area of timber-producing forests since the turn of the century, at global, regional and national scales?

To understand where timber-producing forests are burning, we used two datasets that map global logging activity from Lesiv et al.¹³ and Curtis et al.⁶, focusing primarily on clearcut logging of native forest and timber plantations. We overlaid these logging layers with global data on stand-replacing wildfires⁷, where fire severity was high enough to cause a detectable loss of tree cover, and thus timber (see Methods), and present the results from both layers together.

Since 2001, between 18.5 (s.e.m. 16.5–20.7) and 24.7 Mha (21.8–27.5) of timber-producing forest—an area roughly the size of Great Britain—has been lost due to wildfires (Fig. 1). This equates to between 1.0% (0.9–1.1%) and 1.7% (1.5–1.9%) of global forestry land burned since the turn of the century. Across both logging layers, fire-induced loss of timber-producing forest was particularly high in the northwestern USA and Canada, northeastern Russia, southeastern Australia and Brazil (Fig. 1a,b). Central and northern Europe experienced limited losses, as did parts of southern Asia and sub-Saharan Africa (see Extended Data Fig. 1 for a more spatially detailed map).

a–f, Hotspots of severe burning in timber-producing forests (a,b), total area (c,d) and proportion (e,f) of forestry land severely burned nationally between 2001 and 2021, using the map of global forest management (a,c,e) by Lesiv et al.¹³ and the map of forest loss due to forestry (b,d,f) by Curtis et al.⁶. In a and b, areas of warmer red represent increasing burn, blue represents areas where logging occurs but wildfires did not or were limited (<1,000 ha per cell) and grey represents areas where logging is not prevalent. The western USA and Canada, northeastern Russia, southeastern Australia and Brazil suffered particularly high losses of timber-producing forest to wildfire, while much of Central and northern Europe, and parts of South Asia and sub-Saharan Africa, experienced limited wildfire-induced losses.

Full size image

At the national scale, the three countries with the largest absolute wildfire-induced losses of timber-producing forest were Russia, the USA and Canada, where between 2.9 (2.6–3.4) Mha and 6.5 (5.8–7.2) Mha, 3.8 (3.4–4.1) Mha and 4.3 (3.9–4.8) Mha, and 2.3 (2.2–2.5) Mha and 3.9 (3.7–4.1) Mha were lost, respectively (Fig. 1c,d). This accounts for 1.1% (0.9–1.2%) to 2.6% (2.3–2.9%), 1.6% (1.4–1.7%) to 2.1% (1.9–2.3%) and 1.8% (1.7–2.0%) to 2.8% (2.7–3.0%) of their total national timber-producing forest, respectively. Portugal and Australia were characterized by the highest percentage of their forestry land lost, with 12.5% (11.1–14.0%) to 13.6% (12.1–15.2%) and 6.2% (5.9–6.5) to 10.1% (9.4–10.6%) burning, respectively (Fig. 1e,f). Such high losses of timber stock will leave a shortfall in internal timber supply, threatening timber-related jobs and probably forcing increased reliance on imports or shifts to faster wood production through plantations¹⁴.

While reliable burn data are available since only the turn of the century, timber rotation times across the globe (and especially in much of the Northern Hemisphere) are often far longer than 21 years, indicating even greater timber losses across a whole rotation cycle. For example, rotation times in the boreal forests of Russia and Canada can be ~100 years¹¹, meaning fire-induced losses of timber forests could reach 5.1–12.3% and 8.7–13.5%, respectively, across one rotation cycle under current burn conditions. Similarly, losses in the areas of the USA that employ rotations of 80 years (for example, the northwest¹⁵) could reach 6–8% across a harvest cycle.

Most logging globally entails clearcutting of native forest and timber plantations², which is captured well by both logging layers used (see Extended Data Fig. 2 for a spatial comparison of the two). Additionally, Lesiv et al.¹³ attempted to include selectively logged forest within their forest management map. Using this layer, we identified large areas of burned timber-producing forest across the tropics (~8.1 Mha; 7.0–9.1 Mha), particularly in Latin America (~6.8 Mha; 6.0–7.5 Mha). However, selective logging is far more difficult to detect via satellites than clearcutting or plantation forestry¹⁶, so these results should be interpreted with caution.

While we find a concerningly large area of timber-producing forest has already been lost to wildfire this century (18.5–24.7 Mha), there are limitations in the available forest management data that will affect these results. Curtis et al.⁶ had levels of high accuracy for mapped forestry areas (users: 87%; producers: 91%), but sample-based estimates suggest they slightly overestimated total forestry area. Conversely, the accuracy of the timber-producing forest classes mapped by Lesiv et al.¹³ were lower (users: 58–71%; producers: 35–65%), and sample-based estimates conducted by the authors suggest that forestry areas were underestimated. A third product (Schulze et al.¹⁷) also maps patterns of forest use globally, but is based on Food and Agriculture Organization of the United Nations (FAO) data, where national self-reporting is inconsistent and forest-use classes do not allow for accurate spatial classification of all timber-producing forest. Nevertheless, global hotspots of timber-producing forest loss through wildfire identified using Lesiv et al.¹³ and Curtis et al.⁶ remain largely the same when using Schulze et al.¹⁷ (Supplementary Information and Extended Data Fig. 3).

Globally, the annual area of timber-producing forest lost to stand-replacing wildfires showed an increasing trend between 2001 and 2021 using both Lesiv et al.¹³ (Mann–Kendall test, P = 0.0008, Sen’s slope annual trend size = +68,400 ha yr⁻¹) and Curtis et al.⁶ (P = 0.02, annual trend size = +37,800 ha yr⁻¹) forest management data, with strong correlation between the two layers in annual burned area (r = 0.83). In the past six years, mean annual loss from severe fire was 1.3–2.5 Mha, which is 2–4 times greater than in the previous 15 years (Fig. 2; see Extended Data Fig. 4 for raw data results). This increasing trend post-2015 is much sharper when using the Lesiv et al.¹³ forest management dataset, which is probably due to a combination of widespread fires in Latin America after a strong El Ñino in 2015–2016 that were not identified in the Curtis et al.⁶ layer, and possible under-reporting of burned timber production forest prior to 2014 in the Lesiv et al.¹³ layer (Supplementary Information).

a–f, Using the map of global forest management (red) by Lesiv et al.¹³ and the map of forest loss due to forestry (blue) by Curtis et al.⁶, split by global region: Global (a), North America (b), Latin America (c), Eurasia (d), Australasia (e) and Africa (f). Significant increasing trends in annual area of timber-producing forest burned are present globally and for all regions except Eurasia. Lines represent three-year rolling average, shaded areas represent three-year rolling average ± 1 s.e.m. Annual strength of trend from Sen’s slope analysis and P value from two-sided Mann–Kendall test are as follows: Lesiv et al.¹³ (Global: +68,000 ha yr⁻¹, P = 0.0008; North America: +18,500 ha yr⁻¹, P = 0.05; Latin America: +21,500 ha yr⁻¹, P = 6.8 × 10⁻⁶; Eurasia: no trend, P = 0.22; Australasia: +7000 ha yr⁻¹, P = 0.0003; Africa: +3500 ha yr⁻¹, P = 2.9 × 10⁻⁶); Curtis et al.⁶ (Global: +38,000 ha yr⁻¹, P = 0.002; North America: +18,000 ha yr⁻¹, P = 0.002; Latin America: +1700 ha yr⁻¹, P = 0.003; Eurasia: no trend, P = 0.83; Australasia: +2700 ha yr⁻¹, P = 0.001; Africa: +450 ha yr⁻¹, P = 1.2 × 10⁻⁵).

Full size image

Changes to the forest-loss detection model employed by Global Forest Watch¹⁸ that underpin the fire data used⁷ have improved forest-loss detection in recent years, but we do not believe this impacts our observed increasing trends. First, the fire data our study employs⁷ show no post-2015 spike in forest loss, which has been previously implicated as a clear signal of temporal inconsistencies in loss detection^19,20. Second, the fire data uses v1.7 of the forest-loss data, where loss detection algorithms are consistent between 2011 and 2020. We find no sudden jump in the years after 2010, and while the introduction of the Landsat 8 Operational Land Imager imagery in 2013 may have improved detection rates, we still find increasing trends in annual forestry burned area when considering only the years 2013–2021 (Lesiv et al.¹³: P = 0.029, annual trend size = +237,000 ha yr⁻¹; Curtis et al.⁶: P = 0.076, annual trend size = +85,000 ha yr⁻¹). Finally, Tyukavina et al.⁷ tested their fire data against another dataset that applied the newest forest-loss detection algorithms consistently across the whole period back to 2001, and found both methods demonstrated the same temporal trends⁷. This suggests algorithm inconsistencies between the 2001–2010 and 2011–2021 periods has little impact on observed trends in annual forest loss due to wildfires in the twenty-first century.

We found an increasing trend in annual wildfire-induced loss of timber-producing forest across all regions globally (Mann–Kendall test, P ≤ 0.05), except for Eurasia, which exhibited no trend (Fig. 2; see Extended Data Fig. 5 for free-scaled bar chart of regional annual burn). North America and Eurasia, the two largest timber-producing regions globally²¹, both exhibited the largest total losses across the period, between 6.6 (6.1–7.2) Mha and 7.6 (7.1–8.3) Mha, and 5.8 (4.7–7.1) Mha and 8.6 (7.4–9.9) Mha, respectively. Europe has suffered several major heatwaves and droughts across our study period (for example, in 2003, 2010 and 2017²²), and has experienced strong increasing trends in heatwave frequency and intensity²³. Nevertheless, this is not reflected in Eurasia’s annual fire-induced loss of timber-producing forest, which remained stable across the period (Fig. 2). Similarly, droughts and heatwaves have become prominent in North America (particularly the extreme heatwave of 2021²⁴), and burned area in the western USA and Canada has been steadily increasing across recent decades¹². Unlike Eurasia, annual wildfire-induced losses of timber-producing forest in North America are increasing and were particularly high in the years 2016–2021, equating to a two- to fourfold increase relative to 2001–2015.

Latin America exhibited losses of between 0.8 (0.7–1.0) Mha and 7.9 (6.9–8.7) Mha, with a sharp increase in 2016–2017 after a strong El Ñino event in 2015–2016²⁵. Australia and Oceania lost between 1.3 (1.2–1.4) Mha and 3.6 (3.4–3.7) Mha, the bulk of which occurred in 2019–2020, due to unprecedented wildfires across Australia²⁶, following years of drought²⁷. Africa was characterized by a significant increasing trend in annual timber-producing forest lost due to fire, but total losses remained low compared with other regions at 0.1–0.8 Mha.

Increasing burning trends at global and regional scales suggest that the threat of wildfires to timber production will be exacerbated under future climate change. Wildfires are expected to become more frequent and severe across many regions¹², with regional models supporting increased fire activity in important timber-producing areas such as the western USA²⁸, Canada²⁹, boreal Russia³⁰, the Amazon³¹ and Australia³².

Three of the five largest timber-producing nations— the USA, Canada and Brazil, who together accounted for 33% of industrial roundwood production in 2021²¹— exhibited increasing trends in their annual area of timber-producing forest burned in at least one mapped product (Fig. 3). Increasing trends were also evident in many Latin American countries, Portugal, Italy, Ukraine, Australia, New Zealand, South Africa, Ethiopia, Kenya and Vietnam (Fig. 3).

a,b, Using the map of global forest management (a) by Lesiv et al.¹³ and the map of forest loss due to forestry (b) by Curtis et al.⁶. Countries in orange–red demonstrated significant increasing trends in annual loss to wildfire (two-sided Mann–Kendall test, P < 0.05), with the scale representing the median slope of the trend across the period (in hectares burned per year). Countries in blue showed significant decreasing trends (two-sided Mann–Kendall test, P < 0.05, up to a maximum slope of −650 ha yr⁻¹ in China). Dark grey represents countries with no significant trend in annual burned area across the time period, while countries shown in white have either limited mapped logging activity or forest-loss-inducing wildfires. Strong increasing trends are present in the USA, Brazil and Canada (using Curtis et al.⁶), as well as Australia and many Latin American countries.

Full size image

There was some variation between mapped products in national level patterns. Canada showed a significant increasing trend using Curtis et al.⁶, but no change over time with Lesiv et al.¹³. This result was possibly owing to Curtis et al.⁶ mapping more logging activity in the fire-prone areas of western Canada³³. In addition, many countries in sub-Saharan Africa and Latin America revealed an increasing burn trend using Lesiv et al.¹³ but not Curtis et al.⁶, probably because Curtis et al. mapped limited forestry activity in these areas due to the dominance of forest loss through shifting agriculture and commodity production⁶. Only a small number of countries, including Japan and the UK, were characterized by a decreasing trend across our study duration (Fig. 3a,b). In combination, therefore, countries showing a significant increasing trend in annual losses of timber-producing forest through stand-replacing wildfires together account for ~43–50% of global industrial roundwood production, whereas those that showed a decreasing trend accounted for only ~2–6%²¹.

Stand-replacing wildfires have caused major losses of forest used for timber production, with an increasing trend in annual burned area between 2001 and 2021. Based on FAO data²¹, these findings point to an estimated wildfire-induced loss of ~393–667 million m³ of industrial roundwood timber across the period, which if exported at the 2021 global mean export price of US$115 per m³, would be worth ~US$45–77 billion. Some burned timber may be harvested in post-fire ‘salvage’ logging operations. However, such wood is often of low quality, while salvage logging has strong negative environmental impacts^34,35 and can increase the likelihood of future forest disturbances³⁶.

Greater frequency and severity of wildfires under climate change¹² indicates that threats to global wood production will increase. Overlapping current timber production areas^6,13 with future fire prediction maps⁸ suggests that ~29–62% of current production forest will experience an increase in fire-prone years, and ~44–80% will be subject to longer fire seasons by 2100. To ensure increasing timber demand is met³, timber producers need to minimize the risk of fire to future timber crops today, via better selection of crop species, improved spatial planning, and adoption of new technologies for fire detection and suppression.

Plantations produce ~33% of global industrial roundwood³⁷ on ~3% of forest area² and this will probably increase as fast-growing timber is needed to meet rising demand³, and regions move away from clearcut harvesting of natural forests³⁸. Expanding production through plantations will decrease the risk of fire-induced timber losses, as harvest rotations are typically far shorter, providing a greater chance of producing a crop before it is lost to fire. Where possible, fast-growing timber species should be introduced in plantations, sparing large tracts of old-growth forest elsewhere to support biodiversity and carbon stocks^39,40. However, timber plantations are highly flammable⁴¹. Shifting production in fire-prone regions from monocultures of highly flammable timber species (for example Pinus radiata⁴¹, Eucalyptus globulus⁴²) towards heterogenous mosaics of less-flammable species of varying ages will be vital in reducing timber losses through burning⁴³.

Improved spatial planning of forestry activities can reduce wildfire risk to timber. At the global scale, establishment of plantations in areas of high wildfire risk should be avoided in favour of regions less likely to burn. One concerning example is the recent rapid expansion of highly flammable Eucalyptus plantations in the Brazilian Cerrado⁴⁴, despite globally high wildfire prevalence in this region⁵. At a landscape scale, establishment of plantations on slopes, where fires burn more severely and spread rapidly, should be prevented⁴⁵. Spatial contagion in fire can be reduced by avoiding large areas of contiguous plantations, with more ‘fragmented’ timber plantations interspersed with less flammable land-use types (for example, grazing lands) and ‘green’ firebreaks³⁸.

The timber industry must also adopt emerging technologies to reduce the impact of wildfires where they occur. Improved modelling techniques allow for more accurate spatial prediction of fire ignitions⁴⁶, while adoption of infrared-sensing drones and on-the-ground camera networks could allow for swift, 24-hour detections³⁸. Once detected, autonomous ‘water gliders’ could follow GPS signals to carry water or flame retardants to the ignition source and extinguish the fire before it has time to expand³⁸. In areas that do burn, locally effective post-fire management practices will be required, as salvage logging and replanting post-fire can lead to more severe burning in the future^43,47.

Failure to adopt approaches such as those outlined above could lead to massive future timber losses, with falling timber supply driving higher prices. Concerningly, higher timber prices will increase the opportunity costs of conservation in tropical forests, especially in the Amazon and Congo Basin, making intensive selective logging economically attractive. Such economic feedbacks could make carbon-market payments prohibitively expensive⁴⁸, undermining global climate and biodiversity goals. We must urgently tackle the emerging timber production crisis to meet humanity’s needs and prevent severe unintended environmental feedbacks.

To determine how much timber has already been lost to wildfires in recent years, we overlaid spatial data detailing the global extent of forestry practices^6,13 with annual layers of forest loss due to wildfire from 2001 to 2021⁷. For a summary of the different datasets used within the study, see Extended Data Table 1.

Logging layers

To understand where timber production occurs globally, we used the best available datasets that attempt to map global timber harvesting activity:

Global map of forest management

Lesiv et al.¹³ used GeoWiki and classification algorithms to produce a global map of forest management type at 100 m spatial resolution. To make the classification, GeoWiki users and experts were asked to classify satellite images into different forest management types at 226,000 different points, before the outputs were then used to train classification algorithms to classify all areas of forest, using PROBA-V satellite imagery from the period 2014–2016. Multiple forest management types were included within the map, but to represent possible timber-producing forest, we included only the following three management classes: (1) naturally regenerating forest with signs of management, for example, logging, clear cuts and so on; (2) planted forests (rotation >15 years); and (3) plantation forests (rotation ≤15 years), covering a total of >2.4 billion ha of logged forest and timber plantations (see ref. ¹³ for full definitions). We also performed precautionary re-analyses, whereby we excluded both ‘planted forests’ and ‘plantation forests’ from the analysis. Results showed broadly the same patterns and are displayed in Extended Data Table 2.

Global map of forestry as the dominant driver of forest cover loss

This map from Curtis et al.⁶ used satellite data and machine learning to classify the most dominant driver of forest loss between 2001 and 2019 at a spatial resolution of 10 km. Drivers were attributed as one of the following: urbanization, commodity-driven deforestation (for example, palm oil, soy), shifting agriculture, wildfire or forestry. We included all areas mapped as forestry in our analysis using this layer, which covered a total of ~1.6 billion ha of logged forest. Owing to the coarse spatial resolution of this layer, we applied a mask of forest cover in the year 2000⁴⁹, including only areas that had >10% tree cover (following the FAO definition of forest and the threshold used in Lesiv et al.¹³). This left a mapped area of 1.08 billion ha of clearcut logged forest and plantation area.

Global map of forest use

This third dataset, from Schulze et al.¹⁷, was considered but not included in the main analysis. This map used national and sub-national forest management data and classification models trained with predictor variables to identify types of forest use globally at a 1 km resolution for the year 2000. However, the FAO’s Forest Resources Assessment data used in this study are inconsistent between countries (which self-report their national statistics to the FAO) and the prevalence of the ‘mixed use’ forest class does not allow for accurate classification of all timber-producing forest globally. In addition, we identified multiple regions where timber production classification is inaccurate and large areas of logging concessions and plantations are not included within the map (for example, southeastern Australia⁴⁹). The results from our analysis using this layer should thus be interpreted with caution, and are not included within the main text, but can be found in Extended Data Fig. 3. Owing to the coarse resolution of the map, we again applied a mask of forest cover for the year 2000¹⁸, including only areas that had >10% tree cover (following the FAO definition of forest). Forest-use patterns are grouped into four different classes: production, mixed use, other and unknown. The FAO defined ‘production’ forest as “forest where the management objective is production of wood, fibre, bio-energy and/or non-wood forest products”, while ‘mixed use’ is defined as “forest where the management objective is a combination of several purposes and where none of them is significantly more important than the other”. For the analysis, we assumed ‘production’ forest to represent timber-producing forest. We also included ‘multiple use’ forest in our analysis. While ‘multiple use’ forest may not necessarily be used specifically for timber production, we included this management class as local forest definitions meant that many large timber-producing countries (for example, Canada, Australia, Germany) had limited to no mapped ‘production’ forest.

Fire layer

As our primary interest was in large, stand-replacing fires that deplete an area of forest of its timber stock, we used recent map layers describing stand-replacing wildfires⁷, where fire severity was high enough to cause forest loss (defined as “the removal of woody vegetation exceeding 5 m in height”⁷). This layer is an extension of the work by Curtis et al.⁶ and uses Hansen et al.¹⁸ Landsat-based forest loss data and machine learning to identify whether wildfire or an alternative driver caused a forest loss event in the years 2001–2021, at a 30 m resolution. Classification models were trained on visually collected data, with separate models developed for five global regions (North America, Latin America, Africa, northern Eurasia, and combined South and Southeast Asia, Australia and Oceania). We elected not to implement the widely used MODIS Burned Area dataset⁵⁰ as our measure of fire activity, as these data map burned area without differentiating between stand-replacing fires and low-severity burns. Across many regions, low-severity burns may not cause enough damage to trees to render the majority of their timber unusable, so we focused on high-severity fires that would do so. In producing their map of wildfire-induced forest loss, Tyukavina et al.⁷ undertook a sample-based area estimate and ensured the final map matched the sample-based area estimate for all global regions (except Africa) as well as providing maps containing the estimate ± s.e.m. This allows for estimation of burned area in timber-producing forests that matches sample-based area estimates, as well as inclusion of uncertainty through estimates of burn area ± s.e.m. for all regions except Africa.

Economic cost analyses of high-severity stand-replacing wildfires on timber-producing forests reveal extreme loss of timber stock and economic value. For example, severe wildfires in the Rocky Mountains in 2000 resulted in the burning of ~7 million m³ of timber, of which <5% was salvaged⁵¹. Similarly, two wildfire events in southeastern Australia (2003 and 2006/2007) resulted in >AU$2 billion worth of lost timber, with salvage operations occurring across <4% of burned native forest⁵². In Russia, assessment of forests post-wildfire found that crown fires had dominated the burning, often resulting in total tree mortality⁵³. Given the substantial environmental damage caused by salvage logging^34,35, and associated price depression when poor-quality salvaged timber floods the market⁵¹, we assumed that the stand-replacing wildfires mapped in Tyukavina et al.⁷ result in near-total loss of harvestable timber stock and economic value.

Data analysis

We completed all spatial data operations and analysis using R version 4.2.1. We downloaded the global map of forest management from Lesiv et al.¹³ (https://doi.org/10.5281/zenodo.5879022), the global map of forest-loss by driver (Curtis et al.⁶) through the Global Forest Watch website (https://www.globalforestwatch.org/map/) and the global map of forest loss through wildfire from Tyukavina et al.⁷ (https://glad.umd.edu/dataset/Fire_GFL/). We downloaded the global map of tree cover in 2000¹⁸ through Google Earth Engine and reprojected to a 100 m resolution to match the highest resolution available for our logging data. We projected all spatial data in EPSG:4326 (WGS84) CRS and projected into ESRI:54009 where area calculations or intersections were computed. We used the following R packages: raster (v. 3.5.29)⁵⁴, sf (v. 1.0.8)⁵⁵, terra (v. 1.6.7)⁵⁶, rnaturalearth (v. 0.1.0)⁵⁷, Kendall (v. 2.2.1)⁵⁸ and trend (v. 1.1.5)⁵⁹.

We divided the world into five regions (North America, Latin America, Eurasia, Africa, and Southeast Asia and Australasia), following the regional models of fire-induced forest loss used by Tyukavina et al.⁷ For each region, we created a network of grid cells 0.25° in size for increased computational efficiency.

For each grid, we cropped the logging and fire raster data by the spatial extent of the grid using the terra package, and converted the resulting objects into simple features polygons using the sf package. For the Curtis et al.⁶ logging layer, we then applied a forest mask using global forest cover for 2000¹⁸, and retained only areas of timber-producing forest where tree cover was >10% (following the FAO definition of forest and the same masking threshold used by Lesiv et al.¹³ when creating their map). We aggregated the total area of timber-producing forest in each grid to estimate the total global area of forest being used for timber production in each logging layer (Lesiv et al.¹³: 2.4 billion ha; Curtis et al.⁶: 1.08 billion ha).

Using the sf package, we calculated the total timber-producing forest present in each grid cell (for each logging layer used), as well as the intersection of timber-producing forest with wildfire-induced forest loss in each year of the time series. We then recorded the spatial extent of forest-loss-inducing fires for each year in timber-producing forests within the grid. We repeated this process using different combinations of pixel certainty in the Tyukavina et al.⁷ dataset to also produce uncertainty estimates representing ± s.e.m.

Quantifying the spatial extent of wildfire-induced losses of timber-producing forest

We aggregated total burned area and total timber production area globally, regionally and nationally to create estimates of the global, regional and national area of timber-producing forest burned during our study period. For national level analysis, we used the R package rnaturalearth for country boundaries to determine the total area and proportion of timber production forest burned on a national scale. We calculated total area burned in square metres before conversion to hectares. For estimates of the proportion of timber-producing forest burned, we divided the total burned area of timber-producing forest across all years by the total area of timber-producing forest estimated from each logging dataset.

Temporal trends in annual burned area of timber-producing forests at various scales

We aggregated the total area of timber-producing forest burned by year of burn, to create an estimate of the annual burn area across the time period at global, regional and national scales (2001–2021). To estimate the trend in annual burn area, we used a Mann–Kendall test for monotonic trend in a time series from the Kendall package on the raw data and report these results. To estimate the size of the trend, we conducted a Sen’s slope analysis using the trend package.

1. The State of the World’s Forests 2022: Forest Pathways for Green Recovery and Building Inclusive, Resilient and Sustainable Economies (FAO, 2022); https://doi.org/10.4060/cb9360en

2. Global Forest Resources Assessment 2020: Main Report (FAO, 2020); https://doi.org/10.4060/ca9825en

3. Barua, S. K., Lehtonen, P. & Pahkasalo, T. Plantation vision: potentials, challenges and policy options for global industrial forest plantation development. Int. For. Rev. 16, 117–127 (2014).

   Google Scholar 

4. Mishra, A. et al. Land use change and carbon emissions of a transformation to timber cities. Nat. Commun. 13, 4889 (2022).

   Article  Google Scholar 

5. Andela, N. et al. A human-driven decline in global burned area. Science 356, 1356–1362 (2017).

   Article  Google Scholar 

6. Curtis, P. G., Slay, C. M., Harris, N. L., Tyukavina, A. & Hansen, M. C. Classifying drivers of global forest loss. Science 361, 1108–1111 (2018).

   Article  Google Scholar 

7. Tyukavina, A. et al. Global trends of forest loss due to fire from 2001 to 2019. Front. Remote Sens. 3, 825190 (2022).

   Article  Google Scholar 

8. Senande-Rivera, M., Insua-Costa, D. & Miguez-Macho, G. Spatial and temporal expansion of global wildland fire activity in response to climate change. Nat. Commun. 13, 1208 (2022).

   Article  Google Scholar 

9. Flannigan, M. et al. Global wildland fire season severity in the 21st century. For. Ecol. Manag. 294, 54–61 (2013).

   Article  Google Scholar 

10. Messier, C., Puettmann, K. J. & Coates, K. D. (eds) Managing Forests as Complex Adaptive Systems: Building Resilience to the Challenge of Global Change (Routledge, 2013).

11. Burton, P. J., Kneeshaw, D. D. & Coates, K. D. Managing forest harvesting to maintain old growth in boreal and sub-boreal forests. For. Chron. 75, 623–631 (1999).

    Article  Google Scholar 

12. Bowman, D. M. J. S. et al. Vegetation fires in the Anthropocene. Nat. Rev. Earth Environ. 1, 500–515 (2020).

    Article  Google Scholar 

13. Lesiv, M. et al. Global forest management data for 2015 at a 100 m resolution. Sci. Data. 9, 199 (2022).

    Article  Google Scholar 

14. Cary, G. J., Blanchard, W., Foster, C. N. & Lindenmayer, D. B. Effects of altered fire intervals on critical timber production and conservation values. Int. J. Wildland Fire 30, 322–328 (2021).

    Article  Google Scholar 

15. Hudiburg, T. et al. Carbon dynamics of Oregon and northern California forests and potential land-based carbon storage. Ecol. Appl. 19, 163–180 (2009).

    Article  Google Scholar 

16. Hethcoat, M. G. et al. A machine learning approach to map tropical selective logging. Remote Sens. Environ. 221, 569–582 (2019).

    Article  Google Scholar 

17. Schulze, K., Malek, Ž. & Verburg, P. H. Towards better mapping of forest management patterns: a global allocation approach. For. Ecol. Manag. 432, 776–785 (2019).

    Article  Google Scholar 

18. Hansen, M. C. et al. High-resolution global maps of 21st-century forest cover change. Science 342, 850–853 (2013).

    Article  Google Scholar 

19. Ceccherini, G. et al. Abrupt increase in harvested forest area over Europe after 2015. Nature 583, 72–77 (2020).

    Article  Google Scholar 

20. Palahí, M. et al. Concerns about reported harvests in European forests. Nature 592, E15–E17 (2021).

    Article  Google Scholar 

21. Forestry Production and Trade (FAO, accessed 21 December 2022); https://www.fao.org/faostat/en/#data/FO

22. Teuling, A. J. A hot future for European droughts. Nat. Clim. Change 8, 364–365 (2018).

    Article  Google Scholar 

23. Rousi, E., Kornhuber, K., Beobide-Arsuaga, G., Luo, F. & Coumou, D. Accelerated western European heatwave trends linked to more-persistent double jets over Eurasia. Nat. Commun. 13, 3851 (2022).

    Article  Google Scholar 

24. Thompson, V. et al. The 2021 western North America heat wave among the most extreme events ever recorded globally. Sci. Adv. 8, eabm6860 (2022).

    Article  Google Scholar 

25. Rifai, S. W., Li, S. & Malhi, Y. Coupling of El Niño events and long-term warming leads to pervasive climate extremes in the terrestrial tropics. Environ. Res. Lett. 14, 105002 (2019).

    Article  Google Scholar 

26. Boer, M. M., Resco De Dios, V. & Bradstock, R. A. Unprecedented burn area of Australian mega forest fires. Nat. Clim. Change 10, 171–172 (2020).

    Article  Google Scholar 

27. King, A. D., Pitman, A. J., Henley, B. J., Ukkola, A. M. & Brown, J. R. The role of climate variability in Australian drought. Nat. Clim. Change 10, 177–179 (2020).

    Article  Google Scholar 

28. Anderegg, W. R. L. et al. Future climate risks from stress, insects and fire across US forests. Ecol. Lett. 25, 1510–1520 (2022).

    Article  Google Scholar 

29. Flannigan, M. et al. Future area burned in Canada. Clim. Change 72, 1–16 (2005).

    Article  Google Scholar 

30. De Groot, W. J., Flannigan, M. D. & Cantin, A. S. Climate change impacts on future boreal fire regimes. For. Ecol. Manag. 294, 35–44 (2013).

    Article  Google Scholar 

31. Abatzoglou, J. T., Williams, A. P., Boschetti, L., Zubkova, M. & Kolden, C. A. Global patterns of interannual climate–fire relationships. Glob. Change Biol. 24, 5164–5175 (2018).

    Article  Google Scholar 

32. Canadell, J. G. et al. Multi-decadal increase of forest burned area in Australia is linked to climate change. Nat. Commun. 12, 6921 (2021).

    Article  Google Scholar 

33. Hanes, C. C. et al. Fire-regime changes in Canada over the last half century. Can. J. For. Res. 49, 256–269 (2019).

    Article  Google Scholar 

34. Thorn, S. et al. Impacts of salvage logging on biodiversity: a meta-analysis. J. Appl. Ecol. 55, 279–289 (2018).

    Article  Google Scholar 

35. Leverkus, A. B. et al. Salvage logging effects on regulating ecosystem services and fuel loads. Front. Ecol. Environ. 18, 391–400 (2020).

    Article  Google Scholar 

36. Leverkus, A. B. et al. Tamm review: Does salvage logging mitigate subsequent forest disturbances? For. Ecol. Manag. 481, 118721 (2021).

37. Jürgensen, C., Kollert, W. & Lebedys, A. Assessment of Industrial Roundwood Production from Planted Forests Planted Forests and Trees Working Paper Series No. 48 (FAO, 2014).

38. Lindenmayer, D. B., Yebra, M. & Cary, G. J. Perspectives: Better managing fire in flammable tree plantations. For. Ecol. Manag. 528, 120641 (2023).

    Article  Google Scholar 

39. Edwards, D. P. et al. Land-sharing versus land-sparing logging: reconciling timber extraction with biodiversity conservation. Glob. Change Biol. 20, 183–191 (2014).

    Article  Google Scholar 

40. Betts, M. G. et al. Producing wood at least cost to biodiversity: integrating Triad and sharing–sparing approaches to inform forest landscape management. Biol. Rev. 96, 1301–1317 (2021).

    Article  Google Scholar 

41. Gómez-González, S., Ojeda, F. & Fernandes, P. M. Portugal and Chile: longing for sustainable forestry while rising from the ashes. Environ. Sci. Policy 81, 104–107 (2018).

    Article  Google Scholar 

42. Guerrero, F. et al. Drivers of flammability of Eucalyptus globulus Labill leaves: terpenes, essential oils, and moisture content. Forests 13, 908 (2022).

    Article  Google Scholar 

43. Leverkus, A. B., Thorn, S., Lindenmayer, D. B. & Pausas, J. G. Tree planting goals must account for wildfires. Science 376, 588–589 (2022).

    Article  Google Scholar 

44. Galizia, L. F. D. C. & Rodrigues, M. Modeling the influence of eucalypt plantation on wildfire occurrence in the Brazilian savanna biome. Forests 10, 844 (2019).

    Article  Google Scholar 

45. Lindenmayer, D., Taylor, C. & Blanchard, W. Empirical analyses of the factors influencing fire severity in southeastern Australia. Ecosphere 12, e03721 (2021).

    Article  Google Scholar 

46. Gora, E. M. et al. A mechanistic and empirically supported lightning risk model for forest trees. J. Ecol. 108, 1956–1966 (2020).

    Article  Google Scholar 

47. Thompson, J. R., Spies, T. A. & Ganio, L. M. Reburn severity in managed and unmanaged vegetation in a large wildfire. Proc. Natl Acad. Sci. USA 104, 10743–10748 (2007).

    Article  Google Scholar 

48. Phelps, J., Carrasco, L. R., Webb, E. L., Koh, L. P. & Pascual, U. Agricultural intensification escalates future conservation costs. Proc. Natl Acad. Sci. USA 110, 7601–7606 (2013).

    Article  Google Scholar 

49. Lindenmayer, D. B. & Taylor, C. New spatial analyses of Australian wildfires highlight the need for new fire, resource, and conservation policies. Proc. Natl Acad. Sci. USA 117, 12481–12485 (2020).

    Article  Google Scholar 

50. Giglio, L., Boschetti, L., Roy, D. P., Humber, M. L. & Justice, C. O. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sens. Environ. 217, 72–85 (2018).

    Article  Google Scholar 

51. Prestemon, J. P., Wear, D. N., Stewart, F. J. & Holmes, T. P. Wildfire, timber salvage, and the economics of expediency. For. Policy Econ. 8, 312–322 (2006).

    Article  Google Scholar 

52. Stephenson, C., Handmer, J. & Betts, R. Estimating the economic, social and environmental impacts of wildfires in Australia. Environ. Hazards 12, 93–111 (2013).

    Article  Google Scholar 

53. Kukavskaya, E. A. et al. Influence of logging on the effects of wildfire in Siberia. Environ. Res. Lett. 8, 045034 (2013).

    Article  Google Scholar 

54. Hijmans, R. raster: Geographic Data Analysis and Modeling. R package version 3.5-29 https://CRAN.R-project.org/package=raster (2020).

55. Pebesma, E. Simple features for R: standardized support for spatial vector data. R J. 10, 439–446 (2018).

    Article  Google Scholar 

56. Hijmans, R. terra: Spatial Data Analysis. R package version 1.6-7 https://CRAN.R-project.org/package=terra (2020).

57. South, A. rnaturalearth: World Map Data from Natural Earth. R package version 0.1.0 https://CRAN.R-project.org/package=rnaturalearth (2020).

58. McLeod, A. Kendall: Kendall Rank Correlation and Mann-Kendall Trend Test. R package version 2.2.1 https://CRAN.R-project.org/package=Kendall (2022).

59. Pohlert, T. trend: Non-Parametric Trend Tests and Change-Point Detection. R package version 1.1.5 https://CRAN.R-project.org/package=trend (2023).

Download references

Authors and Affiliations

1. Ecology and Evolutionary Biology, School of Biosciences, University of Sheffield, Sheffield, UK

   Christopher G. Bousfield

2. Department of Plant Sciences and Conservation Research Institute, University of Cambridge, Cambridge, UK

   Christopher G. Bousfield & David P. Edwards

3. Fenner School of Environment and Society, The Australian National University, Canberra, Australian Capital Territory, Australia

   David. B. Lindenmayer

Contributions

C.G.B., D.B.L. and D.P.E. conceived the study and developed the methodology. C.G.B. performed the analysis and drafted the original paper. All authors contributed to reviewing and editing the paper.

Corresponding authors

Correspondence to Christopher G. Bousfield or David P. Edwards.

Competing interests

The authors declare no competing interests.

Peer review information

Nature Geoscience thanks Steffen Fritz, Dave van Wees and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Xujia Jiang, in collaboration with the Nature Geoscience team.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and Permissions