Substantial and increasing global losses of timber-producing forest due to wildfires

Timber is a major global resource, contributing over US$1.5 trillion to economies and with at least one-third of global forest used for timber production. Demand is projected to nearly triple by 2050 due to population growth, urbanization, and decarbonization strategies that substitute wood for more carbon-intensive materials. Concurrently, forest wildfires and fire-driven forest loss have increased, and climate change is projected to lengthen fire seasons and expand burned area. Given long timber rotation times (often 40–100+ years), future crops will mature under different and hotter climates, but a robust global understanding of wildfire impacts on timber-producing forests has been lacking. High-severity, stand-replacing fires pose a substantial threat to timber stocks in managed forests. This study combines global forestry extent datasets with annual wildfire-induced forest loss maps to quantify: (1) how much timber-producing forest has been lost to stand-replacing wildfires since 2001 and where; and (2) temporal trends in annual burned area at global, regional, and national scales.

Prior work shows increases in forest wildfire activity and >110 Mha of forest lost to wildfires between 2001 and 2019. Climate projections indicate substantial increases in fire season length and extent by 2100, heightening risk to forests. Timber rotations in boreal and temperate regions commonly span many decades, implying cumulative exposure to rising fire risk. Global mapping of forest management and drivers of forest loss from satellite data (for example, Lesiv et al. for management types, Curtis et al. for drivers of forest loss) provides a basis for assessing overlap with wildfire-induced losses, while selective logging in the tropics remains challenging to detect remotely. Existing datasets vary in accuracy and spatial resolution, and self-reported FAO-based products have inconsistencies, motivating careful cross-comparison and uncertainty consideration.

The study overlaid maps of global forestry activity with annual maps of stand-replacing wildfire-induced forest loss for 2001–2021. Forestry was mapped primarily using: (1) Lesiv et al. (2015, 100 m), classifying forest management types (naturally regenerating forests with evidence of management; planted forests >15-year rotations; and plantation forests ≤15-year rotations), totaling >2.4 billion ha; and (2) Curtis et al. (drivers of forest loss, 10 km, 2001–2019), where areas with forestry as the dominant loss driver were used (after masking to >10% tree cover in 2000) to represent ~1.08 billion ha of clearcut logging and plantations. A third dataset (Schulze et al., 1 km, FAO-based) was evaluated in supplementary analyses. Wildfire-induced forest loss employed Tyukavina et al. (30 m) that attributes forest loss events to wildfire versus other drivers, capturing stand-replacing fires (defined via Landsat-based forest loss of woody vegetation >5 m in height). The fire dataset provides sample-based area estimates and s.e.m. for most regions, enabling uncertainty propagation. Spatial analyses were conducted in R (raster, sf, terra), projecting data to WGS84 and ESRI:54009 for area computations. The globe was partitioned into 0.25° grid cells; within each cell, the area of timber-producing forest and its intersection with wildfire-induced loss were computed annually. Aggregations were performed globally, regionally (North America, Latin America, Eurasia, Australasia, Africa), and nationally (using rnaturalearth boundaries). Annual trends in burned area were tested using Mann–Kendall tests and quantified with Sen’s slope. Sensitivity analyses included alternative inclusion/exclusion of planted/plantation classes (Lesiv) and use of Schulze et al.

- Total losses 2001–2021: 18.5–24.7 Mha of timber-producing forest burned in stand-replacing wildfires, equivalent to ~1.0%–1.7% of global forestry land. Hotspots include the western USA and Canada, northeastern (Siberian) Russia, southeastern Australia, and Brazil; limited losses in Central/northern Europe, parts of South Asia, and sub-Saharan Africa. - National totals (range across datasets): Russia 2.9–6.5 Mha (1.1%–2.6% of national timber-producing forest), USA 3.8–4.3 Mha (1.6%–2.1%), Canada 2.3–3.9 Mha (1.8%–2.8%). Highest percentage burned: Portugal 12.5%–13.6%; Australia 6.2%–10.1%. - Rotation-scale implications: Given ~100-year rotations in boreal Russia and Canada, fire-induced losses across a rotation could reach 5.1%–12.3% (Russia) and 8.7%–13.5% (Canada) under current conditions; in US regions with ~80-year rotations, 6%–8% losses per rotation are possible. - Tropics (Lesiv dataset including selective logging): ~8.1 Mha (7.0–9.1) of burned timber-producing forest across the tropics, notably Latin America ~6.8 Mha (6.0–7.5), but with higher uncertainty due to detectability challenges. - Global annual trends: Significant increases in annual burned area 2001–2021 in both mapping products. Lesiv: P = 0.0008, Sen’s slope +68,400 ha yr−1; Curtis: P = 0.02, +37,800 ha yr−1; strong inter-product correlation in annual burned area (r = 0.83). In the last six years, mean annual loss was 1.3–2.5 Mha, 2–4× higher than 2001–2015. - Regional totals and trends: North America 6.6–7.6 Mha (increasing); Eurasia 5.8–8.6 Mha (no significant trend); Latin America 0.8–7.9 Mha with a sharp increase after 2015–2016 El Niño (increasing); Australasia 1.3–3.6 Mha with major losses in 2019–2020 (increasing); Africa 0.1–0.8 Mha (increasing but low totals). - National trend patterns: Significant increasing trends in annual burned timber area in key producers including the USA, Canada (in Curtis), Brazil, Australia, and many Latin American countries; decreases in a small set (e.g., Japan, UK). Countries with increasing trends represent ~43%–50% of global industrial roundwood production; those with decreasing trends represent ~2%–6%. - Economic implications: Estimated wildfire-induced loss of ~393–667 million m3 of industrial roundwood (2001–2021), valued at ~US$45–77 billion at 2021 average export prices. Salvage logging recovers little value and has significant ecological costs. - Future exposure: Overlap with projections suggests ~29%–62% of current production forests will face increases in fire-prone years, and 44%–80% will see longer fire seasons by 2100.

The study demonstrates that stand-replacing wildfires have already caused substantial losses in timber-producing forests globally and that annual burned area is increasing in most regions, aligning with expectations of heightened fire activity under climate change. These findings directly answer the posed questions by quantifying the magnitude and locations of losses and documenting robust upward temporal trends. The disproportionate impacts in major timber-producing regions (western USA/Canada, boreal Russia, Australia, Brazil) indicate growing risks to domestic supplies, trade balances, and timber-dependent economies. The results also highlight that even with conservative definitions focused on stand-replacing events, timber stock losses are large and likely underestimated in selectively logged tropical forests due to detection limits. Management implications are significant: reliance on long-rotation, flammable forest types will increasingly expose timber portfolios to fire risk. Strategic responses include shifting to faster rotations where appropriate, diversifying species and age classes, creating less flammable landscape mosaics, and improving spatial planning to avoid high-risk locations (e.g., steep slopes, highly fire-prone regions). Emerging detection and suppression technologies (e.g., predictive ignition models, infrared drones, camera networks, autonomous delivery of water/retardants) can reduce event sizes. However, salvage logging is not a panacea, given low quality returns and ecological drawbacks. Without proactive adaptation, escalating fire losses could raise timber prices, altering conservation economics and potentially incentivizing more intensive logging in tropical forests, undermining climate and biodiversity goals.

Stand-replacing wildfires have already removed 18.5–24.7 Mha of timber-producing forest since 2001, with annual burned area increasing globally and across most regions. Major timber nations show rising trends, and economic losses are substantial. Looking ahead, a large share of current production forests is projected to experience more fire-prone years and longer seasons by 2100, elevating risks to meeting growing timber demand. The paper’s key contribution is a first global quantification of wildfire-driven losses specifically within timber-producing forests, identification of geographic hotspots, and documentation of accelerating trends. To sustain supply, producers should: adopt diversified, less-flammable species and age structures; expand production in lower-risk areas; use spatial planning to reduce contagion; and deploy advanced detection and suppression technologies. Potential future research directions include improving global mapping of selective logging and forest uses, integrating stand-level fuel and flammability metrics into risk models, evaluating the effectiveness and trade-offs of proposed management interventions, and harmonizing multi-source datasets to reduce uncertainty in monitoring trends.

Key limitations stem from input datasets. The Curtis et al. forestry layer has high mapped accuracy but may slightly overestimate total forestry area, while the Lesiv et al. forest management classes have lower user/producer accuracies and likely underestimate forestry extent; the Schulze et al. FAO-based product suffers from inconsistent national reporting and coarse classes (production/mixed-use), reducing suitability for precise mapping of timber-producing forests. Selective logging in the tropics is difficult to detect via satellite, contributing to uncertainty in tropical results. Differences between logging products lead to variation in national/regional patterns (e.g., in Canada, sub-Saharan Africa, Latin America). Although the forest-loss detection algorithm versions are largely consistent post-2011 and independent tests indicate temporal trends are robust, potential under-reporting in specific periods (e.g., pre-2014 in Lesiv) and improvements in sensor inputs (Landsat 8) could affect absolute levels. Sample-based uncertainty (± s.e.m.) is incorporated for most regions except Africa, where such estimates are not available.