Main

Timber is a natural resource of vital importance to the global economy. Production occurs in a third of the forests of the world1 and contributes more than US$1.5 trillion per year to national economies2. Global timber demand is expected to rise by between 54% and 200% by 20503,4, driven by increasing urbanization and the need to replace carbon-intensive construction materials, such as steel or concrete5, as many nations move towards net-zero economies.

Ensuring a sustainable supply of timber throughout the twenty-first century is a key global challenge. Climate change will cause significant stresses to global timber supplies as increasingly severe and frequent extreme weather conditions impact tree survival and drive increased forest disturbances6,7. Wildfires already pose substantial risks, with an area of timber-producing forest the size of Great Britain lost to fire between 2001 and 2021 and a twofold to fourfold increase in the rate of loss from 20168. Pest outbreaks are also expected to increase in extent and severity under climate change9,10, compounding threats to future timber yield. However, an emerging threat to timber production may yet come from a presently overlooked source: competition for land with agriculture.

Concurrent with rising timber demand, meeting the nutritional needs of an increasing and more affluent global population requires agricultural production to double by 2050 compared to 200511. Agriculture is predicted to track suitable growing conditions under climate change12. Thus, by 2060–2080, 1.5 billion ha of previously unsuitable land is predicted to become available for agricultural production13. Conversely, 2.2 billion ha of land is predicted to become less suitable by 2071–210014. Previous research has scarcely considered the implications of meeting rising demand for timber and food on an increasingly hostile planet.

A key unanswered question is the degree to which climate-induced shifts in agricultural suitability will trigger competition with land currently used for timber production. We combine global data on timber production and agricultural suitability to examine the scale of this potential future conflict under two representative concentration pathways (RCP 2.6 and 8.5; Methods) representing best- and worst-case scenarios for global emissions. We have four key objectives: (1) determine the degree to which future changes in agricultural suitability will overlap with timber production; (2) identify hotspots for potential conflict; (3) identify which crops are driving this conflict; and (4) assess the relative pressure of future cropland expansion into forestry land compared to non-forestry land.

Potential for land conflict to increase throughout the twenty-first century

To identify the scale of potential future agriculture and forestry conflict, we overlaid two datasets detailing the global extent of forestry land and predictions of future agricultural suitability for key crops under climate change scenarios. The forestry data come from ref. 15, who combined satellite-derived forest loss data with machine learning to identify areas where forest loss events between 2001 and 2022 were driven by forestry activity, which we use to denote forestry land in this study. The agricultural suitability data come from ref. 14, who used climatic, soil and topographic conditions to predict future agricultural suitability for the 17 most important crops of the world for the global economy, food security and biofuel throughout the twenty-first century14,16 under two climate change scenarios (RCP 2.6 and 8.5). Agricultural suitability is classified by ref. 14 into four categories: not suitable (0), marginally suitable (1–32), moderately suitable (33–75) and highly suitable (76–100).

By the end of the century, 240 and 320 million hectares (Mha) of current forestry land is expected to become more suitable for agriculture under RCP 2.6 and 8.5, respectively (Fig. 1; see Extended Data Fig. 1 for raw suitability values). This equates to 20–26% of current global forestry land—an area almost the size of India. Globally, forestry land demonstrates an average increase in agricultural suitability of 13.3% and 17.9% under RCP 2.6 and 8.5, respectively. Areas of increasing agricultural suitability are dominated by the boreal north, particularly Western Canada and across Siberian Russia, as well as the USA and China (Fig. 1).

Fig. 1: Future change in agricultural suitability of current forestry lands.
figure 1

a–d, Change is relative to the historical period (1990–2009) and is presented for RCP 2.6 (a,c) and RCP 8.5 (b,d) in the time periods 2040–2069 (a,b) and 2070–2099 (c,d). Pixels are aggregated to 50× the original resolution (~1 km2 at the equator) for visualization, with the value presented being the mean change across all aggregated pixels. Agricultural suitability is ranked from 0 to 100 as per ref. 14, with 0 denoting not suitable, 1–32 marginally suitable, 33–75 moderately suitable and 76–100 highly suitable. Thus, a change of +100 indicates that previously unsuitable land will be highly suitable for agriculture. Alaska and northeastern Russia are removed to improve visualization since little forestry land is mapped here.

Full size image

Not all forestry land becomes more suitable for agriculture under climate change. Between 27 and 110 Mha will become less suitable by 2070–2099 under RCP 2.6 and 8.5, respectively. Decreases are predominantly seen across tropical areas, particularly southern and eastern Brazil and southern Europe, with patterns of decreased suitability considerably stronger under RCP 8.5 (Fig. 1). These areas are already experiencing increased frequencies of heatwaves, droughts and wildfires8,17,18,19 with climate change predicted to drive increasingly severe extreme events in the future20,21. Such conditions reflect low future suitability for both agriculture and timber production, further restricting available areas and potentially exacerbating competition for productive land. Across the tropics, large areas of forestry land highly suitable for agriculture remain so into the future (for example, Brazilian Atlantic Forest and Malaysian Peninsula) and these areas are already threatened by agricultural expansion22,23.

The agricultural footprint of the boreal regions is currently low24, primarily because of historically cold temperatures limiting crop growth and productivity25. However, northern latitudes are already warming disproportionately fast26 and will see increased agricultural activity in the future13,14. In part, this expansion will be driven by falling agricultural yields in areas becoming less suitable for crops due to increasing drought conditions. These conditions will simultaneously increase wildfire risk in timber-producing forests8, restricting both timber and agricultural production towards shrinking areas of higher productivity, particularly in northern latitudes10,14.

Focusing on agricultural land that is productive (not marginal, with an agricultural suitability score ≥33 (ref. 14; Methods) and thus assumed to be capable of competing economically with timber production, there will be a net increase of 65 and 84 Mha of agriculturally productive forestry land by the end of the twenty-first century under RCP 2.6 and 8.5, respectively. This represents a 13% and 18% increase relative to the historical period (1990–2009). Europe (including Russia) sees the greatest increase of productive land with 42 Mha of forestry land becoming newly productive by 2070–2099 under RCP 2.6, increasing to 47 Mha under RCP 8.5. Likewise, under RCP 2.6 by 2070–2099, over 16 Mha becomes productive in North America, a 15.1% increase; this rises to 19 Mha under RCP 8.5. Concerningly, the largest increases of agriculturally productive land fall within the largest global timber producers, including the USA, Russia, Canada and China, as well as throughout Scandinavia27.

Key Northern Hemisphere timber producers dominate conflicts

The four nations with the largest gains in agricultural suitability are Russia, the USA, Canada and China, which cumulatively account for ~50% of global roundwood production27 and 65% of intensive forestry land. Across these four countries, between 181 and 243 Mha of forestry land become more suitable for agriculture by 2100, including 60 and 78 Mha of newly productive agricultural land, under RCP 2.6 and 8.5, respectively (Fig. 2). Russia accounts for the largest increase in productive land, seeing a net increase of 40 and 48 Mha, followed by the USA (9 and 13 Mha), Canada (6 and 7 Mha) and China (4 and 11 Mha) under RCP 2.6 and 8.5, respectively. Similar patterns are present when considering the percentage increase in forestry land that is agriculturally productive, with 32% and 38%, 14% and 21%, 17% and 20% and 6% and 16% increases in Russia, the USA, Canada and China under RCP 2.6 and 8.5, respectively.

Fig. 2: Areas of gain, persistence and loss of agriculturally productive land in forestry land in 2070–2099 compared to historical period.
figure 2

a–h, Represented are the top four countries globally in terms of total area gain in productive land: Russia (a,b), mainland USA (excluding Alaska due to minimal forestry currently mapped here) (c,d), Canada (cropped for improved visualization) (e,f) and mainland China (g,h), under RCP 2.6 (a,c,e,g) and RCP 8.5 (b,d,f,h). Areas of gain of productive land in current forestry land (productivity frontiers) are marked yellow, areas of forestry land that are currently productive and will remain so are marked in green, areas of forestry land that are currently productive but become unproductive are marked purple, while forestry land that is not currently productive and will not be in the future is marked grey. Data are aggregated to 10× original resolution (~1 km2 at the equator) for visualization.

Full size image

By the end of this century, forestry land in these four countries will contain 91% of all new productivity frontiers—land that is at present unsuitable or marginal for agriculture but will become moderately or highly suitable in the future (Methods)—under RCP 2.6 and 86% under RCP 8.5. Frontiers of newly cultivable land (land where agriculture is currently not possible but will be in the future) are similarly focused in the Palaearctic and Nearctic realms, in particular Russia and Canada, with 57 and 94 MHa and 17 and 18 Mha of forestry land becoming cultivable by 2100, under RCP 2.6 and RCP 8.5, respectively. These two countries alone account for 86% of all newly cultivable frontiers in forestry land under RCP 2.6 and 85% under RCP 8.5 (Extended Data Fig. 2).

Widespread agricultural productivity gains for key global crops

Of the five crops with the highest global economic value28, soy and potato will probably place the most pressure on timber-producing areas. By the end of the century, the total area of productive agricultural land that falls within current forestry land will increase by 34 and 49 Mha for potato cultivation and 35 and 41 Mha for soy cultivation, under RCP 2.6 and 8.5, respectively (Fig. 3). Rice is likely to place the least pressure on timber production, with only small increases of agriculturally productive land of 1.7 and 2.1 Mha. Wheat suitability increases notably in timber-producing regions under RCP 2.6 but from 2070 onwards under RCP 8.5 it sees minimal net increase of total productive land in forestry regions as a result of significant losses of suitability throughout Eurasia (Extended Data Fig. 3).

Fig. 3: Net changes of agriculturally productive area within current forestry land.
figure 3

Each of the five most valuable crops globally according to the FAO by the end of the century (2070–2099) under RCP 2.6 and 8.5 are shown. a–e, The net changes of agriculturally productive area for all global forestry land (a) and the individual changes for the four countries with the largest extent of forestry land, Russia (b), the USA (c), Canada (d) and China (e). Note: x axis scales vary between plots.

Full size image

Russia will see the greatest increase of agriculturally productive areas within forestry land for several crops, chiefly potato (between 29 and 44 Mha under RCP 2.6 and 8.5, respectively), soy (between 19 and 29 Mha) and wheat (between 7 and 10 Mha). These crops will also place the greatest pressure in Canadian and US forestry lands, whereas timber production in China will come under particular pressure from wheat and maize. Some of these crops are already expanding globally24 and driving major forest loss, for example soy expansion across South America29.

Disproportionate pressure on forestry from growing agricultural suitability

Globally, forestry lands make up 10% of the land surface15, yet by the end of the century they will contain between 21% and 27% of newly productive agricultural land, 2.9 and 2.3 times more than would be expected given their size, under RCP 2.6 and 8.5, respectively (Fig. 4a). Russia, in particular, demonstrates the largest mismatch between the share of global land surface and agriculturally productive frontiers in its forestry land, with its share of productive frontiers being 4.8- and 6.8-times higher than its share of land surface under RCP 2.6 and 8.5, respectively. Similarly, the three countries with the next largest forestry area all have higher shares of productive frontiers in forestry land than would be expected solely by area (USA, 2.2× and 1.9×; Canada, 2.5× and 1.3×; China, 1.9× and 2.9× under RCP 2.6 and 8.5, respectively).

Fig. 4: Relative contribution of forestry land to new productivity frontiers by 2070–2099, under RCP 2.6 and 8.5.
figure 4

a, The global forestry contribution and the four countries with the largest total forestry area shown in descending order. Dashed horizontal line denotes equivalency; that is, the forestry area of a country contains an equal share of the land surface and new agriculturally productive frontiers of the world. A value of two would denote a country that has twice the productivity frontier area in forestry land that would be expected given its size (for example, forestry contributes 10% of all frontiers, on land that is only 5% of the land surface of the world). b, Travel time in hours from new productivity frontiers to population centres of >5,000 people. c, Geographic distance (km) from new productivity frontiers to existing cropland. In b and c, points denote the area weighted means, the thicker error bars denote the 25th–75th percentiles and the thinner error bars denote the 5th–95th percentiles.

Full size image

In addition to productive agricultural frontiers being disproportionately located in present-day forestry lands, they are on average more accessible from urban centres and closer to existing cropland than frontiers in non-forestry land. Mean travel times to urban centres from frontiers in forestry lands were between 231 min (17–671, 5th–95th percentile) and 220 min (10–675), compared to between 547 min (30–2,029) and 887 min (41–2,984) for frontiers located outside forestry lands under RCP 2.6 and 8.5, respectively. Similarly, frontiers in forestry land are on average closer to current cropland compared to frontiers outside of forestry land, with mean distances of 14 km (1–50) and 13 km (1–45) compared to 54 km (1–325) and 114 km (1–568) under RCP 2.6 and 8.5, respectively. Agricultural production typically spreads from current agricultural centres into fertile areas with good access to markets30 and many key crops are already major drivers of deforestation15,31. Therefore, it is likely these frontiers in forestry land will be under greater pressure from agricultural expansion than more isolated frontiers in non-forestry land.

Despite using the best-available data there remains uncertainty that may influence these results. The forestry data described previously15 have high accuracy (users, 87%; producers, 91%) but only span the years 2000–2022. Forests managed for timber that were not cut during this period or had no tree cover in the year 2000 as a result of previous harvests may go undetected. Many tropical plantation rotations are <20 years (ref. 32) with longer rotations of several decades predominantly in the temperate and boreal zones33, which we highlight as areas of high potential conflict. Thus, any missed forestry land would only further increase the possible land conflict between agriculture and timber.

The agricultural suitability layers from ref. 14 account for many topographic, soil and climatic factors but some crop-specific vulnerabilities may be overlooked (for example, high photoperiod sensitivity in soy34) as well as the increasing risk of late frosts at higher latitudes35. Although we assumed ‘moderately and ‘highly’ suitable agricultural land to be capable of competing economically with timber, this may not be the case everywhere, whereas in some regions marginal land may still be competitive. Likewise, the crop suitability approach does not account for the socioeconomic feasibility of crop expansion, particularly into agricultural frontiers which remain poorly understood. Large land-use systems models (for example, PLUMv2)36 offer avenues to understand economic feasibility and implications for some crops but developing these methods to incorporate the societal and political reality of expansion remains a challenge34. Similarly, our approach does not account for future shifts in areas of timber production, which remain a major research frontier. However, multidecadal rotation cycles make timber a far less dynamic commodity than crops. Most of the timber currently growing in future conflict areas (for example, Russia, the USA and Canada) cannot be harvested for decades to come and will thus be subject to increased future competition with agriculture.

Global implications of growing conflict between timber and agriculture

Climate-induced shifts in agricultural production represent a significant threat to the future of timber production. By 2100, 240–320 Mha of land currently used for timber production will increase in agricultural suitability, with 62–80 Mha becoming newly productive. Increased pressure from agriculture is concentrated in the four largest timber-producing nations, who account for 91% of the global forestry land that will become agriculturally productive. Simultaneously, current forestry lands disproportionately contain productive agricultural frontiers that are both closer to human population centres and existing cropland than are frontiers in non-forestry land. This future threat of agricultural expansion into timber lands, alongside other concurrent threats such as wildfires8, pests37 and increasingly extreme climatic events10,38 point towards a severe threat to sustainable timber production this century.

Climate change under both optimistic (RCP 2.6) and worst-case (RCP 8.5) scenarios creates vast tracts of newly available agricultural land predominantly in northern latitudes, offering the potential for agricultural expansion and production to meet growing global nutritional demands11. Several governments, including Canada39, Russia40 and China41 are already supporting policies encouraging agriculture and homesteading in northern latitudes13 and conversion of forestry land to arable land (for example, China42). Unrestrained agricultural expansion into land currently used for forestry risks critical shortages in timber supplies given the key role these areas play in supporting the global timber market.

As agricultural production tracks suitable climates poleward, so too could global centres of timber production. However, this would be extremely detrimental for global climate and biodiversity targets. Forestry expansion northwards would increase exploitation of old-growth boreal forests that account for a large proportion of the few remaining intact forest areas globally43. These forests hold vital global carbon stocks44,45, particularly in their carbon-rich soils46. In addition, encroachment of forestry and its required infrastructure (roads, towns and so on) into some of the few remaining areas of intact wilderness47, which also face intense future competition with agriculture48, would further strain the key role these areas play in sustaining biological and cultural diversity48. Another option for maintaining future timber production would be to increase exploitation of the world’s tropical timber supplies, especially the Amazon and Congo. Tropical forests are irreplaceable for biodiversity49, hold globally significant carbon stocks50, support the livelihoods of 1.6 billion people around the world51 and are a key element of nature-based climate solutions52. Increasing the exploitation of tropical forests must be avoided.

To minimize future land conflict between timber and agricultural production, it is crucial that increased agricultural outputs are predominantly achieved by improving efficiencies within the global food production system30,53,54,55. Closing yield gaps whilst implementing multiple cropping and economically efficient land-use decisions could increase agricultural output by 79–148% (ref. 56), whereas up to a third of the food produced annually is never consumed, with food worth ~US$1 trillion wasted every year57,58. Shifting diets away from ruminant meat would reduce land requirements for food production by up to 70% (ref. 55), allowing some of the ~1 billion ha of land used for ruminant production to be repurposed for the expansion of timber plantations.

In conclusion, widespread expansion of agricultural production into current timber-production systems risks knock-on effects for globally vital timber production and other ecosystem services. Systematic shifts in the supply-side production dynamics of global timber and food production must be evidence-led and preplanned to prevent unforeseen consequences for forests, economies and livelihoods. Beyond bending the emissions curve as early in the twenty-first century as possible, intensifying and optimizing current agricultural and timber-production systems will be critical for avoiding future land conflict and forestry expansion into natural land cover.

Methods

Data

Agricultural suitability

To represent the changing suitability of land for crop production, we used an updated version (v.3) of the agricultural suitability data from ref. 14. These data use a fuzzy logic approach that accounts for local climate, soil and topography conditions to determine the suitability of areas for crop production based on local conditions and crop requirements. The data consider agricultural suitability for each of 17 globally important food, feed, fibre and first-generation bioenergy crops under current global irrigation areas59, at a spatial resolution of 30 arcsec. Suitability scores (0–100) are assigned to each crop on the basis of the lowest suitability score of the various soil, climate and topographic limiters. Each cell is then assigned a suitability score which represents the highest value suitability score of all 17 crops in that cell (that is, the most suitable crop). As per ref. 14, the suitability scores can be broken down into the following categories: 0 (unsuitable); 1–32, marginally suitable; 33–74, moderately suitable; and 75–100, highly suitable.

The data provide suitability layers based on a historical time period (1980–2009), as well as for future time periods (here we use 2040–2069 and 2070–2099) using bias-corrected daily temperature, precipitation and solar radiation data. The suitability data provided are the median predicted values from five different climate models (GFDL, HadGEM2, IPSL, MIROC and NorESM1)60 under two alternative climate change scenarios, RCP 2.6 and 8.5 (ref. 61). The RCP 2.6 scenario represents a future where greenhouse gas emissions are strongly reduced, such that we see a modest temperature rise of 1.6 °C by 2100 compared to the pre-industrial period. However, RCP 8.5 scenario represents a high-emission scenario, where limited policy interventions and greenhouse reductions are implemented, such that warming reaches around 4.3 °C by the end of the century62. Recent research has highlighted that under certain criteria RCP 8.5 actually aligns most closely with current emission trends and likely emissions through the first half of the twenty-first century63. The two scenarios probably represent the best- and worst-case scenarios for future climate change and thus provide reasonable upper and lower estimates of the climate-driven changes in agricultural suitability across the world throughout this century.

Forestry land

To represent the current extent of global timber-production areas, we used a data layer from ref. 15, which used satellite data and machine learning to categorize the dominant drivers of tree-cover loss globally at a resolution of 10 km2 into the following categories: commodity-driven deforestation, shifting agriculture, forestry, wildfire and urbanization. The study covers tree loss between 2000 and 2022, considering drivers of forest loss in any cell where tree cover was detected in the year 2000. If several drivers were detected in a cell, the cell was assigned to whichever driver accounted for the highest proportion of forest loss. We filtered these data to include only the land characterized as tree-cover loss driven by ‘forestry’, defined as ‘large-scale forestry operations occurring within managed forests and tree plantations with evidence of forest regrowth in subsequent years’15. This layer thus provided us with a spatial estimate of regions containing high-intensity timber-production systems across the globe (that is, clearcut and plantations), covering an estimated 1.87 billion ha.

Owing to the relatively coarse nature of this dataset (10 km2), we performed more masking before analysis. We first disaggregated the data to ensure they matched the resolution of the agricultural suitability layer to allow for overlay analysis. Then, following the Food and Agriculture Organization of the United Nations definition of forest as containing tree cover ≥10% (ref. 2), we overlaid the forestry data with data detailing the mean proportion of tree cover per 1 km grid cell64, removing from the analysis any cell with tree cover <10%. Finally, we crossed the data with the ESA Land Cover Map65 and excluded from our forestry layer any cells that were marked as present cropland, urban settlements, bare rock or water. After these extra filtering steps, we were left with a total area of 1.21 billion ha of forestry land.

Spatial analysis

Change in agricultural suitability from present

All spatial analysis was conducted in Google Earth Engine and R66,67. We used the terra68 package to overlay the map of forest loss due to forestry15 with the data on agricultural suitability14 for each time period and each climate change scenario (RCP 2.6 and 8.5). Using the historical time period as a ‘present-day’ baseline, we calculated the change in agricultural suitability in each forestry cell for each future time period and scenario (2040–2069 and 2070–2099 under both RCP 2.6 and 8.5).

Change in agriculturally productive land

Following the original definitions from ref. 14, we defined any area as cultivable if it had an agricultural suitability score >0 and further grouped the data into the following agricultural suitability groups: unsuitable (0), marginally suitable (1–32), moderately suitable (33–74) and highly suitable (75–100). We focused on productivity frontiers, which we define as areas that were marginal or unsuitable (suitability <33) in the historical period that become productive in future periods (suitability ≥33). We focused on these transitions since land that is marginal for agriculture is unlikely to compete economically with land currently used for timber production and thus unlikely to cause much conversion pressure. We also estimated cultivation frontiers as areas that are unsuitable for agriculture of any kind in the historical time period (suitability = 0) but become suitable (suitability > 0) for agriculture in the future under climate change.

To understand which crops were driving the largest changes in agricultural suitability of present-day forestry land, we used the individual crop suitability layers provided in ref. 14. We assessed the changes in agricultural suitability of the top five crops by global production value (rice, maize, wheat, soy and potato; FAOSTAT, 2024) between the historical period (1980–2009) and 2071–2100 under RCP 2.6 and 8.5. To assess the change in total productive land for each crop, we used the same definitions of agricultural suitability and methods as above.

Assessing the pressure of agricultural conversion in forestry and non-forestry land

To assess the proximity and possible threat of agricultural expansion30 of new productivity frontiers in 2070–2099 under RCP 2.6 and 8.5 we used two complementary data sources. First, we used global travel time to urban centres69, specifically, travel time to population centres of at least 5,000 people. We overlaid this with our productivity frontier rasters and extracted the travel time (in minutes) of each frontier raster cell to the nearest population centre. Second, we used the ESA Land Cover Map65 to calculate the minimum distance from each frontier raster pixel to a pixel occupied by existing cropland. The original input layers used in these analyses were not equal-area projections, thus when we calculated summary statistics of travel time and distance we calculated area weighted percentiles and means, for which the area was the individual area of each cell. We did this for all productivity frontiers in areas mapped as forestry land and compared this productivity frontiers in areas of non-forestry land potentially available for agricultural expansion (which we defined as any land that was not mapped as forestry15, current cropland, urban settlements, bare rock or water59).

All analyses were completed at the global level and subsequently aggregated to national levels using the Global Administrative Areas (GADM) v.4.1 database70.