Main

The vast majority of roads are extra-urban roads traversing agricultural, forest and natural landscapes beyond the boundaries of cities1,2. Increases in traffic volume on such extra-urban roads can exacerbate pollution3,4,5, noise6, the spread of invasive species7 and habitat fragmentation8,9,10, affecting species abundance and diversity11, ecosystem dynamics12, organisms’ health13 and food safety14. Although traffic can thus have considerable ecological and environmental impacts in agricultural, forested and other (semi-)natural areas, little is known about the extent of the ecological impact of extra-urban road traffic15. Most previous studies aiming to quantify the impact of extra-urban roads considered only the road network and did not consider traffic volumes16,17,18. While the existence of a road is undoubtedly a prerequisite for any ecological or environmental effects to occur16, traffic volume is a key determinant of the magnitude of these effects3,4,5,6,7,8,9,10,19. However, due to a widespread shortage of extra-urban road traffic data20, spatially explicit estimates of the extent of the ecological impact of extra-urban road traffic are lacking in most regions across the globe1,21.

The ecological impact of a road can be expressed as the area of land surrounding the road that is impacted by the traffic on that road; this area is also known as the road effect zone (REZ)17,18,19. As the ecological and environmental effects of traffic generally decrease with distance to a road5,14,19,22, the REZ is usually delineated by setting a distance threshold from the road beyond which these effects are considered negligible. In studies that do not consider traffic volumes, this threshold is usually a fixed distance to all roads or to certain road types17,18,23. However, this strategy tends to overestimate the extent of the REZ on particularly quiet roads and underestimate it on busy ones. Considering that the ecological and environmental impacts of roads increase with the traffic volume on the road, the threshold distance to delineate REZs is ideally dependent on a road’s traffic volume. Several studies have used kernel density functions to summarize nearby traffic volumes and create traffic exposure maps that quantify the exposure to traffic-related pollutants or disturbances in each grid cell22,24,25,26. From such maps, REZs that are reactive to traffic volume can be derived.

Here we perform a global assessment of the spatial extent of the ecological impact of extra-urban road traffic over the past several decades. We delineated REZs in which ecological and environmental conditions are probably influenced by road traffic. We assessed the degree to which various land uses and key biodiversity areas were exposed to road traffic. Key biodiversity areas27 are areas contributing to the global persistence of biodiversity, defined according to standardized criteria and selected for their unique or threatened biodiversity, ecological intactness or irreplaceability28. Preserving the integrity of these areas is essential for biodiversity conservation. By identifying locations most affected by ecological degradation from road traffic, our analysis supports the spatial prioritization of mitigation strategies to reduce these impacts.

To estimate traffic-induced REZs, we leveraged a global time series of traffic volumes (measured as annual average daily traffic (AADT)) on all extra-urban highways, primary roads and secondary roads2 (that is, main roads) for the years 1975, 1990, 2000 and 20152. Using these AADT estimates, we generated high-resolution traffic exposure maps22, which serve as the foundation for delineating REZs. Our focus is on the direct ecological and environmental effects of traffic, excluding indirect effects such as deforestation due to road construction16,29.

In delineating REZs, our goal was to capture the variability in ecological responses to traffic, recognizing that different environmental and ecological impacts of roads emerge at different traffic volumes. For instance, studies have found that the detrimental effects of roads on numerous animal species begin at traffic volumes of 100 to 1,500 vehicles per day8, that runoff water from roads can surpass environmental quality standards when traffic exceeds 5,000 vehicles per day3 and that exhaust emissions can lead to adverse health effects in humans—and probably also in other organisms—at traffic levels above 10,000 vehicles per day13. We therefore delineated REZs for different traffic volume ranges according to the classification scheme for biologically relevant traffic volumes proposed in ref. 9: 500–5,000 vehicles a day represents a moderate impact, 5,000–10,000 vehicles a high impact and above 10,000 vehicles a very high to extreme impact. We did not calculate REZs for the low impact class below 500 vehicles a day9. Although a few studies have reported ecological effects of roads with traffic volumes below this threshold8, these instances are often limited to specific species. We therefore chose 500 vehicles a day as a conservative and ecologically meaningful lower bound for our analysis, while acknowledging that some impacts may occur below this threshold.

Results

Extent and change of extra-urban REZs

The total extent of extra-urban REZs strongly increased over the past several decades. Between 1975 and 2015, the global extent of the extra-urban REZs caused by moderate to very high traffic (summed across all traffic impact classes) increased by 53%, from 156 to 239 million ha, corresponding to 1.2% to 1.8%, respectively, of the global terrestrial land (excluding Antarctica; Fig. 1a and Supplementary Table 1).

Fig. 1: Development of the global and regional extent of extra-urban REZs, cumulative for three road traffic impact classes.
figure 1

a, Total extent of REZs relative to the extent of terrestrial land (excluding Antarctica; left y axis) and in absolute terms (right y axis). b, For different global regions, the extent of REZs relative to the region’s terrestrial land area. The results for the regions of Melanesia, Micronesia and Polynesia are not displayed, due to very small REZs (Supplementary Table 1).

Full size image

For most global regions, the extent of REZ impact by moderate to very high traffic increased steadily from 1975 to 2015 (Fig. 1b). In contrast, the extent remained fairly constant across all four European regions during the same period, although these regions were also among those with the largest REZ extents (Fig. 1b). Particularly in southern, western and northern Europe, respectively, 7.8%, 12.3% and 5.7% of the terrestrial land was impacted by moderate to very high traffic in 2015 (Fig. 1b and Supplementary Table 1). In central, southern and southeastern Asia as well as northern Africa, the extent of REZs caused by moderate to very high traffic more than doubled between 1975 and 2015 (growth > 111%; Fig. 1b and Supplementary Table 1). The strongest increase occurred in southern Asia, where the total extent of extra-urban REZs grew by 471%, from 0.6% to 3.3% of the terrestrial land surface (Fig. 1b and Supplementary Table 1).

Comparing global maps of the extent of REZs for 1975 and 2015, regional differences can also be observed (Fig. 2 and Extended Data Figs. 1 and 2). Whereas large parts of Europe and North America already had large extents of REZs in 1975, REZs have grown considerably in many Asian countries, such as India, China and Thailand (Fig. 2). Large extents of Latin America, Africa and Australia remain unaffected by moderate to very high extra-urban road traffic (Fig. 2 and Extended Data Figs. 1 and 2).

Fig. 2: Density of REZs caused by moderate to very high (AADT > 500) road traffic for 1975 and 2015.
figure 2

The values represent the area percentage of REZs per 5 km × 5 km raster cell. The colours represent the quartiles of cell values in 1975. Raster cells without an REZ are transparent. Maps for REZs caused by high and very high road traffic volumes are shown in Extended Data Figs. 1 and 2. Country boundaries adapted from ref. 31 under a Creative Commons license CC BY 4.0. Figure created with tmap63 in R v.4.4.2 (ref. 62).

Full size image

Breaking down the extra-urban REZs for the three different traffic impact classes reveals that most global regions are primarily impacted by moderate traffic, while the share of REZs affected by high and very high traffic is lower (Fig. 1b). Of the 239 million ha of REZs in 2015, 187 million ha was caused by moderate traffic, whereas high and very high traffic accounted for 30 and 22 million ha, respectively (Fig. 1a and Supplementary Table 1). In central Asia, Latin America and the Caribbean, and sub-Saharan Africa, less than 6.6% of the total REZ extent was caused by high and very high traffic volumes in 2015, whereas in North America, eastern Asia, and western and southern Europe these shares were more than 30.8% (Fig. 1b). However, for all regions, REZs caused by high and very high traffic are expected to increase. From 1975 to 2015, the average annual expansion rates of REZs caused by moderate traffic were consistently lower than those impacted by high or very high traffic across all global regions (Fig. 3a). The highest difference was found in southern Asia, where REZs caused by moderate traffic grew by 4.3% per year, whereas REZs caused by high traffic expanded by 8.1% per year (Fig. 3a). In southern and western Europe, the extent of REZs caused by moderate traffic even decreased (−0.2 and −0.4% per year, respectively; Fig. 3a), due to an increasing area coming under the influence of high and very high traffic.

Fig. 3: Expansion rates of REZs and their extent in key biodiversity areas.
figure 3

a, Average annual compound growth of REZs between 1975 and 2015 in global regions (red lines) and in key biodiversity areas (blue lines) for moderate, high and very high road traffic impact. The growth rates for moderate and high impact excluded the higher category (for example, the moderate area excluded the areas of high and very high). In southern Asia, there were no REZs caused by very high traffic in key biodiversity areas in 1975, and hence growth could not be calculated. b, The global area percentage of key biodiversity areas covered by REZs, cumulative for moderate, high and very high road traffic impact.

Full size image

The density of extra-urban main roads showed a positive relationship to the density of extra-urban REZs (Fig. 4a), indicating that the largest ecological or environmental impact of road traffic can be found in those locations with the highest road density. However, there is a large variation in REZ extent per kilometre of extra-urban road—a measure of road use intensity—across the globe (Fig. 4b). While most stretches of extra-urban main roads have an REZ caused by moderate to very high road traffic, REZs caused by high and/or very high traffic are absent on most roads (Fig. 4b).

Fig. 4: Density of main roads and REZs and the distribution of REZ extent per kilometre of road.
figure 4

a, Extra-urban road densities plotted against the density of REZs caused by moderate to very high traffic. The x axis has a logarithmic scale. b, Frequency distribution of REZ extent per kilometre of road (an indicator of road use intensity) binned into nine classes for moderate, high and very high road traffic impact. Raster cells containing a road but no REZ are in the class labelled 0. For these plots, only 5 km × 5 km raster cells containing extra-urban main roads were considered (n = 1,242,345).

Full size image

Biodiversity impacts of extra-urban traffic

In key biodiversity areas27, we observed a strong increase in the extent of REZs caused by increasing traffic volumes within or on the periphery of these areas (Fig. 3b). Between 1975 and 2015, the share of key biodiversity areas containing REZs caused by moderate to very high traffic increased from 52% (7,656 areas) to 63% (9,309 areas). For REZs caused by very high traffic, this increase was from 11% to 19%. Between 1975 and 2015, the total extent of key biodiversity areas containing REZs caused by moderate to very high traffic increased by 58%, from 9 to 14 million ha, representing 0.7% to 1.2%, respectively, of the total extent of key biodiversity areas (Fig. 3b and Supplementary Table 2). These percentages were lower than the area percentage of REZs covering terrestrial land (Fig. 1a), but for most global regions, the average yearly growth of REZs within key biodiversity areas was larger than the growth in the whole region (Fig. 3a). Especially for REZs caused by high or very high traffic, the differences between regional growth rates and those within key biodiversity areas were often considerable (Fig. 3a). For instance, in central Asia, REZs caused by very high traffic grew annually by 4.0% in the whole region but by 11.0% within key biodiversity areas (Fig. 3a).

Land-use impacts of extra-urban traffic

Breaking down the total global extent of REZs across different land-use categories revealed that the largest extents of REZs summed across all traffic impact classes were primarily found in cropland, followed by grassland and shrubland (Fig. 5a). In nearly all global regions and for all levels of traffic impact, these land uses were among those hosting the largest extents of REZs (Fig. 5b and Supplementary Figs. 1 and 2). Between 1990 and 2015, the extent of cropland affected by REZs caused by moderate to very high traffic increased from 90 to 110 million ha, representing 4.2% of global cropland in 1990 and 5.0% in 2015 (Supplementary Table 3). Over the same period, REZ coverage in grassland and shrubland rose from 33 to 42 million ha, increasing from 1.2% to 1.5% of their global extent (Fig. 5a and Supplementary Table 3). These trends indicate that the expansion of REZs both in cropland and in grassland and shrubland outpaced the overall growth of these land-use types during this period.

Fig. 5: Extent of land uses covered by REZs impacted by moderate to very high traffic (AADT > 500) at the global and regional levels.
figure 5

a, Extent at the global level. b, Extent at the regional level. These results were created by overlaying the REZ maps with the Climate Change Initiative land-cover maps60. For 1990, land-cover maps from 1992 were used. Permanent snow and ice were excluded. Results for the regions of Melanesia, Micronesia and Polynesia are not displayed due to very small REZs. The values displayed in this figure are summarized in Supplementary Table 3. For similar plots for REZs caused by high and very high road traffic volumes, see Supplementary Figs. 1 and 2.

Full size image

For REZs caused by high and very high traffic, we also observed substantial REZ presence in peri-urban areas (Supplementary Figs. 1 and 2). Relative to the land use’s extent, a large proportion of peri-urban land was covered by REZs in most global regions (Supplementary Fig. 3). In 2015, more than 43% of peri-urban land in northern, southern and western Europe, as well as North America, was affected by REZs from moderate to very high traffic (Supplementary Table 3). Although the absolute extent of REZs in peri-urban areas increased globally between 1990 and 2015, the proportion of peri-urban land affected by REZs from moderate and high traffic declined (for very high traffic, there was a slight increase; Supplementary Tables 3–5). At the global scale, the expansion of REZs in peri-urban areas thus progressed more slowly than the overall growth of peri-urban land.

Although forests (mixed, broad-leaved and coniferous) host a relatively low total extent of REZs globally (Fig. 5a), at the regional scale REZs can cover considerable areas of forest. The largest REZ extents in 2015 were observed in North America’s forests, where 6.3 million ha of mixed forests, 5.6 million ha of broad-leaved forests and 6.4 million ha of coniferous forests were affected by moderate to very high traffic (Supplementary Table 3). Water and wetlands had low REZ coverage both in absolute (Fig. 5) and relative terms (Supplementary Fig. 3).

Discussion

Extent of the ecological impact of extra-urban roads

Our comprehensive assessment of the ecological impact of extra-urban road traffic, measured here as the size of the traffic-induced REZ, revealed that road traffic has a strong and growing ecological impact on land outside of urban centres. The total extent of the REZs caused by moderate to very high traffic expanded by 53% in four decades, indicating increasing traffic-related ecological and environmental degradation in extra-urban areas. While large areas in Europe and North America already had large extents of REZs in 1975, REZs have expanded considerably in many regions in Asia (Figs. 1 and 2). While moderate road traffic continued to cause the largest extents of REZs, the growth rates of these zones were consistently lower than those caused by high or very high traffic (Fig. 3a). This alarming trend will lead to a growing proportion of land shifting from being moderately impacted by road traffic to being heavily impacted, a phenomenon clearly observed in southern and western Europe. The rapid growth of REZs is partly linked to economic growth in many countries, which is often accompanied by increased demand for mobility and motorized transport30. This general trend is supported by the finding that the gross domestic product per capita was a key predictor of traffic volumes used to calculate REZs2.

According to our estimates, the extra-urban REZs caused by moderate, high and very high traffic covered 187, 30 and 22 million ha, respectively, of the Earth’s terrestrial land in 2015, which represent substantial surfaces of land comparable to the sizes of Indonesia (190 million ha), Italy (30 million ha) and Guyana (21 million ha)31. In 2015, the global extent of REZs caused by moderate to very high traffic (239 million ha) was over three times larger than the extent of urban land (65 million ha (ref. 32)) and considerably greater than the sealed surface of extra-urban main roads (9.2 million ha, assuming a total length of extra-urban main roads of 7.2 million km (ref. 2) and an average road width of 12.7 m (ref. 33)). Given the vast extent of extra-urban REZs compared with urban areas and road surfaces, the ecological and environmental impacts of extra-urban road traffic are substantial and cannot be ignored.

The traffic impact classes from this study can guide the prioritization of mitigation measures to reduce the harmful effects of road traffic. While mitigation is particularly urgent in regions heavily affected by high and very high traffic—such as North America, eastern Asia, and western and southern Europe—areas experiencing moderate or low traffic also warrant attention. These regions may still support relatively intact ecosystems that require protection to safeguard them against future degradation. The maps of local REZ densities can be a valuable tool for guiding mitigation and conservation efforts (https://doi.org/10.3929/ethz-b-000745533).

Road traffic in key biodiversity areas

By 2015, 63% of key biodiversity areas were already affected by moderate to very high traffic. Alarmingly, many global regions exhibited higher growth rates of REZs, especially from high and very high traffic, in key biodiversity areas than in surrounding regions. This accelerated growth will narrow the gap between the surface area of key biodiversity areas impacted by REZs and that of all terrestrial land (1.18% versus 1.80%). As 75% of key biodiversity areas already contain roads34, traffic volumes in these ecologically critical areas can rapidly increase if no mitigation measures are implemented and enforced. Key biodiversity areas encompass diverse natural and semi-natural habitats across the globe and are essential for the protection of nature and biodiversity27,28. Without adequate mitigation measures, such as traffic calming, runoff treatment, noise barriers or wildlife bridges35, the traffic-related ecological and environmental damages to these often-sensitive areas are expected to increase fast. Due to a lack of global datasets, our study did not consider any mitigation measures that were already in place. These findings underscore the urgent need for global attention and action to limit traffic-related impacts in key biodiversity areas and safeguard their ecological integrity.

Land uses impacted by extra-urban traffic

We found the largest extents of extra-urban REZs in agricultural land (cropland and grassland/shrubland), and this extent was expanding. This suggests that a considerable share of the cumulative ecological impacts of extra-urban road traffic occurred in agricultural lands, compounding the existing ecological and environmental pressures these areas already face36. For example, in addition to the large impacts that contemporary agricultural practices have on insect populations37, declines caused by road traffic represent a substantial yet understudied threat to insects10. The contamination of crops or livestock with toxic elements or microplastics from traffic is an additional potential risk of high traffic volumes in agricultural land14,38.

Although peri-urban areas account for a relatively small share of global terrestrial land, a large proportion of them was affected by road traffic impacts. These multifunctional spaces not only serve as important residential areas for humans but also can support a higher diversity of plant and animal species than adjacent rural and urban environments39,40. High traffic intensities in peri-urban areas can therefore lead to considerable ecological and environmental degradation, while also posing serious health risks to people living in these areas13,41.

The relatively high impacts of road traffic on agricultural land and peri-urban areas are probably due to the relatively dense road network these areas tend to have17. Indeed, we found a positive relationship between road density and the extent of REZs. The proximity of many agricultural42 and peri-urban areas to urban centres further explains these findings. As most of the extra-urban traffic originates in urban areas, demographic, socio-economic and policy changes in urban areas can impact surrounding regions by altering extra-urban traffic patterns. In this respect, extra-urban traffic can be seen as an important telecoupling between urban and non-urban land uses43. Understanding the processes governing extra-urban road traffic as well as its interaction with land uses outside of cities is thus key for sustainable urban development.

Modelling REZs

Our finding that REZs cover 1.8% of terrestrial land is much lower than the estimate of 20% in ref. 17, which did not account for traffic volume but used a fixed 1-km distance to all roads. Our approach to calculating traffic-influenced REZs involved setting two thresholds: a minimum traffic volume and a maximum distance from roads (Fig. 6). We chose to use three traffic volume thresholds9 to account for varying ecological and environmental impacts, each of which starts to emerge at different traffic volumes. For example, for some species, wildlife–vehicle collisions occur primarily below 5,000 vehicles a day44, whereas avoidance of roads becomes prominent above 10,000 vehicles a day8. Depending on the specific ecological or environmental impact, a traffic volume threshold can be selected that best aligns with the onset of the impact.

Fig. 6: Calculation steps to determine REZs illustrated in a region to the west of Buenos Aires, Argentina.
figure 6

a, Extra-urban roads and urban areas. Each road contains information on the traffic volume (AADT) from which REZs were calculated. The calculation of traffic volumes and the definition of urban areas is described in Van Strien and Grêt-Regamey2. b, Traffic exposure maps calculated (20 m × 20 m cell size) with the kernel density function of ArcGIS Pro v.2.9.5 (Environmental Systems Research Institute). c, Cross-section of the kernel density function53 perpendicular to a single-lane road. For the AADT thresholds of 500, 5,000 and 10,000 vehicles a day, the corresponding kernel density values are calculated with a linear transformation function. Kernel density values above these thresholds determine the REZs caused by moderate (red), high (yellow) and very high (blue) traffic. d, Map of the REZs for different traffic impacts. In the northeastern region of the image, the REZ intersects a key biodiversity area. Road and urban area data in a, b and d from ref. 2, key biodiversity area data from ref. 27. Maps created with ArcGIS Pro v.2.9.5.

Full size image

In this study, the REZ ranged from 0 to 300 metres from roads, with the latter occurring along only the busiest roads, where ‘busy’ is relative to the traffic volume threshold (see Methods for the details; Fig. 6). The 300-m threshold was chosen because most of the ecological and environmental damage of traffic is confined to several hundred metres from roads5,7,17,19,22,38,45,46,47. However, precise information on the distance from roads at which these impacts are detrimental remains limited19 and can be context specific. For example, in a comprehensive meta-analysis, the distance over which roads affect the abundances of mammals, amphibians, reptiles and birds was estimated to be 105, 48, 27 and 655 m, respectively47. How these distances are influenced by traffic volumes remains to be determined. Other factors, such as the terrain5 and land cover47 around roads and the composition of traffic on roads24, can also affect the distance over which roads impact the surrounding landscape. Limiting the REZs to maximally 300 m from roads encompasses the distance over which most ecological and environmental effects have been observed, but we acknowledge that local deviations are likely to occur.

Roads and traffic

Since road traffic occurs on roads, reliable information on the locations of roads is essential for quantifying traffic impacts. Traffic volumes were predicted on the main roads in the GRIP4 road dataset from 2018, one of the most comprehensive road datasets available48. Due to a lack of comparatively reliable global road datasets for earlier years, the road network stayed constant between 1975 and 2015, but road attributes were adapted in an informed way (see Methods for the details). Given the rapid expansion of the road network in some regions over recent decades, maintaining a constant road network may have led to overestimations of roads and REZs in earlier time steps, leading to underestimated REZ growth rates. The 2015 REZ estimates are unaffected by this data limitation. Furthermore, traffic volumes were calculated only on highways, primary roads and secondary roads2, as lower-class roads are often unmapped29,48. Although local roads generally have low traffic volumes1,21, their sheer number48 means their combined impact can be substantial1. Additionally, small, often informal, roads can lead to indirect effects, such as deforestation near the road16,29. Improving the mapping of small and historical roads is essential to better estimate their impacts29.

We found that the REZ extent per kilometre of road varied strongly and that most extra-urban roads had no REZ caused by high or very high traffic. Roads and traffic therefore cannot be used interchangeably for assessing ecological or environmental impacts. Traffic is primarily responsible for the direct ecological and environmental effects of roads. This is supported by the reduced ecological impact observed during the COVID-19 pandemic, when traffic considerably decreased49. In fact, quiet roads can serve as useful corridors for animal species50. Incorporating such effects in ecological studies requires ecological research to focus on traffic volumes instead of simply the presence or absence of roads. The research field of road ecology35, true to its name, has a strong focus on roads. We underline earlier calls for a paradigm shift in road ecology, advocating for a stronger emphasis on traffic dynamics1,2,15,21. By conducting a global assessment of the extent of the ecological and environmental impacts of extra-urban road traffic, we provide compelling evidence that road traffic has a substantial impact and deserves greater attention in both research and practice.

Methods

Global extra-urban traffic volume data

The REZs were calculated from a global dataset of traffic volumes on extra-urban main roads (Fig. 6a), which we described in detail in a previous study2. In summary, we developed a model to predict AADT values on all extra-urban highways, primary roads and secondary roads across the globe for 1975, 1990, 2000 and 2015. The AADT prediction model was trained on a large collection of empirical AADT measurements from 46 countries across all continents (except Antarctica). To calculate explanatory variables for AADT, a global network model was developed with urban areas and intersections as nodes and roads as edges. For each edge, potential explanatory variables were derived from the network, such as betweenness centrality, and from several global datasets, such as gross domestic product, human development index and population density2,51. The empirical AADT values were related to the explanatory variables using a random forest model. Despite the inherent inaccuracies in global datasets, the final model demonstrated a good fit (R2 = 0.74). An independent validation using empirical AADT values from 1975 confirmed its accuracy2.

Spatial data on highways, primary roads and secondary roads were sourced from the comprehensive GRIP4 global road dataset48 (Fig. 6a). Tertiary and local roads in the GRIP4 dataset were not considered, as the mapping completeness of these roads shows strong regional variability29,48, and empirical AADT training data are often not available for such roads21. Due to a lack of reliable historical road data, the road networks for 1975 and 1990 were created by reducing roads in the GRIP4 dataset to a lower grade if they were not present in the Vmap0 dataset predating 19972,52. As Vmap0 is incomplete in many regions48, completely removing roads absent in Vmap0 would have caused a considerable bias in these regions2. Although road types thus changed over time (influencing traffic volumes), the road density stayed constant. To identify which roads were outside of urban areas (that is, extra-urban roads), in our previous study2 we delineated urban areas from the global population density dataset GHS-POP51. Urban areas were defined as clusters of connected raster cells with a population density exceeding 1,500 people per km2 and a total population of at least 2,0002 (Fig. 6a). Roads inside urban areas were not considered.

Traffic exposure maps

The REZs were delineated in traffic exposure maps calculated by applying the kernel density function from ArcGIS Pro v.2.9.5 (Environmental Systems Research Institute) to the predicted median AADT values (Fig. 6b). The function calculates the traffic volume density around roads by fitting a quadratic kernel function53 to the road dataset. The kernel density value is highest at the location of a road and decreases smoothly with distance to the road (Fig. 6c). Traffic on parallel running roads or intersections has an accumulative effect, as can be seen at some intersections in Fig. 6b. Following ref. 22, the search radius of the kernel density function was limited to 300 m from a road22 (Fig. 6c). This distance was chosen as many studies have found that the ecological and environmental damage from traffic is typically restricted to several hundred metres from roads5,17,19,22,38,45,46,47. The output kernel density raster had 20 m × 20 m cells, resulting in a high-resolution traffic exposure map across the globe (Fig. 6b). The calculation of traffic exposure maps and the delineation of REZs were performed in Python v.3.7.11 (ref. 54) with the packages arcpy55, numpy56, pandas57, geopandas58 and rasterio59.

Delineating REZs

The REZs were delineated by selecting all kernel density values in the traffic exposure maps above certain minimum values (Fig. 6b), which correspond to the moderate (AADT ≥ 500), high (AADT ≥ 5,000) and very high (AADT ≥ 10,000) traffic volume thresholds (Fig. 6c). To convert the AADT thresholds to kernel density values, we determined the relationship between the kernel density value and the AADT. As the kernel density value on intersections or parallel lanes can be the result of traffic flows on multiple roads, determining a direct relationship between specific AADT values and kernel density values is difficult at such locations. We therefore focused on single-lane roads to determine this relationship, which yielded a linear transformation function with a perfect fit (R2 = 1.0): kernel density value = AADT/2.9433201904. To delineate, for example, REZs caused by moderate traffic, we selected all kernel density values above 169.88 (500/2.9433201904) from the traffic exposure map. This resulted in global raster maps (20 m × 20 m cell size) of the REZs at three traffic volume thresholds (Fig. 6d) for the years 1975, 1990, 2000 and 20152.

The maximum width of an REZ in our study was thus 600 m, and the actual width depended on the traffic volume threshold as well as the traffic on the road (Fig. 6c). For example, a road with 550 vehicles a day will have a narrow REZ (≪600 m) at the moderate traffic threshold but no REZ at the high or very high traffic thresholds. In contrast, a road with 5,050 vehicles a day will have a relatively wide REZ at the moderate traffic threshold, a narrow REZ at the high traffic threshold (≪600 m) and no REZ at the very high traffic threshold (Fig. 6d).

Postprocessing and land use and biodiversity impacts

As the high-resolution REZ maps (Fig. 6d) were unsuitable to be displayed at a global level, we aggregated them at a local level in a 5 km × 5 km global raster depicting the percentage of area that was covered by the REZs in each raster cell (Fig. 2). For each of these raster cells, we also calculated the length of extra-urban main roads (defined above). For the entire globe and for each global region in the geoBoundaries dataset31, we calculated the total area of the REZs for all traffic impact classes and for all time steps. For the global regions, we calculated the compound annual growth rate of REZs between 1975 and 2015 using the following equation:

$${\rm{Annual}}\; {\rm{average}}\; {\rm{growth}}\; {\rm{of}}\; {\rm{REZs}}=\left({\left(\frac{{\rm{REZ}}\; {\rm{extent}}\; {\rm{in}}\;2015}{{\rm{REZ}}\; {\rm{extent}}\; {\rm{in}}\;1975}\right)}^{\frac{1}{40}}-1\right)\times 100$$

To estimate road traffic impacts on land uses and on areas important for biodiversity conservation, we overlaid the delineated REZs with the Climate Change Initiative (CCI) land-cover time series60 as well as the World Database of Key Biodiversity Areas27. We calculated the share of REZs that covered certain land-use classes and key biodiversity areas. In our analysis, there was no change in the number or extent of key biodiversity areas over time. As the CCI map series goes back to 1992, we did not perform this analysis for the 1975 REZs and combined the 1992 land-cover map with the 1990 REZs. The CCI land-cover classes were aggregated according to ref. 61 to the land-use classes displayed in Fig. 5. The CCI maps also include a class for urban area. However, as the traffic volume dataset excluded high-density urban areas2, we did not calculate REZs within urban areas (see above). We therefore removed the urban areas delineated according to our definition above from the CCI maps and considered the remaining urban land use in the CCI maps to be low-density urban area; we referred to this as peri-urban. As the CCI maps have a coarser resolution than the REZ maps (300 m and 20 m, respectively), the total extent of CCI raster cells covered by an REZ was higher than the extent of the REZ itself. The extents of the different land-cover classes were therefore rescaled to match those of the REZs. The postprocessing of the data and the creation of the figures were performed in R v.4.4.2 (ref. 62) with the packages tmap63, terra64 and readxl65 and packages from the tidyverse66.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.