Roads and their traffic can affect wildlife over large areas and, in regions with dense road networks, may influence a high proportion of the ecological landscape. We assess the abundance of 75 bird species in relation to roads across Great Britain. Of these, 77% vary significantly in abundance with increasing road exposure, just over half negatively so. The effect distances of these negative associations average 700 m from a road, covering over 70% of Great Britain and over 40% of the total area of terrestrial protected sites. Species with smaller national populations generally have lower relative abundance with increasing road exposure, whereas the opposite is true for more common species. Smaller-bodied and migratory species are also more negatively associated with road exposure. By creating environmental conditions that benefit generally common species at the expense of others, road networks may echo other anthropogenic disturbances in bringing about large-scale simplification of avian communities.
Roads are widespread and can impact ecological communities. Cooke et al. use data for 75 bird species across Great Britain to show that common species are disproportionately abundant near roads, whereas rarer, smaller-bodied and migrant species are more likely to be negatively associated with roads.
The ever-expanding environmental footprint of humans is affecting global wildlife populations via a wide range of mechanisms, many of which we are only beginning to understand. Extinctions and population declines are widespread
Known human drivers of population change are numerous and include habitat loss
Roads are a source of noise, wildlife–vehicle collisions, chemical pollution and visual disturbance, including artificial light
Various species characteristics have the potential to affect or predict associations between birds and roads. Communication in smaller-bodied species may be more affected by road noise, due to their typically quieter and higher-frequency songs
Great Britain has one of the densest road networks in the world, with over 80% of land falling within 1 km of a road. We use data from the extensive UK Breeding Bird Survey (BBS) to analyse populations of 75 British bird species in relation to the paved road network, and to assess predictors of these patterns. As potential predictors, we choose three species-level characteristics—mean body mass, migratory tendency and an index of habitat specialisation—and two population-level characteristics—national population size and long-term national population trend. By assessing patterns of bird distribution in relation to roads across the whole of Great Britain, we find evidence to suggest that roads may contribute to broad-scale simplification of avian communities. Our findings provide much-needed information for potential road mitigation and conservation around roads.
We calculated the road exposure of almost 20,000 BBS transect sections using the locations of all paved roads (as mapped in 2013) within a 5-km radius of the midpoint of each transect section. Within these calculations, we estimated the spatial scale of the relationship between distance to road and road exposure (determined by a parameter ‘
Our results show the abundance of 77% ( For each species, the relative effect size was calculated as a composite of the magnitude of the effect size of road exposure and the spatial scale over which the effect could be detected (the latter being determined by the parameter ‘
To estimate the real-world magnitude of the associations between road exposure and bird abundance, we used our models to predict changes in abundance across the ranges of road exposure values recorded for each species. For species with strongly significant associations between abundance and road exposure (i.e., those significant after Bonferroni correction), the mean change in abundance from the 0.25 to 0.75 quartiles of road exposure was −40% for species with negative associations, and +48% for species with positive associations (Fig. Only species for which associations between road exposure and abundance were found to be significant after Bonferroni correction are featured here. The relative effect size of roads (as shown in Fig.
To explain our results in more detail, we use the examples of Eurasian bullfinch The intercept is determined by the coefficient and the rate of decline is determined by the parameter ‘ Bird abundance refers to the number of birds within 100 m of a 200-m BBS transect section. The 0.25 and 0.75 quartiles of road exposure for each species are indicated by the vertical lines, and 95% prediction intervals by the shaded areas. These graphs are available for all species in Supplementary Fig.
Previous studies have suggested differences in the potential impacts of higher and lower traffic-level roads As in Fig.
To assess predictors of the associations we found between road exposure and bird abundance, we analysed the relative effect sizes (of all roads together) in relation to five species characteristics: mean body mass, migratory tendency, an index of habitat specialisation, national population size and long-term national population trend, using a generalised estimating equation. Within this, we accounted for non-independence resulting from similarity within phylogenetic families. We also weighted each species by 1/variance of the effect size of road exposure, to increase the influence of species with more precise association estimates between bird abundance and road exposure.
We found that species with smaller national population sizes had generally lower abundance with increasing road exposure, whereas the opposite was true for more common species (Table Relationships between species characteristics and associations with road exposure. Characteristic Effect size Standard error Mean body mass 0.027 0.009 0.004 Migratory tendency −0.042 0.012 <0.001 Habitat specialisation 0.08 0.10 0.43 National population size 0.092 0.018 <0.001 Long-term national population trend 0.012 0.061 0.84 Black lines/points represent the relationships between relative effect size and each characteristic, from a model in which all five characteristics were included. In all, 95% prediction intervals around each relationship are shown by the shaded grey bars. The grey and red points represent the sum of the predicted effect size and the model residual for each species—those in red are in the top 25% of model weight and thus had the strongest influence on the model.
Our study provides insights into broad-scale associations between paved road exposure and local bird abundance, and considers interspecific variation in these associations in relation to species characteristics. Of the 75 species we tested, 63% showed strongly significant variation in abundance with increasing road exposure, with 53% of these exhibiting reduced abundance. When major and minor roads were analysed separately, of the species with significant associations with major roads, 81% were negative. Finally, we found the effect sizes of road exposure to be more negative for rarer, smaller-bodied and migrant species.
Several smaller-scale studies have shown bird abundance to increase or decrease with proximity to roads
The influence of roadside habitat is particularly difficult to quantify here as, although we incorporated habitat in our models, it was not captured at high enough resolution to account for subtle changes in roadside areas. Roads can create a variety of edge habitat
Our finding of a significant positive relationship between national abundance and road exposure effect size could imply that rarer birds are more inclined to avoid roads. It is possible that roadside habitat is unattractive to rarer species, as their reduced national abundance is, in part, due to their reduced ability to thrive under human disturbance in general. This reduction in competition in areas of higher road exposure could then result in an increase in abundance of species that are more able to tolerate human disturbance and are therefore more common nationally. Smaller-bodied species and migrants may also be found in lower abundances around roads due to increased sensitivity to road-related disturbances such as noise.
As we did not find a significant link between abundance around roads and long-term national population trend, the broader outcome of this lower abundance of some species around roads is difficult to interpret. It could be that road areas act as a sink for these species, or that they are simply avoided by them, but that abundance in areas with lower road exposure has increased enough to stabilise the national population. However, it is important to note that our measures of long-term population trends only began in 1970. Although traffic volume in Great Britain has increased greatly in that time, the total road length has increased by less than 25%
Shifts in species assemblages in areas of high human disturbance have been identified in both urban
Compression of already-vulnerable species into shrinking pockets of low road density may increase future declines and extinctions in countries with high road densities. Our results showed that, for species in reduced abundance with increasing road exposure, this effect extended to a mean of 700 m from a road. Almost three-quarters (72%) of Great Britain’s land surface falls within 700 m of a road (Fig. Blue represents terrestrial protected areas and red represents areas of
We modelled count data from the UK BBS for 75 species in relation to the proximity of nearby roads, whilst also accounting for other potential predictors of bird abundance. In a second step, we then analysed these results with respect to a range of species-specific characteristics to identify predictors of associations between road exposure and bird abundance. We used ArcMap 10.5.1
We obtained bird count data from the UK BBS, a nationwide survey in which experienced volunteers walk two 1-km transects across a 1-km square, each transect being divided into 200-m sections. These transects mostly do not follow roads (64% of the transect sections used in this analysis did not follow a paved road along any part of them). We extracted counts from squares that had been surveyed every year from 2012 to 2014 inclusive. We then calculated the mean bird count for each 200-m transect section across that period, removing any species with a total mean annual count <100. We also extracted the dominant habitat type recorded for each transect section. Our final dataset contained counts from 19,709 transect sections in 2033 squares. Preparation of these data is detailed in
We obtained shapefiles for all road classes (major roads: motorways and A-roads; minor roads: B-, C- and D-roads) in Great Britain, as recorded in 2013. We then used kernel density estimation to calculate a measure of road exposure for the midpoint of every 200-m transect section, using the locations of all roads within a 5-km radius. We optimised the spatial scale of the relationship between the distance from road and road exposure, represented by the parameter
To account for factors other than road exposure that we expected to affect bird abundance, we calculated human population density, temperature and rainfall values for the midpoint of each transect section. We also calculated the following for 5-km buffers around each midpoint: tree cover density, proportion of arable land (as a proxy for yield) and the largest field area (as a proxy of agricultural intensity). For information on data sources and calculation of these data see
Our goal was to understand how bird abundance varies in relation to roads, and to identify the characteristics of species that best predict these associations. We therefore modelled counts of each species, as recorded on BBS transects, as a function of road exposure and other factors that we also expected to affect bird abundance (habitat (as recorded in the BBS), proportion of arable land, largest field area, human population density, temperature, rainfall and tree cover density). We ran Poisson GAMMs for each species separately, using the R package ‘mgcv’
We performed an additional analysis of species that showed significant associations with road exposure (without Bonferroni correction), incorporating major and minor road exposure in separate models. As there are fewer major roads, and fewer BBS squares near major roads (93% and 47% of transect sections were within 1000 and 100 m of a minor road, respectively, and 44% and 9% were within 1000 and 100 m of a major road, respectively), for this analysis, we selected species with total mean annual counts >1000, in a minimum of 100 BBS squares, and only used squares within 5 km of a major road.
Cooke et al.
To assess significance, we calculated confidence limits for each species as the effect size ± standard error multiplied by the appropriate
To test whether species characteristics were associated with different directions and magnitudes of road exposure effects on bird abundance, we modelled the relationships between the relative effect size of road exposure and five chosen characteristics: mean body mass, migratory tendency, an index of habitat specialisation, national population size and long-term national population trend (1970–2016). We extracted mean body masses from Robinson
Further information on research design is available in the
The authors would like to thank Dario Massimino, Rhys E. Green, Simon Gillings, Andrea Manica, William J. Sutherland and Eloy Revilla for their assistance with this study, Tom Finch for providing the agricultural yield estimates and Calum Maney for producing the information on the CBD Sixth National Reports. We also thank all the volunteer BTO fieldworkers. The BBS is jointly funded by the BTO, JNCC and RSPB. Stuart Newson is supported by the BTO’s Young Scientists’ Programme. Sophia C. Cooke is funded by the Natural Environment Research Council (RG81247).
S.C.C. did the analysis and the writing for this paper. A.B., P.F.D., S.E.N. and A.J. helped to shape the ideas, context and analysis of the project, and read and commented on draft versions of the paper. In addition, A.J. provided guidance on the statistics and assisted with the figures, and S.E.N. helped with the obtaining and sorting of the BBS data.
The data analysed in this study are available online through ‘Apollo, the University of Cambridge’s repository.
The codes used in this study are available online through ‘Apollo, the University of Cambridge’s repository.
The authors declare no competing interests.
Supplementary Information Peer Review Reporting Summary