Grassland connectivity by motor vehicles and grazing livestock


A. G. Auffret, Dept of Physical Geography and Quaternary Geology, Stockholm Univ., SE-10691 Stockholm, Sweden. E-mail:


In addition to habitat loss and fragmentation, agricultural change has led to a change in seed dispersal processes in the rural landscape through a loss of structural and functional connectivity. Here, human-mediated dispersal vectors are prevalent, and we explored whether the loss of connectivity via free-ranging livestock could be mitigated by the increase in roads and motor vehicles. We found that structurally, 39% of all valuable semi-natural grassland habitats in southern Sweden are adjacent to public road verges, which in the rural landscape are often considered to be suitable habitat for grassland species. Additionally, by collecting mud attached to cars and farming machinery and manure from livestock (cattle, horse, sheep) grazing semi-natural grassland pasture, we found that motor vehicles are also capable seed dispersers. A similar number of species were dispersed by both vectors, although the composition of samples was quite different. Motor vehicles dispersed more grassland specialists than invasive species, although in much lower abundances than did grazing livestock. Despite these differences, motor vehicles were found to be able to disperse species with the same kinds of dispersal traits as livestock. A high number of seeds, species and specialists in manure samples means that greater movement of livestock is desirable to increase functional grassland connectivity. However, effective management could improve the suitability of roadsides as grassland corridors and increase the availability of seeds for long-distance human-mediated dispersal via cars and tractors. Our results suggest that in many rural landscapes, connectivity by road networks could help mediate habitat loss and fragmentation of grasslands. However, such effects can be context dependent, and the connectivity provided by roads could have serious negative consequences in other regions.

From the local through to the international scale, habitat connectivity is desirable in facilitating seed dispersal for the conservation and survival of plant communities in the face of habitat fragmentation and climate change (Soons and Ozinga 2005, Nathan 2006, Lindenmayer et al. 2008). In open, terrestrial agricultural landscapes, long-distance and human-mediated dispersal are prevalent (Nathan et al. 2008, Wichmann et al. 2009, Auffret 2011), but the grasslands which contribute most to biodiversity are highly fragmented (Krauss et al. 2010), and their inhabitants dispersal limited (Stein et al. 2008). In these environments, it is therefore vital to explore the role of humans in main taining a structural (patches linked by similar habitat) and functional connectivity (by long-distance seed dispersal between patches).

In the European agricultural landscape, fragmentation of semi-natural grassland habitat is leading to biodiversity loss both now and in the future (Krauss et al. 2010). Before agricultural intensification, large pastures, vast common outlands and transhumance meant that grazing livestock could spread seeds within and between landscapes (Bruun and Fritzbøger 2002). Today, local and predicted future extinctions are related both to the loss of grassland habitat and the cessation of long-distance dispersal via animals (Ozinga et al. 2009, Purschke et al. 2012). Consequently, measures to increase the connectivity in the landscape are related to the functional connectivity provided by rotational grazing schemes (Auffret et al. 2012, Wagner et al. 2013), while structural connectivity is thought to be most effectively increased by the (re)creation of continuous habitat corridors (Hodgson et al. 2011a).

Such measures can be costly, and difficult to organise, especially at larger scales. It can therefore be useful to consider other ways in which habitats are, or can be connected. Motor vehicles collect seeds and seed-containing material on the tyres and body, dispersing them at multiple spatial scales (Taylor et al. 2012), as well as dragging seeds shorter distances in their airflow (von der Lippe et al. 2013). However, this vector is usually associated with the spread of invasive species (Lonsdale and Lane 1994, Veldman and Putz 2010). This could be because investigations are con centrated within the urban and suburban environment (Zwaenepoel et al. 2006, von der Lippe and Kowarik 2007). On the other hand, rural roadsides are considered as valuable and diverse habitats worldwide (Cousins 2006, Koyanagi et al. 2009, Spooner and Smallbone 2009, Zeng et al. 2011), also containing grassland species in the seed bank (Milberg and Persson 1994, Berge and Hestmark 1997). Furthermore, farming machinery in the rural setting can accumulate both grassland seeds and seed-containing mud (Strykstra et al. 1997, Mayer 2000), and may provide additional seed dispersal via motor vehicles, while being less confined to road systems. Roads have been found to function as dispersal corridors for plant species (Tikka et al. 2001), so perhaps in rural environments, grassy road verges could provide structural grassland connectivity, with motor-vehicles functionally accelerating the dispersal of target species between habitat patches.

In this study, we explore the novel possibility that the negative impacts of agricultural change might be mitigated by the increase of roads and motor vehicles, with the former maintaining structural and the latter functional connectivity between fragmented semi-natural grasslands. We compare endozoochorous seed dispersal by livestock with seed dispersal potential via attachment to motor- vehicles (hereafter termed ‘dispersal’). Both vectors are concerned with seeds dispersed within a material substrate in which germinable seed content can be examined, and from which seed density can be derived and compared. We use a trait-based approach to ask if motor-vehicles can disperse the kind of species and seeds once dispersed by more free-ranging livestock, and assess the grassland target contra invasive species dispersed by both vectors. We also explore the structural connectivity provided by roadsides between remaining fragments of semi-natural grassland in the whole of southern Sweden, asking if these key habitats are in fact already well connected, facilitating dispersal at multiple scales. Based on our results, we discuss whether grassland connectivity via road verges is helpful or not with regards to functional connectivity by seed dispersal.


Study area

The functional connectivity study was carried out in a 10 km2 agricultural landscape in the county of Södermanland in southern Sweden (58°54′N, 17°00′E). The region has a long history of anthropogenic influence. According to historical maps, most of the landscape was open or semi-open pasture or meadow in 1600, but by 1901, large areas of land had been transformed into arable fields. Today’s landscape consists of a mosaic of deciduous woodland on abandoned pastures, evergreen agroforestry, historical and newly recreated pastures, arable fields, small and linear habitats and settlements. We selected five (of eight) farms, which contained all of the seven remaining dry-mesic semi-natural grasslands in the landscape (mean size: 7.64 ha, range: 2.81–18.54 ha). Five of these grasslands bordered a public road, and the other two private roads. Spread across these seven semi-natural grasslands were seven horses, 31 cattle and 30 sheep during the year of the study. Most of the public roads in the area are present in the map from 1600, which today are a combination of asphalt (9.3 km) and well-maintained gravel roads (14.3 km), all bordered on both sides by grassy verges, running adjacent to all major habitats listed above. In addition to motor vehicles and grazing livestock, other potential long-distance dispersal vectors in the landscape include wild ungulates, especially roe deer Capreolus capreolus and Eurasian elk Alces alces, as well as smaller mammals such as hare Lepus europeaus. Motor vehicle data are not available at this scale, but the 211 smaller households in the area typically own one or two vehicles, with eight farms also owning tractors and other agricultural machinery.

Sample collection and seedling emergence method

Each of the seven semi-natural grasslands in the study area were visited once at the end of each month from May until September (five occasions) 2009, representing the whole of the outdoor grazing season. Manure samples were collected from each species of livestock present in each grassland, 2 l (usually 1–3 deposits) were collected for cattle and horses, and 1 l (several deposits) for sheep. Samples from motor vehicles were collected on the same dates, as well as once in April, before the animals were let outdoors. This allowed us to ‘clean’ the vehicles in order that later sample content could be compared more accurately between vectors for the remainder of the season. At the five farmyards, mud and dust which could be removed using the hands and a washing-up brush was collected from all vehicles which had been in use since the previous sampling date, up to a maximum of 2 l per vehicle. Vehicle samples were almost always smaller than this, but by collecting what was usually all mud from the vehicle, as well as a large volume of livestock manure, we aimed to get an accurate estimation of total dispersal by these vectors during the sampling period. Landowners were consulted to maximise availability of vehicles and grazing animals for sampling, but they were otherwise not instructed manage their vehicles or livestock differently. In total, 31 manure samples from the 68 livestock and 48 samples from 12 motor vehicles (seven tractors and five cars/vans) were collected.

After collection, manure samples were spread out in trays and sun-dried in a greenhouse, before being stored for two weeks in the dark at 4°C. Vehicle samples were dry on collection, and were stratified directly, with one (wet) exception, which was treated in the same way as the manure samples. Samples were kept separate, first concentrated by washing through a 0.2 mm mesh (Ter Heerdt et al. 1996, Wessels and Schwabe 2008) before being spread thinly on trays 50 × 30 cm, which had first been filled with a layer of organic potting soil and a layer of sterilised sand. Samples were greenhouse grown for 13 weeks (median [IQR] temperature: 16.5°C [14–18.5], daylight: 16 h [16:17]).

Samples were checked at least twice a week, when emerging seedlings were identified and removed, and those which could not be identified were transplanted and grown until identification was possible. Seedlings could not always be identified to the species level, and in these cases they were merged into groups (each group counted as one species). Seven trays containing only potting soil and sand were placed among the samples to control for seeds present in the compost and airborne seeds in the greenhouse, from which seven seedlings of four species emerged. The effect of seedlings not present in the original samples is therefore thought to be negligible.

Structural connectivity

The structural connectivity investigation was carried out using Quantum GIS 1.7.3 and ArcGIS 10 (data manipulation and visualisation) and PostGIS 1.53 with the extension PgRouting 1.03 (connectivity analysis). We used the Swedish government’s survey of semi-natural pastures and meadows 2002–2004 (TUVA database – < >) and the 2003 General map of Sweden (Översiktskartan) to test the connectivity of the former by the public roads from the latter. We focussed our study in the agricultural south of Sweden (Fig. 1), which contains 59 259 of the 68 723 grasslands in the TUVA database. We then removed those grasslands labelled as restorable or no longer managed, and those with no description, resulting in 40 407 grasslands (mean ± SD size 5.04 ± 16.49 ha, range 0.05–1124.96 ha). Finally we applied a 20 m buffer to all grasslands in order to merge grasslands which were only separated by a stream or small road, and to ensure that grasslands adjacent to the road would intersect the road layer in the General map. The first, fifth and tenth nearest neighbours (a) following road verges, (b) euclidean distance (all grasslands), and (c) euclidean distance (grasslands connected to the road network –Fig. 1) were calculated. The nearest neighbours were chosen arbitrarily, but represent networks of different sizes within the range of local rotational grazing networks (Auffret et al. 2012), the size of our landscape, and measured distances of zoochorous seed dispersal (Thomson et al. 2011). Edge to edge distance was chosen so that the shortest dispersal routes could be compared, either the Euclidean distance through the matrix or along road verges.

Figure 1.

Illustration of nearest neighbour definitions, where (a) is the nearest neighbour following road verges, (b) is the Euclidean nearest neighbour and (c) is the euclidean nearest neighbour of the grasslands connected to the road network. Inset shows study area for structural connectivity study (shaded) and functional connectivity study (star).

Data analysis

All statistical analyses were carried out using R 2.12.1 (R Development Core Team) with the additional packages car (Fox and Weisberg 2011), labdsv (Roberts 2012), nnet (Venables and Ripley 2002) and vegan (Oksanen et al. 2011). One sample of mud collected from a car where the volume was less than 0.01 l and from which no seedlings emerged was removed before analysis. Total species richness, as well as seedling and species density l−1 collected material were calculated. From the species emerging from the manure and mud samples, we identified which were invasive species in Sweden from the European Network on Invasive Alien Species (NOBANIS – < >), and which were local grassland specialists (Auffret and Cousins 2011). Sample composition was compared between manure and motor-vehicle samples using detrended correspondence analysis (DCA), and permutational multivariate analysis of variance using distance matrices (adonis) for both abundance (Chao distance) and presence–absence data (Raup–Crick distance). Species strongly associated to one of the two dispersal vectors were identified using indicator species analysis (IndVal –Dufrêne and Legendre 1997; modified by Roberts 2012). Monte-Carlo random sampling with 10 000 permutations tested differences in seedling and species density between months within vectors and between vectors within months.

We selected six seed traits which we considered important for dispersal through the landscape in space and time: seed bank (persistent or transient), seed mass (mg), seed morphology (hooked, appendaged or not appendaged), seed number, seed releasing height (m) and terminal velocity (m s−1). Trait data were extracted from the LEDA traitbase (Kleyer et al. 2008), and the available species pool was determined by taking all species found in the two grid cells covering the study area by the inventory for the county plant atlas (Rydberg and Wanntorp 2001). The ability of the above traits to predict dispersal via manure and motor vehicles was tested using multinomial logistic regressions and subsequent chi-square likelihood-ratio tests for each trait individually. Dispersal via ‘manure’, ‘motor vehicles’ and ‘both’ were compared to ‘not-dispersed’ species for the total species pool, and dispersal via ‘manure’ and ‘motor vehicles’ compared to ‘both’ for dispersed species only.


Functional connectivity

We counted a total of 44 411 seedlings of 149 species. These were divided into 31 793 seedlings of 108 species from our 31 manure samples (total volume 57.6 l, mean 538 ± 573 seedlings and 15 ± 7 species l−1), and 12 618 seedlings of 110 species from 48 motor-vehicle mud samples (total volume 35.1 l, mean 396 ± 596 seedlings and 55 ± 70 species l−1). Sixty-nine species were shared between the two vectors. Fifteen of 22 local grassland specialists identified for the area and two national invasive species were dispersed by grazing animals, and motor vehicles dispersed 8 and 4 specialists and invasives respectively. Generally speaking, seeds of grassland specialists were more abundant in the manure samples than the motor-vehicle samples, while the opposite was true for invasive species. Forty species, including 13 of the 15 grassland specialists, were identified as strongly associated with the manure samples. The most prominent indicators were grasses and rushes of the genera Poa, Agrostis, Luzula and Festuca, along with several species of Veronica. Among the 17 species significantly associated to motor-vehicle dispersal were several common ruderals such as Matricaria matricarioides, M. recutita and Plantago major.

Five-hundred and ninety species were available in the local species pool, meaning 25% of the available species emerged from our samples (˜ 18% carried by each vector). Eight dispersed species were not present in the species pool, though several of these are crop or garden plants, which were not registered in the inventories for the plant atlas. These species were dispersed by both livestock and motor vehicles. Supplementary material Appendix 1 contains a full list of dispersed species, with invasives, specialists, characteristic species and those not present in the plant atlas all labelled.

Density and total richness of manure and motor vehicle samples was around the same order, and no significant differences were detected in species density between vectors (Fig. 2). Only in July were the manure samples more seedling-dense (p = 0.02) than the motor vehicle samples, whereas in May, the vehicle samples were more seedling-dense (p = 0.04). Within vectors, species and seedling density of manure samples were highest in the mid-late season samples. Species and seedling density of motor vehicle samples were more steady through the season.

Figure 2.

Density (boxplots) and total species richness per month (line) of (a) manure samples and (b) motor vehicle samples throughout the grazing season. Boxes represent the upper and lower quartiles, thick lines show the median, and the whiskers indicate the range of the dataset without outliers. Outliers (open circles) are where an observation falls outside the quartiles +/– (1.5 × the interquartile range). Letters in common represent no significant difference of species density (a, b) and seedling density (x, z) within vector, while * shows significant differences between vectors for a particular month from the Monte-Carlo analysis.

The DCA plot (Fig. 3) shows a clear separation of the vehicle and manure samples along the first DCA axis. This is supported by the adonis tests for both total seed content and species presence (F = 29.6, R2= 0.28, p = 0.001, and F = 59.3, R2= 0.44, p = 0.001 respectively). The multi nomial logistic regression showed that dispersal traits did to some extent explain how seeds were and were not dispersed. Considering the entire species pool (Table 1), both motor vehicles and livestock were found to preferentially disperse small, persistent seeds, and those from low-growing (releasing) plants. Comparing the two vectors (dispersed species only –Table 2), species which set a lot of seed were more likely to be found in the motor vehicle samples than in the manure. There was also a difference in seed size, with motor vehicle samples more likely to contain the largest seeds, livestock smaller seeds, with the smallest seeds more likely to be dispersed by both vectors.

Figure 3.

DCA plot comparing seedling composition for manure and motor vehicle samples.

Table 1. Results of multinomial logistic regression comparing species traits dispersed in livestock manure and by motor vehicles with those in the available species pool, where species not dispersed by either vector form the base group for comparison. A positive coefficient indicates a positive effect, and a negative coefficient a negative effect. A coefficient further from zero indicates a stronger effect. Significant traits are shown in bold.
 ManureMotor vehiclesBoth  
 CoefficientSECoefficientSECoefficientSEChi squarep
Seed bank persistence 2.12 0.74 2.06 0.74 11.93 73.35 62.16 < 0.001
Seed mass −0.03 0.026 −0.01 0.01 −0.18 0.07 26 < 0.001
Seed morphology      7.860.25
– appendaged−0.360.37−0.140.370.340.28  
– hooked−−0.471.07  
Seed number< –0.001< 0.001< –0.001< 0.001< –0.001< 0.0014.40.22
Seed release height −0.84 0.45 −0.06 0.05 −0.1 0.07 14 0.003
Seed terminal velocity−−
Table 2. Results of multinomial logistic regression comparing species traits dispersed in livestock manure and by motor vehicles, where species dispersed by both vectors form the base group for comparison. A positive coefficient indicates a positive effect, and a negative coefficient a negative effect. A coefficient further from zero indicates a stronger effect. Significant traits are shown in bold.
 ManureMotor vehicles  
 CoefficientSECoefficientSEChi squarep
Seed bank persistence−9.6166.03−9.6666.034.930.084
Seed mass 0.07 0.05 0.1 0.05 7.61 0.02
Seed morphology    7.260.12
– appendaged−0.690.44−0.480.43  
– hooked0.261.441.781.15  
Seed number < –0.001 < 0.001 < 0.001 < 0.001 6.87 0.03
Seed release height−0.670.540.030.083.380.18
Seed terminal velocity0.

Structural connectivity

From a total of 30 950 valuable grasslands in southern Sweden, 12 276 (39.66%) were adjacent to the public road network. The mean Euclidean distance of first nearest neighbours for all grasslands were 414 ± 605 m, and 107 (3.4%) have a nearest neighbour closer than 100 m away. Considering only grasslands adjacent to the road network, the mean distances were 846 ± 1035 m (Euclidean) and 1777 ± 8415 m (along roads) to the nearest grassland. Results were highly skewed, and mean nearest neighbour distances were rather higher than the medians, which are displayed in Table 3.

Table 3. Distances between valuable grassland habitats in southern Sweden and their 1st, 5th and 10th nearest neighbours according to the measurement methods used. Network grasslands are those which are within 20 m of a public road.
 Nearest neighbour median (IQR) 
 Crow distance – all grasslands (n = 30 950)Crow distance – network grasslands (n = 12 276)Distance along roads – network grasslands (n = 12 059*)Time (median at 50 km h−1)
1183 m (62–512)473 m (127–1191)1189 m (540–2140)1′26″
51476 m (932–2261)2728 m (1850–3934)4856 m (3411–6772)5′50″
102501 m (1752–3507)4353 m (3231–5811)7507 m (5659–9880)9′01″


We have found that semi-natural grasslands fragmented by land-use change are in fact well structurally connected by the rural road network, with motor vehicles providing potential functional connectivity via human-mediated long-distance seed dispersal. Roadsides provide structural connectivity between 12 276 grasslands in southern Sweden (39%). Motor vehicles were identified as capable seed dispersers, sharing well over half their dispersed species with livestock, including several specialist species, and dispersed species with traits similar to those dispersed by livestock without acting as a driver for invasions.

The need for such connectivity in these habitats is clear. Only 3% of all valuable semi-natural grasslands in southern Sweden were within 100 Euclidean metres of another, which is the distance reasonably covered by most wind-dispersed species (Schleicher et al. 2011), and livestock are largely confined to their farms for the entire season and unable to fulfil their seed dispersal potential. Road verges have been identified as dispersal corridors for both plant species (Tikka et al. 2001) and insects (Saarinen et al. 2005), and therefore the connection of fragmented grasslands by road verges is potentially important for local and regional scale dispersal of organisms between populations, as well as large-scale migration to track climatic change. Distances between grasslands along roads are around double the Euclidean distances, but the grassy, often species-rich (Cousins 2006) habitat along rural road verges makes this a more likely route for successful dispersal than through the arable or forest matrix. Not all roads have grassy verges (e.g. bridges), but in general we believe our structural connectivity distances between grasslands to be overestimated. Private roads, along with restored pastures and field boundaries can contain grassland communities (Smart et al. 2002, Török et al. 2011), providing potential large stepping-stones and further linear corridors between semi-natural grasslands.

The identification of motor vehicles as potential dispersers of grassland and grassland specialist species in the rural landscape may be significant. In contrast to the slow, unassisted dispersal of grassland plants along corridors (van Dorp et al. 1997), seeds need just to remain attached to motor vehicles for 10 min to be potentially dispersed within the range any of several nearby grasslands. Such distances fall well within detachment distances measured by Taylor et al. (2012). This scenario is more comparable to the long-distance dispersal potential of endozoochory by free-ranging livestock. The species dispersed were, however, quite different between the two different vectors. Despite almost half the dispersed species being shared by both vectors, the DCA (Fig. 3) and adonis tests showed that the composition of samples differed significantly in terms of both total content and species presence. A similar number of species emerged from a smaller volume of motor vehicle than from manure samples, but only half as many grassland specialists, and these in much lower concentrations. Further, the indicator species analysis signalled that where dispersed species did not overlap, they were often typical grassland species in the case of the manure samples, and common ruderals in the case of motor vehicles. We did not, however, find evidence that motor vehicles (or livestock) act as a driver of plant invasions. All four of the invasive species found in this study were introduced to Sweden and present in our study region before the invention of motor vehicles (Rydberg and Wanntorp 2001, NOBANIS – < >). Instead, the high species richness is probably due to the more heterogeneous nature of road verges and their adjacent habitat, meaning that both grassland specialists and generalists were present, as well as the ruderal and pioneer species characteristic of species dispersed by motor vehicles in the urban and sub-urban landscapes (Zwaenepoel et al. 2006, von der Lippe and Kowarik 2008).

Not only were motor vehicles identified as potential dispersers of grassland species, but we also found that they are able to transport species with the similar dispersal traits as livestock do endozoochorously (Table 1). The results from the multinomial trait analysis agreed with previous studies in the literature, whereby small and persistent seeds are dispersed endozoochorously (Pakeman et al. 2002, Bruun and Poschlod 2006) and by motor vehicles (Zwaenepoel et al. 2006, von der Lippe and Kowarik 2012). In addition, both vectors preferentially dispersed low- growing species, reflecting the relatively short vegetation in grassland pastures and road verges. No relationship between dispersal via either vector was found for terminal velocity or presence of appendages, indicating that both are able to disperse seeds apparently specialised to other dispersal vectors. This was also supported by the fact that a quarter of the total species pool in the entire landscape emerged from our samples. Comparing vectors (Table 2), the preferential dispersal of only seed-abundant species by motor vehicles indicates that this vector has the potential to disperse the same species as grazing animals, as long as the seeds are available for attachment to the vehicle. The more stochastic nature of seed attachment to vehicles leads to more seed abundant species being present in these samples, at the expense of some of the less common grassland specialist species, and the different species composition of the samples reflected the different species compositions of road verges and semi-natural grasslands.

Despite containing grassland specialist species (Cousins 2006), road verges are not linear strips of species-rich semi-natural grassland. If motor vehicles are to provide effective functional connectivity between fragmented grasslands, grassland species must be abundant enough and produce enough seeds to allow regular attachment to motor vehicles. As well as being influenced by adjacent habitats (Cousins 2006), road verge populations are a legacy of former land use, with older roads generally more diverse than newer roads (Spooner and Smallbone 2009), although verges can become functionally effective and diverse after only twenty years (García-Palacios et al. 2011, Zeng et al. 2011). Road verges can be managed to maximise flowering and diversity by seed transfer and altering existing mowing regimes (Jantunen et al. 2007, Nordbakken et al. 2010), allowing for dispersal either directly or later via the seed bank.

Due to the sheer number of seeds and species dispersed endozoochorously (including many grassland specialists), plus the complementarity of endo- and epizoochory (Couvreur et al. 2005), rotational grazing networks should be maintained and expanded to ensure effective functional connectivity between fragmented patches. Unpredictable dispersal via motor vehicles cannot substitute this, but it is useful to consider the structural and functional connectivity potential provided by road systems. Endozoochorous dispersal is more dictated by phenology, with more species dispersed later in the summer (see also Bakker et al. 2008, Auffret et al. 2012) whereas motor vehicles provide a more steady stream of seeds and species, which could supplement dispersal by grazing networks. The differences in seed dispersal in terms of species and seasonality indicates that in a wider ecological context, the two vectors can provide complementary long-distance seed dispersal services.

Our study investigated which seeds could become attached to motor-vehicles, but did not consider the subsequent detachment of material. Motor vehicles have been found to deposit seeds at a relatively high concentration for a single dispersal vector (von der Lippe and Kowarik 2007), while also moving seeds in their slipstream (von der Lippe et al. 2013). Seeds have the ability to stay attached to vehicles for long periods and distances, and detachment is quite unpredictable, accelerated by precipitation and wet conditions (Zwaenepoel et al. 2006, Taylor et al. 2012), providing a further stochastic factor regarding dispersal via this vector. Further research into dispersal via motor vehicles is desirable, especially with regard to attachment and detachment patterns, as well as post-dispersal colonisation, which would contribute further to the understanding of the overall effectiveness of dispersal by motor vehicles (Schupp et al. 2010). Further, landscape information, such as how history, management, size and surroundings can influence both roadside plant communities and their potential to facilitate seed dispersal in different regional contexts.

We have for the first time compared motor vehicles and grazing livestock as two long-distance dispersal vectors in the same landscape with the same species pool assessing the potential functional connectivity provided. Linking functional traits to dispersal vectors can be valuable, but human-mediated dispersal has been previously ignored (Thomson et al. 2010). It is often non-standard vectors which provide rare and important long-distance dispersal (Higgins et al. 2003), and it is therefore valuable to know which types of species disperse by human-mediated vectors, as they are likely non-standard but common vectors for many species.

Though unpredictable, we believe that by dispersing grassland species, motor vehicles have the potential to provide connectivity for otherwise dispersal limited species between fragmented grasslands. Connectivity is not the only answer to the problem of fragmentation, size and quality of habitat and the surrounding matrix are also major determinants of biodiversity (Hodgson et al. 2011b, Öckinger et al. 2012). Nevertheless, at least 39% of valuable grasslands in southern Sweden are connected by roads, and if the seeds and species are available, the opportunity exists that they can be delivered to, or at least closer to a target habitat. Managing these existing corridors to maximise target species availability and dispersal potential could be a valuable complement to other measures for increasing connectivity in rural landscapes. Our results suggest that in many rural landscapes, roads and road networks could be playing a part in mediating the habitat loss and frag mentation of grasslands via rare, human-mediated long-distance dispersal events locally and regionally. It is, however, vital to note that these effects are probably context dependent, and that the connectivity we identified as positive could pose serious threats to biodiversity through the spread of invasive species in other regions.


The research was financed by the Swedish Research Council for Environment, Agricultural Sciences and Spatial planning (FORMAS) and the cross-disciplinary project EkoKlim at Stockholm Univ. Our thanks go to I. Wärmé for assistance in the field and in the greenhouse, and P. Litfors and Tyresö Handelsträdgård for greenhouse space and sample care. R. Schmucki initiated the trait database and was a provided valuable technical expertise, while O. Eriksson gave helpful feedback on earlier versions of the manuscript. Finally, a big thank you to the friendly landowners in the Ludgo-Spelvik area, who facilitated successful sample collection.

Supplementary material (Appendix ECOG-00185 at < >). Appendix 1.