Ecol Res (2007) 22: 57–66 DOI 10.1007/s11284-006-0004-1
O R I GI N A L A R T IC L E
Fernando Ascensa˜o Æ Anto´nio Mira
Factors affecting culvert use by vertebrates along two stretches of road in southern Portugal
Received: 4 October 2005 / Accepted: 31 March 2006 / Published online: 21 July 2006 The Ecological Society of Japan 2006
Abstract A major target for environmental managers when trying to minimise the road-barrier effect on wildlife is to improve permeability to animal movements. Previous studies have demonstrated that drainage culverts are used by vertebrates, although knowledge of the main influencing factors remains limited. The use of 34 culverts from two roads in southern Portugal, differing in traffic volume, vehicle speeds and configuration, was evaluated by the analysis of terrestrial vertebrate footprint data (408 passage-operative days). Culvert crossings were related to various explanatory variables by means of canonical ordination techniques. We recorded 901 complete crossings, corresponding to an average of 2.2 crossings/culvert/operative day. Thirteen taxa were detected, all in more than one passage. Animal species included reptiles, small mammals, lagomorphs, carnivores and domestic dogs and cats. Our results suggest that fencing might have a funnelling effect, directing larger animals toward culverts. Also, vegetation covering culvert entrances seems to have a positive effect, particularly on genets; longer passages with entrances far from the pavement were, apparently, avoided by smaller animals; a lower number of crossings was detected on passages with detritus pits; the closest passages to urban areas are more often used by domestic species; forest-living species favour passages with low, open land cover nearby; and smaller species, like lagomorphs and small mammals, appear to use more culverts near the pavement, which probably reflects the importance of road verges as refuges for these species. Although not used by all species present in the study area, constructing numerous passages of different sizes without detritus pits and which are distributed along roads might be an important step in mitigating road fragmentation effects on animal populations.
F. Ascensa˜o (&) Æ A. Mira Unidade de Biologia da Conservac¸a˜o, Departamento de Biologia, Universidade de E´vora, Nu´cleo da Mitra, Apartado 94, 7002-554 E´vora, Portugal E-mail:
[email protected]
Keywords Road ecology Æ Culvert use Æ Conservation Æ Canonical ordination Æ Vertebrates
Introduction Civilisation has always interacted with the landscape, exploiting natural resources and settling human communities throughout the world. This has promoted the fragmentation of the landscape on a vast scale. Landscape fragmentation is now recognised as one of the major threats to the conservation of biodiversity (Saunders et al. 1991; Forman 1995). One major human agent of habitat fragmentation is the ever-increasing network of roads worldwide (Forman et al. 2002), which can be harmful to various faunal groups, including invertebrates (e.g. Haskell 2001), amphibians (e.g. Carr and Fahrig 2001), reptiles (e.g. Gibbs and Shriver 2002), birds (e.g. Kuitunen et al. 1998) and mammals (e.g. Philcox et al. 1999). Roads and traffic can act as barriers which interfere with animal movement and reduce population connectivity, diminishing gene flow and disrupting sink and source population dynamics, thereby, promoting inbreeding and loss of genetic diversity (Ferreras 2001). The resultant isolation might increase the extinction risk of local populations, due to stochastic effects (van der Zande et al. 1980; Saunders et al. 1991; Fahrig and Merrian 1994; Cooper and Walters 2002). Roads also promote high animal–vehicle collision rates (e.g. Hodson 1960; Oxley et al. 1974; Ferreras et al. 1992; Philcox et al. 1999; Gibbs and Shriver 2002; Taylor and Goldingay 2004), one of humankind’s most visible road-related sources of impact upon wildlife. Previous research has provided clear evidence that a wide range of species may be affected by this threat, which can lead to local population extinction (Fahrig et al. 1995; Jones 2000). Additionally, roads may affect animal behaviour, resulting in complete reluctance to cross a road, leading to a barrier effect (Mader 1984; Goosem 2001). For a more detailed review of road ef-
58
fects on wildlife, see Forman and Alexander (1998), Spellerberg (1998), Trombulak and Frissell (2000) and Forman et al. (2002). Therefore, a major target for conservationists, when trying to minimise the negative effects of roads on wildlife, is to improve road permeability to animal movements, in order to promote or establish contact between the populations on either side of the road. As described by Forman (1995), a solution may lie in porous roadbeds with tunnels and underpasses. Specific wildlife passageways have been designed and incorporated into road planning and mitigation programs in numerous countries (Clevenger and Waltho 2000; Goosem et al. 2001; Keller et al. 2002; Cain et al. 2003). However, specifically designed wildlife passages might be too expensive to build and implement on a large scale. Although not designed for animal passage, previous studies have demonstrated that drainage culverts are regularly used by vertebrates to cross roads (e.g. Yanes et al. 1995; Rodrı´ guez et al. 1996; Cain et al. 2003; Taylor and Goldingay 2003; Ng et al. 2004; Mata et al. 2005). These structures are widespread on every road and have a relatively low cost when compared to specific wildlife passages. Nevertheless, knowledge about which factors primarily drive the use of culverts by vertebrates remains limited, particularly in Mediterranean environments (Yanes et al. 1995; Rodrı´ guez et al. 1996; Mata et al. 2005). The present study is the first one conducted in Portugal which addresses the issue of vertebrate use of culverts for road crossing. The main goals of the current study were: (1) to evaluate the extent to which drainage culverts are used by vertebrates for road crossing; (2) to determine which taxa/species use them, along two different types of road; and (3) to determine the influence of culvert characteristics and landscape descriptors on vertebrate use of these structures.
Methods Study area This study took place along two roads, 35 km apart from each other, located in Alentejo, in southern Portugal (Fig. 1). The climate is Mediterranean with an average annual rainfall of 625 mm, an average minimum daily temperature of 9.9C and an average maximum daily temperature of 21.8C. The altitude lies between 100 m and 400 m above sea level. The vicinity of both roads consists of the following: characteristic Mediterranean agro-forest; cork and holm oak tree stands (Quercus suber and Q. rotundifolia), hereafter, referred to as ‘‘Montado’’; open land as pastures, meadows or extensive agriculture (cereal, fodder); and olive groves. The two roads have dissimilar traffic volumes, vehicle speeds and configurations. The first road section, hereafter, identified as M370, is a 16-km national road section with very low traffic density (no official data
available; personal observations). It is 6 m in width, without paved verges, and with two lanes along its entire length. There are no exclusionary fences along the verges. There are two small villages along this section: Escoural, in the extreme south and Giesteira in the middle of the studied section. The second road section, hereafter identified as IP2, is a 30-km stretch of road, having an annual average daily traffic volume of 5,121 vehicles per day, of which, approximately 11% (562 vehicles) circulate during the night (EP, Portuguese National Road Service, unpublished report from 2001). The paved section, including verges, varies between 12 m and 20 m in width, with one lane for each direction or, where a steep slope is present, two lanes directed up the slope. The verges are bordered, along their entire length, by a 1.5-m high exclusionary fence. The fence surrounds all studied culverts but does not enclose the culvert entrance on either side. There are three urban areas along this section of road: two villages, S.Manc¸os and M.Trigo, and one small city, Portel. Human presence is greater along this road than along M370, but is restricted primarily to the vicinity of the three urban areas. There are eight overpasses and eight underpasses for traffic crossing along this section, which connect unpaved trails. Traffic-crossing passages and border trails allowed us to reach both sides of the studied culverts without having to step onto the main road. Albeit the large number of passages present along the studied road sections (52 culverts on M370 and 74 on IP2), for most culverts, we could not easily gain access to either entrance, particularly on IP2. Hence, culvert selection was, basically, dependent upon the degree of accessibility. Culvert use Culvert use was evaluated by analysing animal footprints found inside the passages. Thin layers of marble dust, less than 1-cm thick and 60–100-cm wide, were placed inside each culvert, on both sides, near each entrance. Thirty-four culverts (17 on each road) were sampled regularly. Sampling was carried out during three seasons (spring, summer and autumn of 2004). Each sampling period consisted of four operative days. Each operative day represented two sampling days, given that tracks were checked every second day. We considered an operative day when tracks could be read clearly, i.e. with no damage from water or wind flow. Presence–absence data for each species on each sampling day were used to estimate crossing rates (CR). For each culvert, species’ CRs were computed as the ratio between the number of days with species’ tracks and the number of operative days. CRs vary between zero when no tracks are detected and one when tracks from one given taxon are detected on all operative days. This avoided problems of pseudo-replication associated with the repeated counting of tracks on the same passage.
59
Fig. 1 Map of the studied area. Dark solid lines represent the two road sections, M370 and IP2. White circles represent the studied culverts. Dark thin lines represent the road network and white squares represent the most important localities. Grey-scale areas stand for the altitude (varying from 100 m to 400 m above sea level)
Identification to species level was restricted to mammals. Other vertebrate classes also were considered: reptiles—lizards and ophidians; small mammals not including hedgehogs (Erinaceus europaeus)—rats, mice, voles and shrews; lagomorphs—rabbits (Orictolagus cuniculus) and hares (Lepus granatensis); and wild carnivores. The carnivore community in the study area includes the following species: foxes (Vulpes vulpes), weasels (Mustela nivalis), polecats (Mustela putorius), stone martens (Martes foina), Eurasian badgers (Meles meles), otters (Lutra lutra), common genets (Genetta genetta), Egyptian mongooses (Herpestes ichneumon) and wildcats (Felis silvestris). Amphibians were excluded from analysis, due to the low number of tracks recorded. Explanatory variables We evaluated the relationship between culvert use and 10 explanatory variables related to attributes of passage design and location (Table 1). We defined two classes for culvert width (CW); class 1: under 0.8 m and class 2: between 0.8 m and 1 m. Culvert openness (CO) was derived from the formula: culvert cross-section area (CCS) divided by culvert length (CL) (adapted from Yanes et al. 1995). Detritus pits (DP) are designed to collect debris that otherwise may block culverts and are located at the entrance, across the full width of the culverts. None of the passages studied had detritus pits at both entrances. Land cover surrounding each culvert was assessed for a radius of 300 m by means of orthophotomap analysis (data provided by Associac¸a˜o de Municı´ pios do Distrito de E´vora), corrected by means of fieldwork observations. Although several classes of land cover could be identified along both road sections, we grouped them into four main categories (open, olive groves, Montado
with shrubs and Montado without shrubs), which covered about 95% of the total circle areas (Appendices 1 and 2). The density of the crossing structures along the road corridor may be one confounding variable when the main aim is to identify the major characteristics of culverts and the landscape that influence culvert use. As stated by Clevenger and Waltho (2005), high crossing rates might be a consequence of the nonexistence of nearby passages, and lower use might be a result of there being numerous nearby passages. Although these authors applied this correction to specifically designed macro fauna passages, we also incorporated it in our study. For each culvert, we defined the ‘‘probability of use,’’ taking into account its distance to adjacent culverts, underpasses and overpasses. It was defined as the average distance to the nearest passage on each side (DST). We assume that passages have a higher probability of being used when they are more isolated; that is, distanced from other crossing structures. The road type (RT), i.e. M370 or IP2, was considered to be a nominal variable. This variable represents differences in road attributes, including traffic volumes and the presence/absence of exclusionary fences. The season (spring, summer and autumn) also was considered to be a nominal variable. All GIS data were managed using ArcView GIS 3.2 (ESRI, 1999). Data analysis We assumed that each studied culvert was independent of all the others. However, this assumption may not always have been justified, because some passages were less than 600 m apart (n=9 on M370 and n=5 on IP2), and, therefore, shared adjoining landscapes. Also, different passages might have been visited by the same animals. The mean distance between studied culverts was 802 m on M370 and 1,717 m on IP2. Differences in culvert use, related to categorical variables, were evaluated by means of Mann–Whitney U tests (MW) or Kruskal–Wallis H tests (KW) (Zar 1999). For continuous variables, we performed Spearman rank correlations (rs). Non-parametric tests were preferred because of small sample sizes. For multivariate analysis, we used canonical ordination techniques. Detrended correspondence analysis (DCA) was performed to evaluate the grouping of culverts from the two roads, with the down-weighting of rare species and detrending by segment (Jongman et al. 1995). A direct gradient analysis (canonical correspondence analysis: CCA) was then executed, with crossing rate data and all spatial explanatory variables. Again, the down-weighting of rare species and detrending by segment was selected. We used the manual process for forward selection of environmental variables. The significance of each variable was evaluated by Monte Carlo tests (with 499 permutations). Variables with an
60 Table 1 Explanatory variables for the 34 culverts (17 on each road) surveyed in this study, including code, variable type and median and range values, for the stretches of road designated M370 and IP2
Design attributes
Location attributes
Variable
Code
Type
M370
IP2
Culvert length (m) Culvert width (m) (number of passages per width class) Culvert cross-section area (m2) Culvert openness (·100) (cm) Detritus pits (number of passages with attribute) Distance of the culvert from the pavement (m) Vegetation covering the passage entrances (number of passages) Distance to urban areas (m) Open land (%) Montado with shrubs (%) Montado without shrubs (%) Olive groves (%) Average distance to nearest passages (m)
CL CW
Numerical Ordinal
CCS CO DP
Numerical Numerical Ordinal
9.0 (8.0–25.0) class 1, n=9; class 2, n=8 0.7 (0.24–1.20) 7.11 (2.67–13.33) n=2
19.5 (8.0–37.0) class 1, n=11; class 2, n=6 0.5 (0.38–0.79) 2.72 (1.37–13.09) n=3
DFP
Numerical
2.5 (1.0–7.0)
2.5 (1.7–15.0)
VC
Ordinal
n=14
n=7
DU OP MSh MNT OL DST
Numerical Numerical Numerical Numerical Numerical Numerical
901 (90–4,116) 10.06 (0.00–20.89) 14.10 (0.00–28.12) 0.00 (0.00–12.52) 0.00 (0.00–15.18) 291 (110–760)
2,310 (657–4,265) 13.44 (0.00–28.12) 0.00 (0.00–14.09) 1.17 (0.00–28.12) 0.00 (0.00–25.50) 285 (154–666)
estimated p value lower than 0.1 were included in the model. Ordination axes were interpreted using intraset correlations that allow for inferences with respect to the relative importance of each variable in predicting community composition (ter Braak 1986). Except for seasonal variation analysis, the CR was derived from the three sample periods (spring, summer and fall). CANOCO for Windows, version 4.5 (ter Braak and Sˇmilauer 2002) was used for DCA and CCA.
M370 culvert was similar to the average of the nine culverts along IP2; that is, tracks were detected on more than one-third of the total operative days, and reached a maximum average CR among carnivores for both roads, denoting frequent and continuous use of this passage. Also, fox tracks were detected in only two M370 passages, versus 10 passages along IP2. In contrast, hedgehogs exhibited generalised use of passages on M370 (n=12), but on IP2, they crossed using only three culverts.
Results
Seasonal variation
Culvert usage
Vertebrate crossings were mostly observed during the spring (Table 3). Culvert use did not differ significantly between seasons for any animal group, except for small mammals (KW: K=14.784; p
