- Email: [email protected]
Do unpaved, low-traffic roads affect bird communities?
Acta Oecologica 71 (2016) 14e21
Contents lists available at ScienceDirect
Acta Oecologica journal homepage: www.elsevier.com/locate/actoec
Original article
Do unpaved, low-traffic roads affect bird communities? Christos Mammides a, b, *, Constantinos Kounnamas a, Eben Goodale c, Costas Kadis a a
Nature Conservation Unit, Frederick University, P.O. Box 24729, 1303, Nicosia, Cyprus Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Mengla, Yunnan Province, 666303, China c College of Forestry, Guangxi University, Nanning, Guangxi Province, 530005, China b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 November 2015 Received in revised form 9 January 2016 Accepted 10 January 2016 Available online xxx
Unpaved, low traffic roads are often assumed to have minimal effects on biodiversity. To explore this assertion, we sampled the bird communities in fifteen randomly selected sites in Pafos Forest, Cyprus and used multiple regression to quantify the effects of such roads on the total species richness. Moreover, we classified birds according to their migratory status and their global population trends, and tested each category separately. Besides the total length of unpaved roads, we also tested: a. the site's habitat diversity, b. the coefficient of variation in habitat (patch) size, c. the distance to the nearest agricultural field, and d. the human population size of the nearest village. We measured our variables at six different distances from the bird point-count locations. We found a strong negative relationship between the total bird richness and the total length of unpaved roads. The human population size of the nearest village also had a negative effect. Habitat diversity was positively related to species richness. When the categories were tested, we found that the passage migrants were influenced more by the road network while resident breeders were influenced by habitat diversity. Species with increasing and stable populations were only marginally affected by the variables tested, but the effect of road networks on species with decreasing populations was large. We conclude that unpaved and sporadically used roads can have detrimental effects on the bird communities, especially on vulnerable species. We propose that actions are taken to limit the extent of road networks within protected areas, especially in sites designated for their rich avifauna, such as Pafos Forest, where several of the affected species are species of European and global importance. © 2016 Elsevier Masson SAS. All rights reserved.
Keywords: Forest management Bird richness Unpaved road networks Habitat heterogeneity Protected areas
1. Introduction Protected areas are considered vital for the conservation of species worldwide (Gaston et al., 2008). A key aspect to the successful management of protected areas is understanding how human-induced habitat changes affect the biodiversity of a region. The type and level of anthropogenic activities within protected areas depend on a range of factors, such as the geographic location of the site and the socio-economic factors that operate there. For example, in the European Union, sites located in southern European countries are more likely to be affected by hunting (Tsiafouli et al., 2013). Yet, inarguably one of the most common and widespread activity, evident in almost all areas, is the establishment and the use of road networks (Trombulak and Frissell, 2000; Husby and Husby,
* Corresponding author. Nature Conservation Unit, Frederick University, P.O. Box 24729, 1303, Nicosia, Cyprus. E-mail address: [email protected] (C. Mammides). http://dx.doi.org/10.1016/j.actao.2016.01.004 1146-609X/© 2016 Elsevier Masson SAS. All rights reserved.
2014). For that reason, along with the fact that new road networks are being planned and constructed on a continuing basis (Laurance and Balmford, 2013), assessing the effects of road networks on species is of high conservation importance. We know from the plethora of the studies available in the field of road ecology that the effects of roads vary according to the area and species studied (Husby and Husby, 2014; Mammides et al., 2015). Morelli et al. (2014), in the review of ninety-two studies published over the last four decades, showed that roads and other transportation related structures can have positive effects on certain species in some cases. For example, Helldin and Seiler (2003) found that in south-central Sweden roadsides were associated with higher habitat heterogeneity, which benefited forest birds. On the other hand, many studies have shown that roads often have significant negative effects on many species, by resulting for example in habitat loss and fragmentation or increased disturbance due to higher presence of humans (Forman and Alexander, 1998; Fahrig and Rytwinski, 2009; Kociolek et al., 2011). The effects can be direct, such as increased mortality due to collisions with
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
vehicles, or indirect, such as behavioral changes resulting from higher levels of noise and pollution (Brotons and Herrando, 2001; pez et al., 2010). The mechanisms by which roads Benítez-Lo affect bird communities can be multileveled and complex. For example, they can affect the reproductive cycles of birds by reducing nest success, and at the same time reduce the fitness of individuals by interfering with foraging calls and prey detection (reviewed in Kociolek et al., 2011). The impacts of roads may also have different time scales, ranging from mortality caused during road construction, to long-term community changes due to the alteration of the physical and the chemical environment (Trombulak and Frissell, 2000). The exact effects of roads also depend on the extent and the type of the road network within an area (Laurance and Balmford, 2013; Robson and Blouin-Demers, 2013). So far, most of the studies on the effects of roads have focused on high-traffic paved roads (S alek et al., 2010), presumably because their effects can be potentially more adverse (Laurance and Balmford, 2013). Apart from resulting in habitat loss, highly-used paved roads are also likely to affect species in additional ways, as they make an area accessible to a higher number of people, they result in higher levels of pollution and may be more difficult to be crossed by animals (Robson and Blouin-Demers, 2013), thus affecting their movement to a larger extent. Unpaved roads on the other hand, especially low-traffic roads, are unlikely to affect species in these ways, so they are often assumed to have weaker effects, caused mostly by habitat loss and increased edge effects (Ortega and Capen, 2002; Laurance et al., 2004). In this study we test and quantify the extent to which low-traffic unpaved road networks affect the bird communities in Pafos Forest, Cyprus, using multiple linear regression modeling. In conjunction to the road networks we also include in our models habitat heterogeneity, which has been shown to have a strong positive effect on the bird communities in protected areas in Cyprus (Mammides et al., 2015). Furthermore, we examine the effects of human presence (based on the population size of the nearest village) and distance to the nearest agricultural field, along with a number of other landscape metrics that have been shown to have a significant effect on birds in similar ecosystems (Schindler et al., 2013). Following the results of previous studies (Hargis et al., 1998; Morelli et al., 2013; Schindler et al., 2013), which have shown that the performance of landscape metrics depends on the scale measured, we assess the effects of our predictor variables at six different distances from the bird point-count locations: 100 m, 250 m, 500 m, 750 m, 1 km and 1.25 km. We chose Pafos Forest as the study area because it is a nature reserve of high ornithological importance, which has been designated as a Special Protection Area (SPA), i.e. a Natura 2000 site (Hadjipanayiotou, 2012) and has a particularly dense network of unpaved roads. Many of these roads, which are maintained on a regular basis, are closed to the public and are mostly used by the Department of Forests for management purposes, such as for access in case of fire and for patrolling. 2. Material and methods 2.1. Study area Pafos Forest is located in the northwestern part of Cyprus (Fig. 1), an island of high ornithological importance (Iezekiel et al., 2004). More than 370 species of birds have been recorded in Cyprus so far, of which about 85% are non-resident species (Iezekiel et al., 2004). Pafos Forest covers an area of 602.68 km2 (Hadjipanayiotou, 2012). Its altitude ranges from near sea level to 1352 m, receiving an annual mean precipitation of 1060 mm (Hadjipanayiotou, 2012). The site is covered mostly by Pinus brutia forest, with its understory
15
consisting mainly of thermo-Mediterranean sclerophyllous vegetation such as Quercus alnifolia, Arbutus andrachne and Cistus spp. (Hadjipanayiotou, 2012). The high ecological significance of the forest stems from the fact that it is an important area for birds (Iezekiel et al., 2004), supporting populations of 22 species listed in the Annex I of the European Union's Birds Directive, including the four endemic species and subspecies of Cyprus (Oenanthe cypriaca, Sylvia melanothorax, Certhia brachydactyla dorothea and Parus ater cypriotes; Hadjipanayiotou, 2012). Additionally, the forest is one of the most important areas in Cyprus for nesting raptors such as Hieraaetus fasciatus and Accipiter gentilis. According to Hellicar et al. (2014), a major threat to Pafos Forest's biodiversity is habitat degradation, resulting in fragmentation, caused by the establishment of firebreaks and the dense network of roads. The majority of the roads are low-traffic unpaved roads built by the authorities for better accessing and managing the forest. For the purposes of this study we sampled twenty sites, all located next to existing roads. Due to the forest's steep topography it was not feasible to select sites at varying distances from the road. The sites were spread throughout the west-central part of the forest (Fig. 1). Areas in the very far north and the east were avoided for logistic purposes since, due to the forest's large size and topography, it was not practical to also cover those within the amount of time available. We selected and surveyed twenty sites randomly using Hawth's Analysis Tools (Beyer, 2004) in ArcMap (Version 9.3.1), by first dividing the chosen part of the forest into 2 km 2 km grid cells to ensure adequate distance between the sites and to avoid spatial autocorrelation. Note, however, that during the analysis we had to disregard five of the sites because they were located less than 1.25 km from the forest's boundary; data on the road network, habitat diversity, and the rest of the landscape metrics were only available for areas within the forest boundaries and consequently these sites had to be excluded due to missing information. Our results are therefore based on fifteen sites. Results from analyses that included more sites, but with smaller radii around them, were qualitatively similar, but explained less of the variation. 2.2. Bird censuses We conducted bird censuses, along the roadside, during the spring migration of 2012, between early March to mid-May, using 10-min point counts (Bibby et al., 1992). We decided to sample birds during the migratory season, as many species in the forest (including several of the high conservation importance) are migratory (Iezekiel et al., 2004). We visited each site six times to ensure adequate sampling. We carried out all surveys between sunrise and 10:00 AM. We rotated the order by which we surveyed the sites to make sure that no site was constantly sampled in early or late morning. During the surveys we identified and recorded all species seen or heard within a distance of 100 m from the sampling point. To account for species that were possibly missed during the surveys, we estimated species richness at each site using the firstorder Jackknife richness estimator (Magurran, 1998), using the specpool function in “vegan” package in R (R Core Team, 2015). We used the species migratory status list provided to us by BirdLife Cyprus, to classify birds into “Resident Breeders” and “Passage Migrants”. We also classified species according to their global population trends, using the corresponding data from BirdLife International, which are available online (BirdLife International, 2014). 2.3. Sites characteristics Using ArcMap, we established six nested circular buffer zones of
16
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Fig. 1. Map of Pafos Forest showing the fifteen point-count locations, the network of unpaved roads (gray lines) and the different habitats within the forest's boundaries. For illustrative purposes, the road networks within four of the study sites is shown, representing the full gradient of road density recorded within the 1 km buffer zone.
incremental radii (i.e. 100 m, 250 m, 500 m, 750 m, 1 km and 1.25 km) around each point-count location. Within each buffer zone we measured the extent of the biotic and the anthropogenic variables of interest. 2.3.1. Biotic variables We estimated habitat heterogeneity at each site using spatial data provided by the Department of Forests of Cyprus. The Department is the responsible authority for managing Pafos Forest and for that purpose has surveyed and mapped the forest's vegetation at the 0.002 km2 level. The data provided consisted of 9 vegetation categories: 1. stands of P. brutia, 2. stands of Cedrus brevifoliae, 3. stands of Q. alnifolia or other species, 4. mixed stands of P. brutia and Q. alnifolia, 5. mixed stands of P. brutia and C. brevifoliae, 6. mixed stands of P. brutia and other species, 7. mixed stands of Q. alnifolia and other species, 8. mixed stands of other species, and 9. riparian vegetation. Using FRAGSTATS v.4.2.1 (McGarigal et al., 2012), we measured the ShannoneWeiner Habitat Diversity Index (SHDI). Additionally, following the findings of Schindler et al. (2013), who studied the relationships between various landscape metrics and the species richness of multiple taxa (including birds), in a mosaic of similar habitats in a protected forested area in Greece, we measured, also using FRAGSTATS: the patch density (PD), the area-weighted mean patch fractal dimension (FRAC_AM), the area-weight patch shape (SHAPE_AM) and the coefficient of variation in patch area (AREA_CV) within each buffer zone (McGarigal et al., 2012). 2.3.2. Anthropogenic variables The total length of unpaved roads, within each buffer zone, was
measured using ArcMap and the forest's road map, provided by the Department of Forests. Small stretches of paved roads, representing 11.58% of the total roads recorded in our largest buffer zones (1.25 km), found at six of the sites, were excluded from the analyses. Results of an analysis that included these roads in measuring total road length were qualitatively similar. The distance to the nearest agricultural field was calculated using the CORINE Land Cover 2006 (CLC) map, a European wide land cover map, depicting, among other land cover categories, the spatial distribution of the agricultural land in each participating country (EEA, 2014). For the purposes of this analysis, we used ArcMap to measure the distance between each study site and the nearest agricultural field. Data on the human population size of each nearby village were extracted from the 2011 census data, made available online by the Statistical Service of the Republic of Cyprus (Statistical Service, 2014). 2.4. Data analysis We used bivariate correlation tests to assess collinearity between the independent variables. PD, FRAC_AM, and SHAPE_AM were found to be highly correlated with habitat diversity (>0.65) so they were subsequently removed from the analyses. The effects of the rest of the five explanatory variables (i.e. total road length, habitat diversity, distance to the nearest agricultural field, human population size, and variation in patch size), were tested using linear multiple regression models, and following a multimodel inference approach (Burnham and Anderson, 2002). All possible subset models for each bird category tested were ran, using the “MuMIn” package (Barton, 2014). Models were ranked in ascending
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
order, according to their corrected Akaike Information Criterion (AICc) value. All models with a DAICc of less than 7 were selected (Burnham and Anderson, 2002; Nakagawa and Freckleton, 2011). To assess the effect of each predictor on the response variables the model.avg function in the “MuMIn” package was used to calculate the averaged unstandardized and standardized regression coefficients, using the zero-average method (Burnham and Anderson, 2002; Nakagawa and Freckleton, 2011). The relative effect of each predictor on each bird category tested was evaluated using the standardized regression coefficients. To assess the predictive power of our models, we calculated the adjusted R2 values for each category tested within each buffer zone, based on the corresponding full model (i.e. when all independent variables were included). To show trends of the different variables individually, we present graphs of partial residual graphs. This is offered as a visual, we did not do further analysis on partial residuals (Darlington and Smulders, 2001). To check whether our data were spatial autocorrelated or nonnormally distributed, we tested the residuals of all the selected models, using the Moran's I test in the “ape” package and the ShapiroeWilk normality test. To confirm that there was no collinearity in our models, we used the vif function in the “car” package to check the variance inflation factors in each of the selected models (Fox and Weisberg, 2011). To check whether there was a phylogenetic signal in the two species' traits we used to classify birds (i.e. migratory status and population trends), we performed an Abouheif test (Pavoine and Ricotta, 2013) on the phylogenetic trees we obtained from www.birdtree.org (Jetz et al., 2012), using the abouheif.moran function in the “adephylo” package in R. Since www.birdtree.org does not provide a consensus tree, we downloaded 1000 trees for each trait, sourced from Hackett (all species; 10,000 trees with 9993 OTUs each; Jetz et al., 2012) and tested each tree for phylogenetic signal. We then examined the distribution of the resulting p-values to assess the strength of the signal. 3. Results A total of sixty-four species were recorded in all twenty sites; fifty-eight of those were recorded in the 15 sites that were included in the analysis. Species richness in those 15 sites ranged from 13 to 26 (mean 19.67, sd 3.15). Fourteen species were only recorded at one site while three species, one of them being the endemic subspecies P. ater cypriotes, were found in all 15. For a complete list of all the species recorded at each site please see Appendix I. In total, within our largest buffer zones, we recorded 287.68 km of roads, of which 33.31 km were paved roads (found at six sites) and were excluded from the analysis. On average, 19.18 km of unpaved roads were recorded at each site within the 1.25 km buffer, ranging from 8.08 km to 29.33 km (standard deviation ¼ 5.46). The average habitat diversity, within the largest buffer zones, was 0.51 (sd ¼ 0.20), ranging from 0.23 to 0.92. The coefficient of variation in path size ranged from 163.9 to 339.8 (mean ¼ 238.3, sd ¼ 52.64). The shortest distance to the nearest agricultural field was 2.23 km and the longest 8.44 km (mean ¼ 4.36, sd ¼ 1.91). The average human population size was 145 (sd ¼ 159) ranging from 0 (in two villages which are no longer inhabited) to 473. In total, 576 regression models were selected for the six bird categories tested at the six buffer zone radii (i.e. total species richness [91 models], richness of passage migrants [77], richness of resident breeders [106], richness of species with increasing [110], stable [97] and decreasing global population trends [95]). The results of the spatial autocorrelation and normality tests showed that none of our models violated the assumptions of independence and normality (all p-values were higher than 0.05). There was no collinearity either; all inflation factors recorded were less than 4.00.
17
The phylogenetic tests on the two traits showed that there was a weak signal in the population status (11.5% of the p-values were <0.05) and a stronger signal in the migratory status (0.44% of the pvalues were <0.05). The effects of the variables tested varied according to the size of the buffer zone used. Fig. 2 shows the bivariate Pearson correlation coefficients between total bird species richness and the five explanatory variables for each of the buffer zone sizes. The correlation between species richness and habitat diversity and the coefficient of variation in patch size appeared to be scale dependent, something which is also reflected in the averaged regression coefficients of the corresponding linear models (Table 1). The effect of habitat diversity, although overall positive, peaked at 500 m and then dropped as the buffer size increases. Interestingly, the effect of variation in patch size started as positive (albeit very weak) and then turned negative as the buffer size increased (Table 1). The effect of road networks, however, was consistently negative, and so was the effect of the human population size of the nearest village (Table 1; Fig. 2). The effect of the distance to the nearest agricultural field was also negative, but overall very weak, except in the smallest buffer zone where the variable ranked third in importance, with an effect size similar to that of the road networks (Table 1). The R2 values for the total species richness ranged from 0.55 (at 250 m) to 0.73 (at 1 km), showing that our models explain a satisfactory amount of the observed variance in bird species richness. At the 1 km, the most important variable was human population size, which had a strong negative effect, followed by the road networks, which also affected species negatively (Fig. 3). Habitat diversity, the third important variable, had a positive effect, while the effect of variation in patch area was negative (Fig. 3). The effect of distance to the nearest agricultural field was negligible (Table 1). When passage migrants were tested the effects of the variables also varied according to the buffer zone tested and so did the R2 values. Overall, the R2 values were lower than those of the total species richness, ranging from 0.31 at 100 m to 0.58 at 500 m (Table 2). At 500 m, road networks was the most important variable, showing strong negative effect, while variation in patch size was the second most important variable, with a positive effect. The effect of the rest of the three variables was weak (Table 2). Resident breeders appear to be affected by road networks only when the smallest buffer zone is considered (Table 2), but the R2 value of that model was particularly low (0.26). The highest R2 value for resident breeders (0.52) was also at 500 m. At that level, resident breeders were mostly positively affected by the habitat diversity, and to a lesser degree by the distance to the agriculture (positive effect) and the variation in patch size (negative effect). The R2 values for the species with increasing and stable populations were consistently very low (Table 3) showing that other variables are affecting the presence of these species. On the other hand, the richness of the species with decreasing population size appears to be explained, at least partly, by the variables included in the model. R2 values ranged from 0.22 at 250 m to 0.57 at 1 km. At the 1 km distance, the most important factor was road networks followed by human population size both of which have a negative effect (Table 3). Habitat diversity had a positive effect albeit weaker, while the rest two variables had no effects. 4. Discussion In the case of Pafos Forest, unpaved low-traffic road networks have consistently negative effects on the bird communities, especially on vulnerable species such as species with decreasing population status. Several of these species are of high conservation importance, such as the endemic Cyprus Warbler (S. melanothorax), and are included in the Annex I of the European Union's Birds
18
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Fig. 2. Pearson correlation coefficients between overall species richness (first-order Jackknife estimator) and each of the five explanatory variables at each of the radii used in the analyses.
Table 1 Results of the linear regression models for each buffer size tested, for the overall bird species richness, showing the explanatory variables' averaged unstandardized (b) and standardized (beta) coefficients, along with the variables' relative importance (i), calculated as the percentage of the selected models they appear in. The R2 values of the full models are also shown. RD: Road Density, SHDI: Habitat Diversity, DA: Distance to the nearest agricultural field, PS: Human population size of the nearest village, PA: Coefficient of variation in patch size. 100 m (R2 ¼ 0.64)
RD SHDI DA PS PA
250 m (R2 ¼ 0.55)
500 m (R2 ¼ 0.60)
750 m (R2 ¼ 0.65)
1000 m (R2 ¼ 0.73)
1250 m (R2 ¼ 0.62)
b
beta
i
B
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
10.23 2.03 0.85 1.53 0.00
0.29 0.14 0.29 0.60 0.02
0.48 0.31 0.42 0.87 0.11
3.34 5.31 0.08 0.71 0.01
0.28 0.31 0.03 0.28 0.06
0.58 0.61 0.12 0.51 0.21
0.53 7.66 0.01 0.89 0.01
0.11 0.44 0.00 0.35 0.06
0.36 0.83 0.11 0.74 0.19
0.67 5.08 0.02 1.24 0.00
0.32 0.29 0.01 0.49 0.01
0.75 0.69 0.09 0.92 0.08
0.50 3.42 0.01 1.74 0.02
0.37 0.18 0.00 0.69 0.14
0.79 0.46 0.03 1.00 0.42
0.25 2.62 0.03 1.73 0.01
0.26 0.15 0.01 0.68 0.14
0.59 0.38 0.07 1.00 0.42
Directive, which lists all species requiring special conservation attention (Donald et al., 2007). The road network within the forest is particularly dense, with an average of 6.5 km of roads per km2, compared to the mean density of roads in the Natura 2000 sites in Cyprus, which is 1.94 km/km2 (Zomeni and Vogiatzakis, 2014). 88.5% percent of the roads in our sites, as measured within the largest buffer zones, were unpaved low-traffic roads. Many of them are closed to the public, a few even with locked roadbarriers. Considering this fact, and also that paved roads (excluded from the analysis) only represented a small percentage of the total recorded road network, we argue that it is highly unlikely that the recorded effects on the birds were caused by either traffic, noise pollution or direct mortality associated with road usage, as found in other studies which looked at the effects of roads (Brotons and Herrando, 2001; Peris and Pescador, 2004; pez et al., 2010). Mitchell et al., 2006; Benítez-Lo It is probable that the adverse effects we found are due to habitat loss, and associated changes in the vegetation structure and composition, such as increased canopy openness, which can alter the surrounding microhabitat (Delgado et al., 2007). Although the percentage of the habitat lost to unpaved roads, within the sampled areas, only ranged from about 1.3 to 4.8%, it is likely that the long stretches of roads and their resulting edge effects have converted those areas into less suitable habitats for certain species (Ortega and Capen, 2002; Laurance et al., 2004). Another plausible explanation is that some bird species may avoid unpaved roads, even if there is suitable habitat along the roadside, because the area is considered to be of lower quality, for example due to higher perceived predation risk (Whittingham et al., 2006; Zanette et al., 2011). It's worth noting that habitat heterogeneity was also an important determinant of the bird species richness, especially of
migrant birds. We need to acknowledge that our phylogenetic analysis showed that there is possibly a signal in the migratory status, and thus the pattern may be driven by the particular group of species tested. Nonetheless, for conservation purposes, each species has an intrinsic value. A previous study in protected areas in Cyprus (Mammides et al., 2015) and studies elsewhere have shown that bird presence relates positively to habitat heterogeneity because more diverse habitats usually provide more foraging and nesting opportunities for a wider range of species (MacArthur and MacArthur, 1961; Santos et al., 2002; Brotons et al., 2005; Loehle et al., 2005; Mitchell et al., 2006; Gil-Tena et al., 2007; Santos et al., 2010). For the Pafos Forest, this result suggests that the areas of the reserve which are dominated by only one tree species, usually P. brutia, support lower numbers of birds species. Such areas are more likely to have impoverished bird communities if they also contain a dense network of roads. Somewhat surprisingly, distance to the nearest agricultural field was found to have a negative or negligible effect on the bird categories tested, although it was originally hypothesized that shorter distances would probably mean higher human pressure and disturbance and therefore less species. It turns out from our analysis that this is not the case. This is probably due to the type of the agriculture practiced in the region, which is small-scale, low intensity, and usually involving a variety of crops that has been shown to be beneficial especially to farmland birds, which most probably comprised part of our sample (Pino et al., 2000; Su arezSeoane et al., 2002). The population size of the nearest village, on the other hand, had consistently a negative effect (except on few of the resident breeders models) indicating that higher human presence relates to lower bird species richness. The effects of variation in patch size varied widely both in terms of the bird categories tested and the buffer sizes examined. This
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
19
Fig. 3. Partial residual plots for each of the five explanatory variables, showing their effect on the total species richness (first-order Jackknife estimator) when the rest of the variables are held constant. Plots are based on the averaged model within the 1 km buffer zone.
metric is a measure of the spread of the patch size relative to the mean patch size at any given site. When tested using the total bird communities it had a weak positive effect within the smallest buffers, but as the size increased the effect became negative and grew stronger. This suggests that at small scales variation in patch size may indicate heterogeneous habitat, which may translate into higher niche availability, but at a larger scales the metric may represent discontinuous habitat, possibly of lower quality.
In general, our results concur with previous findings showing that the performance of various landscape metrics depend on the scale measured. Although in previous studies it was found that landscape metrics, similar to the ones we used in this study, correlate strongest with bird indices when measured at short distances (e.g. 150e250 m; Morelli et al., 2013; Schindler et al., 2013), in our case results were strongest at larger distances (i.e. within a radius of 500 m or 1 km). This suggests that the effects of the
20
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Table 2 Results of the linear regression models for each buffer size tested, for the species richness of passage migrants (A) and residents breeders (B). Statistics and abbreviations as Table 1. A
100 m (R2 ¼ 0.31)
250 m (R2 ¼ 0.45)
500 m (R2 ¼ 0.58)
750 m (R2 ¼ 0.42)
1000 m (R2 ¼ 0.44)
1250 m (R2 ¼ 0.39)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
4.85 0.25 0.20 0.26 0.00
0.22 0.03 0.11 0.17 0.02
0.43 0.15 0.25 0.35 0.12
3.73 0.36 0.06 0.06 0.02
0.50 0.03 0.03 0.04 0.23
0.82 0.16 0.16 0.16 0.49
1.43 1.12 0.01 0.02 0.02
0.48 0.10 0.01 0.01 0.36
0.91 0.30 0.07 0.09 0.72
0.87 0.62 0.00 0.13 0.00
0.67 0.06 0.00 0.08 0.01
0.99 0.22 0.11 0.29 0.10
0.58 0.04 0.04 0.42 0.00
0.69 0.00 0.02 0.27 0.00
1.00 0.08 0.14 0.65 0.09
0.38 0.04 0.06 0.48 0.00
0.66 0.00 0.04 0.30 0.00
0.98 0.08 0.15 0.67 0.08
B
100 m (R2 ¼ 0.26)
RD SHDI DA PS PA
250 m (R2 ¼ 0.47)
500 m (R2 ¼ 0.52)
750 m (R2 ¼ 0.15)
1000 m (R2 ¼ 0.18)
1250 m (R2 ¼ 0.04)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
2.41 1.62 0.04 0.00 0.00
0.19 0.33 0.04 0.00 0.08
0.44 0.60 0.19 0.13 0.23
0.25 3.51 0.07 0.02 0.00
0.06 0.59 0.07 0.02 0.04
0.22 0.91 0.23 0.12 0.16
0.02 3.78 0.30 0.04 0.00
0.01 0.63 0.29 0.04 0.12
0.09 0.91 0.60 0.18 0.31
0.01 1.60 0.19 0.01 0.00
0.01 0.26 0.18 0.02 0.01
0.12 0.54 0.43 0.15 0.12
0.01 0.84 0.21 0.03 0.00
0.01 0.13 0.21 0.03 0.04
0.13 0.33 0.44 0.16 0.18
0.00 0.31 0.17 0.03 0.00
0.01 0.05 0.17 0.04 0.06
0.13 0.20 0.39 0.17 0.21
Table 3 Results of the linear regression models for each buffer size tested, for the species richness of birds with increasing (C), stable (D) and decreasing population status (E). Statistics and abbreviations as Table 1. C
100 m (R2 ¼ 0.21)
250 m (R2 ¼ 0.01)
500 m (R2 ¼ 0.17)
750 m (R2 ¼ 0.07)
1000 m (R2 ¼ 0.15)
1250 m (R2 ¼ 0.14)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
0.52 0.22 0.12 0.29 0.02
0.03 0.03 0.09 0.24 0.25
0.15 0.16 0.24 0.50 0.54
0.33 0.51 0.18 0.23 0.00
0.06 0.06 0.13 0.19 0.05
0.19 0.20 0.32 0.40 0.19
0.12 0.34 0.17 0.26 0.00
0.05 0.04 0.12 0.21 0.12
0.20 0.16 0.31 0.44 0.30
0.02 0.15 0.19 0.32 0.00
0.02 0.02 0.14 0.26 0.02
0.12 0.12 0.32 0.51 0.12
0.02 0.02 0.19 0.36 0.00
0.04 0.00 0.13 0.29 0.06
0.16 0.09 0.31 0.55 0.22
0.03 0.03 0.16 0.33 0.00
0.06 0.00 0.12 0.27 0.04
0.22 0.10 0.29 0.53 0.17
D
100 m (R2 ¼ 0.28)
250 m (R2 ¼ 0.32)
500 m (R2 ¼ 0.21)
750 m (R2 ¼ 0.13)
1000 m (R2 ¼ 0.00)
1250 m (R2 ¼ 0.20)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
0.76 0.06 0.01 0.07 0.00
0.05 0.01 0.01 0.07 0.01
0.20 0.14 0.14 0.23 0.14
0.13 0.37 0.01 0.07 0.00
0.03 0.05 0.01 0.07 0.01
0.16 0.20 0.14 0.22 0.14
0.21 0.26 0.01 0.06 0.00
0.11 0.04 0.01 0.05 0.07
0.30 0.17 0.13 0.19 0.22
0.19 0.14 0.02 0.06 0.00
0.22 0.02 0.01 0.06 0.01
0.48 0.14 0.12 0.20 0.11
0.08 0.18 0.03 0.08 0.00
0.15 0.02 0.03 0.08 0.03
0.37 0.14 0.15 0.24 0.16
0.03 0.11 0.02 0.08 0.00
0.08 0.01 0.02 0.08 0.03
0.24 0.14 0.15 0.24 0.15
E
100 m (R2 ¼ 0.48)
RD SHDI DA PS PA
250 m (R2 ¼ 0.22)
500 m (R2 ¼ 0.45)
750 m (R2 ¼ 0.53)
1000 m (R2 ¼ 0.57)
1250 m (R2 ¼ 0.54)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
3.87 0.62 0.36 0.39 0.00
0.16 0.06 0.18 0.22 0.02
0.32 0.19 0.32 0.41 0.12
1.05 1.34 0.11 0.21 0.01
0.13 0.12 0.06 0.12 0.10
0.32 0.30 0.19 0.29 0.28
1.16 5.12 0.01 0.04 0.00
0.36 0.43 0.00 0.03 0.00
0.73 0.80 0.08 0.13 0.11
0.71 4.28 0.01 0.09 0.00
0.49 0.35 0.01 0.05 0.00
0.90 0.74 0.08 0.20 0.07
0.56 1.68 0.00 0.44 0.00
0.61 0.13 0.00 0.25 0.00
0.94 0.36 0.12 0.63 0.07
0.43 0.49 0.01 0.59 0.00
0.67 0.04 0.01 0.34 0.01
0.96 0.15 0.11 0.76 0.08
landscape metrics are often area specific and depend also on the bird guilds examined. At this juncture we should also note limitations of the study. In particular, we have only used species richness as a proxy of bird diversity. Further studies are needed to explore whether the effects examined remain the same when other diversity metrics are tested, and other analyses would be useful to incorporate the abundance and density of birds. In conclusion, the message of our study is twofold. Firstly, it shows that even unpaved and infrequently used roads can have negative effects on birds, highlighting also the importance of habitat heterogeneity for the presence of birds in protected areas. Based on these findings, we suggest that management authorities consider possible options for reducing the adverse impacts of roads on the avifauna in forested areas, for example by restoring the vegetation in roads which are not essential for accessing and managing the forest. Secondly, our study contributes to the growing literature which emphasizes that the effects of various landscape metrics on bird communities can be scale and guild specific.
Authors' contributions Christos Mammides, Constantinos Kounnamas and Costas Kadis designed the study. Christos Mammides and Constantinos Kounnamas collected the data. Christos Mammides, in collaboration with Eben Goodale, analyzed the data. All authors were involved in writing the manuscript. Acknowledgments We are thankful to BirdLife Cyprus, the Department of Forests, the Department of Land and Surveys, and the Statistical Service for making available the data necessary for this analysis, and to Salindra Kasun Dayananda for his help with the first figure. We are also thankful to the associate editor of the journal, Patrick Doncaster, and two anonymous reviewers for their highly constructive suggestions. The project was funded by the Research Promotion Foundation (Protocol No. PENEK/SUPPORT/0308/42), with cofunding from the European Union's Structural Funds.
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.actao.2016.01.004. References Barton, K., 2014. MuMIn: Multi-model Inference. R Package Version 1.12.1. http:// CRAN.R-project.org/package¼MuMIn. pez, A., Alkemade, R., Verweij, P.A., 2010. The impacts of roads and other Benítez-Lo infrastructure on mammal and bird populations: a meta-analysis. Biol. Conserv. 143, 1307e1316. Beyer, H.L., 2004. In: Hawth's Analysis Tools for ArcGIS. Available at: http://www. spatialecology.com/htools. Bibby, C., Burgess, N., Hill, D., 1992. Bird census techniques. In: British Trust for Ornithology and the Royal Society for the Protection of Birds. Academic Press, London. BirdLife International, 2014. IUCN Red List for Birds. Downloaded from. http://www. birdlife.org. Brotons, L., Herrando, S., 2001. Reduced bird occurrence in pine forest fragments associated with road proximity in a Mediterranean agricultural area. Landsc. Urban Plan. 57, 77e89. Brotons, L., Herrando, S., Martin, J.L., 2005. Bird assemblages in forest fragments within Mediterranean mosaics created by wild fires. Landsc. Ecol. 19, 663e675. Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: a Practical Information-theoretic Approach. Springer Science & Business Media, New York, USA. Darlington, R.B., Smulders, T.V., 2001. Problems with residual analysis. Anim. Behav. 62, 599e602. valo, J.R., Ferna ndez-Palacios, J.M., 2007. Edge effects Delgado, J.D., Arroyo, N.L., Are of roads on temperature, light, canopy cover, and canopy height in laurel and pine forests (Tenerife, Canary Islands). Landsc. Urban Plan. 81, 328e340. Donald, P.F., Sanderson, F.J., Burfield, I.J., Bierman, S.M., Gregory, R.D., Waliczky, Z., 2007. International conservation policy delivers benefits for birds in Europe. Science 317, 810e813. EEA, 2014. CORINE Land Cover 2006 (CLC 2006). http://www.eea.europa.eu/dataand-maps/data/corine-land-cover-2000-clc2000-seamless-vector-database (accessed Oct 2014). Fahrig, L., Rytwinski, T., 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecol. Soc. 14, 21. Forman, R.T.T., Alexander, L.E., 1998. Roads and their major ecological effects. Annu. Rev. Ecol. Syst. 29, 207eC202. Fox, J., Weisberg, S., 2011. An {R} Companion to Applied Regression, second ed. Sage, Thousand Oaks CA http://socserv.socsci.mcmaster.ca/jfox/Books/Companion. ~o n, G., 2008. The ecological Gaston, K.J., Jackson, S.F., Cantú-Salazar, L., Cruz-Pin performance of protected areas. Annu. Rev. Ecol. Evol. Syst. 39, 93e113. Gil-Tena, A., Saura, S., Brotons, L., 2007. Effects of forest composition and structure on bird species richness in a Mediterranean context: Implications for forest ecosystem management. For. Ecol. Manag. 242, 470e476. Hadjipanayiotou, C., 2012. DASOS PAFOU e CY2000006. In: Natura 2000-Standard Data Form. European Environment Agency. Hargis, C.D., Bissonette, J.A., David, J.L., 1998. The behavior of landscape metrics commonly used in the study of habitat fragmentation. Landsc. Ecol. 13, 167e186. Helldin, J.O., Seiler, A., 2003. Effects of roads on the abundance of birds in Swedish forest and farmland. In: Proceedings of the IENE Conference on “Habitat Fragmentation Due to Transport Infrastructure”, 13e14 November 2003, Brussels. Hellicar, M.A., Anastasi, V., Beton, D., Snape, R., 2014. Important Bird Areas of Cyprus. BirdLife Cyprus. Cyprus, Nicosia. Husby, A., Husby, M., 2014. Interspecific analysis of vehicle avoidance behavior in birds. Behav. Ecol. 25, 504e508. Iezekiel, S., Makris, C., Antoniou, A., 2004. In: Cyprus, B. (Ed.), Important bird areas of European Union importance in Cyprus. Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K., Mooers, A.O., 2012. The global diversity of birds in space and time. Nature 491, 444e448. Kociolek, A.V., Clevenger, A.P., St Clair, C.C., Proppe, D.S., 2011. Effects of road networks on bird populations. Conserv. Biol. 25, 241e249. Laurance, W.F., Balmford, A., 2013. Land use: a global map for road building. Nature 495, 308e309. Laurance, S.G., Stouffer, P.C., Laurance, W.F., 2004. Effects of road clearings on movement patterns of understory rainforest birds in central Amazonia. Conserv. Biol. 18, 1099e1109.
21
Loehle, C., Wigley, T.B., Rutzmoser, S., Gerwin, J.A., Keyser, P.D., Lancia, R.A., Reynolds, C.J., Thill, R.E., Weih, R., White Jr., D., Wood, P.B., 2005. Managed forest landscape structure and avian species richness in the southeastern US. For. Ecol. Manag. 214, 279e293. MacArthur, R.H., MacArthur, J.W., 1961. On bird species diversity. Ecology 42, 594e598. Magurran, A.E., 1998. Ecological Diversity and its Measurement. Princeton University Press, Princeton, NJ, USA. Mammides, C., Kadis, C., Coulson, T., 2015. The effects of road networks and habitat heterogeneity on the species richness of birds in Natura 2000 sites in Cyprus. Landsc. Ecol. 30, 67e75. McGarigal, K., Cushman, S.A., Ene, E., 2012. FRAGSTATS v4: Spatial Pattern Analysis Program for Categorical and Continuous Maps. Computer software program produced by the authors at the University of Massachusetts, Amherst. Available at the following web site: http://www.umass.edu/landeco/research/fragstats/ fragstats.html. Mitchell, M.S., Rutzmoser, S.H., Wigley, T.B., Loehle, C., Gerwin, J.A., Keyser, P.D., Lancia, R.A., Perry, R.W., Reynolds, C.J., Thill, R.E., Weih, R., White, D., Wood, P.B., 2006. Relationships between avian richness and landscape structure at multiple scales using multiple landscapes. For. Ecol. Manag. 221, 155e169. Morelli, F., Pruscini, F., Santolini, R., Perna, P., Benedetti, Y., Sisti, D., 2013. Landscape heterogeneity metrics as indicators of bird diversity: determining the optimal spatial scales in different landscapes. Ecol. Indic. 34, 372e379. Morelli, F., Beim, M., Jerzak, L., Jones, D., Tryjanowski, P., 2014. Can roads, railways and related structures have positive effects on birds? e a review. Transp. Res. Part D Transp. Environ. 30, 21e31. Nakagawa, S., Freckleton, R.P., 2011. Model averaging, missing data and multiple imputation: a case study for behavioural ecology. Behav. Ecol. Sociobiol. 65, 103e116. Ortega, Y.K., Capen, D.E., 2002. Roads as edges: effects on birds in forested landscapes. For. Sci. 48, 381e390. Pavoine, S., Ricotta, C., 2013. Testing for phylogenetic signal in biological taitrs: the ubiquity of cross-product statistics. Evolution 67, 828e840. Peris, S.J., Pescador, M., 2004. Effects of traffic noise on paserine populations in Mediterranean wooded pastures. Appl. Acoust. 65, 357e366. Pino, J., Rod a, F., Ribas, J., Pons, X., 2000. Landscape structure and bird species richness: implications for conservation in rural areas between natural parks. Landsc. Urban Plan. 49, 35e48. R Core Team, 2015. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. Robson, L.E., Blouin-Demers, G., 2013. Eastern hognose snakes (Heterodon platirhinos) avoid crossing paved roads, but not unpaved roads. Copeia 2013, 507e511. , J., Zasadil, P., 2010. Edge effect of low-traffic forest roads on S alek, M., Svobodova bird communities in secondary production forests in central Europe. Landsc. Ecol. 25, 1113e1124. Santos, T., Tellería, J.L., Carbonell, R., 2002. Bird conservation in fragmented Mediterranean forests of Spain: effects of geographical location, habitat and landscape degradation. Biol. Conserv. 105, 113e125. Pe rez Tris, J., Carbonell, R., Tellería, J.L., 2010. Santos, T., Galarza, A., Ramírez, A., Vegetational versus topographical effects on forest bird communities: a test in the cantabrian mixed forest ecoregion (Spain). Ardeola 57, 285e302. Schindler, S., von Wehrden, H., Poirazidis, K., Wrbka, T., Kati, V., 2013. Multiscale performance of landscape metrics as indicators of species richness of plants, insects and vertebrates. Ecol. Indic. 31, 41e48. Statistical Service, 2014. Population Census 2001. Downloaded from: http://www. cystat.gov.cymof/cystat/statistics.nsf/index_en/index_en?OpenDocument. rez-Seoane, S., Osborne, P.E., Baudry, J., 2002. Responses of birds of different Sua biogeographic origins and habitat requirements to agricultural land abandonment in northern Spain. Biol. Conserv. 105, 333e344. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conserv. Biol. 14, 18e30. Tsiafouli, M.A., Apostolopoulou, E., Mazaris, A.D., Kallimanis, A.S., Drakou, E.G., Pantis, J.D., 2013. Human activities in natura 2000 sites: a highly diversified conservation network. Environ. Manag. 1e9. Whittingham, M.J., Devereux, C.L., Evans, A.D., Bradbury, R.B., 2006. Altering perceived predation risk and food availability: management prescriptions to benefit farmland birds on stubble fields. J. Appl. Ecol. 43, 640e650. Zanette, L.Y., White, A.F., Allen, M.C., Clinchy, M., 2011. Perceived predation risk reduces the number of offspring songbirds produce per year. Science 334, 1398e1401. Zomeni, M., Vogiatzakis, I., 2014. Roads and roadless areas in Cyprus: implications for the Natura 2000 network. J. Landsc. Ecol. 7, 75e90.
Contents lists available at ScienceDirect
Acta Oecologica journal homepage: www.elsevier.com/locate/actoec
Original article
Do unpaved, low-traffic roads affect bird communities? Christos Mammides a, b, *, Constantinos Kounnamas a, Eben Goodale c, Costas Kadis a a
Nature Conservation Unit, Frederick University, P.O. Box 24729, 1303, Nicosia, Cyprus Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences, Mengla, Yunnan Province, 666303, China c College of Forestry, Guangxi University, Nanning, Guangxi Province, 530005, China b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 30 November 2015 Received in revised form 9 January 2016 Accepted 10 January 2016 Available online xxx
Unpaved, low traffic roads are often assumed to have minimal effects on biodiversity. To explore this assertion, we sampled the bird communities in fifteen randomly selected sites in Pafos Forest, Cyprus and used multiple regression to quantify the effects of such roads on the total species richness. Moreover, we classified birds according to their migratory status and their global population trends, and tested each category separately. Besides the total length of unpaved roads, we also tested: a. the site's habitat diversity, b. the coefficient of variation in habitat (patch) size, c. the distance to the nearest agricultural field, and d. the human population size of the nearest village. We measured our variables at six different distances from the bird point-count locations. We found a strong negative relationship between the total bird richness and the total length of unpaved roads. The human population size of the nearest village also had a negative effect. Habitat diversity was positively related to species richness. When the categories were tested, we found that the passage migrants were influenced more by the road network while resident breeders were influenced by habitat diversity. Species with increasing and stable populations were only marginally affected by the variables tested, but the effect of road networks on species with decreasing populations was large. We conclude that unpaved and sporadically used roads can have detrimental effects on the bird communities, especially on vulnerable species. We propose that actions are taken to limit the extent of road networks within protected areas, especially in sites designated for their rich avifauna, such as Pafos Forest, where several of the affected species are species of European and global importance. © 2016 Elsevier Masson SAS. All rights reserved.
Keywords: Forest management Bird richness Unpaved road networks Habitat heterogeneity Protected areas
1. Introduction Protected areas are considered vital for the conservation of species worldwide (Gaston et al., 2008). A key aspect to the successful management of protected areas is understanding how human-induced habitat changes affect the biodiversity of a region. The type and level of anthropogenic activities within protected areas depend on a range of factors, such as the geographic location of the site and the socio-economic factors that operate there. For example, in the European Union, sites located in southern European countries are more likely to be affected by hunting (Tsiafouli et al., 2013). Yet, inarguably one of the most common and widespread activity, evident in almost all areas, is the establishment and the use of road networks (Trombulak and Frissell, 2000; Husby and Husby,
* Corresponding author. Nature Conservation Unit, Frederick University, P.O. Box 24729, 1303, Nicosia, Cyprus. E-mail address: [email protected] (C. Mammides). http://dx.doi.org/10.1016/j.actao.2016.01.004 1146-609X/© 2016 Elsevier Masson SAS. All rights reserved.
2014). For that reason, along with the fact that new road networks are being planned and constructed on a continuing basis (Laurance and Balmford, 2013), assessing the effects of road networks on species is of high conservation importance. We know from the plethora of the studies available in the field of road ecology that the effects of roads vary according to the area and species studied (Husby and Husby, 2014; Mammides et al., 2015). Morelli et al. (2014), in the review of ninety-two studies published over the last four decades, showed that roads and other transportation related structures can have positive effects on certain species in some cases. For example, Helldin and Seiler (2003) found that in south-central Sweden roadsides were associated with higher habitat heterogeneity, which benefited forest birds. On the other hand, many studies have shown that roads often have significant negative effects on many species, by resulting for example in habitat loss and fragmentation or increased disturbance due to higher presence of humans (Forman and Alexander, 1998; Fahrig and Rytwinski, 2009; Kociolek et al., 2011). The effects can be direct, such as increased mortality due to collisions with
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
vehicles, or indirect, such as behavioral changes resulting from higher levels of noise and pollution (Brotons and Herrando, 2001; pez et al., 2010). The mechanisms by which roads Benítez-Lo affect bird communities can be multileveled and complex. For example, they can affect the reproductive cycles of birds by reducing nest success, and at the same time reduce the fitness of individuals by interfering with foraging calls and prey detection (reviewed in Kociolek et al., 2011). The impacts of roads may also have different time scales, ranging from mortality caused during road construction, to long-term community changes due to the alteration of the physical and the chemical environment (Trombulak and Frissell, 2000). The exact effects of roads also depend on the extent and the type of the road network within an area (Laurance and Balmford, 2013; Robson and Blouin-Demers, 2013). So far, most of the studies on the effects of roads have focused on high-traffic paved roads (S alek et al., 2010), presumably because their effects can be potentially more adverse (Laurance and Balmford, 2013). Apart from resulting in habitat loss, highly-used paved roads are also likely to affect species in additional ways, as they make an area accessible to a higher number of people, they result in higher levels of pollution and may be more difficult to be crossed by animals (Robson and Blouin-Demers, 2013), thus affecting their movement to a larger extent. Unpaved roads on the other hand, especially low-traffic roads, are unlikely to affect species in these ways, so they are often assumed to have weaker effects, caused mostly by habitat loss and increased edge effects (Ortega and Capen, 2002; Laurance et al., 2004). In this study we test and quantify the extent to which low-traffic unpaved road networks affect the bird communities in Pafos Forest, Cyprus, using multiple linear regression modeling. In conjunction to the road networks we also include in our models habitat heterogeneity, which has been shown to have a strong positive effect on the bird communities in protected areas in Cyprus (Mammides et al., 2015). Furthermore, we examine the effects of human presence (based on the population size of the nearest village) and distance to the nearest agricultural field, along with a number of other landscape metrics that have been shown to have a significant effect on birds in similar ecosystems (Schindler et al., 2013). Following the results of previous studies (Hargis et al., 1998; Morelli et al., 2013; Schindler et al., 2013), which have shown that the performance of landscape metrics depends on the scale measured, we assess the effects of our predictor variables at six different distances from the bird point-count locations: 100 m, 250 m, 500 m, 750 m, 1 km and 1.25 km. We chose Pafos Forest as the study area because it is a nature reserve of high ornithological importance, which has been designated as a Special Protection Area (SPA), i.e. a Natura 2000 site (Hadjipanayiotou, 2012) and has a particularly dense network of unpaved roads. Many of these roads, which are maintained on a regular basis, are closed to the public and are mostly used by the Department of Forests for management purposes, such as for access in case of fire and for patrolling. 2. Material and methods 2.1. Study area Pafos Forest is located in the northwestern part of Cyprus (Fig. 1), an island of high ornithological importance (Iezekiel et al., 2004). More than 370 species of birds have been recorded in Cyprus so far, of which about 85% are non-resident species (Iezekiel et al., 2004). Pafos Forest covers an area of 602.68 km2 (Hadjipanayiotou, 2012). Its altitude ranges from near sea level to 1352 m, receiving an annual mean precipitation of 1060 mm (Hadjipanayiotou, 2012). The site is covered mostly by Pinus brutia forest, with its understory
15
consisting mainly of thermo-Mediterranean sclerophyllous vegetation such as Quercus alnifolia, Arbutus andrachne and Cistus spp. (Hadjipanayiotou, 2012). The high ecological significance of the forest stems from the fact that it is an important area for birds (Iezekiel et al., 2004), supporting populations of 22 species listed in the Annex I of the European Union's Birds Directive, including the four endemic species and subspecies of Cyprus (Oenanthe cypriaca, Sylvia melanothorax, Certhia brachydactyla dorothea and Parus ater cypriotes; Hadjipanayiotou, 2012). Additionally, the forest is one of the most important areas in Cyprus for nesting raptors such as Hieraaetus fasciatus and Accipiter gentilis. According to Hellicar et al. (2014), a major threat to Pafos Forest's biodiversity is habitat degradation, resulting in fragmentation, caused by the establishment of firebreaks and the dense network of roads. The majority of the roads are low-traffic unpaved roads built by the authorities for better accessing and managing the forest. For the purposes of this study we sampled twenty sites, all located next to existing roads. Due to the forest's steep topography it was not feasible to select sites at varying distances from the road. The sites were spread throughout the west-central part of the forest (Fig. 1). Areas in the very far north and the east were avoided for logistic purposes since, due to the forest's large size and topography, it was not practical to also cover those within the amount of time available. We selected and surveyed twenty sites randomly using Hawth's Analysis Tools (Beyer, 2004) in ArcMap (Version 9.3.1), by first dividing the chosen part of the forest into 2 km 2 km grid cells to ensure adequate distance between the sites and to avoid spatial autocorrelation. Note, however, that during the analysis we had to disregard five of the sites because they were located less than 1.25 km from the forest's boundary; data on the road network, habitat diversity, and the rest of the landscape metrics were only available for areas within the forest boundaries and consequently these sites had to be excluded due to missing information. Our results are therefore based on fifteen sites. Results from analyses that included more sites, but with smaller radii around them, were qualitatively similar, but explained less of the variation. 2.2. Bird censuses We conducted bird censuses, along the roadside, during the spring migration of 2012, between early March to mid-May, using 10-min point counts (Bibby et al., 1992). We decided to sample birds during the migratory season, as many species in the forest (including several of the high conservation importance) are migratory (Iezekiel et al., 2004). We visited each site six times to ensure adequate sampling. We carried out all surveys between sunrise and 10:00 AM. We rotated the order by which we surveyed the sites to make sure that no site was constantly sampled in early or late morning. During the surveys we identified and recorded all species seen or heard within a distance of 100 m from the sampling point. To account for species that were possibly missed during the surveys, we estimated species richness at each site using the firstorder Jackknife richness estimator (Magurran, 1998), using the specpool function in “vegan” package in R (R Core Team, 2015). We used the species migratory status list provided to us by BirdLife Cyprus, to classify birds into “Resident Breeders” and “Passage Migrants”. We also classified species according to their global population trends, using the corresponding data from BirdLife International, which are available online (BirdLife International, 2014). 2.3. Sites characteristics Using ArcMap, we established six nested circular buffer zones of
16
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Fig. 1. Map of Pafos Forest showing the fifteen point-count locations, the network of unpaved roads (gray lines) and the different habitats within the forest's boundaries. For illustrative purposes, the road networks within four of the study sites is shown, representing the full gradient of road density recorded within the 1 km buffer zone.
incremental radii (i.e. 100 m, 250 m, 500 m, 750 m, 1 km and 1.25 km) around each point-count location. Within each buffer zone we measured the extent of the biotic and the anthropogenic variables of interest. 2.3.1. Biotic variables We estimated habitat heterogeneity at each site using spatial data provided by the Department of Forests of Cyprus. The Department is the responsible authority for managing Pafos Forest and for that purpose has surveyed and mapped the forest's vegetation at the 0.002 km2 level. The data provided consisted of 9 vegetation categories: 1. stands of P. brutia, 2. stands of Cedrus brevifoliae, 3. stands of Q. alnifolia or other species, 4. mixed stands of P. brutia and Q. alnifolia, 5. mixed stands of P. brutia and C. brevifoliae, 6. mixed stands of P. brutia and other species, 7. mixed stands of Q. alnifolia and other species, 8. mixed stands of other species, and 9. riparian vegetation. Using FRAGSTATS v.4.2.1 (McGarigal et al., 2012), we measured the ShannoneWeiner Habitat Diversity Index (SHDI). Additionally, following the findings of Schindler et al. (2013), who studied the relationships between various landscape metrics and the species richness of multiple taxa (including birds), in a mosaic of similar habitats in a protected forested area in Greece, we measured, also using FRAGSTATS: the patch density (PD), the area-weighted mean patch fractal dimension (FRAC_AM), the area-weight patch shape (SHAPE_AM) and the coefficient of variation in patch area (AREA_CV) within each buffer zone (McGarigal et al., 2012). 2.3.2. Anthropogenic variables The total length of unpaved roads, within each buffer zone, was
measured using ArcMap and the forest's road map, provided by the Department of Forests. Small stretches of paved roads, representing 11.58% of the total roads recorded in our largest buffer zones (1.25 km), found at six of the sites, were excluded from the analyses. Results of an analysis that included these roads in measuring total road length were qualitatively similar. The distance to the nearest agricultural field was calculated using the CORINE Land Cover 2006 (CLC) map, a European wide land cover map, depicting, among other land cover categories, the spatial distribution of the agricultural land in each participating country (EEA, 2014). For the purposes of this analysis, we used ArcMap to measure the distance between each study site and the nearest agricultural field. Data on the human population size of each nearby village were extracted from the 2011 census data, made available online by the Statistical Service of the Republic of Cyprus (Statistical Service, 2014). 2.4. Data analysis We used bivariate correlation tests to assess collinearity between the independent variables. PD, FRAC_AM, and SHAPE_AM were found to be highly correlated with habitat diversity (>0.65) so they were subsequently removed from the analyses. The effects of the rest of the five explanatory variables (i.e. total road length, habitat diversity, distance to the nearest agricultural field, human population size, and variation in patch size), were tested using linear multiple regression models, and following a multimodel inference approach (Burnham and Anderson, 2002). All possible subset models for each bird category tested were ran, using the “MuMIn” package (Barton, 2014). Models were ranked in ascending
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
order, according to their corrected Akaike Information Criterion (AICc) value. All models with a DAICc of less than 7 were selected (Burnham and Anderson, 2002; Nakagawa and Freckleton, 2011). To assess the effect of each predictor on the response variables the model.avg function in the “MuMIn” package was used to calculate the averaged unstandardized and standardized regression coefficients, using the zero-average method (Burnham and Anderson, 2002; Nakagawa and Freckleton, 2011). The relative effect of each predictor on each bird category tested was evaluated using the standardized regression coefficients. To assess the predictive power of our models, we calculated the adjusted R2 values for each category tested within each buffer zone, based on the corresponding full model (i.e. when all independent variables were included). To show trends of the different variables individually, we present graphs of partial residual graphs. This is offered as a visual, we did not do further analysis on partial residuals (Darlington and Smulders, 2001). To check whether our data were spatial autocorrelated or nonnormally distributed, we tested the residuals of all the selected models, using the Moran's I test in the “ape” package and the ShapiroeWilk normality test. To confirm that there was no collinearity in our models, we used the vif function in the “car” package to check the variance inflation factors in each of the selected models (Fox and Weisberg, 2011). To check whether there was a phylogenetic signal in the two species' traits we used to classify birds (i.e. migratory status and population trends), we performed an Abouheif test (Pavoine and Ricotta, 2013) on the phylogenetic trees we obtained from www.birdtree.org (Jetz et al., 2012), using the abouheif.moran function in the “adephylo” package in R. Since www.birdtree.org does not provide a consensus tree, we downloaded 1000 trees for each trait, sourced from Hackett (all species; 10,000 trees with 9993 OTUs each; Jetz et al., 2012) and tested each tree for phylogenetic signal. We then examined the distribution of the resulting p-values to assess the strength of the signal. 3. Results A total of sixty-four species were recorded in all twenty sites; fifty-eight of those were recorded in the 15 sites that were included in the analysis. Species richness in those 15 sites ranged from 13 to 26 (mean 19.67, sd 3.15). Fourteen species were only recorded at one site while three species, one of them being the endemic subspecies P. ater cypriotes, were found in all 15. For a complete list of all the species recorded at each site please see Appendix I. In total, within our largest buffer zones, we recorded 287.68 km of roads, of which 33.31 km were paved roads (found at six sites) and were excluded from the analysis. On average, 19.18 km of unpaved roads were recorded at each site within the 1.25 km buffer, ranging from 8.08 km to 29.33 km (standard deviation ¼ 5.46). The average habitat diversity, within the largest buffer zones, was 0.51 (sd ¼ 0.20), ranging from 0.23 to 0.92. The coefficient of variation in path size ranged from 163.9 to 339.8 (mean ¼ 238.3, sd ¼ 52.64). The shortest distance to the nearest agricultural field was 2.23 km and the longest 8.44 km (mean ¼ 4.36, sd ¼ 1.91). The average human population size was 145 (sd ¼ 159) ranging from 0 (in two villages which are no longer inhabited) to 473. In total, 576 regression models were selected for the six bird categories tested at the six buffer zone radii (i.e. total species richness [91 models], richness of passage migrants [77], richness of resident breeders [106], richness of species with increasing [110], stable [97] and decreasing global population trends [95]). The results of the spatial autocorrelation and normality tests showed that none of our models violated the assumptions of independence and normality (all p-values were higher than 0.05). There was no collinearity either; all inflation factors recorded were less than 4.00.
17
The phylogenetic tests on the two traits showed that there was a weak signal in the population status (11.5% of the p-values were <0.05) and a stronger signal in the migratory status (0.44% of the pvalues were <0.05). The effects of the variables tested varied according to the size of the buffer zone used. Fig. 2 shows the bivariate Pearson correlation coefficients between total bird species richness and the five explanatory variables for each of the buffer zone sizes. The correlation between species richness and habitat diversity and the coefficient of variation in patch size appeared to be scale dependent, something which is also reflected in the averaged regression coefficients of the corresponding linear models (Table 1). The effect of habitat diversity, although overall positive, peaked at 500 m and then dropped as the buffer size increases. Interestingly, the effect of variation in patch size started as positive (albeit very weak) and then turned negative as the buffer size increased (Table 1). The effect of road networks, however, was consistently negative, and so was the effect of the human population size of the nearest village (Table 1; Fig. 2). The effect of the distance to the nearest agricultural field was also negative, but overall very weak, except in the smallest buffer zone where the variable ranked third in importance, with an effect size similar to that of the road networks (Table 1). The R2 values for the total species richness ranged from 0.55 (at 250 m) to 0.73 (at 1 km), showing that our models explain a satisfactory amount of the observed variance in bird species richness. At the 1 km, the most important variable was human population size, which had a strong negative effect, followed by the road networks, which also affected species negatively (Fig. 3). Habitat diversity, the third important variable, had a positive effect, while the effect of variation in patch area was negative (Fig. 3). The effect of distance to the nearest agricultural field was negligible (Table 1). When passage migrants were tested the effects of the variables also varied according to the buffer zone tested and so did the R2 values. Overall, the R2 values were lower than those of the total species richness, ranging from 0.31 at 100 m to 0.58 at 500 m (Table 2). At 500 m, road networks was the most important variable, showing strong negative effect, while variation in patch size was the second most important variable, with a positive effect. The effect of the rest of the three variables was weak (Table 2). Resident breeders appear to be affected by road networks only when the smallest buffer zone is considered (Table 2), but the R2 value of that model was particularly low (0.26). The highest R2 value for resident breeders (0.52) was also at 500 m. At that level, resident breeders were mostly positively affected by the habitat diversity, and to a lesser degree by the distance to the agriculture (positive effect) and the variation in patch size (negative effect). The R2 values for the species with increasing and stable populations were consistently very low (Table 3) showing that other variables are affecting the presence of these species. On the other hand, the richness of the species with decreasing population size appears to be explained, at least partly, by the variables included in the model. R2 values ranged from 0.22 at 250 m to 0.57 at 1 km. At the 1 km distance, the most important factor was road networks followed by human population size both of which have a negative effect (Table 3). Habitat diversity had a positive effect albeit weaker, while the rest two variables had no effects. 4. Discussion In the case of Pafos Forest, unpaved low-traffic road networks have consistently negative effects on the bird communities, especially on vulnerable species such as species with decreasing population status. Several of these species are of high conservation importance, such as the endemic Cyprus Warbler (S. melanothorax), and are included in the Annex I of the European Union's Birds
18
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Fig. 2. Pearson correlation coefficients between overall species richness (first-order Jackknife estimator) and each of the five explanatory variables at each of the radii used in the analyses.
Table 1 Results of the linear regression models for each buffer size tested, for the overall bird species richness, showing the explanatory variables' averaged unstandardized (b) and standardized (beta) coefficients, along with the variables' relative importance (i), calculated as the percentage of the selected models they appear in. The R2 values of the full models are also shown. RD: Road Density, SHDI: Habitat Diversity, DA: Distance to the nearest agricultural field, PS: Human population size of the nearest village, PA: Coefficient of variation in patch size. 100 m (R2 ¼ 0.64)
RD SHDI DA PS PA
250 m (R2 ¼ 0.55)
500 m (R2 ¼ 0.60)
750 m (R2 ¼ 0.65)
1000 m (R2 ¼ 0.73)
1250 m (R2 ¼ 0.62)
b
beta
i
B
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
10.23 2.03 0.85 1.53 0.00
0.29 0.14 0.29 0.60 0.02
0.48 0.31 0.42 0.87 0.11
3.34 5.31 0.08 0.71 0.01
0.28 0.31 0.03 0.28 0.06
0.58 0.61 0.12 0.51 0.21
0.53 7.66 0.01 0.89 0.01
0.11 0.44 0.00 0.35 0.06
0.36 0.83 0.11 0.74 0.19
0.67 5.08 0.02 1.24 0.00
0.32 0.29 0.01 0.49 0.01
0.75 0.69 0.09 0.92 0.08
0.50 3.42 0.01 1.74 0.02
0.37 0.18 0.00 0.69 0.14
0.79 0.46 0.03 1.00 0.42
0.25 2.62 0.03 1.73 0.01
0.26 0.15 0.01 0.68 0.14
0.59 0.38 0.07 1.00 0.42
Directive, which lists all species requiring special conservation attention (Donald et al., 2007). The road network within the forest is particularly dense, with an average of 6.5 km of roads per km2, compared to the mean density of roads in the Natura 2000 sites in Cyprus, which is 1.94 km/km2 (Zomeni and Vogiatzakis, 2014). 88.5% percent of the roads in our sites, as measured within the largest buffer zones, were unpaved low-traffic roads. Many of them are closed to the public, a few even with locked roadbarriers. Considering this fact, and also that paved roads (excluded from the analysis) only represented a small percentage of the total recorded road network, we argue that it is highly unlikely that the recorded effects on the birds were caused by either traffic, noise pollution or direct mortality associated with road usage, as found in other studies which looked at the effects of roads (Brotons and Herrando, 2001; Peris and Pescador, 2004; pez et al., 2010). Mitchell et al., 2006; Benítez-Lo It is probable that the adverse effects we found are due to habitat loss, and associated changes in the vegetation structure and composition, such as increased canopy openness, which can alter the surrounding microhabitat (Delgado et al., 2007). Although the percentage of the habitat lost to unpaved roads, within the sampled areas, only ranged from about 1.3 to 4.8%, it is likely that the long stretches of roads and their resulting edge effects have converted those areas into less suitable habitats for certain species (Ortega and Capen, 2002; Laurance et al., 2004). Another plausible explanation is that some bird species may avoid unpaved roads, even if there is suitable habitat along the roadside, because the area is considered to be of lower quality, for example due to higher perceived predation risk (Whittingham et al., 2006; Zanette et al., 2011). It's worth noting that habitat heterogeneity was also an important determinant of the bird species richness, especially of
migrant birds. We need to acknowledge that our phylogenetic analysis showed that there is possibly a signal in the migratory status, and thus the pattern may be driven by the particular group of species tested. Nonetheless, for conservation purposes, each species has an intrinsic value. A previous study in protected areas in Cyprus (Mammides et al., 2015) and studies elsewhere have shown that bird presence relates positively to habitat heterogeneity because more diverse habitats usually provide more foraging and nesting opportunities for a wider range of species (MacArthur and MacArthur, 1961; Santos et al., 2002; Brotons et al., 2005; Loehle et al., 2005; Mitchell et al., 2006; Gil-Tena et al., 2007; Santos et al., 2010). For the Pafos Forest, this result suggests that the areas of the reserve which are dominated by only one tree species, usually P. brutia, support lower numbers of birds species. Such areas are more likely to have impoverished bird communities if they also contain a dense network of roads. Somewhat surprisingly, distance to the nearest agricultural field was found to have a negative or negligible effect on the bird categories tested, although it was originally hypothesized that shorter distances would probably mean higher human pressure and disturbance and therefore less species. It turns out from our analysis that this is not the case. This is probably due to the type of the agriculture practiced in the region, which is small-scale, low intensity, and usually involving a variety of crops that has been shown to be beneficial especially to farmland birds, which most probably comprised part of our sample (Pino et al., 2000; Su arezSeoane et al., 2002). The population size of the nearest village, on the other hand, had consistently a negative effect (except on few of the resident breeders models) indicating that higher human presence relates to lower bird species richness. The effects of variation in patch size varied widely both in terms of the bird categories tested and the buffer sizes examined. This
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
19
Fig. 3. Partial residual plots for each of the five explanatory variables, showing their effect on the total species richness (first-order Jackknife estimator) when the rest of the variables are held constant. Plots are based on the averaged model within the 1 km buffer zone.
metric is a measure of the spread of the patch size relative to the mean patch size at any given site. When tested using the total bird communities it had a weak positive effect within the smallest buffers, but as the size increased the effect became negative and grew stronger. This suggests that at small scales variation in patch size may indicate heterogeneous habitat, which may translate into higher niche availability, but at a larger scales the metric may represent discontinuous habitat, possibly of lower quality.
In general, our results concur with previous findings showing that the performance of various landscape metrics depend on the scale measured. Although in previous studies it was found that landscape metrics, similar to the ones we used in this study, correlate strongest with bird indices when measured at short distances (e.g. 150e250 m; Morelli et al., 2013; Schindler et al., 2013), in our case results were strongest at larger distances (i.e. within a radius of 500 m or 1 km). This suggests that the effects of the
20
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Table 2 Results of the linear regression models for each buffer size tested, for the species richness of passage migrants (A) and residents breeders (B). Statistics and abbreviations as Table 1. A
100 m (R2 ¼ 0.31)
250 m (R2 ¼ 0.45)
500 m (R2 ¼ 0.58)
750 m (R2 ¼ 0.42)
1000 m (R2 ¼ 0.44)
1250 m (R2 ¼ 0.39)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
4.85 0.25 0.20 0.26 0.00
0.22 0.03 0.11 0.17 0.02
0.43 0.15 0.25 0.35 0.12
3.73 0.36 0.06 0.06 0.02
0.50 0.03 0.03 0.04 0.23
0.82 0.16 0.16 0.16 0.49
1.43 1.12 0.01 0.02 0.02
0.48 0.10 0.01 0.01 0.36
0.91 0.30 0.07 0.09 0.72
0.87 0.62 0.00 0.13 0.00
0.67 0.06 0.00 0.08 0.01
0.99 0.22 0.11 0.29 0.10
0.58 0.04 0.04 0.42 0.00
0.69 0.00 0.02 0.27 0.00
1.00 0.08 0.14 0.65 0.09
0.38 0.04 0.06 0.48 0.00
0.66 0.00 0.04 0.30 0.00
0.98 0.08 0.15 0.67 0.08
B
100 m (R2 ¼ 0.26)
RD SHDI DA PS PA
250 m (R2 ¼ 0.47)
500 m (R2 ¼ 0.52)
750 m (R2 ¼ 0.15)
1000 m (R2 ¼ 0.18)
1250 m (R2 ¼ 0.04)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
2.41 1.62 0.04 0.00 0.00
0.19 0.33 0.04 0.00 0.08
0.44 0.60 0.19 0.13 0.23
0.25 3.51 0.07 0.02 0.00
0.06 0.59 0.07 0.02 0.04
0.22 0.91 0.23 0.12 0.16
0.02 3.78 0.30 0.04 0.00
0.01 0.63 0.29 0.04 0.12
0.09 0.91 0.60 0.18 0.31
0.01 1.60 0.19 0.01 0.00
0.01 0.26 0.18 0.02 0.01
0.12 0.54 0.43 0.15 0.12
0.01 0.84 0.21 0.03 0.00
0.01 0.13 0.21 0.03 0.04
0.13 0.33 0.44 0.16 0.18
0.00 0.31 0.17 0.03 0.00
0.01 0.05 0.17 0.04 0.06
0.13 0.20 0.39 0.17 0.21
Table 3 Results of the linear regression models for each buffer size tested, for the species richness of birds with increasing (C), stable (D) and decreasing population status (E). Statistics and abbreviations as Table 1. C
100 m (R2 ¼ 0.21)
250 m (R2 ¼ 0.01)
500 m (R2 ¼ 0.17)
750 m (R2 ¼ 0.07)
1000 m (R2 ¼ 0.15)
1250 m (R2 ¼ 0.14)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
0.52 0.22 0.12 0.29 0.02
0.03 0.03 0.09 0.24 0.25
0.15 0.16 0.24 0.50 0.54
0.33 0.51 0.18 0.23 0.00
0.06 0.06 0.13 0.19 0.05
0.19 0.20 0.32 0.40 0.19
0.12 0.34 0.17 0.26 0.00
0.05 0.04 0.12 0.21 0.12
0.20 0.16 0.31 0.44 0.30
0.02 0.15 0.19 0.32 0.00
0.02 0.02 0.14 0.26 0.02
0.12 0.12 0.32 0.51 0.12
0.02 0.02 0.19 0.36 0.00
0.04 0.00 0.13 0.29 0.06
0.16 0.09 0.31 0.55 0.22
0.03 0.03 0.16 0.33 0.00
0.06 0.00 0.12 0.27 0.04
0.22 0.10 0.29 0.53 0.17
D
100 m (R2 ¼ 0.28)
250 m (R2 ¼ 0.32)
500 m (R2 ¼ 0.21)
750 m (R2 ¼ 0.13)
1000 m (R2 ¼ 0.00)
1250 m (R2 ¼ 0.20)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
RD SHDI DA PS PA
0.76 0.06 0.01 0.07 0.00
0.05 0.01 0.01 0.07 0.01
0.20 0.14 0.14 0.23 0.14
0.13 0.37 0.01 0.07 0.00
0.03 0.05 0.01 0.07 0.01
0.16 0.20 0.14 0.22 0.14
0.21 0.26 0.01 0.06 0.00
0.11 0.04 0.01 0.05 0.07
0.30 0.17 0.13 0.19 0.22
0.19 0.14 0.02 0.06 0.00
0.22 0.02 0.01 0.06 0.01
0.48 0.14 0.12 0.20 0.11
0.08 0.18 0.03 0.08 0.00
0.15 0.02 0.03 0.08 0.03
0.37 0.14 0.15 0.24 0.16
0.03 0.11 0.02 0.08 0.00
0.08 0.01 0.02 0.08 0.03
0.24 0.14 0.15 0.24 0.15
E
100 m (R2 ¼ 0.48)
RD SHDI DA PS PA
250 m (R2 ¼ 0.22)
500 m (R2 ¼ 0.45)
750 m (R2 ¼ 0.53)
1000 m (R2 ¼ 0.57)
1250 m (R2 ¼ 0.54)
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
b
beta
i
3.87 0.62 0.36 0.39 0.00
0.16 0.06 0.18 0.22 0.02
0.32 0.19 0.32 0.41 0.12
1.05 1.34 0.11 0.21 0.01
0.13 0.12 0.06 0.12 0.10
0.32 0.30 0.19 0.29 0.28
1.16 5.12 0.01 0.04 0.00
0.36 0.43 0.00 0.03 0.00
0.73 0.80 0.08 0.13 0.11
0.71 4.28 0.01 0.09 0.00
0.49 0.35 0.01 0.05 0.00
0.90 0.74 0.08 0.20 0.07
0.56 1.68 0.00 0.44 0.00
0.61 0.13 0.00 0.25 0.00
0.94 0.36 0.12 0.63 0.07
0.43 0.49 0.01 0.59 0.00
0.67 0.04 0.01 0.34 0.01
0.96 0.15 0.11 0.76 0.08
landscape metrics are often area specific and depend also on the bird guilds examined. At this juncture we should also note limitations of the study. In particular, we have only used species richness as a proxy of bird diversity. Further studies are needed to explore whether the effects examined remain the same when other diversity metrics are tested, and other analyses would be useful to incorporate the abundance and density of birds. In conclusion, the message of our study is twofold. Firstly, it shows that even unpaved and infrequently used roads can have negative effects on birds, highlighting also the importance of habitat heterogeneity for the presence of birds in protected areas. Based on these findings, we suggest that management authorities consider possible options for reducing the adverse impacts of roads on the avifauna in forested areas, for example by restoring the vegetation in roads which are not essential for accessing and managing the forest. Secondly, our study contributes to the growing literature which emphasizes that the effects of various landscape metrics on bird communities can be scale and guild specific.
Authors' contributions Christos Mammides, Constantinos Kounnamas and Costas Kadis designed the study. Christos Mammides and Constantinos Kounnamas collected the data. Christos Mammides, in collaboration with Eben Goodale, analyzed the data. All authors were involved in writing the manuscript. Acknowledgments We are thankful to BirdLife Cyprus, the Department of Forests, the Department of Land and Surveys, and the Statistical Service for making available the data necessary for this analysis, and to Salindra Kasun Dayananda for his help with the first figure. We are also thankful to the associate editor of the journal, Patrick Doncaster, and two anonymous reviewers for their highly constructive suggestions. The project was funded by the Research Promotion Foundation (Protocol No. PENEK/SUPPORT/0308/42), with cofunding from the European Union's Structural Funds.
C. Mammides et al. / Acta Oecologica 71 (2016) 14e21
Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.actao.2016.01.004. References Barton, K., 2014. MuMIn: Multi-model Inference. R Package Version 1.12.1. http:// CRAN.R-project.org/package¼MuMIn. pez, A., Alkemade, R., Verweij, P.A., 2010. The impacts of roads and other Benítez-Lo infrastructure on mammal and bird populations: a meta-analysis. Biol. Conserv. 143, 1307e1316. Beyer, H.L., 2004. In: Hawth's Analysis Tools for ArcGIS. Available at: http://www. spatialecology.com/htools. Bibby, C., Burgess, N., Hill, D., 1992. Bird census techniques. In: British Trust for Ornithology and the Royal Society for the Protection of Birds. Academic Press, London. BirdLife International, 2014. IUCN Red List for Birds. Downloaded from. http://www. birdlife.org. Brotons, L., Herrando, S., 2001. Reduced bird occurrence in pine forest fragments associated with road proximity in a Mediterranean agricultural area. Landsc. Urban Plan. 57, 77e89. Brotons, L., Herrando, S., Martin, J.L., 2005. Bird assemblages in forest fragments within Mediterranean mosaics created by wild fires. Landsc. Ecol. 19, 663e675. Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference: a Practical Information-theoretic Approach. Springer Science & Business Media, New York, USA. Darlington, R.B., Smulders, T.V., 2001. Problems with residual analysis. Anim. Behav. 62, 599e602. valo, J.R., Ferna ndez-Palacios, J.M., 2007. Edge effects Delgado, J.D., Arroyo, N.L., Are of roads on temperature, light, canopy cover, and canopy height in laurel and pine forests (Tenerife, Canary Islands). Landsc. Urban Plan. 81, 328e340. Donald, P.F., Sanderson, F.J., Burfield, I.J., Bierman, S.M., Gregory, R.D., Waliczky, Z., 2007. International conservation policy delivers benefits for birds in Europe. Science 317, 810e813. EEA, 2014. CORINE Land Cover 2006 (CLC 2006). http://www.eea.europa.eu/dataand-maps/data/corine-land-cover-2000-clc2000-seamless-vector-database (accessed Oct 2014). Fahrig, L., Rytwinski, T., 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecol. Soc. 14, 21. Forman, R.T.T., Alexander, L.E., 1998. Roads and their major ecological effects. Annu. Rev. Ecol. Syst. 29, 207eC202. Fox, J., Weisberg, S., 2011. An {R} Companion to Applied Regression, second ed. Sage, Thousand Oaks CA http://socserv.socsci.mcmaster.ca/jfox/Books/Companion. ~o n, G., 2008. The ecological Gaston, K.J., Jackson, S.F., Cantú-Salazar, L., Cruz-Pin performance of protected areas. Annu. Rev. Ecol. Evol. Syst. 39, 93e113. Gil-Tena, A., Saura, S., Brotons, L., 2007. Effects of forest composition and structure on bird species richness in a Mediterranean context: Implications for forest ecosystem management. For. Ecol. Manag. 242, 470e476. Hadjipanayiotou, C., 2012. DASOS PAFOU e CY2000006. In: Natura 2000-Standard Data Form. European Environment Agency. Hargis, C.D., Bissonette, J.A., David, J.L., 1998. The behavior of landscape metrics commonly used in the study of habitat fragmentation. Landsc. Ecol. 13, 167e186. Helldin, J.O., Seiler, A., 2003. Effects of roads on the abundance of birds in Swedish forest and farmland. In: Proceedings of the IENE Conference on “Habitat Fragmentation Due to Transport Infrastructure”, 13e14 November 2003, Brussels. Hellicar, M.A., Anastasi, V., Beton, D., Snape, R., 2014. Important Bird Areas of Cyprus. BirdLife Cyprus. Cyprus, Nicosia. Husby, A., Husby, M., 2014. Interspecific analysis of vehicle avoidance behavior in birds. Behav. Ecol. 25, 504e508. Iezekiel, S., Makris, C., Antoniou, A., 2004. In: Cyprus, B. (Ed.), Important bird areas of European Union importance in Cyprus. Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K., Mooers, A.O., 2012. The global diversity of birds in space and time. Nature 491, 444e448. Kociolek, A.V., Clevenger, A.P., St Clair, C.C., Proppe, D.S., 2011. Effects of road networks on bird populations. Conserv. Biol. 25, 241e249. Laurance, W.F., Balmford, A., 2013. Land use: a global map for road building. Nature 495, 308e309. Laurance, S.G., Stouffer, P.C., Laurance, W.F., 2004. Effects of road clearings on movement patterns of understory rainforest birds in central Amazonia. Conserv. Biol. 18, 1099e1109.
21
Loehle, C., Wigley, T.B., Rutzmoser, S., Gerwin, J.A., Keyser, P.D., Lancia, R.A., Reynolds, C.J., Thill, R.E., Weih, R., White Jr., D., Wood, P.B., 2005. Managed forest landscape structure and avian species richness in the southeastern US. For. Ecol. Manag. 214, 279e293. MacArthur, R.H., MacArthur, J.W., 1961. On bird species diversity. Ecology 42, 594e598. Magurran, A.E., 1998. Ecological Diversity and its Measurement. Princeton University Press, Princeton, NJ, USA. Mammides, C., Kadis, C., Coulson, T., 2015. The effects of road networks and habitat heterogeneity on the species richness of birds in Natura 2000 sites in Cyprus. Landsc. Ecol. 30, 67e75. McGarigal, K., Cushman, S.A., Ene, E., 2012. FRAGSTATS v4: Spatial Pattern Analysis Program for Categorical and Continuous Maps. Computer software program produced by the authors at the University of Massachusetts, Amherst. Available at the following web site: http://www.umass.edu/landeco/research/fragstats/ fragstats.html. Mitchell, M.S., Rutzmoser, S.H., Wigley, T.B., Loehle, C., Gerwin, J.A., Keyser, P.D., Lancia, R.A., Perry, R.W., Reynolds, C.J., Thill, R.E., Weih, R., White, D., Wood, P.B., 2006. Relationships between avian richness and landscape structure at multiple scales using multiple landscapes. For. Ecol. Manag. 221, 155e169. Morelli, F., Pruscini, F., Santolini, R., Perna, P., Benedetti, Y., Sisti, D., 2013. Landscape heterogeneity metrics as indicators of bird diversity: determining the optimal spatial scales in different landscapes. Ecol. Indic. 34, 372e379. Morelli, F., Beim, M., Jerzak, L., Jones, D., Tryjanowski, P., 2014. Can roads, railways and related structures have positive effects on birds? e a review. Transp. Res. Part D Transp. Environ. 30, 21e31. Nakagawa, S., Freckleton, R.P., 2011. Model averaging, missing data and multiple imputation: a case study for behavioural ecology. Behav. Ecol. Sociobiol. 65, 103e116. Ortega, Y.K., Capen, D.E., 2002. Roads as edges: effects on birds in forested landscapes. For. Sci. 48, 381e390. Pavoine, S., Ricotta, C., 2013. Testing for phylogenetic signal in biological taitrs: the ubiquity of cross-product statistics. Evolution 67, 828e840. Peris, S.J., Pescador, M., 2004. Effects of traffic noise on paserine populations in Mediterranean wooded pastures. Appl. Acoust. 65, 357e366. Pino, J., Rod a, F., Ribas, J., Pons, X., 2000. Landscape structure and bird species richness: implications for conservation in rural areas between natural parks. Landsc. Urban Plan. 49, 35e48. R Core Team, 2015. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/. Robson, L.E., Blouin-Demers, G., 2013. Eastern hognose snakes (Heterodon platirhinos) avoid crossing paved roads, but not unpaved roads. Copeia 2013, 507e511. , J., Zasadil, P., 2010. Edge effect of low-traffic forest roads on S alek, M., Svobodova bird communities in secondary production forests in central Europe. Landsc. Ecol. 25, 1113e1124. Santos, T., Tellería, J.L., Carbonell, R., 2002. Bird conservation in fragmented Mediterranean forests of Spain: effects of geographical location, habitat and landscape degradation. Biol. Conserv. 105, 113e125. Pe rez Tris, J., Carbonell, R., Tellería, J.L., 2010. Santos, T., Galarza, A., Ramírez, A., Vegetational versus topographical effects on forest bird communities: a test in the cantabrian mixed forest ecoregion (Spain). Ardeola 57, 285e302. Schindler, S., von Wehrden, H., Poirazidis, K., Wrbka, T., Kati, V., 2013. Multiscale performance of landscape metrics as indicators of species richness of plants, insects and vertebrates. Ecol. Indic. 31, 41e48. Statistical Service, 2014. Population Census 2001. Downloaded from: http://www. cystat.gov.cymof/cystat/statistics.nsf/index_en/index_en?OpenDocument. rez-Seoane, S., Osborne, P.E., Baudry, J., 2002. Responses of birds of different Sua biogeographic origins and habitat requirements to agricultural land abandonment in northern Spain. Biol. Conserv. 105, 333e344. Trombulak, S.C., Frissell, C.A., 2000. Review of ecological effects of roads on terrestrial and aquatic communities. Conserv. Biol. 14, 18e30. Tsiafouli, M.A., Apostolopoulou, E., Mazaris, A.D., Kallimanis, A.S., Drakou, E.G., Pantis, J.D., 2013. Human activities in natura 2000 sites: a highly diversified conservation network. Environ. Manag. 1e9. Whittingham, M.J., Devereux, C.L., Evans, A.D., Bradbury, R.B., 2006. Altering perceived predation risk and food availability: management prescriptions to benefit farmland birds on stubble fields. J. Appl. Ecol. 43, 640e650. Zanette, L.Y., White, A.F., Allen, M.C., Clinchy, M., 2011. Perceived predation risk reduces the number of offspring songbirds produce per year. Science 334, 1398e1401. Zomeni, M., Vogiatzakis, I., 2014. Roads and roadless areas in Cyprus: implications for the Natura 2000 network. J. Landsc. Ecol. 7, 75e90.