Introduction
The management of suckler cattle in European mountain conditions is based on long winter housing phases, imposed by the climatic conditions, followed by summer grazing on high-altitude pastures during the period of higher availability and quality of grass (Casasus et al., Reference Casasus, Sanz, Villalba, Ferrer and Revilla2002). However, in mountainous areas, cattle are particularly exposed to ectoparasites, especially horseflies (Diptera: Tabanidae). Horseflies are large haematophagous insects that feed on the blood of ungulates and can be a serious nuisance for livestock (Mullen & Durden, Reference Mullen and Durden2002). Previous studies have shown that grazing animals are likely to be particularly bothered by tabanids, and when the fly load is high, they may move to habitats with maximal exposure to wind or with mixed vegetation, for example, bushy areas, where tabanids are less active (Hughes et al., Reference Hughes, Duncan and Dawson1981; Raymond & Rousseau, Reference Raymond and Rousseau1987). In addition, tabanids are mechanical vectors of pathogens such as Besnoitia besnoiti, a sporozoan parasite enzootic to the Pyrenees, but currently considered an emerging disease in Europe (Foil, Reference Foil1989; EFSA, 2010; Jacquiet et al., Reference Jacquiet, Lienard and Franc2010; Gentile et al., Reference Gentile, Militerno, Schares, Nanni, Testoni, Bassi and Gollnick2012). High-altitude pastures are areas with a high risk of transmission, as sick livestock can mix with healthy livestock in the herd.
The occurrence and activity of tabanids depend on the presence of the host animal, as well as on environmental parameters such as climatic conditions (temperature, humidity and wind) and landscape parameters (water, vegetation and altitude) (Chvála et al., Reference Chvála, Lyneborg and Moucha1972; Sheppard & Wilson, Reference Sheppard and Wilson1977; Hackenberger et al., Reference Hackenberger, Jaric and Krčmar2009; Van Hennekeler et al., Reference Van Hennekeler, Jones, Skerratt, Muzari and Fitzpatrick2011). Insect communities may be influenced by habitat variation at the scale of a patch and/or the landscape, depending on the focal taxa's body size, home range area and dispersal distances. For example, species richness and abundance of bumblebees in mountainous areas is influenced by both patch and landscape factors (Hatfield & LeBuhn, Reference Hatfield and LeBuhn2007).
In Europe, most ecological studies dealing with tabanids have been conducted in plains habitats, with only a few studies carried out in mountainous areas (Chvála, Reference Chvála1979; Hackenberger et al., Reference Hackenberger, Jaric and Krčmar2009). However, Koerner (Reference Koerner2007) underlined the importance of altitude in ecological research, and altitude has been virtually proven to influence the distribution of insects (McCoy, Reference McCoy1990; Hodkinson, Reference Hodkinson2005; Pyke et al., Reference Pyke, Inouye and Thomson2011). Moreover, the richness of Tabanidae fauna in European mountains (the Alps and the Pyrenees) was shown by Leclercq (Reference Leclercq1971, Reference Leclercq1977) who has attempted to characterize the altitudinal distribution of species.
The eastern Pyrenees are an ideal study site for evaluating the influence of environmental variables on diversity and abundance of tabanids. The climate of this region is influenced by both Mediterranean and oceanic climatic conditions (Esteban et al., Reference Esteban, Ninyerola and Prohom2009). The transversal aspect of the main valleys in relation to the north–south axis results in two clearly contrasted slopes, one dry and one wet. Furthermore, agricultural and pastoral activities have created a structured mosaic of landscapes, modified by the expansion of forest areas over the last few decades (Gracia et al., Reference Gracia, Meghelli, Comas and Retana2011). The selection of this area for our study allowed us to consider the effects of different altitudes along high-elevation gradients, the effects of distance from water bodies – which are known as breeding sites for tabanids – and the effects of landscape factors on the distribution of tabanids. Seasonality and daily activity patterns of the main species present were also recorded.
Materials and methods
Study sites
Two sampling valleys (A and B) of summer pasture grounds on opposite slopes were selected in the eastern Pyrenees, France (fig. 1). Valley A is located in the summer pastures of La Mouline, ranging from 850 to 2300 m above sea level on the southeast slope of the Madres massif (2°14′E, 42°37′N). The climate is influenced by Mediterranean and alpine conditions, with a mean annual temperature of 10.8 °C and annual rainfall of 753 mm (Nohèdes village, 1000 m a.s.l., at a distance of 4 km) (Roux et al., Reference Roux, Masson, Bricaud, Coste and Poumarat2011). This area is in the centre of the Madres-Coronat Natura 2000 site (FR9101473) and includes Evol Lake (‘Gorg Negre’) at 2083 m a.s.l. It is characterized by dense grassland (Isoetes echinospora) around the lake, peat bogs, forests of mountain pine (Pinus mugo) and granitic scree.
Studied valleys (A and B) with vertical profiles of the trapping locations. Trapping locations were numbered from 1 to 9.
Valley B is located in the Mantet Natural Reserve, ranging from 1450 to 2700 m a.s.l. on the northwest slope of the Costabona massif (2°18′E, 42°28′N). The climate is influenced by both moderate Atlantic and orogenic continental conditions with a mean annual temperature of 9.5 °C and annual rainfall of 855 mm (Mantet village, 1545 m a.s.l.) (Mantet Natural Reserve data, unpublished). This area is characterized by metamorphic rocks from the Mantet–Fillos rift. It is a complex mosaic of woodlands (Pinus uncinata and Betula spp.), moorland (Cytisus oromediterraneus), rocks, pastures and grasslands. Wetlands and aquatic habitats are poorly represented.
Sampling
Figure 1 represents the two valleys with the vertical profile of the trapping locations. To evaluate the altitudinal effect, three transects at different altitude levels were designated in each valley: low (A1–A3: 1373–1554 m, B1–B3: 1580–1732 m), medium (A4–A6: 1757–1921 m, B4–B6: 1824–1932 m) and high (A7–A9: 2087–2258 m, B7–B9: 2101–2205 m). Along each transect, three locations were fixed at an increasing distance from the nearest river (0 m, 200 m and 400 m) to evaluate the influence of the proximity of water bodies. In both valleys, the respective rivers were considered the main water body, except for the highest altitude level of valley A, where the transect starts at the shore of the lake.
Mapping
To evaluate the influence of landscape factors, aerial photographs were analysed using ArcMap 9.3 (ESRI Inc., California, USA). Radii of 200 m were overlaid on trapping locations (fig. 2). The radii were quite large in size because tabanids are strong, fast fliers. Studies conducted on tabanid dispersal have shown that the mean daily flying distance is less than 1–2 km, although unfed tabanids may disperse over distances of more than 6 km (Sheppard & Wilson, Reference Sheppard and Wilson1976; Cooksey & Wright, Reference Cooksey and Wright1987; Konstantinov, Reference Konstantinov1993; Barros & Foil, Reference Barros and Foil2007). In each radius, polygon-shaped patches of homogeneous habitat were delineated according to the vegetation cover using aerial orthophotography from 2009 and classified into two land-cover classes: forested and unforested (fig. 2). Forested patches corresponded to conifer forests, and unforested patches corresponded to open areas such as grasslands, pastures, moors, bushes, rock, water or mixed vegetation.
Example of 200 m radii around three trapping locations. Polygon-shaped patches of forested and unforested habitats were delineated using aerial photographs.
Landscape metrics were calculated in each radius using the patch analyst extension for ArcGis 9.x. In each radius, forested and unforested patches were characterized by their total surface area (CA, ha), the mean patch size (MPS, ha) and their mean perimeter–area ratio (MPAR, m ha−1) as an indicator of the patch-shape complexity (table 1).
Landscape parameters extracted from thematic maps for each radius with the mean and standard deviation for each class (forested vs unforested).
Around trapping locations, the landscape was largely dominated by unforested habitat. The number of forested and unforested patches was quite similar, but the MPS was four times higher for unforested patches. Conversely, the MPAR was two times higher for forested patches.
Trapping
One Nzi trap (Mihok, Reference Mihok2002) and one Vavoua trap (Laveissiere & Grebaut, Reference Laveissiere and Grebaut1990) were used at each trapping location. The distance between traps was about 10 m. The traps were made from blue and black components (SuperMaine 300 g cotton/polyester 65/35, TDV Industries, France) and polyester mosquito netting. Aged cow urine (50 ml) was used as the attractant and placed under each trap in plastic vials with a 2.5-cm aperture. The front of each Nzi trap was positioned facing west. A total of eight collections in each valley were made over the period of May to September 2011. Each collection was conducted over a continual 48-hour period. Tabanids were identified using Chvála's key (Chvála Reference Chvála, Lyneborg and Moucha1972). Each sampling unit (SU) consisted of the total catches obtained in a pair of Nzi-Vavoua traps during a 48-hour session. Then, all the abundance data per species from the 72 SUs (8 collections×9 trapping locations per valley) were used to analyse the spatial and temporal distribution of tabanids.
Daily activity pattern
To investigate the daily activity pattern of tabanids, collections were made in selected locations on selected days and times. In valley A, collections were made at three locations (A3, A4 and A7) on 1 August 2011, and in valley B at five locations (B1, B2, B4, B5 and B6) on 11 August 2011 and at three locations (B7, B8 and B9) on 7 July 2012. Species were collected during six consecutive 2-hour periods from 7:00 to 19:00 h (GMT+1). Only one Nzi trap baited with aged cow urine was used at each location.
The results took account of all trapping locations in valley A and B during the three collection days. The percentages of the total number of each species were calculated for each 2-hour period.
Statistical analyses
Differences in catches between Nzi traps and Vavoua traps were compared using the non-parametric Wilcoxon signed-rank test (W). Species accumulation curves for each valley were constructed using EstimateS (Colwell et al., Reference Colwell, Chao, Gotelli, Lin, Mao, Chazdon and Longino2012). A correspondence analysis and a canonical correspondence analysis (CCA) were performed per trapping location on abundance log-transformed data using Canoco 4.5 (ter Braak & Smilauer, Reference ter Braak and Smilauer2002). For the CCA, the environmental data-set consisted of a matrix of nine variables: the slope (valley A vs valley B), the altitude, the distance from the river and the three landscape metrics (CA, MPS and MPAR) calculated for forested (F) and unforested (UF) patches (table 1). During the CCA, Monte Carlo permutations (n = 499) were performed in order to identify the subset of environmental variables that exerted a significant influence on tabanids distribution at P<0.01. Correlations between the total number of species or the abundance of main species and (i) the altitude, (ii) the distance from the river and (iii) the landscape metrics were evaluated using Spearman's rank-order correlations coefficient (r s ) with Monte Carlo permutations (n = 9999). Statistical tests were performed with PAST 2.12 (Hammer et al., Reference Hammer, Harper and Ryan2001).
Results
Effectiveness of traps
A total of 2020 tabanids (2019 females and one male) belonging to 19 species in six genera (Philipomyia, Tabanus, Hybomitra, Atylotus, Haematopota and Chrysops) were captured over the period of the study (table 2). Most of the insects, including the only male, were Philipomyia aprica (Meigen 1820) (55.8%). Other abundant species were Tabanus bromius Linnaeus 1758 (13.4%), Hybomitra auripila (Meigen 1820) (9.8%) and Tabanus glaucopis Meigen 1820 (9.2%).
List of collected Tabanidae species and the number of individuals of each species.
1 Specimen not collected in a trap.
The Nzi traps consistently caught more tabanids than the Vavoua traps, with a total of 1824 specimens (90%): Philipomyia spp. (88%, W = 36, P = 0.01), Tabanus spp. (92%, W = 21, P = 0.04), Hybomitra spp. (88%, W = 15, P = 0.09) and Atylotus spp. (83%, W = 12, P = 0.27). Nzi traps seemed less effective for Chrysops spp. (63%, W = 3, P = 0.6) and Haematopota spp. (28%, W = 6, P = 0.33), and very few specimens of these two genera were collected.
Spatial distribution
During the study, we collected more species but less specimens in valley A (N = 18, n = 731) than in valley B (N = 13, n = 1289). Catches of the four main species were more numerous in valley B. Species accumulation curves show that the sampling effort was quite sufficient for estimating species richness in both valleys and that species richness was higher in valley A (fig. 3).
Species accumulation curves for valleys A and B (each is shown with its 95% confidence interval).
The initial correspondence analysis revealed that trapping location A7, close to the lake, was an outlier (fig. 4). Therefore, a second correspondence analysis was done without this sample. Locations in valleys A and B were separated along the first axis, while higher locations (A8, A9, B8 and B9) were clustered along the second axis. To investigate community composition in relation to environmental variables, a CCA was performed (fig. 5). The A7 sample was not included in the analysis and two samples were downweighted as they were detected as outliers with a high influence. The results of the CCA were significant, explaining 66.2% (P = 0.002) of tabanids distribution. The results indicated that three variables significantly influenced fauna: slope, altitude and the mean patch size of unforested patches (P < 0.01).
Plot of trapping locations (without A7) on axes 1–2 of the correspondence analysis (the size of each circle is proportional to the total abundance of catches at each location). The plot of trapping locations including A7 is represented in the inset.
CCA biplot of the species and the environmental variables (inset). The size of each circle is proportional to the total abundance of catches of each species; complete names in table 2. The environmental variables are the slope of the valley (SLOPE), the altitude (ALT), the distance from the river (DIS), the class area (CA), the mean patch size (MPS) and the mean perimeter area ratio (MPAR) for forested (F) and unforested (UF) patches. The conditional effect (F and P values) after Monte Carlo permutation is reported for each variable. Significant levels are in bold (P < 0.01).
The relationships between the number of taxa or the abundance of the main species and (i) the altitude, (ii) the distance from the river and (iii) the landscape context were also assessed by Spearman's correlation analysis. All the significant results are summarized in table 3. They show that (i) no correlation occurred between the diversity of tabanids and altitude. However, the abundance of P. aprica and T. bromius were negatively correlated with altitude in valleys A and B, whereas H. auripila was positively correlated with altitude in valley B. (ii) Distance from the river had no perceptible effect on the catches. (iii) Species richness was positively correlated with the size of forested patches, and negatively correlated with the surface covered by unforested habitats. The shape complexity of unforested patches also had a positive effect on the number of taxa. In both valleys, the abundance of P. aprica was positively correlated with the shape complexity of unforested patches, but negatively correlated with the size of unforested patches. In valley B, the size of unforested patches had a negative effect on the abundance of T. bromius and T. glaucopis, and the shape complexity of unforested patches had a positive effect on the abundance of T. bromius.
Landscape parameters with significant positive or negative effects on species richness and abundance of tabanids using Spearman's rank-order correlation (P ≤ 0.05, Monte Carlo permutation test based on 9999 random replicates)
Seasonal abundance and daily activity
In both valleys, tabanids were active from early June to early September – only a few specimens were collected in late May and late September. The largest number of individuals and species were collected in early August in valley A and in mid-August in valley B (fig. 6). There was a seasonal succession of the four main species. H. auripila was the first species collected in abundance, with a peak in late June in both valleys; P. aprica was active from late July to mid-August, with the shortest flight period in valley A; T. bromius had the longest flight period, from late June to mid-August, with a peak in abundance in early August in valley A, extending to mid-August in valley B. The last to appear was T. glaucopis, which peaked in abundance in mid-August in valley B.
Catches of the four main species from June to September 2011 in valleys A and B.
Tabanids were active from 9:00 to 19:00 h (fig. 7). The diurnal activity of H. auripila, P. aprica, T. bromius and T. glaucopis was characterized by a one-peak curve which reached a maximum between 11:00 and 13:00 h (GMT+1). During this period, 40–50% of tabanids were collected.
Daily activity of the four main species between 7:00 and 19:00 h (GMT+1). The figures show the percentages of the total number of specimens collected during six consecutive 2-hour periods.
Discussion
Efficiency of traps and daily flight activity
Our results showed that Nzi traps were more effective than Vavoua traps for sampling Tabanidae species. The efficiency of these traps has been previously demonstrated in Africa, North America and Australia (Mihok, Reference Mihok2002; Mihok et al., Reference Mihok, Carlson, Krafsur and Foil2006; Van Hennekeler et al., Reference Van Hennekeler, Jones, Skerratt, Fitzpatrick, Reid and Bellis2008). In studies carried out in southeastern France, Nzi traps captured a large number of tabanids in the plain (Baldacchino et al., Reference Baldacchino, Cadier, Porciani, Buatois, Dormont and Jay-Robert2012). Nzi traps seem to be particularly attractive to flies as opposed to other insects, as in addition to horseflies, Haematobosca stimulans (Meigen 1824) (Diptera: Muscidae) and other muscid flies that annoy livestock, were also caught (Baldacchino, unpublished data).
Given the high efficiency of the Nzi traps, they were used to estimate the flying activity of tabanids during the day. In the eastern Pyrenees, the four main species, P. aprica, T. bromius, T. glaucopis and H. auripila, were all active from 9:00 to 19:00 h, with a marked peak between 11:00 and 13:00 h (GMT+1). In studies carried out in southern Bulgaria, the major peak of activity for T. bromius was recorded between 12:00 and 15:00 h (Ganeva, Reference Ganeva1999), while in the eastern Caucasus (about 1000 m a.s.l.), Tabanus sp. and Philipomyia sp. had a conspicuous peak between 16:00 and 18:00 h (Chvála, Reference Chvála1979). Chvála suggested that temperature, relative humidity and thermal winds in the mountains were decisive factors influencing the late afternoon peak of tabanid activity, yet our findings contrasted with these observations. Since meteorological conditions influence the flying activity of tabanids (Van Hennekeler et al., Reference Van Hennekeler, Jones, Skerratt, Muzari and Fitzpatrick2011), one might hypothesize that cold temperatures induced by high elevations (our study ranged from 1300 to 2300 m a.s.l.) restricted the favourable period for flight. Our findings revealed that the four main species had a similar pattern of daily activity with a peak at midday, but other tabanid species are known to be active in early or late afternoon, or early in the evening (Chvála, Reference Chvála1979; McElligott & Galloway, Reference McElligott and Galloway1991; Oliveira et al., Reference Oliveira, Ferreira and Rafael2007). Furthermore, some species have a crepuscular activity, such as Tabanus equalis Hine 1923 in Oklahoma (Hollander & Wright, Reference Hollander and Wright1980) or Tabanus paradoxus Jaennicke 1866 in Europe (Chvála et al., Reference Chvála, Lyneborg and Moucha1972), or exhibit a bimodal pattern of activity, such as Haematopota sp. in Uganda (Harley, Reference Harley1965).
Influence of elevation on species richness and abundance
Insects vary in their population response to altitude (Hodkinson, Reference Hodkinson2005), yet the influence of elevation on tabanids has been poorly studied. In Tennessee, studies have shown that many species are prevalent only at higher altitudes (600 m a.s.l.), while most species common at lower altitudes (250 m a.s.l.) were also fairly common in upland locations (Mullens & Gerhardt, Reference Mullens and Gerhardt1980). In Croatia, research has found that tabanid abundance and species richness are partially determined by altitude or vegetation along a transect rising from sea level to 1157 m (Hackenberger et al., Reference Hackenberger, Jaric and Krčmar2009). In Armenia and the mountainous Nakhitchevan region (1400–1700 m a.s.l.), the flight activity of Tabanus infestus Bogatchev & Samedov 1949 has also been found to depend on altitude (Dolin & Andreeva, Reference Dolin and Andreeva1983).
In our study, the CCA showed that altitude had a significant influence on tabanids distribution, as well as a significant effect on the abundance of three out of four main species (positively for H. auripila, negatively for P. aprica and T. bromius), but had no significant effect on species richness. These findings were congruent with published data. H. auripila, observed from Scandinavia to Spain, occurs at low altitude in northern latitudes, but in southern Europe is restricted to mountains. On the other hand, P. aprica is a typical species of hilly regions, occurring from Spain to Belgium, while T. bromius is a ubiquitous species widely distributed throughout the western Palaearctic from Scandinavia to North Africa (Chvála et al., Reference Chvála, Lyneborg and Moucha1972). All these species have previously been observed up to 2000 m a.s.l. in European mountains (Leclercq, Reference Leclercq1977).
Slope orientation and seasonal dynamics
The mix of species was different in the two valleys. The CCA showed that community composition of tabanids species was significantly influenced by slope. In the south-facing valley (A), species richness was high, but P. aprica largely dominated (64% of catches). Conversely, the evenness was high among the 17 remaining species (Shannon evenness = 0.82). Some typical species were observed close to the lake, such as Chrysops and Haematopota species, which are known to occur in this type of biotope and do not fly far from the places where the larvae live (Chvála et al., Reference Chvála, Lyneborg and Moucha1972). This explains why trapping location A7 was an outlier with respect to community composition.
The north-facing valley was characterized both by lower diversity and by a lesser degree of domination of P. aprica (51%), which was associated with three other abundant species (T. bromius, T. glaucopis and H. auripila). Together, these four species represented more than 93% of the catches. Such a contrast in fauna related to exposure to sunlight has been previously shown at lower elevations in the eastern Mediterranean basin, in a study carried out in the Velika Kapela mountains in Croatia (Hackenberger et al., Reference Hackenberger, Jaric and Krčmar2009).
A seasonal succession of the main species was observed and may be associated with a strategy to reduce interspecific competition (Barros, Reference Barros2001). H. auripila was active very early in the summer, T. bromius and P. aprica were active between late July and early August and T. glaucopis was active starting from early August. The propitious period for tabanid activity was shorter in the south-facing valley, where 66% of insects were caught during the fifth trapping session (30 July). Conversely, the apparent abundance of tabanids in the north-facing valley (ratio of total catches in valley B/valley A = 1.8) can be explained by the length of the period of activity.
The seasonal dynamics of the most abundant species were consistent with previous studies. In North America, Hybomitra species were most active from late May to early July (Mullens & Gerhardt, Reference Mullens and Gerhardt1980; McElligott & Lewis, Reference McElligott and Lewis1998). In the Czech Republic, H. auripila emerged first, in June, whereas T. bromius emerged in July (Dusbabek, Reference Dusbabek1986). In Turkey, Hybomitra sp. were also the first species to emerge, in late May, whereas T. bromius and P. aprica peaked in July, with T. glaucopis emerging in late July and reaching its peak in mid-August (Kilic, Reference Kilic1993; Altunsoy & Kilic, Reference Altunsoy and Kilic2012). In Croatia, T. bromius was most active in mid-July or early August, depending on whether the location was pasture or forest (Krčmar, Reference Krčmar2005). In 2007, Hackenberger et al. (Reference Hackenberger, Jaric and Krčmar2009) observed even earlier activity of T. bromius in the Velika Kapela mountains (peak of activity in June) and suggested the explanation might have been that the temperature was 3 °C above the seasonal average. Hence, seasonal meteorological variability may have a significant influence on the duration of tabanid flight activity, on when they emerge and on their peak of abundance (Krčmar, Reference Krčmar2005).
Influence of water bodies and landscape structure
Our study indicated that distance from water (in this case, a river) did not significantly influence the community composition of tabanids. For most tabanid species, the ecological requirements of larvae are poorly known. Wet hollows or somewhat drier edaphic conditions should be suitable laying places for some species, and one might expect that the 19 species observed during the study belong to different ecological groups (Chvála et al., Reference Chvála, Lyneborg and Moucha1972; Andreyeva, Reference Andreyeva1982; Andreeva et al., Reference Andreeva, Kilic and Altunsoy2009). Some species, such as P. aprica or T. glaucopis, do not depend on aquatic biota for larval development, however, others, such as T. bromius, are semi-hydrophilous at larval stage and lay their eggs in slow-moving water bodies (Chvála et al., Reference Chvála, Lyneborg and Moucha1972). Similarly, the presence of H. auripila adults near biota such as peat bogs and marshes suggest that the larvae may be hydrophilous (Chvála et al., Reference Chvála, Lyneborg and Moucha1972). However, given their strong flying ability, (semi-) hydrophilous species are seemingly able to disperse widely, and adults can be collected far from breeding sites.
Although our study did not reveal any significant effect of the presence of a river on the prevalence of tabanids, the community composition was found to be significantly influenced by the distribution of forested and unforested patches in the mosaic of habitats (MPS unforested, P < 0.01). Specifically, landscape parameters showed that species richness and abundance of P. aprica, T. bromius and T. glaucopis (78% of total catches) decreased in relation to an increase in the size of clearings, and was enhanced by patch-shape complexity. Therefore, pastures dominated by grasslands and shrubs with patches of forested areas appeared to favour tabanids.
Forested areas are preferential resting sites for tabanids (Okiwelu, Reference Okiwelu1977). In a mark–recapture study in Texas, Tabanus abactor Philip 1936 was found in greater numbers in habitats with cover (e.g. dense leaves and branches) than in areas lacking cover (Kingston et al., Reference Kingston, Wangberg and Sanders1986). In contrast, open areas are more suitable for host-seeking activity, as a target host tends to attract over a wider range in an open habitat than in a forested one (Barros, Reference Barros2001).
A study in southern Louisiana showed that Tabanus sp. had the greatest host-seeking activity near the edge of a wood, and that this activity decreased significantly at over 200 m from the edge of the wood or more than 20 m into the wood (Sheppard & Wilson, Reference Sheppard and Wilson1977). This preference for pasture–forest ecotone habitats may explain the higher abundance of tabanids trapped in forest clearings than in primary forest in a study in the Amazon (Ferreira-Keppler et al., Reference Ferreira-Keppler, Rafael and Guerrero2010), and that, in an Ivory Coast study, the largest numbers of tabanids were caught in gallery forests, and the fewest in primary forest (Acapovi, Reference Acapovi2001).
Conclusion
The size and physical arrangement of habitat patches in landscapes play a fundamental role in determining the abundance and diversity of insect fauna (Hunter, Reference Hunter2002). Moreover, the complexity of the border between patches at any scale may be as important as the content of the patch, from ‘a fly's point of view’ (Haslett, Reference Haslett2001). Our study demonstrates that tabanid distribution was influenced by altitude and forest cover, and particularly by the size and shape complexity of the unforested patches. Our findings indicate that horseflies’ preferential habitats are at the edges of their breeding sites, host-seeking areas and resting sites. In the Pyrenees Mountains, the decrease in agro-sylvo-pastoral activities during the last century has favoured the expansion of forest areas and a reduction in open habitats (Gracia et al., Reference Gracia, Meghelli, Comas and Retana2011). This change may have an influence on the distribution of tabanids in the future, potentially extending their range of activity.
During their period of activity, tabanids reached the maximal peak of abundance in the middle of the summer. Hence, management of grazing areas relative to seasonal occurrence may be an element to consider in integrated management of tabanids (Foil & Hogsette, Reference Foil and Hogsette1994). Selective grazing in large open areas, well away from wooded habitats and at high elevation levels, may reduce biting activity on livestock during mid-summer. This management could be associated with mechanical control of the population of horseflies using an effective method. Nzi traps have been shown to collect great numbers of tabanids. The setting of such traps close to the pasture–forest ecotone habitats may decrease the populations of tabanids. Such an approach is a potential alternative to the use of insecticides, which are not considered to be very effective against horseflies (Mullen & Durden, Reference Mullen and Durden2002), are ecotoxic and have a negative impact on populations of dung-dependent arthropods (Wardhaugh, Reference Wardhaugh2005).
Acknowledgements
We would like to thank Claude Guisset of the Mantet Nature Reserve, Karine Geslot of the Jujols Nature Reserve and Alain Mangeot of the Nohèdes Nature Reserve for their invaluable help and useful input. We thank also Hélène Guis, Centre de coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), for useful discussions and anonymous reviewers for critical and helpful comments on the manuscript. We are grateful to those who gave us technical assistance in the field: Alain, Lyse, Marion, Jean-Philippe, Lucile and Steve.
