Conservation value of linear woody remnants for two forest carnivores in a Mediterranean agricultural landscape


Correspondence author. E-mail:


1. The loss of biodiversity caused by agricultural expansion can be countered by adopting wildlife-friendly farming strategies and by expanding the network of nature reserves. The potential benefits of agricultural extensification, represented in Europe by agri-environmental schemes, still remain unclear. In particular, the effectiveness of preserving linear woody vegetation to retain forest carnivores in farmland has received limited attention. We document the value of hedgerows and narrow strips of riparian forest for the Egyptian mongoose Herpestes ichneumon and the common genet Genetta genetta.

2. In an agricultural mosaic of southern Spain containing 4·7% of woody vegetation, we tested hypotheses about the role of linear elements and three farmland types differing in the amount, quality and structure of woody cover. We analysed the influence of linear elements on the placement and utilization of home ranges by combining compositional analysis and numerical methods.

3. Mongooses and genets strongly selected linear woody vegetation. All types of farmland, including open fields, dehesa (savanna-like pastureland or arable land rich in oak trees Quercus ilex and Q. suber) and olive Olea europaea groves, were avoided, suggesting that both species strictly depend upon native woody cover.

4. Most individuals made regular use of hedgerows and some individuals used hedgerows as the only source of woody cover in their home ranges.

5. The distribution of home ranges suggested that individuals made up a continuous, rather than discrete, population in a spatially structured habitat. An evenly distributed hedgerow network across the intervening agricultural matrix could prevent population fragmentation.

6.Synthesis and applications. A suitable network of linear cover allows some forest carnivores to survive in agricultural landscapes containing a low proportion (<10%) of native woody vegetation. Length (>0·5 km), width (4–10 m), quality (continuous native shrubs, scattered native trees, low levels of human disturbance), density (10–50 m ha−1) and a regular distribution of linear features are key elements in the conservation or restoration of agricultural landscapes where the preservation of small forest carnivores is an objective.


The increasing need for food by the growing human population makes the conversion of natural areas into agriculture a major threat to biodiversity (Tilman et al. 2002). This can be compensated for by increasing the amount of land protected in nature reserves or by adopting wildlife-friendly farming (Green et al. 2005). Extensive farming methods may alleviate species loss (Knop et al. 2006), but may fail to conserve sensitive species (Donald 2004; Kleijn et al. 2006) and net biodiversity gains may be limited (Kleijn et al. 2001; Feehan, Gillmor & Culleton 2005; Whittingham et al. 2007). Moreover, the reduced yield of extensive farming may result in more land being converted to agricultural production (Green et al. 2005). More research is needed to assess properly the potential of low-intensity agriculture as a global conservation option (Kleijn & Sutherland 2003).

In Europe, over the last decades, extensification has been widely applied in the form of agri-environment schemes (AES) (Kleijn & Sutherland 2003). Among other measures, AES include the restoration of woody cover in farmland, often in the form of hedgerows or similar linear elements, typical of extensive agricultural landscapes in the temperate region (Fritz & Merriam 1996; Baudry, Bunce & Burel 2000; Herzog et al. 2005). Conserved or restored lines of woody cover often increase the diversity and abundance of a variety of taxa (Hinsley & Bellamy 2000; Aviron et al. 2005; Hannon & Sisk 2009), but this has not been found everywhere (Bates & Harris 2009).

Implementation of AES at small spatial scales (single farms or small groups of farms) may partly explain their limited performance in restoring species of conservation concern (Whittingham 2007). The benefits of linear woody vegetation have been reported for small organisms including plants, invertebrates and vertebrates with low mobility (Corbit, Marks & Gardescu 1999; Jehle & Arntzen 2000; Thomas et al. 2001). However, little information exists about the role of woody linear features for larger and more mobile organisms that are unlikely to find sufficient resources within a single hedgerow or similar feature (Redpath 1995).

We studied the habitat use of the Egyptian mongoose Herpestes ichneumon L. and the common genet Genetta genetta L. in a Mediterranean agroecosystem where woody cover is scarce and occurs mainly as linear remnants (riparian forest and hedgerows). These species are considered forest carnivores, but their ecology remains almost unknown in agricultural mosaic landscapes. Therefore, we first tested the hypothesis that resident mongooses and genets must obtain their resources in woody vegetation and that adjacent farmland was unsuitable because of low prey abundance or insufficient cover. We expected a positive selection of woody cover and a negative selection of farmland patches.

Secondly, taking into account the scansorial habits of common genets (Larivière & Calzada 2001), we tested the hypothesis that resident genets, but not mongooses, would exploit and positively select farmland with high tree cover.

Thirdly, we examined the value of linear landscape features, including hedgerows, tree rows and grassy field margins. We tested the null hypothesis that these structures do not offer enough resources for resident mongooses and genets to use them regularly. This hypothesis predicts that (1) the density of linear landscape features within home ranges will be, at best, proportional to their availability in the landscape; (2) the proportion of animal locations outside riparian forest that fall in linear landscape features will be equal to, or lower than, the availability of such structures within home ranges; and (3) the spatial distribution of animal locations in open farmland will be independent of the proximity of linear landscape features.

Fourthly, we measured the amount of woody vegetation within the home ranges of resident individuals to establish a threshold and to explore whether hedgerows may substitute for riparian forest as a source of woody cover. If cover quality was similar in hedgerows and riparian forest, we would expect home ranges including only hedgerows (lower cover density) to be larger than those including also riparian forest. An inverse correlation between the extent of hedgerows and that of riparian forest would also be expected within home ranges.

Finally, we tested two null hypotheses reflecting the quality of linear landscape features: (1) hedgerows and grassy lines were used with similar intensity and (2) the intensity of use was independent of the length and width of the linear element, as well as of the degree of human disturbance in their surroundings.

Materials and methods

Model species

In Europe, the Egyptian mongoose (mean adult weight: 2·9 kg) is restricted to the south-west of the Iberian Peninsula, has diurnal habits, prefers dense cover for resting and actively avoids open areas (Palomares & Delibes 1993). The common genet (mean adult weight: 1·8 kg) also occurs in south-western Europe, exhibits nocturnal activity, forages both on the ground and in the tree canopy and has been reported to select dense cover for breeding and resting (Palomares & Delibes 1994). Both species feed upon small vertebrates, mostly mammals (Palomares & Delibes 1991).

Study area and landscape structure

We conducted our study in the lower Guadiamar basin, SW Spain (37º23′N, 6º13′W; Fig. 1). We corrected the position, size, shape and content of polygons in a land cover layer (Junta de Andalucía 1999) through comparison with orthophotographs and field surveys. We then simplified cover types that shared a similar vegetation structure into four categories (Table 1). All landscape measurements were made using ArcView GIS 3·2 and ArcMap 9·0 (ESRI, Redlands, California, USA).

Figure 1.

 Location of the study area (square) in the Guadiamar agroecosystem and habitat map. Home ranges (nearest-neighbour convex hulls) for seven Egyptian mongooses (H, solid line) and six common genets (G, broken line) are shown.

Table 1.   Proportion of each habitat type in the study area. Habitat availability (urban excluded) is compared with mean (±SE) habitat proportion in nearest-neighbour convex hull (NNCH) estimates of home range for adult resident Egyptian mongooses (= 7) and common genets (= 6). Likewise, mean (±SE) habitat availability within minimum convex polygon (MCP) estimates of home range is compared with the mean (±SE) proportion of radio locations in each habitat type
 Study areaAvailability
study area
NNCHMean availability
Egyptian mongoose
 Olive groves0·5460·5680·24 ± 0·140·21 ± 0·090
 Crops0·2360·2450·29 ± 0·140·26 ± 0·110·05 ± 0·03
 Dehesa0·1310·1370·15 ± 0·110·26 ± 0·090·05 ± 0·02
 Scrubland0·0470·0490·32 ± 0·120·27 ± 0·120·90 ± 0·03
Common genet
 Olive groves0·5460·5680·44 ± 0·170·38 ± 0·160·16 ± 0·04
 Crops0·2360·2450·02 ± 0·020·07 ± 0·050
 Dehesa0·1310·1370·26 ± 0·090·36 ± 0·100·06 ± 0·03
 Scrubland0·0470·0490·28 ± 0·160·20 ± 0·120·83 ± 0·04

Remnant woody vegetation was structurally similar in all landscape elements (including hedgerows): a continuous association of tall shrubs (Rubus, Pistacia, Phyllirea, Myrtus) interspersed with trees (Quercus, Fraxinus, Salix, Populus). As shrub was the dominant vegetation layer, we used ‘scrubland’ to denote patches, riparian strips of native woody cover and hedgerows. Scrubland made up 1·6% of the landscape. We distinguished three types of farmland: crops (cereals and sunflower Helianthus annuus L.), olive Olea europaea L. groves and dehesa. The last one (dehesa) is an agroforestry system that combines pastures or cereal with scattered holm oaks Quercus ilex L. and cork oaks Q. suber L. and little or no understorey (Joffre et al. 1988). The distribution of farmland types was not homogeneous. Cereal crops abound in the northern half of the study area, where very little scrubland remains, while in the south, olive groves dominate the landscape mosaic and the proportion of scrubland is above the average (Fig. 1). Preliminary mammal surveys reported mongooses and genets in the south, but rarely in the north (Rodríguez & Delibes 2003). Therefore, we defined our study area as a square landscape sample of 79·2 km2 in the southern half of the Guadiamar agroecosystem (Fig. 1). This area contains 4·7% of scrubland (Table 1), most of which is associated with three streams that run from north-west to south-east (Fig. 1). Some fields, groves and dehesas are bounded by hedgerows (Fig. 2), which made up 6·7% of total scrubland.

Figure 2.

 The distribution of scrubland (shaded) in the study area. Black lines: hedgerows; broken: tree lines; grey: grassy lines. Dots: independent radio locations of seven Egyptian mongooses (light, = 51) and six common genets (dark, = 92) >30 m away from scrubland.

On orthophotographs, we detected, digitized and characterized 187 linear elements that were subsequently checked in the field. Their overall length was 73 km (density: 0·92 km km−2), of which 55% were hedgerows (0·51 km km−2), 38% grassy lines (0·35 km km−2) and 7% tree lines. We estimated the mean width of linear elements on orthophotographs by taking 3–16 measurements at random points along each line. Most hedgerows were narrow (frequency in width classes: <5 m, 31%; 5–10 m, 57%; 10–15 m, 6%; >15 m, 6%) and did not have physical connections with scrubland patches. The mean width (±SE; n = 20) of the riparian forest along the eastern, central and western streams was 49 ± 3 m, 133 ± 12 m and 41 ± 4 m, respectively (Fig. 2).

Field methods

Seventeen mongooses (13 adults) and 17 genets (nine adults) were caught with boxtraps (2·0 × 0·5 × 0·5 m) from June 2005 to March 2007. We immobilized animals with tiletamine-zolazepam (Zoletil, Virbac, Spain) and determined their age from body weight, tooth wear and signs of reproductive activity. All adults were equipped with radiocollars (Biotrack, Wareham, UK). We used the homing technique (Mech 1983) and a GPS unit (Garmin, Olathe, Kansas, USA) to locate tagged animals. We accurately established whether animals were inside or outside linear elements. Animals were located on average every 3 days (range 1–17 days), at random times, and we assumed that these positions represented independent samples of habitat use during activity and inactivity periods.

Home range estimates

Tracking periods were evenly distributed throughout the year and their mean length was 15 weeks (range 6–24 weeks). We calculated minimum convex polygons (MCP) with the ArcView extension Animal Movement (Hooge & Eichenlaub 2000) as estimates of home ranges. We considered an animal as resident if the increase of MCP area plotted against the number of radio locations decelerated and reached a plateau (mean = 28 radio locations), and analyses were restricted to resident individuals. Home ranges were also estimated with Nearest-Neighbour Convex Hulls (NNCH; parameter = 5), an extension of the MCP technique that merges a set of local, smaller MCPs constructed with clusters of locations (Getz & Wilmers 2004). Compared with other estimates, convex hulls reduce the amount of unused areas within the home range and are useful when the configuration of habitat elements may force nonconvex ranges (Getz & Wilmers 2004), as may be the case with the elongated structures of our landscape. NNCH were calculated with the ArcView extension LoCoh (Getz & Wilmers 2004).

Habitat selection

We used compositional analysis (Aebischer, Robertson & Kenward 1993) to determine habitat selection in the placement of home ranges. Availability was the proportion of each habitat type in the study area and animal usage was the respective habitat proportions included within individual NNCH home ranges. Compositional analysis was also employed to analyse habitat use within home ranges. At this level of selection, availability was defined as the proportion of each habitat type within each MCP home range and animal usage was the proportion of independent radio locations that fell in each habitat type. Zero usage was replaced by the value 0·001. We performed compositional analysis with the free software Resource Selection (Fred Leban, University of Idaho, Moscow, USA). We calculated the geometric mean selection ratio to estimate selection at the population level (Pendleton et al. 1998).

Use of linear elements

We used Monte Carlo simulations to assess the value of linear elements. To test whether the density of linear elements within home ranges was proportional to their availability in the landscape, for each NNCH home range, we generated 99 random convex hulls with identical area and shape, the centroids of which were random points within the study area. Simulated convex hulls were rotated at a random angle. In actual and simulated ranges, we calculated the density of hedgerows, tree lines and grassy lines. For each home range, the 100 values were ordered, the highest value being assigned rank 1. The rank value/100 was the probability that the density of linear elements was equal to or lower than its availability (ranking test, Manly 1997). We took the individual as the sampling unit and compared the mean density of linear elements in simulated ranges with the observed density using the Wilcoxon matched pairs test to examine whether a pattern of selection emerges at the population level.

We computed the number of radio locations outside riparian forest and generated 99 groups of the same number of random locations for each MCP to test whether linear elements of different type were used proportionally to their availability. In actual and simulated ranges, we calculated the number of positions that fell within hedgerows, tree lines and grassy lines and the distance from each position to the nearest linear structure of each type. We also tested whether hedgerow quality influenced its usage. We counted how many radio locations fell in hedgerows, simulated 99 sets of the same number of locations randomly placed along available hedgerows within each MCP and measured hedgerow width. The ranking test was used to assess the significance of selection. Means were compared across individuals using the Wilcoxon test.

We estimated disturbance levels at linear elements by considering their distance to the nearest paved road. We used generalized linear mixed models (GLMMs) to examine the effects of type, length, mean width and disturbance of linear elements contained in at least one MCP. Individual identity was included as a random variable. We analysed whether each linear element was ever used (models with binomial error) and the number of radio locations that it contained (models with Poisson error). Species identity was included in all models as a fixed factor.


Dependence upon woody cover

We found considerable variability in habitat composition of mongoose home ranges (Fig. 1, Table 1), in spite of which standardized selection ratios indicated that scrubland was strongly preferred over any farmland habitat type at the home range level (-N ln(λ)=9·911, = 7, = 0·019; Table 2 & Fig. S1 in Supporting Information). Positive selection of scrubland was also evident when habitat use, indicated by radio locations, was compared with its availability within home ranges (-N ln(λ)=10·338, = 7, = 0·016; Fig. S1). On average, mongooses were located in scrubland 90% of the time (Table 1) and scrubland use was significantly higher than the use of other farmland habitat (Table 2). Within home ranges, individual variability in habitat use was quite low (Table 1).

Table 2.   Compositional analysis of habitat preference by Egyptian mongooses (= 7) and common genets (= 6) in the Guadiamar agroecosystem, south-western Spain. Habitat ranking matrices represent selection ratios at two levels: (1) habitat content of NNCH estimates of home range vs. habitat availability in the landscape sample and (2) habitat at animal radio locations vs. habitat availability in MCP estimates of home ranges. Signs denote a positive or negative departure from random use between habitat pairs, followed in brackets by the P-value of a t-test (Aebischer et al. 1993). Ranks range from 3 (most used) to 0 (least used)
UseHabitat content of home rangesHabitat at radio locations
Availability in the landscapeAvailability in the home range
DehesaOlive grovesCropsRankDehesaOlive grovesCropsRank
Egyptian mongoose
 Dehesa +(0·133)+(0·275)2 +(0·306)−(0·481)1
 Olive groves  +(0·926)1  −(0·041)0
 Crops   0   2
Common genet
 Dehesa +(0·192)+(0·007)2 −(0·398)+(0·280)1
 Olive groves  +(0·240)1  +(0·053)2
 Crops   0   0

Common genets included little or no crops within their ranges, but variability in the proportion of other habitats was substantial (Table 1). Standardized selection ratios showed that scrubland and dehesa were preferred by common genets over olive groves and crops at the landscape scale (-N ln(λ)=13·228, = 6, = 0·004; Fig. S1). Within their home ranges, all genets were located in scrubland over 71% of the time, indicating a clear preference for this habitat, whereas <12% of radio locations were recorded in farmland habitats (Table 1). No genet radio location was found in crops. The positive selection of scrubland and the negative selection of farmland within home ranges were significant (-N ln(λ)=15·006, = 6, = 0·002; Fig. S1). The differences in preference were significant for the scrubland-dehesa and scrubland-crops pairs (Table 2).

Use of farmland with trees by common genets

Olive groves were used less than expected from their availability at the two levels of selection examined by us. Out of 210 independent genet locations, only 5% were assigned to dehesa, whose availability in MCPs ranged between 10% and 71%. This avoidance of dehesa was unexpected given that genets included relatively large amounts of dehesa within their home ranges. It is possible that this result could be an artefact of the landscape structure in our study area, i.e. the spatial association between dehesa and scrubland (Fig. 1). To test this hypothesis, we calculated the proportion of dehesa in 1-km circles around 99 random points and the distance of its centre to the nearest scrubland. This demonstrated that the proportion of dehesa and the distance to scrubland were negative and significantly correlated (rs=−0·353, < 0·001), whereas the association with scrubland was positive and weak for the proportion of olive groves (rs=0·141, = 0·163) and crops (rs=0·181, = 0·073) in simulated home ranges.

Use of linear elements by Egyptian mongooses

With the exception of H12, the home range of which contained riparian forest almost exclusively (Fig. 1), the proportion of scrubland in mongoose convex hulls did not reach 30% (Table 3). The density of hedgerows in mongoose home ranges was 6–22 times higher than the mean density of hedgerows in simulated, random home ranges (Table 3) and these differences were significant (Wilcoxon test, = 2·201, = 6, = 0·028). No tree row was observed within mongoose home ranges and mean density of tree lines in simulated convex hulls was in the range 0·6–0·8 m ha−1. The density of grassy linear elements was negligible in the home ranges of three mongooses, but significantly higher than in random ranges for three other individuals (Table 3). As the densities of grassy lines and hedgerows were independent in observed (rs=0·522, = 6, = 0·288) or simulated convex hulls (mean values, rs=0·574, = 6, = 0·234), the positive selection of grassy lines could be ecologically meaningful. However, the overall differences between observed and available density of grassy lines within mongoose home ranges were not significant (Wilcoxon test, = 0·943, = 6, = 0·345).

Table 3.   Proportion of scrubland and density of hedgerows and grassy lines in NNCH estimates of home ranges. The mean density of hedgerows and grassy lines in 99 randomly placed convex hulls of the same size and shape is given. Probability (P) that the density of linear elements in observed home ranges was lower than that of simulated ranges (ranking test)
IndividualScrublandDensity of hedgerows (m ha−1)Density of grassy lines (m ha−1)
Egyptian mongoose
 H30·2414·02·40·01 0·01·61·00
 H50·1718·33·20·01 0·02·11·00
 H100·2948·92·20·01 6·51·30·03
 H120·97 0·01·11·00 0·00·91·00
 H130·0137·52·00·01 0·41·80·40
Common genet
 G10·04 8·42·80·12 5·84·20·14
 G20·10 6·03·50·22 0·93·90·82
 G30·63 0·01·81·00 0·01·01·00
 G50·89 0·03·31·00 2·01·70·28
 G80·0116·03·40·01 0·51·90·61
 G90·0120·02·20·01 0·41·70·44

The size of mongoose convex hulls increased significantly as the proportion of scrubland decreased (rs=−0·786, = 7, = 0·036). All mongooses the ranges of which contained <30% scrubland used farmland with a density of hedgerows ≥14 m ha−1 (Table 3); such high densities occurred only in 16% of simulated home ranges (Fig. 3). The relationship between the proportion of scrubland and hedgerow density was indeed negative (Fig. 3). Mongoose convex hulls contained 0·5–17·4 ha of woody vegetation, of which 0·0–0·9 ha corresponded to hedgerows. One resident mongoose (H13) did not use the riparian forest at all and lived exclusively in hedgerows covering 0·53 ha (Table 3, Fig. 3).

Figure 3.

 Relationship between density of hedgerows and proportion of scrubland in observed (large symbols) and simulated (small symbols) home ranges (convex hulls) for Egyptian mongooses (top) and common genets (bottom). Y-axis was truncated at 50 m ha−1 (a few larger simulated values occurred at X = 0).

We recorded 43 independent mongoose locations in farmland >30 m away from the scrubland edge. For mongooses the ranges of which contained linear elements, locations fell in them more often than expected. Farmland radio locations were significantly closer to linear elements than expected from their availability (Table 4) and these differences were significant across individuals (Wilcoxon test, = 2·201, = 6, = 0·028).

Table 4.   Number of radio locations in linear elements, mean width of hedgerows at the radio location and mean distance of radio locations in farmland to the nearest linear element. Mean values are given for 99 replicates of an equal number of random radio locations within MCPs or, for widths, the hedgerows they contain. Probability (P) that observed values were lower (number of radio locations, width) or higher (distance) than in simulated ranges. Radio locations within 30 m of scrubland were excluded from the analyses
IndividualRadio locations in linear elementsHedgerow widthDistance to hedgerow (m) for radio locations in farmland
Egyptian mongoose
 H300·0     110·0189·00·02
 H920·90·19    3108·3228·40·07
Common genet
 G100·5     4386·3228·00·98
 G300·0     2465·5411·60·79

Five mongooses were found 47 times in 12 different hedgerows the mean (±SE) length of which was 574 ± 165 m and mean width was 8·1 ± 2·1 m. During the study period, H13 lived in a set of one grassy linear element and three hedgerows 1·2–1·5 km long and 3·1–4·7 m wide. This individual selected stretches of hedgerow wider than expected from random (mean width: 7·8 m; Table 4). Mongooses that used hedgerows occasionally also chose spots wider than expected (6–24 m; Table 4; Wilcoxon test, = 1·826, = 4, = 0·068).

Use of linear elements by common genets

Two genets (G3 and G5) lived mostly in the riparian forest (Fig. 1). Their home ranges did not contain hedgerows (Table 3). Farmland was the dominant habitat (≥90%) within the ranges of the other genets, but hedgerow density was 2–10 times higher than the mean density in home ranges distributed randomly (Table 3). Genets that used large portions of farmland tended to place their home ranges in areas where hedgerow density was higher than random (Wilcoxon test, = 1·826, = 4, = 0·068). Tree lines within genet convex hulls were scarce or absent, while in simulated ranges, mean tree row density ranged between 0·7 and 1·0 m ha−1. Densities of grassy lines did not differ significantly between observed and simulated home ranges (Table 3; Wilcoxon test, = 0·730, = 4, = 0·465).

The size of genet home ranges tended to increase as the proportion of scrubland decreased (rs=−0·493, = 6, = 0·321). Genets the ranges of which contained <10% of scrubland used farmland with a density of hedgerows ≥6 m ha−1 (Table 3). This threshold density was found in only 24% of simulated ranges with <10% scrubland (Fig. 3). A significant negative correlation existed between hedgerow density and the proportion of scrubland within home ranges (Fig. 3; rs=−0·971, = 6, = 0·001). This relationship held even when two individuals that did not use hedgerows (Table 1) were removed from the analysis (rs=−0·949, = 4, = 0·051). The amount of woody vegetation within genet convex hulls varied in the range 0·4–32·1 ha, of which 0·0–0·7 ha appeared as hedgerows. Two genets (G8 and G9) placed their home ranges along hedgerows (0·48 and 0·42 ha respectively).

We recorded 92 genet radio locations in farmland beyond 30 m of the scrubland boundary. Three genets contained linear elements in their ranges and their radio locations in farmland were located in linear elements significantly more often than expected (Table 4). When in farmland, these genets were recorded at distances significantly closer to linear elements than expected (Table 4). Excluding genets G3 and G5, recorded mostly in riparian forest, farmland radio locations tended to be close to linear elements in spite of the small sample size (Wilcoxon test, = 1·826, = 4, = 0·068).

Three genets were found 67 times in 5 different hedgerows the mean (±SE) length of which was 1200 ± 170 m, and mean width was 3·9 ± 0·3 m. During the study period, G8 and G9 lived almost exclusively in three hedgerows 1·2–1·5 km long and 3·1–4·7 m wide. These genets selected stretches of hedgerow wider than expected from random (mean width ≥7·0 m). G2 was also found in spots wider than expected (Table 4).

Attributes of linear elements

The sign and significance of predictors were consistent in mixed models of occurrence and intensity of use (Table 5). Genets and mongooses did not differ in their selection of linear elements. Hedgerows were likely to be used at least once and were used with higher frequency than grassy lines (Table 5). Longer and broader linear elements were preferred within the range of values available (Table 5). Greater use was made of linear elements relatively close to paved roads (Table 5); this result can be explained by the location of a single highly used hedgerow about 100 m from a secondary road suggesting that moderate disturbance by traffic did not affect the use of hedgerows by both carnivore species.

Table 5.   Generalized linear mixed models of linear element use (left; binomial error) and use frequency (right; Poisson error) by Egyptian mongoose and common genet. Reference levels for factors Species and Type were ‘mongoose’ and ‘hedgerow’, respectively
PredictorOccurrence in linear elementsNumber of radio locations in linear elements
Intercept−13·194·90  −6·712·48  
Species (genet) −3·613·750·92  0·342−0·110·49 0·05  0·823
Type (grassy) −8·141·7920·69<0·001−1·930·71 7·44  0·009
Width0·600·1130·66<0·0010·100·03 9·12  0·004
Distance to road −1·360·4011·46  0·001−0·490·1121·36<0·001


Egyptian mongooses and common genets depended upon the scarce remnants of woody cover still present in the Guadiamar agroecosystem. These carnivores were able to establish enlarged home ranges outside riparian forest (the main source of woody cover) provided that a sufficient density of hedgerows was available. Both species used hedgerows preferentially suggesting that they provide valuable cover.

A variety of farmland habitats were avoided, including dehesa and olive groves that are rich in old trees, but contain little understorey. Open farmland with high canopy cover may not provide enough food or shelter, when compared with shrub cover. Therefore, matrix or unsuitable habitats for mongooses and genets can be defined as the absence of a layer of native shrubs in this landscape. Our results agree with circumstantial evidence indicating that genet latrines occur mostly in dehesas with understorey (Virgós & Casanovas 1997; Costa & Santos Reis 2002), where grazing might have been temporarily abandoned. In the south-west of the Iberian peninsula, dehesa is a widespread agroforestry system covering >3 million ha (Díaz, Campos & Pulido 1997). Dehesa may preserve some mammal species of Mediterranean forests (Díaz, Pulido & Marañón 2003), but its value for forest carnivores remains unclear.

The scarcity of nonlinear scrubland did not prevent genets and mongooses from inhabiting the Guadiamar agroecosystem. Whereas spatially structured habitats often result in spatially structured populations (Thomas & Kunin 1999), these carnivores were able to establish a continuous population, with adjacent home ranges, in a landscape containing as little as 4·7% of suitable habitat. We propose four explanations for this observation.

First, resident adults of both species included a high density of linear elements within their home ranges. Small vertebrates abound in riparian vegetation and hedgerows (Hinsley & Bellamy 2000; Maisonneuve & Rioux 2001) and therefore provide suitable foraging habitats for carnivores. Although many carnivores use woody, elongate landscape features regularly (Beier 1995; Tigas, van Vuren & Sauvajot 2002), they rarely rest or stay in them for long, suggesting that they may not provide sufficient shelter. This clearly was not the case for genets and mongooses in our study, the home ranges of which were entirely in hedgerows. Human activity in our study area was limited to farming and hunting, with little disturbance from traffic because of the distances (>100 m) of the occupied hedgerows from roads. Negative edge effects (risk of predation by dogs or humans) might be similar in hedgerows and riparian strips, despite their mean widths differing by one order of magnitude (4–40 m). The preference for relatively broad segments along hedgerows may be related to better refuge quality.

Secondly, most individuals in our study used several isolated linear elements suggesting that they could travel across the farmland matrix and may perceive the landscape in a fine-grained manner. This indicates that the agricultural mosaic was perceived by our study animals as functionally continuous.

Thirdly, physical connectivity or the relative proximity of linear woody features in the landscape may also play a role. In the study area, streams and associated riparian forest are <4 km apart and individuals could move between them, often making use of the hedgerow network. Hedgerows were distributed quite evenly across the area used by the study animals. The spatial distribution of hedgerows could be as important as their overall density as it determines the distance of open farmland gaps that animals have to cross. A regular distribution of hedgerows may allow a fairly even distribution of home ranges across the agricultural landscape and therefore continuous rather than structured carnivore populations. In turn, an even distribution of home ranges would facilitate a quick detection and refill of territory vacancies and the maintenance of a continuous occupation of the landscape. It remains unclear whether a more clumped hedgerow distribution, while keeping constant hedgerow density, would allow animals to occupy adjacent home ranges. Neither species occurs in the northern sector of the Guadiamar plain where hedgerows are absent and riparian vegetation is scarce (Rodríguez & Delibes 2003).

Finally, at the regional scale, genets and mongooses occupy a pine forest block 5–10 km south-west of the study landscape (Rodríguez & Delibes 2003). We found no evidence that the study population was sustained by immigration, but this forest could be a source of immigrant animals.

The length and density of hedgerows are constrained by field size. As in other Mediterranean agroecosystems (Concepción, Díaz & Baquero 2008), in our study area, hedgerows were longer (up to 1·9 km) and field sizes larger (up to 541 ha) than in traditional bocage landscapes of northern Europe with average field sizes of 0·5–3·0 ha and mean hedgerow lengths of 0·1–0·2 km (Deckers, Hermy & Muys 2004; Aviron et al. 2005). Therefore, in the typical agricultural landscapes of central and southern Spain, hedgerow density cannot reach the high values recorded in temperate Europe by simply restoring hedges along every field margin. However, our results may guide the design of specific agri-environment schemes for carnivores in Mediterranean agricultural landscapes characterized by large fields. Specifically, in a landscape with <5% of native woody vegetation, mostly in the form of riparian strips, an overall hedgerow density of 5 m ha−1 can support resident populations of common genet and Egyptian mongoose. A hedgerow density of 10–50 m ha−1 within landscapes containing <20% of riparian forest allows mongooses and genets to cross open farmland and to establish stable home ranges. Suitable hedgerows ranged between 0·5 and 2·0 km long and between 4 and 10 m wide, provided that stretches broader than 7 m occur in most hedge lines. Hedgerows should consist of native shrubs and may contain native trees, but tree lines without understorey are not suitable. A regular spatial distribution of hedgerows is preferable to an aggregated distribution (and may be crucial).

Hedge width can be manipulated more easily than length. Mean width of well preserved hedgerows approach 10 m (Fritz & Merriam 1996; Deckers et al. 2004), but those created under AES are seldom broader than 3 m (Tattersall et al. 2002; Bates & Harris 2009; Lye et al. 2009). While narrow hedges may favour arthropods and small vertebrates (Jehle & Arntzen 2000; Thomas et al. 2001; Tattersall et al. 2002), broader hedges were inhabited by resident genets and mongooses in our study.

The protection and restoration of linear remnants of native woody cover in farmland benefits some carnivores and many smaller organisms (Herlin & Fry 2000; van der Ree & Bennett 2003; Herzog et al. 2005), probably without reducing habitat quality for open land species. We conclude that, if enough linear elements are retained and their quality preserved, resident populations of two forest carnivores could live in an agroecosystem with a proportion of woody cover well below 10%. The tolerance of genets and mongooses to such open agricultural landscapes has not been reported previously. We note, however, that the survival of wild carnivores in agroecosystems requires habitat management at the landscape scale. Hedgerow management over large tracts of farmland may therefore resemble the ‘protected area’ approach (Whittingham 2007) within the wildlife-friendly farming solution to the global problem of agricultural expansion.


This research was funded by Consejería de Innovación, Ciencia y Empresa (grant P06-RNM-1903) and Consejería de Medio Ambiente, Junta de Andalucía, which also authorized animal capture, handling and tagging in compliance with regulations in force. MP was supported by the Ministry of Education through an FPU fellowship (AP2003-2370). We thank Miguel Solís and the Lazo family who let us work in their properties, Jolies Dortland and Luis León for field assistance, and the Associate Editor and two reviewers for helpful comments.