Conservation of European hares Lepus europaeus in Britain: is increasing habitat heterogeneity in farmland the answer?

Authors


Present address and correspondence: R. K. Smith, Department of Anthropology, University of Durham, 43 Old Elvet, Durham DH1 3HN, UK (fax: +44 (0)191 334 6175; e-mail: r.k.smith@durham.ac.uk).

Summary

  • 1Agricultural intensification has had dramatic effects on farmland biodiversity and has caused declines in many taxa. Habitat changes are thought to be the main cause of the decline in numbers of European hares, Lepus europaeus, throughout Europe. In Britain there is greater potential to increase hare numbers in pastural landscapes than in arable landscape. Hares in pasture have lower population densities, poorer body condition and participate less in breeding than in arable habitats. We aimed to investigate habitat selection and home range size in a mainly pastural area in order to reveal why the habitat is suboptimal, and how it could be managed to benefit the species.
  • 2A seasonal radio-tracking study was used to determine the importance of heterogeneity at the between- and within-habitat scales. Habitat selection by active and resting hares was quantified. Selection was investigated by categorizing habitats by type, and by structure in terms of vegetation height.
  • 3Mean home range size was 34 ha. Winter and spring ranges were larger than summer and autumn ranges. Hares selected fallow land and pasture grazed by cattle in preference to arable crops throughout the year, except during the winter when crops were suitable as forage. Pasture grazed by sheep was avoided in all seasons but winter. Heterogeneity at the between-habitat scale was less important to hares than heterogeneity at the within-habitat scale in the pastural landscape studied.
  • 4Hares selected habitats with taller vegetation during the spring and summer. Many of the habitats selected were heterogeneous in structure mainly due to cattle grazing, and hares avoided short homogeneous vegetation in all seasons. Hares are more likely to be limited by habitat in terms of cover than food in these landscapes.
  • 5Synthesis and applications. Increasing habitat heterogeneity at the farm scale may benefit hares, especially in highly homogeneous, intensively managed landscapes. However, managers of pastural farmland should aim to increase habitat heterogeneity at the within-habitat (or within-field) scale in particular, to provide better cover throughout the year. Agri-environment schemes should target the regeneration of heterogeneity in pastural landscapes, by encouraging changes such as an increase in fallow land and a reduction in livestock density. Such shifts in management are likely to benefit both hares and farmland biodiversity in general.

Introduction

Farmland biodiversity has declined in Europe over recent decades because of agricultural intensification. The intensification of farming, owing to an increase in efficiency and incentives to improve productivity, has brought with it homogeneity, both temporally and spatially, at landscape, farm and field scales (Benton, Vickery & Wilson 2003). In Britain there has been a large reduction in landscape diversity as agriculture has polarized; arable farming now dominates in the east and pastural farming in the west. Average field sizes have increased as hedgerows have been removed (Westmacott & Worthington 1997), and the intensive use of agrochemicals ensures a more uniform crop. Intensively managed grassland is structurally uniform throughout the year, owing to height-based management guidelines and buffer feeding strategies (offering hay, silage or straw mixtures in times of grass shortage), which are adopted to sustain maximum utilization and animal yield (Frame, Baker & Henderson 1995). On a temporal scale, factors such as the improved efficiency of machinery and the increase in farm size have brought about a synchronization of management practices, reducing heterogeneity between fields. Since habitat heterogeneity is associated with high biodiversity, it is likely that the change to more homogeneous landscapes has played an important role in the decrease in farmland biodiversity (Benton et al. 2003).

Farmland is the primary habitat of European hares, Lepus europaeus Pallas 1778, throughout Europe (Meriggi & Alieri 1989; Harris et al. 1995; Marboutin & Aebischer 1996). The hare is one of the many species that can be affected by the changes in habitats caused by agricultural intensification (Tapper & Parsons 1984; Slamečka 1991; McLaren, Hutchings & Harris 1997). Records of numbers of hares shot suggest that populations have declined throughout Europe, particularly since the 1960s (Pielowski & Pucek 1976; Tapper & Parsons 1984; Mitchell-Jones et al. 1999). As a result, the hare is protected under Appendix III of the Convention on the Conservation of European Wildlife and Natural Habitats (Bern Convention; Anonymous 1979). It is classed as a ‘priority species of conservation concern’ by the UK government, and has a Biodiversity Action Plan (BAP; Anonymous 1995). The BAP states the main factors causing the decline as (i) the conversion of grassland to arable, (ii) a loss of biodiversity in agricultural landscapes, and (iii) changes in cropping practices, including planting cereal crops in the autumn, and the move from hay to silage (Anonymous 1995). However, evidence is limited and other explanations for the decline of the hare include increases in predator populations (Tapper & Barnes 1986; Panek & Kamieniarz 1999) and changes in climate (e.g. increased precipitation; Hackländer, Arnold & Ruf 2002). The effects of changes in habitat, predation and climate, individually or collectively, on hare populations need to be quantified if the objective of the BAP ‘to maintain and expand existing populations, doubling spring numbers in Britain by 2010’ is to be met.

Throughout its wide geographical range in Europe, the hare is common in arable landscapes, and less common in non-arable areas such as pasture, uplands and woodland (Tapper & Parsons 1984; Hutchings & Harris 1996; Klansek et al. 1998; Vaughan et al. 2003). It is unclear why hares have declined in pastural landscapes, but there is greater potential to increase hare numbers in these areas than in arable landscapes (McLaren et al. 1997). Knowledge about the habitat requirements of hares in pastural landscapes is limited, as studies have mostly taken place in arable areas, where numbers are higher. Broad-scale studies have shown that higher numbers of hares are often associated with arable crops, in both arable and pastural landscapes (Hutchings & Harris 1996; Vaughan et al. 2003). On pastural farms hares are also associated with improved grassland and woodland (Vaughan et al. 2003). Broad-scale studies show which habitats are important to hares, but fine-scale studies are needed to gain an understanding of why particular habitats are important to the species, and how to implement conservation measures. On a mixed farm, hares only selected crops when the plants had recently emerged, and autumn hare numbers were positively related to landscape diversity (Tapper & Barnes 1986). Although diversity at the landscape and farm scale are considered to be important for hare numbers (e.g. Frylestam 1980a; Tapper & Barnes 1986; Lewandowski & Nowakowski 1993), diversity within fields has not yet been investigated.

In this paper we aim to identify the importance of specific habitats to hares in a predominantly pastural landscape. We investigate seasonal home range size and habitat selection by hares in relation to the heterogeneity of habitats at two spatial scales. In particular we test the hypotheses that heterogeneity is important (i) at the between-habitat, or farm scale, by investigating whether hares select arable and fallow land within a pastural landscape, and (ii) at the within-habitat, or field scale, by investigating the structure of fields selected by hares, in terms of vegetation height. The overall goal is to provide information to aid the management of pastural landscapes to reverse the decline in hare numbers.

Methods

study area

The study area consisted of mixed farmland (415 ha) in the mainly pastural county of Somerset, south-west England (Ordnance Survey grid reference of centre of site: ST 275425). It lies within the Bridgwater Bay Site of Special Scientific Interest (SSSI), an area designated by the Government because of its diverse flora and fauna. The density of hares at the study site was 15·9 ± 1·5 hares 100 ha−1 based on point-sampling spotlight counts (mean ± standard deviation; Péroux et al. 1998) during October each year, 2000–2002. The density of hares at the site is high compared with that in other pastural areas (mean based on day counts during October to January: 3 hares 100 ha−1; Hutchings & Harris 1996). The study population is subject to low levels of hunting with dogs in the form of coursing (hunting with sight hounds) and beagling (hunting with packs of scent hounds).

On average between August 2000 and September 2002, 23% of the study area was used for arable crops (wheat, barley, linseed, field beans, oilseed rape and fodder beet; mostly winter grown and harvested in summer), 8% for grass ley (temporary grass cut up to three times annually for silage), 6% for fallow land (including grass margins around arable fields and set-aside which is land taken out of production in return for European Union subsidies), and 63% was semi-improved grassland (mostly used for livestock grazing, some cut for hay in summer). The mean field size was 7·4 ± 2·7 ha for arable fields, 6·2 ± 1·6 ha for grass leys and 6·3 ± 2·7 ha for pastures (total n = 56 fields). These are relatively small; the average field size for pastural landscapes in Somerset and Dorset is 9·5 ha (Westmacott & Worthington 1997). Field enlargement may have been impeded as field boundaries consist primarily of ditches.

Detailed data on the habitat, including crop types, measurements of vegetation height (estimated average minimum and maximum) in each field and dates of harvest for crops, hay and silage, were collected every 2 weeks during intensive tracking periods and every month throughout the rest of the year. Livestock in each field were counted every week during intensive tracking periods, and once a month during the rest of the year. Cattle were the dominant livestock from spring to autumn; few were present during the winter. The area of grassland classed as sheep pasture increased from 11% in spring and summer to 18% in autumn and 34% in winter. Livestock units (LSU) were calculated using the average LSU for each category of stock (age and reproductive state; Nix 2001) weighted according to the number in each category kept in England and Wales, as stock categories were not known during the study. Livestock densities at the site (median 0·9, range 0·0–8·2 LSU ha−1) were typical for grassland in England and Wales (median 1·0, range 0·0–8·5 LSU ha−1; Vaughan et al. 2003). Crops and other habitat features were plotted using digital maps (1:10 000; Ordnance Survey scale raster maps), digital aerial photographs taken in July 1999 (Counties Revealed, Geoinformation Group, Cambridge, UK), and Geographical Information System software (ArcView GIS, Environmental Systems Research Institute, Aylesbury, UK).

capture techniques

Hares were captured between May 2000 and August 2002. We used static nets and cage traps to minimize apparent sex and age biases which existed in each method of capture, and because the methods were suitable in different habitat types. When using static nets, 5–30 beaters flushed hares from cover into nets 2–30 m long set across gateways or other routes used by hares. This method could be used only in fields without livestock or mature crops. In total 32 females and 10 males were caught, on average 3·2 hares per day (n = 13 days).

Cage traps (single catch, single entry, spring door traps, 100 × 40 × 40 cm3; Albi-traps, Wymondham, UK) were set in fields without livestock. Traps were unbaited, but the floor was covered with hay. They were checked at least once every 24 h, soon after dawn, i.e. immediately following the period of activity of hares. In total, 45 females and 42 males were caught using this method, averaging one hare per 16·6 trap-nights. A lower proportion was adults, 60% compared with 83% when using static nets. Adults were defined as hares > 7·7 months old and subadults 4–7·7 months; juveniles and leverets (< 4 months) are not considered here. Ages were determined by the presence of an epiphyseal protrusion of the lateral ulnar knob in animals aged ≤ 234 days, i.e. 7·7 months (Stroh 1931), the hind foot length, the skull length and the body weight (Bray, Champely & Soyez 2002; Hackländer et al. 2002).

Hares captured were placed in cloth bags, sexed, aged and weighed. Each was fitted with two coloured and numbered plastic ear tags (Jumbotag or Rototag; Dalton, Henley-on-Thames, UK) and a radio collar (Biotrack, Wareham, UK). The combined weight of ear tags and collar was less than 5% of the body weight. Hares were released at the site of capture within 5 min of removal from the net or cage trap. The capture and handling of animals were carried out in accordance with the Animals (Scientific Procedures) Act 1986, following advice from the University of Bristol Ethical Review Committee. During the study 15–30% of the population were radio-collared at any one time.

data collection

Individual radio-collared hares were located at least 30 times during their active period (mainly at night) and 30 times during their inactive period (day), during each seasonal intensive tracking period. Active and inactive periods were defined following Holley (2001). Seasons were defined as spring: intensive breeding (2 May to 12 June 2002); summer: some hares breeding (1 August to 5 September 2000 and 2002); autumn: no breeding (1 November to 13 December 2000); winter: early breeding (Lincoln 1974; 1 February to 13 March 2002). Different hares were tracked in each season; spring: 6 females and 5 males; summer: 8 females and 3 males; autumn: 8 females and 3 males; winter: 5 females and 5 males. Seventy per cent of hares were collared during the month leading up to the intensive tracking periods, the rest were collared between 1 and 6 months before intensive tracking commenced. Hares were located using a Suretrack STR 1000 receiver (Lotek Engineering, Ontario, Canada) and a three-element VHF Yagi antenna (Biotrack). Triangulation methods were used to locate each hare once during each inactive period and usually once but up to a maximum of three times during each active period (each separated by at least 3 h). Radio locations (fixes) were therefore discontinuous and independent (Harris et al. 1990), and were accurate to a 50 × 50 m2 grid. For each radio fix, information was collected about the vegetation type and height, and the number and type of livestock in the field. If a hare was not resting in its form when located during the day (<1%) the radio fix was omitted from analysis. All radio fixes collected during the active period were included in analyses.

statistical analysis

Home range size

For each hare, the 100% minimum convex polygon (MCP; Ranges V software, CEH, Dorchester, Dorset, UK) was used to describe home range size, to allow comparison with earlier studies. These 100% MCPs will be referred to as ‘home ranges’. Kernel methods, using least squares cross-validation to select the smoothing parameter, were employed to estimate separate active and inactive 95% kernel range sizes (95% of radio fixes), which are referred to as ‘ranges’, and core areas (50% of radio fixes), which are referred to as ‘core areas’ throughout. These kernel estimates were used for all analyses, as this method provides robust estimates (Worton 1989; Seaman & Powell 1996; Animal Movement, USGS – BRD, Alaska Science Center, Anchorage, Alaska; Arcview GIS 3·2). Range sizes were compared using repeated measures analysis of variance (anovas), in which activity (active or inactive) was a within-subject factor, and season and sex were between-subject factors. Variables were transformed if necessary to conform to the assumptions of anova (Zar 1999) and were checked for sphericity (Mauchly's test) and homogeneity of variances (Levene's test) (Field 2000). All interaction terms were initially included, then where possible non-significant interaction terms were omitted and the analyses were repeated; results from the final models are shown here. Tukey post-hoc tests were used to compare means for seasons. Statistical analysis was carried out with a significance level of 5% unless stated otherwise (SPSS for Windows, Release 10, SPSS Inc., Chicago, IL).

Habitat selection

Compositional analysis was used to compare utilized and available habitats at three levels during each season: (i) range selection (active and inactive) within the study area, (ii) core area selection (active and inactive) within the study area, and (iii) habitat selection (active and inactive radio fixes) within ranges. Compositional analysis allows proportions describing the utilization of each habitat category (which sum to one) to be analysed as independent data by log-ratio transformation. Tests for departure from random habitat utilization were carried out using likelihood ratios. Habitats were then ranked in order of relative utilization, and significant differences were identified (Aebischer, Robertson & Kenward 1993). Analysis was carried out using the ‘Compositional Analysis Add-In Tool’ (Version 4·1; Peter Smith, Abergavenny, Wales, UK). Proportions of ‘utilized’ habitats equal to zero were substituted by 0·001 (Aebischer et al. 1993). When comparing the availability and utilization of habitats within ranges, if the availability for a habitat was zero for ≥50% of the individuals in a particular season it was excluded from analysis (Aebischer et al. 1993; approximately two were excluded per season). Not all individuals had all remaining habitats within their ranges, and hares with >60% of habitats unavailable to them were also removed from analysis. Wilks’ lambda was calculated as a weighted mean for remaining proportions of ‘available’ habitats equal to zero (Aebischer et al. 1993). Data randomization was used to calculate levels of significance because of the departure from multivariate normality of log-ratio difference distributions (Aebischer et al. 1993). Randomization was based on 1000 iterations (Manly 1997).

For the investigation into habitat utilization, habitats were classified in two ways. To investigate selection at the between-habitat scale, habitats were categorized by vegetation type (or use in the case of pasture) using the following categories: pasture with sheep (0·06–1·15 LSU ha−1), pasture with a low (0–0·99 LSU ha−1), medium (1·00–1·99 LSU ha−1) and high (>2·00 LSU ha−1) density of cattle, grass ley, cereal crops, non-cereal crops and fallow land. To investigate selection at the within-habitat scale, habitats were categorized by habitat structure in terms of vegetation height. Homogeneous vegetation was classified as: short (<70 mm), medium (70–220 mm) and tall (>220 mm). For heterogeneous vegetation, where the vegetation tended to be at two different heights, we recorded the minimum and maximum heights: short/medium, short/tall and medium/tall. In 67% of cases the maximum vegetation height covered the greatest area in the field, and the median percentage of vegetation at the maximum height was 75%.

Log-ratio data (transformed utilization proportions) calculated during the compositional analysis were analysed using repeated measures anovas in which activity was a within-subject factor, and season and sex were between-subject factors. Variables for which activity, season or sex had significant effects were then analysed using a doubly multivariate repeated measures model (Field 2000). Pillai's trace test statistic was used, as it is robust to deviations from the assumption of equal population variances (Quinn & Keough 2002).

Results

home range size

Home ranges calculated as 100% MCPs using both active and inactive radio fixes were similar in size to active ranges, as inactive ranges tended to fall within active ranges (Fig. 1). Active ranges were significantly larger than inactive ranges (95% kernel ranges; active 36 ± 26 ha, inactive 19 ± 13 ha; Table 1, Fig. 1). Ranges were significantly larger in winter and spring than in summer and autumn (Table 1, Figs 1 and 2).

Figure 1.

Mean range sizes of adult and subadult hares in the four seasons (n = 43). Grey bars are 100% MCP home ranges, black bars are 95% kernel active ranges, and white bars are 95% kernel inactive ranges. Vertical bars show one standard deviation. Range sizes that are not significantly different from one another are indicated with the same lower-case letter.

Table 1.  Repeated measures anova on range size for within-subject effect: activity (active and inactive) and between-subject effects: sex and season
Source of variationMSFd.f.P
  1. d.f. = degrees of freedom, MS = mean square, * = interaction term.

Activity50·73233·149 10·000
Activity * season 3·735 2·440 30·079
Activity * sex 2·522 1·648 10·207
Error (within-subject) 1·530 38 
Sex 3·043 2·480 10·124
Season15·40212·552 30·000
Error (between-subject) 1·227 38 
Figure 2.

Home ranges (thick black lines = 100% MCPs, active and inactive radio fixes) of adult and subadult hares in the four seasons (n = 43): (a) spring, (b) summer, (c) autumn, (d) winter. Arable crops are shown as light grey, grass leys as dark grey, and pastures (with and without livestock) as white. Field boundaries are shown as thin black lines.

habitat selection

Between-habitat scale

The percentage habitat composition of ranges (utilized) compared with that of the study area (available) was significantly different from random for active and inactive ranges in all seasons (Table 2). Habitats were ranked in order of relative utilization, and significant differences between habitats were identified (Table 3). Cattle pasture and fallow land were ranked higher than all other habitats for active and inactive hares in all seasons except winter, when sheep pasture ranked above fallow land, and grass ley ranked highest for active hares. Sheep pasture was one of the least selected habitats during all other seasons. There was no selection for pasture with a particular density of livestock (low including no cattle, medium or high).

Table 2.  Compositional analysis (results of manova). The habitat composition of 95% kernel ranges (active and inactive), in terms of vegetation type, was different from random within the study area in all seasons
  Wilks’Λχ2d.f.P
ActiveSpring0·02938·95460·004
Summer0·07229·01970·046
Autumn0·07428·63550·007
Winter0·03932·41550·010
InactiveSpring0·02939·14960·006
Summer0·06629·85070·033
Autumn0·09825·55250·012
Winter0·11022·11150·019
Table 3.  Habitat types ranked according to relative utilization, most utilized is at the top of the list: (a) 95% kernel ranges vs. total study area (b) 50% core areas vs. total study site (c) radio fixes vs. 95% kernel ranges. Habitat categories that are not significantly different from one another are indicated with the same lower-case letter
ActiveInactive
SpringSummerAutumnWinterSpringSummerAutumnWinter
  1. Non-cereal = non-cereal arable fields, NS = not significantly different from random utilization.

(a) 95% kernel ranges vs. total study area
High stockaLow stockaLow–medium stockaLeyaHigh stockaLow stockaLow–medium stockaLow stocka
FallowabFallowaFallowaLow stockabLow–medium stockabMedium stockaFallowaSheepab
Low–medium stockabcMedium stockabLeyabSheepabFallowabHigh stockaLeyabFallowb
Non-cerealbcdHigh stockabcCerealbcFallowabNon-cerealbcFallowabCerealbNon-cerealab
CerealcdCerealabcNon-cerealcNon-cerealabCerealbcLeyabNon-cerealbCerealb
LeycdLeybcSheepcCerealbLeycCerealabSheepbLeyb
SheepdSheepbc  SheepcSheepab  
 Non-cerealc   Non-cerealb  
(b) 50% core areas vs. total study area
  LeyaLow stockaHigh stockaMedium stockabLow–medium stockaLow stocka
  Low–medium stockabFallowabLow–medium stockabHigh stockaLeyabSheepabc
NSNSFallowaCerealabNon-cerealabFallowaFallowabNon-cerealabc
Non-cerealbLeyabCerealabLow stockabNon-cerealbFallowb
CerealbcSheepabFallowaSheepaSheepbCerealabc
SheepcNon-cerealbSheepLeybCerealab Leya Non-cerealbCerealbLeyc
(c) Radio fixes vs. 95% kernel ranges
  Low–medium stockaLeyaHigh stocka   
  LeyaCerealaLow–medium stockab  NS
NSNSFallowaLow stockbFallowbNSNS 
CerealaSheepb Fallowb Non-cerealb    

Habitat composition of core areas compared with the study area showed similar results to those for range selection. The main differences were that the composition of core areas was not different from random for active hares in spring and summer, and that ley was ranked higher than pasture and/or fallow land in autumn rather than winter (Table 3). Since results for ranges and core areas were similar, further analyses and discussion of habitat selection focus on ranges, as use of the wider habitat is considered important for management implications.

To investigate whether there was an effect of activity, season or sex on habitat selection for ranges, the three log-ratio variables for which one or more factors showed significant effects (sex showed none; significance level 10%) were entered into a doubly multivariate repeated measures model. Low, medium and high cattle density categories were combined, so that seasons were comparable. The model showed a significant overall effect of activity (Pillai's trace = 0·221, F3 = 3·412, P = 0·028) and season (Pillai's trace = 0·514, F9 = 2·618, P = 0·009), but no effect of sex (Pillai's trace = 0·067, F3 = 0·862, P = 0·470), the activity * season interaction (Pillai's trace = 0·096, F9 = 0·417, P = 0·924) or the activity * sex interaction term (Pillai's trace = 0·060, F3 = 0·764, P = 0·522). Therefore, selection of habitats for ranges depended on whether the hares were active or inactive, and on the season.

The comparison of radio fixes (utilized) with habitat composition of ranges (available) showed that habitat utilization was significantly different from random only for active hares in autumn and winter, and for inactive hares in spring (Table 4). Where selection was significantly different from random, habitats were ranked in order of relative utilization, and significant differences between habitats were identified (Table 3). Selection of habitats in autumn (active) and spring (inactive) were similar to the selection of ranges described above. However, in winter cereal, which was selected least by active hares for ranges, was ranked second highest when selecting habitats within these ranges.

Table 4.  Compositional analysis (results of manova). Seasonal habitat utilization (active and inactive radio fixes), in terms of vegetation type, was at random within 95% kernel ranges in all but three cases
  Weighted mean Λd.f.P
ActiveSpring0·80920·570
Summer0·36740·732
Autumn0·16030·031
Winter0·02850·033
InactiveSpring0·39920·041
Summer0·03240·266
Autumn0·10330·054
Winter0·50130·835

Within-habitat scale

Habitat composition of ranges, in terms of habitat structure, compared with that of the study area was significantly different from random for active and inactive ranges in all seasons (Table 5). Habitat structures were ranked in order of relative utilization, and significant differences between habitat structures were identified (Table 6). In spring and summer, medium/tall habitat ranked in the top two for active and inactive hares, but had one of the lowest ranks in autumn and winter. In autumn, medium and short/medium habitats ranked in the top two for active and inactive hares, and for inactive hares in winter. In winter, tall habitat ranked higher than short/medium habitat.

Table 5.  Compositional analysis (results of manova). The composition of 95% kernel ranges (active and inactive), in terms of habitat structure, was different from random within the study area in all seasons
  Wilks’Λχ2d.f.P
ActiveSpring0·06230·61750·007
Summer0·07129·04950·006
Autumn0·06829·55650·011
Winter0·03134·70350·011
InactiveSpring0·12223·11950·015
Summer0·00461·02750·002
Autumn0·19318·09150·043
Winter0·00356·94950·002
Table 6.  Habitats categorized by vegetation height (minimum/maximum) ranked according to relative utilization, most utilized is at the top of the list. Habitat categories that are not significantly different from one another are indicated with the same lower-case letter. Results for core areas vs. total study area are not shown as they are the same as for 95% kernel ranges except that habitat utilization was not different from random in spring
ActiveInactive
SpringSummerAutumnWinterSpringSummerAutumnWinter
  1. NS = not significantly different from random utilization.

(a) 95% kernel ranges vs. total study area
Medium/tallaMedium/tallaMediumaMediumaMedium/tallaShort/tallaShort/mediumaMediuma
TallabShort/tallabShort/mediumaTallaShort/mediumabMedium/tallabMediumaShort/mediumab
Short/mediumabTallbcShortabShort/mediumabTallabTallbcTallabTallac
MediumabMediumbcTallbShortbcMediumabcShort/mediumbcdShortbShort/tallcd
Short/tallbShort/mediumcdMedium/tallbShort/tallcShort/tallbcMedcdMedium/tallbShortbd
ShortbShortdShort/tallbMedium/tallcShortcShortdShort/tallbMedium/talld
(b) Radio fixes vs. 95% kernel ranges
  Mediuma Mediuma    
NSNSShort/mediumabNSTallabNSNSNS
  Shortb Medium/tallb   
    Short/mediumb   

Habitat composition of core areas compared to the study area showed almost identical results to those for range selection (see Table 6), except that habitat composition of core areas was not different from random for hares in spring. Further analyses focus on habitat selection for ranges rather than core areas.

Four log-ratio variables for which activity or season had significant effects (sex showed none; significance level of 10%) on habitat selection for ranges were entered into a doubly multivariate repeated measures model. The model showed a significant overall effect of activity (Pillai's trace = 0·283, F4 = 3·460, P = 0·017) and season (Pillai's trace = 1·481, F12 = 9·016, P < 0·001), but no effect of sex (Pillai's trace = 0·018, F4 = 0·160, P = 0·957), the activity * season interaction (Pillai's trace = 0·376, F12 = 1·324, P = 0·215) or the activity * sex interaction (Pillai's trace = 0·050, F4 = 0·460, P = 0·764). Therefore, selection of habitats, in terms of structure, for ranges depended on whether the hares were active or inactive, and on the season.

The comparison of radio fixes with habitat composition of ranges showed that habitat utilization was significantly different from random only for active hares in autumn, and for inactive hares in spring (Table 7). Where selection was significantly different from random, habitats were ranked in order of relative utilization, and significant differences between habitats were identified (Table 6). Selection of habitats in autumn (active) was similar to the selection of ranges described above. However, the ranks of habitat structures within ranges in spring (inactive) were almost the opposite of those for the range within the study site for the same season.

Table 7.  Compositional analysis (results of manova). Seasonal habitat utilization (active and inactive radio fixes), in terms of habitat structure, was at random within 95% kernel ranges in all but two cases
  Weighted mean Λd.f.P
ActiveSpring0·34330·189
Summer0·31430·142
Autumn0·27920·039
Winter0·44320·104
InactiveSpring0·19430·019
Summer0·45620·098
Autumn0·94920·807
Winter0·46830·370

Discussion

heterogeneity at the between-habitat scale

Our data on seasonal changes in habitat selection suggest that hares select a variety of habitats for both feeding and resting throughout the year, as found by Tapper & Barnes (1986). When habitats were suitable for both, hares often used the same one for feeding and resting, as found by Reitz & Léonard (1994) and Marboutin & Aebischer (1996). For most of the year hares did not select arable habitats over other habitat types; instead they selected pasture grazed by cattle (0·0–8·2 LSU ha−1) and fallow land, both when foraging and resting. Of the 43 home ranges, 14% of ranges and 40% of core areas were made up entirely of pasture and fallow land, even though arable crops were available nearby, and covered 31% of the site. Ranges without arable crops did not occur during the winter, but 30% of winter core areas contained no arable crops. Hares selected arable habitats (cereal and grass ley crops) above others only when crops were short and suitable as forage, a preference also found by Tapper & Barnes (1986).

In arable landscapes hares are associated with set-aside (Frylestam 1992; Hutchings & Harris 1996; Vaughan et al. 2003), although the presence of set-aside has no effect on hare numbers in pastural landscapes (Hutchings & Harris 1996). In this study hares selected cattle pasture and fallow land (including set-aside) equally.

Although broad-scale studies have shown that higher numbers of hares are associated with arable crops within pastural landscapes (Hutchings & Harris 1996; Vaughan et al. 2003), at the finer scale hares do not appear to be dependent on these habitats. This suggests that heterogeneity at the between-habitat scale is of limited importance to hares in the pastural landscape studied. However, where grassland is managed more intensively than at our site, the vegetation is often short and homogeneous throughout the year (Stoate 1996; Vickery et al. 1999). In this case the diversity of habitat types available may become more important in providing hares with the resources they require, meaning that they rely on arable crops to a greater extent in these areas. In intensively farmed arable landscapes, where fields are very large, hares have much larger home ranges than in less intensive arable areas (Reitz & Léonard 1994; Marboutin & Aebischer 1996), suggesting that individuals increase their home ranges to include a diversity of habitats.

A need for heterogeneity at the between-habitat scale in intensively managed pastural landscapes may be due to requirements for food. Our mean home range size fits the trend shown by other studies in agricultural landscapes, for home ranges to decrease in size as the landscape changes from intensively managed arable land to pasture (Table 8). We found that ranges at our study site were largest during winter, as did Reitz & Léonard (1994), and during spring. During the winter, vegetation growth is likely to be limited, and hares selected arable crops in addition to pasture and fallow land for forage. Hares may have needed to move over a larger area to gain access to arable crops as these cover just 31% of the area of the site. However, in spring, hares selected pasture and fallow land, not arable crops, and so are unlikely to have needed to range as far to find food. In the summer and autumn, arable crops are not suitable as forage, and in autumn, as vegetation growth slows, food availability is likely to be at its lowest, and yet ranges were small. Although changes in food availability do not seem to explain variations in range size, behavioural changes may do. At the beginning of the breeding season (winter), and during peak breeding (spring), females are searching for suitable sites in which to give birth, and males are searching for females. Low levels of mating activity occur in summer and autumn (Lincoln 1974), when we found range sizes to be at their minimum.

Table 8.  Home range sizes of hares in agricultural landscapes determined by radio-tracking
LocationHabitat% grassMethod of calculationMonthsnMean home range size (ha)Source
  1. Months = number of months animals were tracked, n = number of animals tracked, A = active home range, A + I = active and inactive radio fixes, i.e. total home range.

FranceIntensive arable land<11% 95% MCP< 520138 (A)Marboutin & Aebischer (1996)
FranceIntensive arable land<6%100% MCP> 621 61 (A)Reitz & Léonard (1994)
The NetherlandsArable land100% MCP1–4 3 39 (A)Broekhuizen & Maaskamp (1982)
HungaryAgricultural land100% MCP1–612 37 (A + I)Kovács & Búza (1988)
OxfordshireAgricultural land25%100% MCP285 25 (A)Bradshaw (1993)
EnglandAgricultural land≈50% 90% isopleth1–715 38 (A + I)Tapper & Barnes (1986)
EnglandAgricultural land63%100% MCP143 29 (A)This study
The NetherlandsCattle pastureMajority100% MCP1–1410 26 (A)Broekhuizen & Maaskamp (1982)

heterogeneity at the within-habitat scale

A need for heterogeneity at all habitat scales may be due to requirements for cover. In this study hares selected specific habitat structures over others, and this depended on whether they were foraging or resting, and on the time of year. During spring and summer, hares tended to select habitat structures with taller minimum and maximum vegetation heights than those selected during the autumn and winter. Many of the habitats selected were heterogeneous in structure with different minimum and maximum vegetation heights, although tall homogeneous habitats ranked fairly highly in all seasons. In all seasons except winter, but particularly during the spring and summer, hares avoided both short, even vegetation and pasture grazed by sheep (characterized by a short homogeneous structure; Fuller & Gough 1999; Vickery et al. 1999; Benton et al. 2003). This suggests that during the main part of the breeding season, hares select habitat structures that provide more cover from predators and unfavourable weather conditions, which is particularly important for the survival of leverets (Tapper & Parsons 1984). In Poland, foxes Vulpes vulpes Linnaeus consume more hares during the spring, when leverets are abundant, than during other months (Goszczyński & Wasilewski 1992). In addition, very low temperatures or moderate temperatures combined with high precipitation cause mortality in leverets, particularly during the first 2 weeks of life (Hackländer et al. 2002). Weather conditions also influence the selection of resting habitats by adults; in sunny weather hares are twice as likely to use long crops as when it is raining (Tapper & Barnes 1986).

Our data suggest that heterogeneity at the within-habitat scale is important to hares in a pastural landscape. The heterogeneous vegetation structure produced by different levels of cattle grazing pressure (up to densities of 8·2 LSU ha−1) and by fields used for hay provided the habitat structures preferred by hares throughout the year. The exception was tall homogeneous habitat, which was also selected but which was provided by arable crops. Hares selected fields based on the sward structure created by cattle grazing, and not on the cattle themselves. Of all the radio fixes recorded in fields in which cattle were present for some of the intensive tracking period but not continuously, 39% of those in spring, and 70% of those in summer were recorded when cattle were absent (spring, 49%; summer, 58% of the time). In both seasons 60% of all radio fixes recorded in cattle pasture were recorded in these fields (60% of cattle fields). Hares selected pastures regardless of cattle density, and did not appear to avoid cattle as suggested by previous studies (Pielowski & Pucek 1976; Barnes, Tapper & Williams 1983; Pépin 1985). In intensively managed pastural land, where habitats tend to be more homogeneous throughout the year, hares are likely to be limited by habitat quality in terms of the availability of cover. In Poland, autumn densities of hares increased with the number of permanent cover areas (Panek & Kamieniarz 1999). At our site hares may have been limited by the availability of cover to some extent as tall homogeneous habitats, which were selected in all seasons but summer, were not readily available to hares during the autumn and winter (mean, 8% area of site).

implications for management

At our study site hares maintained relatively small home ranges, and made use of the diversity of different vegetation types and vegetation heights available in mixed farming. The fact that hares had access to heterogeneous habitat structures, providing cover throughout the year, may help to explain why hare numbers at this site are relatively high for pastural landscapes. Results suggest that habitat heterogeneity at the between-habitat scale is of some benefit to hares, and may be more important in intensively managed pastural landscapes. Although hares tended to select pasture and fallow land throughout the year, arable crops did provide good cover during the breeding season, and forage during the winter. This may explain why hare numbers are relatively high in pastural areas where some arable crops are present (Hutchings & Harris 1996). However, data collected during this study did not support the view that hares are limited by the distribution and abundance of arable habitats in pastural landscapes (Hutchings & Harris 1996; Mclaren et al. 1997). This suggests that increasing heterogeneity at the between-habitat, or farm scale, by increasing arable land within pastural landscapes, may not have significant effects on hare numbers.

Our results showed that hares in the pastural landscape studied select for heterogeneity at the within-habitat, or field scale. Throughout the year hares tended to select heterogeneous habitat structures, such as pasture grazed by cattle and fallow land, in preference to homogeneous structures, such as arable crops and sheep-grazed pasture. In arable landscapes, numbers of hares are higher with the presence of set-aside (Hutchings & Harris 1996), which increases heterogeneity at the between- and within-habitat scale. Unimproved grassland, which has a heterogeneous structure, is strongly associated with high hare numbers in both arable and pastural landscapes (Hutchings & Harris 1996). Further research is now required to investigate habitat utilization, in terms of vegetation structure, and hare abundance in pastural landscapes with different levels of heterogeneity. Hares not only required varying habitat structures when foraging and resting, but also selected habitat structures depending on season, suggesting that both spatial and temporal heterogeneity are important to the species at the within-habitat scale.

Hares in pastural landscapes are in poorer condition than those in arable landscapes (Frylestam 1980b). They are therefore likely to be more susceptible to mortality by predation, disease, exposure and hunting by humans. Further studies need to be conducted in order to identify the most common causes of mortality in adult hares in pastural areas, and the effects of vegetation structure on these, so that conservation measures can be implemented in order to increase survival. Programmes of habitat change that increased permanent cover and diversity for European hares, or food availability for snowshoe hares, Lepus americanus Erxleben, showed that increased numbers of hares can result from improving habitat quality without manipulation of predator numbers (Slamečka 1991; O'Donoghue & Krebs 1992).

In order to slow the decline in hare numbers, management strategies need to be developed which will start to undo some of the adverse effects of agricultural intensification (Smith, Jennings & Harris 2005). Farmland needs to be managed in a way that will re-establish both temporal and spatial heterogeneity. Hare abundance is positively associated with habitat diversity throughout Europe (Smith et al. 2005). Habitat heterogeneity is important to farmland biodiversity as a whole as it ensures resources are available to a wide variety of species throughout the year (Vickery et al. 2001; Robinson & Sutherland 2002; Benton et al. 2003).

In pastural landscapes there needs to be a move away from the management of grassland for maximum utilization, which produces structural uniformity between and within fields throughout the year. High-intensity grazing by cattle and particularly sheep causes homogeneity within grassland in terms of structure and composition (Vickery et al. 1999). Support should be given to farmers to make the restoration of lower-input, more extensive, livestock systems more feasible (Vickery et al. 2001). Silage making affects grassland structure and biodiversity, and frequent cutting causes higher levels of disturbance than more traditional haymaking. In arable landscapes the management of fallow land, including grass margins, for biodiversity is common practice. This may also be of benefit in intensive pastural systems (Vickery et al. 2001; Mcintyre, Heard & Martin 2003). At present agri-environment schemes do not target the creation and management of heterogeneity on farmland; if they did it would be an important step towards reversing the decline of farmland biodiversity (Benton et al. 2003). This study suggests that the hare may be one of the species that would benefit from such a change in farmland management.

Acknowledgements

We thank all landowners who allowed us access to their land. We are very grateful to the many volunteers who helped us to catch and track hares including: Herbie and Simon Wharton, Jennie Allen, Emily Bennitt, Tom Burditt, Jo Clifton, Jane DeGabriel, Fiona Luckhurst and Pierre Péron. We also thank Tony Holley, Eric Marboutin and Piran White. This research was funded by the Dulverton Trust and DEFRA (grant number BD1436).

Ancillary