Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Bird species turnover and stochastic extinction in woodland fragments

Bird species turnover and stochastic extinction in woodland fragments Hinsley, S A , Bellamy, P E and Newton, I 1995 Bird species turnover and stochdsuc extinction in woodland fragments - Ecography 18 41-50 Year-to-year turnover in bird species composiuon was recorded across, the whole size range (0 02-30 ha) of 146 woods studied The mean number of resident breeding species both lost and gained per wood between consecutive breeding seasons was 2 (range 0-8) No relationship was found between this absolute turnover rate and woodland area, or any other of 24 predictor vanables (descnbing woodland structure, isolation, connectedness and surrounding land use) Extmcuon and colonisation rates (in terms of numbers of species lost and gained) were also unrelated to woodland area In all sizes of woods, the species most likely to show local extinctions and colonisations were those with small populaUons within those woods, but the ldenuty of the species concerned changed as woodland area increased In the smallest woods, the majonty of turnover involved common species, such as wren and dunnock, which occurred in only small numbers in these small woods As woodland area increased, these species attained sufficient numbers to usually avoid stochastic extincUon The majonty of turnover was then due to more specialist (and less numerous) woodland species, such as great-spotted woodpecker and marsh Ut, which were usually lacking in small woods In Bntain, much existing broadleaved woodland falls within the size range studied Thus the numbers of many bird species are liable to be small enough for yearly turnover in woodland bird communities to be appreciable, and for the long-term persistence of individual species in particular woods to depend on dispersal S A Hinsley. P E Bellamy and 1 Newton. NERC, Inst of Terrestnal Ecology. Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire, UK PEI7 2LS TntrnHiiptinn iniroaucuon In this paper, we examine turnover in bird spiecies composition in woods of different sizes and degrees of lsolation, set within an agricultural landscape in eastern England, We also investigate the influence of the charactenstics of the woods themselves and of the surrounding landscape on the stability of species composition and on local extinction and colonisation rates Habitat loss is one of the major causes of a worldwide decline in biodiversity (lUCN/UNEPAVWF 1991, Roistad 1991) This process frequently begins with the fragmentation of a natural habitat, to be followed sooner or later by the degradation and eventual loss of the firagments themselves Many studies have examined the efAccepted 23 November 1994 Copynght © ECOGRAPHY 1995 ISSN 0906-7590 Pnnted in Denmark - all nghts reserved ECOGRAPHY 18 I (1995) ^^'''^ °^ fragmentation on forest plant and animal commu^^^^ ^^ ^ Weaver and Kellman 1981, Hams 1984, Angelstam 1992), although the forests concerned have mostly been large, of the order of hundreds or thousands of hectares Some of the most frequently cited studies concern the bird communities of the North Amencan eastern deciduous forest, in which a long-term decline of forest lntenor bird species has been attnbuted to forest fragmentation (Robbins 1980, Butcher et al 1981, Whitcomb et al, 1981, Temple and Cary 1988, Memam and Wegner 1992) In Bntain, the process of woodland fragmentation, viewed from a North Amencan or Scandinavian perspective, is extreme, there being few remaining extensive tracts of semi-natural woodland, especially in the low- lands (Wilcove et al 1986) In England, woodland once covered almost the whole country, but now accounts for only c 7% of the total area, of which 40% consists mainly of conifers (Locke 1987) A large prof)ortion of the avifauna of the onginal broad-leaved woodland in Bntain now occurs in relatively small habitat fragments, up to some tens of hectares in area For example, in Cambndgeshire, excluding managed conifer plantations, < 10% of woodland exists in patches of > 100 ha and c 50% IS in patches of 10 ha or less (Forestry Commission 1980 unpubl data) The small woods of our study, compnsing a size range of 0 02-30 ha, represented this extreme case of fragmentation In general, a small habitat fragment supports fewer individuals of a given species than a large fragment and small populations have a greater nsk of extinction for stochastic reasons alone Thus the persistence of bird species in small habitat patches may depend largely on efficient dispersal between patches, and turnover in the species composition of individual patches could be expected to be high Bird observations The woods were surveyed in three years (1990-1992) to determine their annual complement of resident breeding species (McCoy 1982) Resident species were defined as those in which the majonty of the population was normally resident in southern England throughout the year Species, such as blackcap Sylvta atncapilla and chiffchaff Phylloscopus colybita, in which extremely small numbers overwintered, were considered to be migrants Analyses were restncted to resident species because their distributions and abundances were likely to be influenced entirely by factors operating locally whereas migrants were also affected by conditions in wintenng areas and dunng migration Within the study woods, there were 46 species of year-round residents compared to 13 species of tropical migrants (Table 2), of which only three species were common, occumng in c 30% of the woods Birds were censused using a mapping technique based on the methodology of the Bntish Thist for Omithology's Common Bird Census (Marchant 1983) Each wood was searched systematically by walking a route designed to encounter every potential bird breeding temtory In the four woods larger than c 9 ha, the census routes were more widely spaced, but every habitat type present was included Dunng the search, all birds encountered, and their activities at the time, were recorded on a map of the wood However, mapping was not attempted m the two largest woods (Table 1), nor in four where dense thickets impeded movement In these cases, species presence and activities only were recorded In all woods, the presence and activities of woodpigeons Columba palumbus, pheasants Phasianus colchicus and red-legged partndges Alectons rufa were also recorded, but not mapjied Colonies of spiecies such as starling Stumus vulgans and rook Corvus fnigilegus were recorded as a single presence The time spent searching in each wood was determined not only by the area of the wood, but also by the conditions of bird activity and the detectability of different species at the time, the aim being to achieve a high census efficiency on each visit (Svensson 1978) Particular attention was paid to recording any behaviour indicative of breeding Actively seaching for nests was included as pan of the census rouune This intensive method was found to be necessary because singing activity in small woods was often infrequent, and skulking species, such as dunnock Prunella modulans, could otherwise have been missed The woods were visited alternately by two independent observers Search routes were vaned between visits, but always included walking around the outside of each wood The woods were censused four times dunng the breeding season, on visits evenly spaced between the end of March and the beginning of August. For 15 of the woods (size range 0.14-0.95 ha, x = 0.48 ± 0 24 [SD] ha), between three and five additional visits of about seven hours each were also made (in the course of a trapping and marking program to be reported elsewhere) Over the ECOGRAPHY 18 I (1995) Methods Study area The study area was situated in East Anglia (in the counties of Cambndgeshire and Lincolnshire) in lowland ea,stem England, roughly centred on Monks Wood (52°24'N, 0°14'W) The area was c 70 km from north to south (but with a gap of c. 17 km) and 35 km from east to west, and straddled the boundary between flat, drained fenland to the east and slightly higher claylands to the west Within this area, 146 woods, ranging in size from 0 02 to 30 ha and of vanous degrees of isolation, were investigated (Table 1) They compnsed c 20% by number (and c 10% by area) of the woods in the study area. They were selected as being representative of the size range of woods in the area in general (except for a few larger woods up to c 150 ha, which were too large to census efficiently in the time available). Selection was also influenced by factors such as accessibility and the agreement of landowners The majonty of the woods, although comprising a wide range of woody vegetation types, were broadleaved. A small number also including areas of intermixed conifers and broad-leaved trees Ponds and waterfilled ditches both within and on the edges of the woods were fairly common. Because birds associated with these water bodies also spread into the woods (e g mooorhens Galltnula chloropus nesting in trees and mallards Anas platyrhynchos in nettle beds) such ponds and ditches were censused as parts of the woods The landscape surrounding the woods was largely devoted to intensive arable agnculture, pasture and farm stock were uncommon and localised The main crops were cereals, rape, beans, sugar beet and potatoes 42 Table I Size distnbution and examples of measures of isolation of the study woods Size class, ha n Mean area of woods in each size class x ± 1 SD 0 02- 0 24 0 25- 049 0 50- 0 74 0 75- 0 99 100-199 2 0 0 - 3 99 4 0 0 - 5 99 6 0 0 - 999 10 00-30 00 29 30 20 14 21 13 9 7 3 0 14±0 06 O,33±O 07 0 63±0 09 0 86±0 09 1 43±0 32 291±0 55 4 97±0 75 7 87±0 85 19 88±9 80* Mean and range of distance to nearest wood, km x ± 1 SD 0 35±O 38 0 36±0 26 0 44±0 43 O5I±O57 0 29±0 23 021±0I7 0 23±0 11 0 15±0 15 O31±O19 Range 0 01-1 48 0 02-0 98 0 02-162 0 01-2.10 0 01-0 82 001-0 55 0 02-0 38 001-045 0 10-0 48 Mean and range of area of woodland within 0 5 km, ha X ± 1 SD 2 11±3 68 1 96±3 49 I 70±l 97 2 78±7 08 3 96±6 07 9 80±9 74 4 43±4 08 7 12+6 92 2 38±1 79 Range 0 00-12 96 0 00-15 78 0 0 0 - 7 28 0 00-27 09 0 00-23 21 0 00-26 83 108-12 86 1 89-21 75 0 70- 4 27 *lndi vidual areas of these three largest woods were 10 30, 19 42 and 29 92 ha In 1990-91, three woods of 0 47, 0 57 and 1 28 ha, and in 1991-92, five different woods of 0 72, 0 84, 1 18, 1 23 and 5 85 ha were omitted from analyses because of gross habitat changes three years of the study, the numbers of breeding species in these 15 woods recorded by the four survey visits did not differ significantly (paired t-test, t^ = 1 76, NS) from those obtained using the total number of visits, which amounted to an additional effort of c 21-35 h f)er wood Nor were any additional species consistently encountered or caught that were not recorded during the main visits A similar companson was not available for the larger woods, but because such a large increase in survey time did not detect significantly more breeding species in small woods, the lnten'sive survey method with four visiLs was considered adequate for detecting breeding species (Haila and Kuusela 1982, Haila and Hanski 1984) Visits started at dawn and ended when bu'd activity declined. usually before midday Woods were not censused dunng weather conditions likely to depress bird activity, namely strong winds and rain Sjsecies were considered to be breeders if a) they were present on three of the four visits, b) they were present on two consecutive visits, c) singing was recorded on any two or more occasions, d) a pair was recorded on any visit, e) breeding behaviour other than singing (e g nest building, carrying food, the presence of recently fledged dependent young) was recorded on any visit, 0 a nest was found Species recorded which did not meet one or more of these cntena were classed as transient and omitted from analyses All owls and woodcock Scolopax rwsttcola Table 2 Resident species breeding in the study woods Grey hero" Ardea cinerea Mute swan Cygnus olor Mallard Anas platyrhynchos Sparrowhawk Accipiter ntsus Kestrel Falco tinnunculus Red-legged partndge Alectoris rufa Grey panndge Perdix perdix Pheasant Phasianus colchicus Moorhen Galtinula chloropus Woodpigeon Columba palumbus Stock dove Columba oenas Collared dove Streptopelia decaocto Green woodpecker Picus viridis Great-spotted woodpecker Dendrocopos major Pied wagtail Motacilla alba Wren Troglodytes troglodytes Dunnock Prunella modulans Goldcrest Regulus regulus Robin Erithacus rubecula Blackbird Turdus merula Song thrush Turdus philomelos Mistle thrush Turdus viscivorus Long-tailed Ut Aegithalos caudatus Marsh tit Parus paliLstris Willow tit Parus montanus Coal tit Parus ater Blue tit Parus caeruleus Great tit Parus major Nuthatch Sitta europaea Treecreeper Certhia familians Com bunung Emberiza calandra Yellowhammer Embenza cttrmella Reed bunung Emberiza schoeniclus Chaffinch Frmgilla coelebs Greenfinch Carduelis chlorts Goldfinch Carduelis carduelis Linnet Carduelis cannabina Bullflnch Pyrrhula pyrrhula House sparrow Passer domesticus Tree sparrow Passer montanus Starling Stumus vulgaris Jay Garrulus glandarius Magpie Pica pica Jackdaw Corvus monedula Rook Corvus frugilegus Camon crow Corvus corone corone Migrant breeding species also present were hobby Falco subbuteo, turtle dove Streptopelia turtur, cuckoo Cucutus canorus, nightingale Luscinia megarhynchos, sedge warbler Acrocephalus schoenobaenus, reed warbler Acrocephalus scirpaceus, lesser whitethroat Sylvia curruca, whitethroat Sylvia commums, garden warbler Sylvia borin, blackcap Sylvia atrtcapilla, chiffchaff Phylloscopus collybita, willow warbler Phylloscopus trochilus and spotted flycatcher Muscicapa strtata ECOGRAPHY 18 I (1995) Table 3 Vanables, descnbing woodland area and strueture, isolauon, connectedness and surrounding land use, used in the stepwise multiple regression analyses Vanable Area and structure 1 Woodland area, ha 2 Penmeter, m 3 Shape 4 Density of canopy layer' 5 Density of shrub layer' 6 Density of field layer' 7 Number of habitats' Isolauon 8 Area of woodland within 0 5 km, ha 9 Area of woodland within 1 0 km, ha 10 Length of hedgerow within 0 5 km, km 11 Length of hedgerow within 1 0 km, km 12 Distance to nearest wood, km 13 Distanse to nearest wood of '22 ha, km 14 Distance to nearest village/town, km Connectedness 15 Number of hedges connected to wood 16 Number of ditches connected to wood 17 Number of dirt roads to/from wood 18 Total number of linear connections Surrounding land use 19 % of penmeter adjoined by cereal' 20 % of penmeter adjoined by rape and beans' 21 22 23 24 % of penmeter % of penmeter % of penmeter % of penmeter usage' adjoined adjoined adjoined adjoined by by by by Abbreviation Area Penmeter Shape Canopy density Shrub density Field density No habitats Wood 0 5 km Wood 1 0 km Hedge 0 5 km Hedge 1 0 km Dist wood Dist wood 2 ha Dist vill Hedge connect No ditch No dirt road Linear connect % cereal % rape and beans root erops' % roots other crops' % other crops grass' % grass non-crop % non-crops 'Vanables calculated separately for each year of the study ^Habitat categones were a) deciduous woodland, b) coniferous woodland, c) mixed woodland, d) dense shrub layer, e) thin shrub layer, f) mature scrub, > 2 m tall, g) young scrub/trees, < 2 m tall, h) open glades, i) cultivated ground, j) hedge as an intergral part of the wood (part or all of the penmeter of some woods consisted of a hedge, these hedges were integral parts of the woods and disUnct from the connecting hedgerows compnsing vanable 15), k) water body (pond, dyke or nver), 1) building (s) within wood (all uninhabited/derelict buildings), m) building (s) near wood (buildings within 1(X) m of wood, both inhabited and uninhabited) were also omitted because the census technique could not detect them reliably Woods in which gross habitat changes occurred between years were excluded from analysis (McCoy 1982) These compnsed three woods m 1990-91 and five different woods in 1991-92 (Table 1), Analyses Absolute turnover in sjjecies composition between successive breeding seasons was calculated, following Williamson (1978), as (E - C)/2, where E and C are the H numbers of species extinctions and colonisations respec44 tively between one year and the next For an extinction to be recorded, the species had to have bred at the site concerned in the previous breeding season Similarly, for a colonisauon to be recorded, the species had to have been absent from the site concerned for at least the previous breeding season Relative turnover was calculated, following Diamond (1969), as (E + C)/(S1 + S2) x 100%, where E and C are as above and S1 and S2 are the numbers of breeding species present in the two years Note that, with absolute turnover, maximum possible values increase with increasing numbers of spiecies in the wood, but with relative turnover, the values are expressed as a proportion of the number of species present. Extinction and colonisation rates were calculated as the numbers of species lost and gained respectively between one breeding season and the next With three consecutive annual surveys, species turnover, and extinction and colonisauon rates, were calculated between the breeding seasons of 1990-91 and 1991-92 The individual species responsible for the majonty of turnover in a sample of small woods (1 02 ± 0,18 ha, n = 24) and a sample of larger woods (5 93 ± 2 25 ha, n = 16) were compared The small woods were selected as those with areas closest to 1 ha and the large woods compnsed all those > 4 0 ha in which species presence was mapped For all the woods in each sample, the number of extinctions plus colonisations (E -i- C) due to each species separately was calculated as a percentage of the total number of extinctions and colonisauons recorded for all the species in all the woods m each sample For example, for pheasant m the sample of small woods, between the 1990 and 1991 breeding seasons, the species disappeared from three woods and appeared in six, making a total E -f C = 9 For all species combined in the small woods, the total number of extinctions plus colonisations was 98 Thus the contnbution of pheasant to the turnover for 1990-91 was (9/98) x 100% = 9,2% The results for each species were expressed as the mean of the two sets of turnover measurements For all 24 small woods and 16 large woods, the number of pairs of each species breeding in each wood was estimated from the census maps for 1990 (Marchant 1983) Four visits, although too few to give an accurate count of the number of temtones, were sufficient to provide an estimate of the relaove species population sizes in the two samples of woods The estimates were made using the data for 1990 because in eastern England generally, woodland bird populations were highest m this year of the study (Marchant and Musty 1992) The effects on species turnover of woodland area, structure and location m the landscape, were investigated by stepwise multiple regression analysis (Minitab 7,2) using the vanables described below. Woodland area was log transformed to lineanse the relaUonships with species numbers and turnover. This was not necessary for any of the other variables and no other transformauons were made ECOGRAPHY 18 1 (1995) Absoluts turnover rate y - 2 03 (±0 085) -f 0 317 (±0 136) log area r 2 . 0 030, n • 143, p < 0 05 -100 -10 L o g , , woodland area 100 • • Relative turnover rate y . 24 9 (11 38) - 19 6 (12 22) log area r 2 - 0 350, n - 143. p < 0 001 IT -100 -10 0 L o g , , woodland area Fig I Relationships between turnover rates of resident breeding species and woodland area for 1990-91 Absolute turnover rate (A) was calculated as (E + C)/2 where E and C were the the numbers of species extinctions and colonisations respectively between one year and the next Relative turnover rate (B) was calculated as (E + C)/(S1 + S2) X 100% where E and C were as above and S1 and S2 were the numbers of species present in each of the two years There was no significant difference in either relationship between 1990-91 and 1991-92 Woodland and landscape variables The woods themselves, their position in the landscape and the surrounding land use, were descnbed using the 24 vanables listed in Table 3 Woodland area and penmeter were measured directly in the field for the smaller woods and from 1:25 O X Ordnance Survey maps for the larger C) woods The shape of each wood was descnbed using an index calculated as Pm/Pc. where P^ was the measured penmeter of the wood and P^ was the penmeter of a circular wood of the same area. Thus the greater the edge to area ratio, the larger the index The number of "habitats" present within each wood was assessed independently by two observers. In the small number of instances when the assessments differed, consensus was reached after discussion. Habitats included features such ECOGRAPHY 18 I (1995) as ponds, dense thickets and open glades (Table 3), In general, the number of habitats increased with increasing woodland area (y = 4 13 (± 0,106) + 1 67 (± 0,172) log area; n = 146, r^ = 0 393, p < 0,001), The densiues of the canopy, shrub and field layers of each wood were scored each year on an arbitary scale of 0 to 4, where 0 = no canopy/shrub/field layer, and 4 = a completely closed canopy or a dense shrub/field layer Each vegetation layer was scored by assessing what proportion in the whole wood was attnbutable to each of the five possible scores These scores were then combined to give an overall index for each layer as* index = S (score x proportion) The minimum and maximum values for each layer were therefore 0 and 4 respiecuvely The layers were scored independently by two observers and the mean of the two assessments used m the analyses The following seven vanables were used as measures of woodland isolation The area of woodland within 0 5 and 1 0 km of the penmeter of each wood was obtained from aenal photographs (Cambndge Univ collection) taken in 1989 The positions of all hedgerows within 0,5 and 1,0 km were determined by ground surveys, plotted on 1:25000 Ordnance Survey maps, and their lengths then measured (Hedgerows could not be reliably distinguished from overgrown ditches on the aenal photographs ) Distances to the nearest wood of any size, to the nearest wood of > 2 ha and to the nearest active centre of human habitation, such as a farm or village, were also measured using 1,25000 Ordnance Survey maps The size of 2 ha was selected because, in our study area, bird species numbers accumulated most rapidly with increasing woodland area up to c 2 ha, woods of this size contained c 65% of the number of species occumng in our largest woods and c 25% of the total number of resident species found across all our woods (Bellamy et al in press) Four vanables were used as measures of connectedness (Baudry and Memam 1988), The numbers of hedges, ditches and dirt roads connected to each wood were counted in the field The fourth vanable, the total number of linear connections was calculated from the above three by combining linear features which occupied a common position For example, a hedge with a ditch alongside it or a dirt road with a hedge on one side and a ditch on the other would both be counted as single linear connections Surrounding land use was assessed annually and was expressed as the percentage of the penmeter of each wood which was adjouied by the categones shown in Table 3 Root crops compnsed sugar beet, potatoes and carrots, while "other crops" included linseed, peas and onions The non-crop category included all other land uses, such as buildings, yards, waste ground, gardens, scattered scrub and trees and woods. Cereals, rape and beans (combined) and root crops were by far the most common categones. Table 4 Stepwise multiple regression analyses of extinction and colonisaUon rates of resident species in relation to woodland structure, isolation, connectedness and sunounding land use Time interval Step Vanables entered Additional Signif% vanance icance explained ExUnction rate 1990-91 1 Hedge connect 2 Wood 0 5 km 3 % roots 1990, -ve* Total 127 <0 001 <0 05 <0 05 ColonisaUon rate 1990-91 1 2 3 4 Total Wood 0 5 km, -ve* Shrub density, -ve* % grass 1990 Shape <0 001 <0 05 <0 05 <0 05 <OOOI <0 05 1991-92 1 Hedge connect 2 No habitats Total tween 1990 and 1992 (t|37 = 1 40, NS), Thus, losses outnumbered gains in 1990-91, and gains outnumbered losses in 1991-92 Species tumover, resulting from local extincuons and colonisations, occurred across the whole size range of woods surveyed TTie mean number of species lost jier wood in 1991 was 2 2 ± 1.6 (range 0-8) and in 1992 it was 1 5 ± 1 2 (range 0-5) The mean number of species gained per wood in 1991 was 1 7 ± 1 4 (range 0-6) and in 1992 It was 2 3 ± 1 6 (range 0-7) Tumover of species was complete in four woods for 1990-91 and m one (different) wood for 1991-92 Three of these woods were very small, at 0 1 ha or less, and one was a little larger at 0 24 ha All four held no more than three breeding species in the years concemed No tumover was recorded in two woods for 1990-91 (n = 143) and in five woods for 1991-92 (n = 141), one of which was the same in both sets of years These woods ranged in size from 0 18 to 0,84 ha, except for one of 9 05 ha No meaningful relationship was found between absolute tumover rate of species and any of the predictor vanables, including woodland area (Fig la), for either 1990-91 or 1991-92 However, the relationship with area was statistically significant because of the large number of woods used and because the tumover rates in the smallest woods, with very few breeding species, were less than those in slightly larger woods, even when tumover was complete Individually, none of the vanables used explained > 10% of the vanation m absolute tumover rate and no vanables entered the stepwise regression analysis For extinction and colonisation rates (Table 4), the two most significant vanables to emerge from the analyses (area of woodland within 0 5 km and number of hedges connected to a wood) were the same for both, but the relationship between colonisation rate and area of Abbreviations as in Table 3 * denotes direction of the relationship, all are +ve unless otherwise indicated For extinction rate in 1991-92, no vanables entered the analysis Results The numbers of resident breeding species vaned from less than four per wood in the smallest woods (< 0 1 ha) to > 16 per wood in woods of 10 ha or more (Bellamy et al in press) In general, the numbers of resident species in each wood decreased between 1990 and 1991 (pan-ed t-test, ti42 = - 2 55, p < 0 05) and then increased between 1991 and 1992 (t,4o = 3 78, p < 0 001), such that there was no significant difference in the numbers of species be- Table 5 Contnbution of individual species to tumover in two size classes of woods "Small" woods (mean area±l SD=1 02±0 18 ha, n=24) Species % contnbution to absolute tumover 92 57 57 52 52 47 47 47 47 42 42 58 2 Species "Urge" woods (mean area±l SD=5,93±2 25 ha, n=16) % contnbution to absolute tumover 108 81 66 61 57 57 56 46 46 57 8 Pheasant* Wren Camon crow* Dunnock Song thrush Robin Blue Ut Great tit Red-legged partndge* Greenfinch Moorhen Total Great-spotted woodpecker Camon crow* Marsh ut Sparrowhawk Magpie* Stock dove Long-tailed tit Bullflnch Red-legged partndge* Total The percentages are expressed as the mean of each species conUibuUon to tumover in 1990-91 and 1991-92 For the small woods, the remaining 41 8% of tumover was contnbuted by a further 25 species and for the large woods, the remaining 42 2% of tumover was contnbuted by a further 22 species The species marked * are those which were regularly hunted, trapped and/or eaptive reared and released (see text) ECOGRAPHY 18 1 (1995) Table 6 Mean numbers of pairs of the species conlnbuung most to turnover in two sizes of woods Data for 1990 Species "Small" wood,s (mean area±l SD=1 02±0 18 ha, n=24) Mean no pairs (x±SD) Wren Dunnock Song thnish Robin Blue tit Great Ut G-s woodpecker Marsh tit Sparrowhawk Long-t tit Bullfinch 2 4±1 2 1 4±1 0 0 4±0 5 I4±12 13±10 0 5±0 7 0 04±0 2 0 04±0 2 0 1±0 3 0 3±0 6 0 l±05 Range 1-5 0-3 0-1 0^ 0^ 0-2 O-I 0-1 0-1 0-2 0-1 % of sample woods occupied in 1990 100 79 42 71 79 38 4 4 13 17 8 "Large" woods (mean area±l SD=5 93±2,25 ha, n=16) Mean no pairs (x+SD) 10 4±4 5 4 0±2 4 1 l±09 5 5±3 3 4 8±2 5 2 8±17 0 5±0 5 0 6±0 9 O3±O5 1 1±1 4 1 3±1 5 Range 5-18 1-9 0-3 1-11 1-10 0-7 0-1 0-2 0-1 0^ 2^ % of sample woods occupied in 1990 100 100 69 Numbers of the following species were not estimated red-legged partndge, stock dove, moorhen, greenfinch, camon crow and magpie Long-t lit = long-tailed tit, G-s woodpecker = great-spotted woodpecker woodland within 0 5 km was negative However, the relationships of extinction and colonisation rates with the individual vanables were weak and the total percentage of the vanance explained in both cases was small Despite the lack of a strong relationship between absolute turnover rate and woodland area, with the exception of game birds and corvids whose abundance and distnbution were strongly influenced by human hunting activity, the spyecies which contnbuted most to turnover changed as area increased (Table 5) The species which contnbuted most to turnover in the sample of small woods were represented by only 1-5 pairs (Table 6) In the larger woods, these same species had larger populations, were less prone to extinction and so contnbuted little or nothing to the turnover rate In the sample of large woods, the species contnbuting most to turnover seldom occurred in the small woods and were again those represented by only a small number of pairs (Table 6) The numbers of species breeding in our study woods increased with woodland area (c 70% of the vanation in resident breeding species numbers was accounted for by area alone, Bellamy et al in press) and because the numbers of species changing between one year and the next was the same regardless of woodland area, relative turnover inevitably emerged as greater in the smaller woods (Fig, lb) Despite the inevitability of this relaUonship, relative turnover did emphasize the instability of the bird populations of the smallest woods There was no difference in either intercept (F^si =2,21, NS) or slope (F| 280 = 3,76, NS) between the fitted lines for both sets of years Discussion The year-to-year changes m breeding species compwsition, resulting firom local extinctions and colonisations. ECOGRAPHY 18 1 (1995) clearly demonstrated the occurrence of species turnover in the resident bird populations of our woods Other work has also documented species turnover m birds, both for true islands (e g Diamond 1969, Jones and Diamond 1976, Wnght 1985) and for habitat islands on land (e g Beven 1976, Whitcomb et al 1977, Woolhouse 1987), but the validity and magnitude of the turnover rates recorded in some of these studies have been questioned (Gilbert 1980, McCoy 1982) The commonest causes of spurious turnover include census errors, inappropnate census intervals, the inclusion of non-breeding vagrants and changes in the avifauna directly attnbutable to human acuvity (Lynch and Johnson 1974) In our study, census inadequacies were the most important ]X)tential source of error, affecting mainly ,species in which pairs were present for only a small part of the breeding season Census deficiencies can cause turnover to be both over-estimated and under-estimated, although the former is likely to be commonest (Lynch and Johnson 1974) For example, a species present in three years, but missed in the rmddle one, will generate a spunous extinction in that year and a spunous colonisation in the next, increasing the estimate of turnover m both years. Failing to record a colonisation in one year will underestimate turnover in that year, but if the species remains, it may then produce a spunous colonisation in the following year In our study, errors were most likely to have been made in the largest woods through the occasional failure to record a species as present and breeding This would have resulted in an over-estimate of turnover in large woods relative to small ones In this case, the real trend in relative turnover with woodland area might have been even more marked than Fig lb suggests, and absolute turnover might have been more sensitive to increasing woodland area than indicated by Fig, la However, the numbers of species lost and gained between years in our study were, on average, small, suggesung that census errors were not common. 47 Small woods support only small numbers of individuals of the species present, and such small populations have the greatest nsk of stochastic extinction Because the nsk of stochastic extinction should decrease as population size increases, it might have been expected that absolute turnover, and especially extinction rate, would have been related to woodland area (m general, large woods should contain larger numbers of individuals of any given species than should small woods), but this was not the case (Fig la. Table 4) In both small and large woods. It was the species represented by only a few breeding pairs which contnbuted most to turnover and, m the main, the species concerned were different in small and large woods (Table 5 and 6) The only species contnbutmg substantially to turnover irrespective of woodland size were those in which extinction and colonisation rates were artificially increased by shooting and trapping, and in some cases (pheasant and red-legged partndge), by captive reanng and release Withm our study area, different species occurred at different regional densities and may also have had different minimum habitat area requirements (van Dorp and Opdam 1987) Hence, as woodland area increased, the population sizes of some sfwcies reached levels where extinction due to stochastic processes became uncommon, while others were still only represented by a small number of individuals. Thus a decrease in extinction rate among the most abundant species was offset by the high rates among the rarer ones resulting m no overall relationship between extinction rate, and hence turnover rate, and area This process should continue until the woodland is of sufficient area to support populations of the whole species pool large enough to usually avoid stochastic extinction. The absence of a relationship between absolute turnover and any of the 24 predictor vanables also suggested that stochastic processes were paramount The above argument is valid for homogeneous woodland habitat In reality, species may be missing from an adequate area of woodland if their specific habitat requirements are lacking. Area and habitat diversity are closely correlated (as shown for our woods), often so closely that it is hard to separate their effects (Lack 1976), Thus, habitat mfiuences could generate the same sort of pattern of local extinctions in species with small population sizes as suggested above for area. In this case, species would have small populations because their habitat, rather than area per se, was limiting In the absence of a strong influence of isolation on colonisation rates, the form of the relationship between turnover rate and woodland area may be determined by the interaction of habitat availabibty and area, and the autecological requirements of the individual species present in the locality Several other studies have found that turnover is due mainly to species with small populations (Williamson 1981, Wnght 1985), which come and go against a background of large populations with apparent long-teim stabihty (Williamson 1983, Schoener and SpiUer 1987) 48 This has led to the suggestion that the majonty of turnover IS "ecologically tnvial" (Williamson 1989, Hamson 1991) In conservation terms, this could be a dangerous assertion for species in which the majonty of the population now exists in fragmented habitat, especially when population density withm the fragments is low (Doak and Mills 1994) Lx)cal extinction of the populations of some fragments may increase the isolation of the remainder beyond the ability of the species to recolonise through dispersal, resulting m regional extinction (Barinaga 1990, Rolstad 1991, Fahng and Memam 1994) When a species has been reduced to small populations in scattered habitat patches, the overall fate of the species will be dominated by the stochastic processes affecting the individual small populations The movement of birds between our woods was not of the mainland-island type descnbed by the equilibnum theory of MacArthur and Wilson (1967) Movement appeared to be diffuse, with individuals moving between the woods themselves and between the woods and other habitats such as gardens and hedgerows Specifically, extinction rates (expressed as numbers of sfiecies lost) in our study were not related to woodland area (= island area), the main cause of extinction being small population size as descnbed above Colonisation rates did not increase with decreasing degree of woodland isolation Indeed, for 1990-91, colonisation rates were inversely related to the area of woodland within 0.5 km, suggesting the opposite trend This same trend was also suggested by the positive relationship between extinction rates and the area of woodland within 0 5 km also found for 1990-91 The lnfiuence of surrounding woodland on extinction and colonisation rates may have ansen as a result of space becoming available in larger woods (assuming that such large woods are preferred habitat) following a penod of severe weather in February 1991 which caused a reduction in the overall size of woodland bird populations in eastern England in 1991 (Marchant and Musty 1992) The winter of 1991-92 was mild and bird populations in general were higher in 1992 than m 1991 Unlike 1990-91, more species were gamed in our woods in 1991-92 than were lost and therefore it was not surpnsing that the small infiuence of woodland and landscape vanables on extinction rates found in 1990-91 was not apparent in 1991-92 (Table 4) The result for colonisation rates in 1991-92 suggested a possible role for hedgerows as comdors (Hinsley et al. 1995), but the relationship was weak. Overall, the small amount of vanation in extinction and colonisation rates explained by the woodland and landscape vanables, and the lack of consistency in the results for the two sets of years, emphasized the lmpwrtance of stochastic processes in this highly fragmented habitat Discrete populations linked by dispersal compnse the basic stiTicture of a metapopulation (sensu Levins 1970) However, if dispersal rates are high, a fragmented population can still function essentially as a single population (Hamson 1991) Of the species in our woods, those most ECOGRAPHY 18 I (1995) Angelstam, P 1992 Conservation of communities - the importance of edges, surroundings and landscape mosaic structure - In Hansson, L (ed ), Ecological principles of nature conservation application in temperate and boreal environments Elsevier Applied Science, London, pp 9-70 Bannaga, M 1990 Where have all the froggies gone''- Science 247 I033-I034 Baudry, J and Memam, G 1988 Connecuvity and connectedness functional versus structural patterns in landscapes In Schreiber, K F (ed ), Proc 2nd Int Seminar, Int Ass Landscape Ecol, Munstersche Geogr Arbeiten 29 23-28 Bellamy, P E , Hinsley, S A and Newton, I In press Factors influencing bird species numbers in small woods in southeast England - J Appl Ecol Beven, G 1976 Changes in breeding bird populaUons of an oak-wood on Bookham Common, Surrey, over twentyseven years - Lond Nat 55 23-42 Butcher, G S , Nienng, W A , Barry, W J and Goodwin, R H 1981 Equihbnum biogeography and the size of nature reserves an avian case study - Oecologia 49 29-37 Diamond, J M 1969 Avifaunal equilibria and species turnover rates on the Channel Islands of California - Proc Nat Acad SCI USA 69 3199-3203 Doak, D F and Mills, L S 1994 A useful role for theory in conservation - Ecology 75 615-626 Fahng, L and Memam, G 1994 Conservation of fragmented populations - Cons Biol 8 50-59 Gilbert, F S 1980 The equilibrium theory of island biogeography fact or fiction'' - J Biogeogr 7 209-235 Haila, Y and Kuusela, S 1982 Efficiency of one-visit censuses of bird communiues breeding on small islands - Omis Scand 13 17-24 - and Hanski, I K 1984 Methodology for studying the effecLs of habitat fragmentation on land birds - Ann Zool Fennici 21 393-397 - , Hanski, 1 K and Raivio, S 1993 Turnover of breeding birds in small forest fragments the "sampling" colonizauon hypothesis corroborated - Ecology 74 714-725 Hanski, I K and Haila, Y 1988 Singing teentones and home ranges of breedmg chaffinches visual observation vs radiotraeking - Omis Fennica 65 97-103 Hams, L D 1984 The fragmented forest - Univ Chicago Overall, the difficulties of scale and of movements Press, Chicago discussed above are unlikely to alter the conclusion that Hamson, S 1991 Local exUnction in a metapopulauon context an empincal evaluaUon - In Gilpin, M and Hanski, I species turnover rates in small, fragmented woods are (eds), Metapopulauon dynamics empirical and theoreucal high, and that woodlands of up to at least 10 ha are invesugauons Reprinted fi^m the Biol J Linn Soc Vol insufficient to maintain populations of certain woodland 42 Academic Press, London, pp 73-88 species large enough to avoid purely stochastic extinc- Hinsley, S A , Bellamy, P E , Newton, I and Sparks, T H 1995 Habitat and landscape factors influencing the presence tion In Bntain and elsewhere, a large proportion of many of individual breeding bird species in woodland fragments species populations exist in fragments where their numJ Avian Biol in press bers are liable to be small enough to show appreciable IUCNAJNEPAVWF 1991, Canng for the earth A strategy for turnover and where species long-term persistence will sustainable living - Gland, Switzerland Jones, H L and Diamond, J M 1976 Short-time-base studies depend on continued and effective dispersal of tumover in breedmg bind populations on the Califomia Channel Islands - Condor 78 526-549 Acknowledgements - We would like to thank the many lan- Lack, D 1954 The natural regulauon of animal numbers downers, farmers, managers, tenants and gamekeepers who have Oxford Univ Press, Oxford kindly allowed us to work on their land and without whose help - 1976 Island biology - Univ of Califomia Press, Los Anand co-operation this project would not have been possible geles Thanks are also due to the Bedfordshire and Cambridgeshire Levins, R 1970 Exunction - In Gerstenhaber, M. (ed ), LecWildhfe Trust, English Nature and Cambndgeshire County tures on mathematics in the life sciences, Vol 2 Some 3 ECOGRAPHY 18 1 (1995) likely to show a metapopulation strucmre would be species such as marsh tit which are restncted to woodland and have short dispersal distances However, whether or not a species can make the transition from contiguous habitat to a metapopulation would depend not only on its dispersal charactenstics, but also on the quality of the available habitat fragments and the landscape between them For example, nuthatches Sitta europaea, which have been reported to have a metapopulation structure in small woods in the Netherlands (Verboom et al 1991) are rare or absent in our study area (one pair bred in one study wo(xi in 1990 only), even in suitable woodland of over 100 ha This suggests that the degree of isolation of these woods from surrounding potential source populations IS sufficient to prevent colonisation and the development of a metapopulation The mobility of some species, such as great-spotted woodpecker Dendrocopus major, may alter the scale at which turnover should be considered for that species (Haila et al 1993) Apparent turnover may result from pairs nesting in one wood rather than another withm the same home range However, most of the species in our woods, and certainly most of those responsible for the majonty of turnover, were thought to be more or less site faithful (Lack 1954) The results of a marking program to be reported elsewhere, have so far indicated that movements between woods of breeding adults of the species typical of our small woods are uncommon (Hinsley unpubl data) Other evidence suggests that breeding individuals may range more widely than is apparent from temtory size determined by song post location (e g Hanski and Haila 1988) Indeed, we have occasionally seen some of our marked birds in woods adjacent to their "own" However, such individuals do not sing and generally keep a low profile and withm the context of our study would probably be recorded as transient Such movements by small birds are also likely to be less common in fragmented than in continuous woodland because of the costs (energy and time) and nsks (predators such as sparrowhawks) involved in regularly crossing open ground Council for access to woodland, to T Sparks for statisucal advice and to two referees for comments improving the manuscnpt This work was initiated as part of the Joint Agncultural and Environment Program (JAEP) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Ecography Wiley

Bird species turnover and stochastic extinction in woodland fragments

Hinsley, Shelley A.; Bellamy, Paul E.; Newton, Ian
Ecography , Volume 18 (1) – Mar 1, 1995
Download PDF
11 pages

Loading next page...
 
/lp/wiley/bird-species-turnover-and-stochastic-extinction-in-woodland-fragments-924EKyN4gL

References (45)

Publisher
Wiley
Copyright
Copyright © 1995 Wiley Subscription Services, Inc., A Wiley Company
ISSN
0906-7590
eISSN
1600-0587
DOI
10.1111/j.1600-0587.1995.tb00117.x
Publisher site
See Article on Publisher Site

Abstract

Hinsley, S A , Bellamy, P E and Newton, I 1995 Bird species turnover and stochdsuc extinction in woodland fragments - Ecography 18 41-50 Year-to-year turnover in bird species composiuon was recorded across, the whole size range (0 02-30 ha) of 146 woods studied The mean number of resident breeding species both lost and gained per wood between consecutive breeding seasons was 2 (range 0-8) No relationship was found between this absolute turnover rate and woodland area, or any other of 24 predictor vanables (descnbing woodland structure, isolation, connectedness and surrounding land use) Extmcuon and colonisation rates (in terms of numbers of species lost and gained) were also unrelated to woodland area In all sizes of woods, the species most likely to show local extinctions and colonisations were those with small populaUons within those woods, but the ldenuty of the species concerned changed as woodland area increased In the smallest woods, the majonty of turnover involved common species, such as wren and dunnock, which occurred in only small numbers in these small woods As woodland area increased, these species attained sufficient numbers to usually avoid stochastic extincUon The majonty of turnover was then due to more specialist (and less numerous) woodland species, such as great-spotted woodpecker and marsh Ut, which were usually lacking in small woods In Bntain, much existing broadleaved woodland falls within the size range studied Thus the numbers of many bird species are liable to be small enough for yearly turnover in woodland bird communities to be appreciable, and for the long-term persistence of individual species in particular woods to depend on dispersal S A Hinsley. P E Bellamy and 1 Newton. NERC, Inst of Terrestnal Ecology. Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire, UK PEI7 2LS TntrnHiiptinn iniroaucuon In this paper, we examine turnover in bird spiecies composition in woods of different sizes and degrees of lsolation, set within an agricultural landscape in eastern England, We also investigate the influence of the charactenstics of the woods themselves and of the surrounding landscape on the stability of species composition and on local extinction and colonisation rates Habitat loss is one of the major causes of a worldwide decline in biodiversity (lUCN/UNEPAVWF 1991, Roistad 1991) This process frequently begins with the fragmentation of a natural habitat, to be followed sooner or later by the degradation and eventual loss of the firagments themselves Many studies have examined the efAccepted 23 November 1994 Copynght © ECOGRAPHY 1995 ISSN 0906-7590 Pnnted in Denmark - all nghts reserved ECOGRAPHY 18 I (1995) ^^'''^ °^ fragmentation on forest plant and animal commu^^^^ ^^ ^ Weaver and Kellman 1981, Hams 1984, Angelstam 1992), although the forests concerned have mostly been large, of the order of hundreds or thousands of hectares Some of the most frequently cited studies concern the bird communities of the North Amencan eastern deciduous forest, in which a long-term decline of forest lntenor bird species has been attnbuted to forest fragmentation (Robbins 1980, Butcher et al 1981, Whitcomb et al, 1981, Temple and Cary 1988, Memam and Wegner 1992) In Bntain, the process of woodland fragmentation, viewed from a North Amencan or Scandinavian perspective, is extreme, there being few remaining extensive tracts of semi-natural woodland, especially in the low- lands (Wilcove et al 1986) In England, woodland once covered almost the whole country, but now accounts for only c 7% of the total area, of which 40% consists mainly of conifers (Locke 1987) A large prof)ortion of the avifauna of the onginal broad-leaved woodland in Bntain now occurs in relatively small habitat fragments, up to some tens of hectares in area For example, in Cambndgeshire, excluding managed conifer plantations, < 10% of woodland exists in patches of > 100 ha and c 50% IS in patches of 10 ha or less (Forestry Commission 1980 unpubl data) The small woods of our study, compnsing a size range of 0 02-30 ha, represented this extreme case of fragmentation In general, a small habitat fragment supports fewer individuals of a given species than a large fragment and small populations have a greater nsk of extinction for stochastic reasons alone Thus the persistence of bird species in small habitat patches may depend largely on efficient dispersal between patches, and turnover in the species composition of individual patches could be expected to be high Bird observations The woods were surveyed in three years (1990-1992) to determine their annual complement of resident breeding species (McCoy 1982) Resident species were defined as those in which the majonty of the population was normally resident in southern England throughout the year Species, such as blackcap Sylvta atncapilla and chiffchaff Phylloscopus colybita, in which extremely small numbers overwintered, were considered to be migrants Analyses were restncted to resident species because their distributions and abundances were likely to be influenced entirely by factors operating locally whereas migrants were also affected by conditions in wintenng areas and dunng migration Within the study woods, there were 46 species of year-round residents compared to 13 species of tropical migrants (Table 2), of which only three species were common, occumng in c 30% of the woods Birds were censused using a mapping technique based on the methodology of the Bntish Thist for Omithology's Common Bird Census (Marchant 1983) Each wood was searched systematically by walking a route designed to encounter every potential bird breeding temtory In the four woods larger than c 9 ha, the census routes were more widely spaced, but every habitat type present was included Dunng the search, all birds encountered, and their activities at the time, were recorded on a map of the wood However, mapping was not attempted m the two largest woods (Table 1), nor in four where dense thickets impeded movement In these cases, species presence and activities only were recorded In all woods, the presence and activities of woodpigeons Columba palumbus, pheasants Phasianus colchicus and red-legged partndges Alectons rufa were also recorded, but not mapjied Colonies of spiecies such as starling Stumus vulgans and rook Corvus fnigilegus were recorded as a single presence The time spent searching in each wood was determined not only by the area of the wood, but also by the conditions of bird activity and the detectability of different species at the time, the aim being to achieve a high census efficiency on each visit (Svensson 1978) Particular attention was paid to recording any behaviour indicative of breeding Actively seaching for nests was included as pan of the census rouune This intensive method was found to be necessary because singing activity in small woods was often infrequent, and skulking species, such as dunnock Prunella modulans, could otherwise have been missed The woods were visited alternately by two independent observers Search routes were vaned between visits, but always included walking around the outside of each wood The woods were censused four times dunng the breeding season, on visits evenly spaced between the end of March and the beginning of August. For 15 of the woods (size range 0.14-0.95 ha, x = 0.48 ± 0 24 [SD] ha), between three and five additional visits of about seven hours each were also made (in the course of a trapping and marking program to be reported elsewhere) Over the ECOGRAPHY 18 I (1995) Methods Study area The study area was situated in East Anglia (in the counties of Cambndgeshire and Lincolnshire) in lowland ea,stem England, roughly centred on Monks Wood (52°24'N, 0°14'W) The area was c 70 km from north to south (but with a gap of c. 17 km) and 35 km from east to west, and straddled the boundary between flat, drained fenland to the east and slightly higher claylands to the west Within this area, 146 woods, ranging in size from 0 02 to 30 ha and of vanous degrees of isolation, were investigated (Table 1) They compnsed c 20% by number (and c 10% by area) of the woods in the study area. They were selected as being representative of the size range of woods in the area in general (except for a few larger woods up to c 150 ha, which were too large to census efficiently in the time available). Selection was also influenced by factors such as accessibility and the agreement of landowners The majonty of the woods, although comprising a wide range of woody vegetation types, were broadleaved. A small number also including areas of intermixed conifers and broad-leaved trees Ponds and waterfilled ditches both within and on the edges of the woods were fairly common. Because birds associated with these water bodies also spread into the woods (e g mooorhens Galltnula chloropus nesting in trees and mallards Anas platyrhynchos in nettle beds) such ponds and ditches were censused as parts of the woods The landscape surrounding the woods was largely devoted to intensive arable agnculture, pasture and farm stock were uncommon and localised The main crops were cereals, rape, beans, sugar beet and potatoes 42 Table I Size distnbution and examples of measures of isolation of the study woods Size class, ha n Mean area of woods in each size class x ± 1 SD 0 02- 0 24 0 25- 049 0 50- 0 74 0 75- 0 99 100-199 2 0 0 - 3 99 4 0 0 - 5 99 6 0 0 - 999 10 00-30 00 29 30 20 14 21 13 9 7 3 0 14±0 06 O,33±O 07 0 63±0 09 0 86±0 09 1 43±0 32 291±0 55 4 97±0 75 7 87±0 85 19 88±9 80* Mean and range of distance to nearest wood, km x ± 1 SD 0 35±O 38 0 36±0 26 0 44±0 43 O5I±O57 0 29±0 23 021±0I7 0 23±0 11 0 15±0 15 O31±O19 Range 0 01-1 48 0 02-0 98 0 02-162 0 01-2.10 0 01-0 82 001-0 55 0 02-0 38 001-045 0 10-0 48 Mean and range of area of woodland within 0 5 km, ha X ± 1 SD 2 11±3 68 1 96±3 49 I 70±l 97 2 78±7 08 3 96±6 07 9 80±9 74 4 43±4 08 7 12+6 92 2 38±1 79 Range 0 00-12 96 0 00-15 78 0 0 0 - 7 28 0 00-27 09 0 00-23 21 0 00-26 83 108-12 86 1 89-21 75 0 70- 4 27 *lndi vidual areas of these three largest woods were 10 30, 19 42 and 29 92 ha In 1990-91, three woods of 0 47, 0 57 and 1 28 ha, and in 1991-92, five different woods of 0 72, 0 84, 1 18, 1 23 and 5 85 ha were omitted from analyses because of gross habitat changes three years of the study, the numbers of breeding species in these 15 woods recorded by the four survey visits did not differ significantly (paired t-test, t^ = 1 76, NS) from those obtained using the total number of visits, which amounted to an additional effort of c 21-35 h f)er wood Nor were any additional species consistently encountered or caught that were not recorded during the main visits A similar companson was not available for the larger woods, but because such a large increase in survey time did not detect significantly more breeding species in small woods, the lnten'sive survey method with four visiLs was considered adequate for detecting breeding species (Haila and Kuusela 1982, Haila and Hanski 1984) Visits started at dawn and ended when bu'd activity declined. usually before midday Woods were not censused dunng weather conditions likely to depress bird activity, namely strong winds and rain Sjsecies were considered to be breeders if a) they were present on three of the four visits, b) they were present on two consecutive visits, c) singing was recorded on any two or more occasions, d) a pair was recorded on any visit, e) breeding behaviour other than singing (e g nest building, carrying food, the presence of recently fledged dependent young) was recorded on any visit, 0 a nest was found Species recorded which did not meet one or more of these cntena were classed as transient and omitted from analyses All owls and woodcock Scolopax rwsttcola Table 2 Resident species breeding in the study woods Grey hero" Ardea cinerea Mute swan Cygnus olor Mallard Anas platyrhynchos Sparrowhawk Accipiter ntsus Kestrel Falco tinnunculus Red-legged partndge Alectoris rufa Grey panndge Perdix perdix Pheasant Phasianus colchicus Moorhen Galtinula chloropus Woodpigeon Columba palumbus Stock dove Columba oenas Collared dove Streptopelia decaocto Green woodpecker Picus viridis Great-spotted woodpecker Dendrocopos major Pied wagtail Motacilla alba Wren Troglodytes troglodytes Dunnock Prunella modulans Goldcrest Regulus regulus Robin Erithacus rubecula Blackbird Turdus merula Song thrush Turdus philomelos Mistle thrush Turdus viscivorus Long-tailed Ut Aegithalos caudatus Marsh tit Parus paliLstris Willow tit Parus montanus Coal tit Parus ater Blue tit Parus caeruleus Great tit Parus major Nuthatch Sitta europaea Treecreeper Certhia familians Com bunung Emberiza calandra Yellowhammer Embenza cttrmella Reed bunung Emberiza schoeniclus Chaffinch Frmgilla coelebs Greenfinch Carduelis chlorts Goldfinch Carduelis carduelis Linnet Carduelis cannabina Bullflnch Pyrrhula pyrrhula House sparrow Passer domesticus Tree sparrow Passer montanus Starling Stumus vulgaris Jay Garrulus glandarius Magpie Pica pica Jackdaw Corvus monedula Rook Corvus frugilegus Camon crow Corvus corone corone Migrant breeding species also present were hobby Falco subbuteo, turtle dove Streptopelia turtur, cuckoo Cucutus canorus, nightingale Luscinia megarhynchos, sedge warbler Acrocephalus schoenobaenus, reed warbler Acrocephalus scirpaceus, lesser whitethroat Sylvia curruca, whitethroat Sylvia commums, garden warbler Sylvia borin, blackcap Sylvia atrtcapilla, chiffchaff Phylloscopus collybita, willow warbler Phylloscopus trochilus and spotted flycatcher Muscicapa strtata ECOGRAPHY 18 I (1995) Table 3 Vanables, descnbing woodland area and strueture, isolauon, connectedness and surrounding land use, used in the stepwise multiple regression analyses Vanable Area and structure 1 Woodland area, ha 2 Penmeter, m 3 Shape 4 Density of canopy layer' 5 Density of shrub layer' 6 Density of field layer' 7 Number of habitats' Isolauon 8 Area of woodland within 0 5 km, ha 9 Area of woodland within 1 0 km, ha 10 Length of hedgerow within 0 5 km, km 11 Length of hedgerow within 1 0 km, km 12 Distance to nearest wood, km 13 Distanse to nearest wood of '22 ha, km 14 Distance to nearest village/town, km Connectedness 15 Number of hedges connected to wood 16 Number of ditches connected to wood 17 Number of dirt roads to/from wood 18 Total number of linear connections Surrounding land use 19 % of penmeter adjoined by cereal' 20 % of penmeter adjoined by rape and beans' 21 22 23 24 % of penmeter % of penmeter % of penmeter % of penmeter usage' adjoined adjoined adjoined adjoined by by by by Abbreviation Area Penmeter Shape Canopy density Shrub density Field density No habitats Wood 0 5 km Wood 1 0 km Hedge 0 5 km Hedge 1 0 km Dist wood Dist wood 2 ha Dist vill Hedge connect No ditch No dirt road Linear connect % cereal % rape and beans root erops' % roots other crops' % other crops grass' % grass non-crop % non-crops 'Vanables calculated separately for each year of the study ^Habitat categones were a) deciduous woodland, b) coniferous woodland, c) mixed woodland, d) dense shrub layer, e) thin shrub layer, f) mature scrub, > 2 m tall, g) young scrub/trees, < 2 m tall, h) open glades, i) cultivated ground, j) hedge as an intergral part of the wood (part or all of the penmeter of some woods consisted of a hedge, these hedges were integral parts of the woods and disUnct from the connecting hedgerows compnsing vanable 15), k) water body (pond, dyke or nver), 1) building (s) within wood (all uninhabited/derelict buildings), m) building (s) near wood (buildings within 1(X) m of wood, both inhabited and uninhabited) were also omitted because the census technique could not detect them reliably Woods in which gross habitat changes occurred between years were excluded from analysis (McCoy 1982) These compnsed three woods m 1990-91 and five different woods in 1991-92 (Table 1), Analyses Absolute turnover in sjjecies composition between successive breeding seasons was calculated, following Williamson (1978), as (E - C)/2, where E and C are the H numbers of species extinctions and colonisations respec44 tively between one year and the next For an extinction to be recorded, the species had to have bred at the site concerned in the previous breeding season Similarly, for a colonisauon to be recorded, the species had to have been absent from the site concerned for at least the previous breeding season Relative turnover was calculated, following Diamond (1969), as (E + C)/(S1 + S2) x 100%, where E and C are as above and S1 and S2 are the numbers of breeding species present in the two years Note that, with absolute turnover, maximum possible values increase with increasing numbers of spiecies in the wood, but with relative turnover, the values are expressed as a proportion of the number of species present. Extinction and colonisation rates were calculated as the numbers of species lost and gained respectively between one breeding season and the next With three consecutive annual surveys, species turnover, and extinction and colonisauon rates, were calculated between the breeding seasons of 1990-91 and 1991-92 The individual species responsible for the majonty of turnover in a sample of small woods (1 02 ± 0,18 ha, n = 24) and a sample of larger woods (5 93 ± 2 25 ha, n = 16) were compared The small woods were selected as those with areas closest to 1 ha and the large woods compnsed all those > 4 0 ha in which species presence was mapped For all the woods in each sample, the number of extinctions plus colonisations (E -i- C) due to each species separately was calculated as a percentage of the total number of extinctions and colonisauons recorded for all the species in all the woods m each sample For example, for pheasant m the sample of small woods, between the 1990 and 1991 breeding seasons, the species disappeared from three woods and appeared in six, making a total E -f C = 9 For all species combined in the small woods, the total number of extinctions plus colonisations was 98 Thus the contnbution of pheasant to the turnover for 1990-91 was (9/98) x 100% = 9,2% The results for each species were expressed as the mean of the two sets of turnover measurements For all 24 small woods and 16 large woods, the number of pairs of each species breeding in each wood was estimated from the census maps for 1990 (Marchant 1983) Four visits, although too few to give an accurate count of the number of temtones, were sufficient to provide an estimate of the relaove species population sizes in the two samples of woods The estimates were made using the data for 1990 because in eastern England generally, woodland bird populations were highest m this year of the study (Marchant and Musty 1992) The effects on species turnover of woodland area, structure and location m the landscape, were investigated by stepwise multiple regression analysis (Minitab 7,2) using the vanables described below. Woodland area was log transformed to lineanse the relaUonships with species numbers and turnover. This was not necessary for any of the other variables and no other transformauons were made ECOGRAPHY 18 1 (1995) Absoluts turnover rate y - 2 03 (±0 085) -f 0 317 (±0 136) log area r 2 . 0 030, n • 143, p < 0 05 -100 -10 L o g , , woodland area 100 • • Relative turnover rate y . 24 9 (11 38) - 19 6 (12 22) log area r 2 - 0 350, n - 143. p < 0 001 IT -100 -10 0 L o g , , woodland area Fig I Relationships between turnover rates of resident breeding species and woodland area for 1990-91 Absolute turnover rate (A) was calculated as (E + C)/2 where E and C were the the numbers of species extinctions and colonisations respectively between one year and the next Relative turnover rate (B) was calculated as (E + C)/(S1 + S2) X 100% where E and C were as above and S1 and S2 were the numbers of species present in each of the two years There was no significant difference in either relationship between 1990-91 and 1991-92 Woodland and landscape variables The woods themselves, their position in the landscape and the surrounding land use, were descnbed using the 24 vanables listed in Table 3 Woodland area and penmeter were measured directly in the field for the smaller woods and from 1:25 O X Ordnance Survey maps for the larger C) woods The shape of each wood was descnbed using an index calculated as Pm/Pc. where P^ was the measured penmeter of the wood and P^ was the penmeter of a circular wood of the same area. Thus the greater the edge to area ratio, the larger the index The number of "habitats" present within each wood was assessed independently by two observers. In the small number of instances when the assessments differed, consensus was reached after discussion. Habitats included features such ECOGRAPHY 18 I (1995) as ponds, dense thickets and open glades (Table 3), In general, the number of habitats increased with increasing woodland area (y = 4 13 (± 0,106) + 1 67 (± 0,172) log area; n = 146, r^ = 0 393, p < 0,001), The densiues of the canopy, shrub and field layers of each wood were scored each year on an arbitary scale of 0 to 4, where 0 = no canopy/shrub/field layer, and 4 = a completely closed canopy or a dense shrub/field layer Each vegetation layer was scored by assessing what proportion in the whole wood was attnbutable to each of the five possible scores These scores were then combined to give an overall index for each layer as* index = S (score x proportion) The minimum and maximum values for each layer were therefore 0 and 4 respiecuvely The layers were scored independently by two observers and the mean of the two assessments used m the analyses The following seven vanables were used as measures of woodland isolation The area of woodland within 0 5 and 1 0 km of the penmeter of each wood was obtained from aenal photographs (Cambndge Univ collection) taken in 1989 The positions of all hedgerows within 0,5 and 1,0 km were determined by ground surveys, plotted on 1:25000 Ordnance Survey maps, and their lengths then measured (Hedgerows could not be reliably distinguished from overgrown ditches on the aenal photographs ) Distances to the nearest wood of any size, to the nearest wood of > 2 ha and to the nearest active centre of human habitation, such as a farm or village, were also measured using 1,25000 Ordnance Survey maps The size of 2 ha was selected because, in our study area, bird species numbers accumulated most rapidly with increasing woodland area up to c 2 ha, woods of this size contained c 65% of the number of species occumng in our largest woods and c 25% of the total number of resident species found across all our woods (Bellamy et al in press) Four vanables were used as measures of connectedness (Baudry and Memam 1988), The numbers of hedges, ditches and dirt roads connected to each wood were counted in the field The fourth vanable, the total number of linear connections was calculated from the above three by combining linear features which occupied a common position For example, a hedge with a ditch alongside it or a dirt road with a hedge on one side and a ditch on the other would both be counted as single linear connections Surrounding land use was assessed annually and was expressed as the percentage of the penmeter of each wood which was adjouied by the categones shown in Table 3 Root crops compnsed sugar beet, potatoes and carrots, while "other crops" included linseed, peas and onions The non-crop category included all other land uses, such as buildings, yards, waste ground, gardens, scattered scrub and trees and woods. Cereals, rape and beans (combined) and root crops were by far the most common categones. Table 4 Stepwise multiple regression analyses of extinction and colonisaUon rates of resident species in relation to woodland structure, isolation, connectedness and sunounding land use Time interval Step Vanables entered Additional Signif% vanance icance explained ExUnction rate 1990-91 1 Hedge connect 2 Wood 0 5 km 3 % roots 1990, -ve* Total 127 <0 001 <0 05 <0 05 ColonisaUon rate 1990-91 1 2 3 4 Total Wood 0 5 km, -ve* Shrub density, -ve* % grass 1990 Shape <0 001 <0 05 <0 05 <0 05 <OOOI <0 05 1991-92 1 Hedge connect 2 No habitats Total tween 1990 and 1992 (t|37 = 1 40, NS), Thus, losses outnumbered gains in 1990-91, and gains outnumbered losses in 1991-92 Species tumover, resulting from local extincuons and colonisations, occurred across the whole size range of woods surveyed TTie mean number of species lost jier wood in 1991 was 2 2 ± 1.6 (range 0-8) and in 1992 it was 1 5 ± 1 2 (range 0-5) The mean number of species gained per wood in 1991 was 1 7 ± 1 4 (range 0-6) and in 1992 It was 2 3 ± 1 6 (range 0-7) Tumover of species was complete in four woods for 1990-91 and m one (different) wood for 1991-92 Three of these woods were very small, at 0 1 ha or less, and one was a little larger at 0 24 ha All four held no more than three breeding species in the years concemed No tumover was recorded in two woods for 1990-91 (n = 143) and in five woods for 1991-92 (n = 141), one of which was the same in both sets of years These woods ranged in size from 0 18 to 0,84 ha, except for one of 9 05 ha No meaningful relationship was found between absolute tumover rate of species and any of the predictor vanables, including woodland area (Fig la), for either 1990-91 or 1991-92 However, the relationship with area was statistically significant because of the large number of woods used and because the tumover rates in the smallest woods, with very few breeding species, were less than those in slightly larger woods, even when tumover was complete Individually, none of the vanables used explained > 10% of the vanation m absolute tumover rate and no vanables entered the stepwise regression analysis For extinction and colonisation rates (Table 4), the two most significant vanables to emerge from the analyses (area of woodland within 0 5 km and number of hedges connected to a wood) were the same for both, but the relationship between colonisation rate and area of Abbreviations as in Table 3 * denotes direction of the relationship, all are +ve unless otherwise indicated For extinction rate in 1991-92, no vanables entered the analysis Results The numbers of resident breeding species vaned from less than four per wood in the smallest woods (< 0 1 ha) to > 16 per wood in woods of 10 ha or more (Bellamy et al in press) In general, the numbers of resident species in each wood decreased between 1990 and 1991 (pan-ed t-test, ti42 = - 2 55, p < 0 05) and then increased between 1991 and 1992 (t,4o = 3 78, p < 0 001), such that there was no significant difference in the numbers of species be- Table 5 Contnbution of individual species to tumover in two size classes of woods "Small" woods (mean area±l SD=1 02±0 18 ha, n=24) Species % contnbution to absolute tumover 92 57 57 52 52 47 47 47 47 42 42 58 2 Species "Urge" woods (mean area±l SD=5,93±2 25 ha, n=16) % contnbution to absolute tumover 108 81 66 61 57 57 56 46 46 57 8 Pheasant* Wren Camon crow* Dunnock Song thrush Robin Blue Ut Great tit Red-legged partndge* Greenfinch Moorhen Total Great-spotted woodpecker Camon crow* Marsh ut Sparrowhawk Magpie* Stock dove Long-tailed tit Bullflnch Red-legged partndge* Total The percentages are expressed as the mean of each species conUibuUon to tumover in 1990-91 and 1991-92 For the small woods, the remaining 41 8% of tumover was contnbuted by a further 25 species and for the large woods, the remaining 42 2% of tumover was contnbuted by a further 22 species The species marked * are those which were regularly hunted, trapped and/or eaptive reared and released (see text) ECOGRAPHY 18 1 (1995) Table 6 Mean numbers of pairs of the species conlnbuung most to turnover in two sizes of woods Data for 1990 Species "Small" wood,s (mean area±l SD=1 02±0 18 ha, n=24) Mean no pairs (x±SD) Wren Dunnock Song thnish Robin Blue tit Great Ut G-s woodpecker Marsh tit Sparrowhawk Long-t tit Bullfinch 2 4±1 2 1 4±1 0 0 4±0 5 I4±12 13±10 0 5±0 7 0 04±0 2 0 04±0 2 0 1±0 3 0 3±0 6 0 l±05 Range 1-5 0-3 0-1 0^ 0^ 0-2 O-I 0-1 0-1 0-2 0-1 % of sample woods occupied in 1990 100 79 42 71 79 38 4 4 13 17 8 "Large" woods (mean area±l SD=5 93±2,25 ha, n=16) Mean no pairs (x+SD) 10 4±4 5 4 0±2 4 1 l±09 5 5±3 3 4 8±2 5 2 8±17 0 5±0 5 0 6±0 9 O3±O5 1 1±1 4 1 3±1 5 Range 5-18 1-9 0-3 1-11 1-10 0-7 0-1 0-2 0-1 0^ 2^ % of sample woods occupied in 1990 100 100 69 Numbers of the following species were not estimated red-legged partndge, stock dove, moorhen, greenfinch, camon crow and magpie Long-t lit = long-tailed tit, G-s woodpecker = great-spotted woodpecker woodland within 0 5 km was negative However, the relationships of extinction and colonisation rates with the individual vanables were weak and the total percentage of the vanance explained in both cases was small Despite the lack of a strong relationship between absolute turnover rate and woodland area, with the exception of game birds and corvids whose abundance and distnbution were strongly influenced by human hunting activity, the spyecies which contnbuted most to turnover changed as area increased (Table 5) The species which contnbuted most to turnover in the sample of small woods were represented by only 1-5 pairs (Table 6) In the larger woods, these same species had larger populations, were less prone to extinction and so contnbuted little or nothing to the turnover rate In the sample of large woods, the species contnbuting most to turnover seldom occurred in the small woods and were again those represented by only a small number of pairs (Table 6) The numbers of species breeding in our study woods increased with woodland area (c 70% of the vanation in resident breeding species numbers was accounted for by area alone, Bellamy et al in press) and because the numbers of species changing between one year and the next was the same regardless of woodland area, relative turnover inevitably emerged as greater in the smaller woods (Fig, lb) Despite the inevitability of this relaUonship, relative turnover did emphasize the instability of the bird populations of the smallest woods There was no difference in either intercept (F^si =2,21, NS) or slope (F| 280 = 3,76, NS) between the fitted lines for both sets of years Discussion The year-to-year changes m breeding species compwsition, resulting firom local extinctions and colonisations. ECOGRAPHY 18 1 (1995) clearly demonstrated the occurrence of species turnover in the resident bird populations of our woods Other work has also documented species turnover m birds, both for true islands (e g Diamond 1969, Jones and Diamond 1976, Wnght 1985) and for habitat islands on land (e g Beven 1976, Whitcomb et al 1977, Woolhouse 1987), but the validity and magnitude of the turnover rates recorded in some of these studies have been questioned (Gilbert 1980, McCoy 1982) The commonest causes of spurious turnover include census errors, inappropnate census intervals, the inclusion of non-breeding vagrants and changes in the avifauna directly attnbutable to human acuvity (Lynch and Johnson 1974) In our study, census inadequacies were the most important ]X)tential source of error, affecting mainly ,species in which pairs were present for only a small part of the breeding season Census deficiencies can cause turnover to be both over-estimated and under-estimated, although the former is likely to be commonest (Lynch and Johnson 1974) For example, a species present in three years, but missed in the rmddle one, will generate a spunous extinction in that year and a spunous colonisation in the next, increasing the estimate of turnover m both years. Failing to record a colonisation in one year will underestimate turnover in that year, but if the species remains, it may then produce a spunous colonisation in the following year In our study, errors were most likely to have been made in the largest woods through the occasional failure to record a species as present and breeding This would have resulted in an over-estimate of turnover in large woods relative to small ones In this case, the real trend in relative turnover with woodland area might have been even more marked than Fig lb suggests, and absolute turnover might have been more sensitive to increasing woodland area than indicated by Fig, la However, the numbers of species lost and gained between years in our study were, on average, small, suggesung that census errors were not common. 47 Small woods support only small numbers of individuals of the species present, and such small populations have the greatest nsk of stochastic extinction Because the nsk of stochastic extinction should decrease as population size increases, it might have been expected that absolute turnover, and especially extinction rate, would have been related to woodland area (m general, large woods should contain larger numbers of individuals of any given species than should small woods), but this was not the case (Fig la. Table 4) In both small and large woods. It was the species represented by only a few breeding pairs which contnbuted most to turnover and, m the main, the species concerned were different in small and large woods (Table 5 and 6) The only species contnbutmg substantially to turnover irrespective of woodland size were those in which extinction and colonisation rates were artificially increased by shooting and trapping, and in some cases (pheasant and red-legged partndge), by captive reanng and release Withm our study area, different species occurred at different regional densities and may also have had different minimum habitat area requirements (van Dorp and Opdam 1987) Hence, as woodland area increased, the population sizes of some sfwcies reached levels where extinction due to stochastic processes became uncommon, while others were still only represented by a small number of individuals. Thus a decrease in extinction rate among the most abundant species was offset by the high rates among the rarer ones resulting m no overall relationship between extinction rate, and hence turnover rate, and area This process should continue until the woodland is of sufficient area to support populations of the whole species pool large enough to usually avoid stochastic extinction. The absence of a relationship between absolute turnover and any of the 24 predictor vanables also suggested that stochastic processes were paramount The above argument is valid for homogeneous woodland habitat In reality, species may be missing from an adequate area of woodland if their specific habitat requirements are lacking. Area and habitat diversity are closely correlated (as shown for our woods), often so closely that it is hard to separate their effects (Lack 1976), Thus, habitat mfiuences could generate the same sort of pattern of local extinctions in species with small population sizes as suggested above for area. In this case, species would have small populations because their habitat, rather than area per se, was limiting In the absence of a strong influence of isolation on colonisation rates, the form of the relationship between turnover rate and woodland area may be determined by the interaction of habitat availabibty and area, and the autecological requirements of the individual species present in the locality Several other studies have found that turnover is due mainly to species with small populations (Williamson 1981, Wnght 1985), which come and go against a background of large populations with apparent long-teim stabihty (Williamson 1983, Schoener and SpiUer 1987) 48 This has led to the suggestion that the majonty of turnover IS "ecologically tnvial" (Williamson 1989, Hamson 1991) In conservation terms, this could be a dangerous assertion for species in which the majonty of the population now exists in fragmented habitat, especially when population density withm the fragments is low (Doak and Mills 1994) Lx)cal extinction of the populations of some fragments may increase the isolation of the remainder beyond the ability of the species to recolonise through dispersal, resulting m regional extinction (Barinaga 1990, Rolstad 1991, Fahng and Memam 1994) When a species has been reduced to small populations in scattered habitat patches, the overall fate of the species will be dominated by the stochastic processes affecting the individual small populations The movement of birds between our woods was not of the mainland-island type descnbed by the equilibnum theory of MacArthur and Wilson (1967) Movement appeared to be diffuse, with individuals moving between the woods themselves and between the woods and other habitats such as gardens and hedgerows Specifically, extinction rates (expressed as numbers of sfiecies lost) in our study were not related to woodland area (= island area), the main cause of extinction being small population size as descnbed above Colonisation rates did not increase with decreasing degree of woodland isolation Indeed, for 1990-91, colonisation rates were inversely related to the area of woodland within 0.5 km, suggesting the opposite trend This same trend was also suggested by the positive relationship between extinction rates and the area of woodland within 0 5 km also found for 1990-91 The lnfiuence of surrounding woodland on extinction and colonisation rates may have ansen as a result of space becoming available in larger woods (assuming that such large woods are preferred habitat) following a penod of severe weather in February 1991 which caused a reduction in the overall size of woodland bird populations in eastern England in 1991 (Marchant and Musty 1992) The winter of 1991-92 was mild and bird populations in general were higher in 1992 than m 1991 Unlike 1990-91, more species were gamed in our woods in 1991-92 than were lost and therefore it was not surpnsing that the small infiuence of woodland and landscape vanables on extinction rates found in 1990-91 was not apparent in 1991-92 (Table 4) The result for colonisation rates in 1991-92 suggested a possible role for hedgerows as comdors (Hinsley et al. 1995), but the relationship was weak. Overall, the small amount of vanation in extinction and colonisation rates explained by the woodland and landscape vanables, and the lack of consistency in the results for the two sets of years, emphasized the lmpwrtance of stochastic processes in this highly fragmented habitat Discrete populations linked by dispersal compnse the basic stiTicture of a metapopulation (sensu Levins 1970) However, if dispersal rates are high, a fragmented population can still function essentially as a single population (Hamson 1991) Of the species in our woods, those most ECOGRAPHY 18 I (1995) Angelstam, P 1992 Conservation of communities - the importance of edges, surroundings and landscape mosaic structure - In Hansson, L (ed ), Ecological principles of nature conservation application in temperate and boreal environments Elsevier Applied Science, London, pp 9-70 Bannaga, M 1990 Where have all the froggies gone''- Science 247 I033-I034 Baudry, J and Memam, G 1988 Connecuvity and connectedness functional versus structural patterns in landscapes In Schreiber, K F (ed ), Proc 2nd Int Seminar, Int Ass Landscape Ecol, Munstersche Geogr Arbeiten 29 23-28 Bellamy, P E , Hinsley, S A and Newton, I In press Factors influencing bird species numbers in small woods in southeast England - J Appl Ecol Beven, G 1976 Changes in breeding bird populaUons of an oak-wood on Bookham Common, Surrey, over twentyseven years - Lond Nat 55 23-42 Butcher, G S , Nienng, W A , Barry, W J and Goodwin, R H 1981 Equihbnum biogeography and the size of nature reserves an avian case study - Oecologia 49 29-37 Diamond, J M 1969 Avifaunal equilibria and species turnover rates on the Channel Islands of California - Proc Nat Acad SCI USA 69 3199-3203 Doak, D F and Mills, L S 1994 A useful role for theory in conservation - Ecology 75 615-626 Fahng, L and Memam, G 1994 Conservation of fragmented populations - Cons Biol 8 50-59 Gilbert, F S 1980 The equilibrium theory of island biogeography fact or fiction'' - J Biogeogr 7 209-235 Haila, Y and Kuusela, S 1982 Efficiency of one-visit censuses of bird communiues breeding on small islands - Omis Scand 13 17-24 - and Hanski, I K 1984 Methodology for studying the effecLs of habitat fragmentation on land birds - Ann Zool Fennici 21 393-397 - , Hanski, 1 K and Raivio, S 1993 Turnover of breeding birds in small forest fragments the "sampling" colonizauon hypothesis corroborated - Ecology 74 714-725 Hanski, I K and Haila, Y 1988 Singing teentones and home ranges of breedmg chaffinches visual observation vs radiotraeking - Omis Fennica 65 97-103 Hams, L D 1984 The fragmented forest - Univ Chicago Overall, the difficulties of scale and of movements Press, Chicago discussed above are unlikely to alter the conclusion that Hamson, S 1991 Local exUnction in a metapopulauon context an empincal evaluaUon - In Gilpin, M and Hanski, I species turnover rates in small, fragmented woods are (eds), Metapopulauon dynamics empirical and theoreucal high, and that woodlands of up to at least 10 ha are invesugauons Reprinted fi^m the Biol J Linn Soc Vol insufficient to maintain populations of certain woodland 42 Academic Press, London, pp 73-88 species large enough to avoid purely stochastic extinc- Hinsley, S A , Bellamy, P E , Newton, I and Sparks, T H 1995 Habitat and landscape factors influencing the presence tion In Bntain and elsewhere, a large proportion of many of individual breeding bird species in woodland fragments species populations exist in fragments where their numJ Avian Biol in press bers are liable to be small enough to show appreciable IUCNAJNEPAVWF 1991, Canng for the earth A strategy for turnover and where species long-term persistence will sustainable living - Gland, Switzerland Jones, H L and Diamond, J M 1976 Short-time-base studies depend on continued and effective dispersal of tumover in breedmg bind populations on the Califomia Channel Islands - Condor 78 526-549 Acknowledgements - We would like to thank the many lan- Lack, D 1954 The natural regulauon of animal numbers downers, farmers, managers, tenants and gamekeepers who have Oxford Univ Press, Oxford kindly allowed us to work on their land and without whose help - 1976 Island biology - Univ of Califomia Press, Los Anand co-operation this project would not have been possible geles Thanks are also due to the Bedfordshire and Cambridgeshire Levins, R 1970 Exunction - In Gerstenhaber, M. (ed ), LecWildhfe Trust, English Nature and Cambndgeshire County tures on mathematics in the life sciences, Vol 2 Some 3 ECOGRAPHY 18 1 (1995) likely to show a metapopulation strucmre would be species such as marsh tit which are restncted to woodland and have short dispersal distances However, whether or not a species can make the transition from contiguous habitat to a metapopulation would depend not only on its dispersal charactenstics, but also on the quality of the available habitat fragments and the landscape between them For example, nuthatches Sitta europaea, which have been reported to have a metapopulation structure in small woods in the Netherlands (Verboom et al 1991) are rare or absent in our study area (one pair bred in one study wo(xi in 1990 only), even in suitable woodland of over 100 ha This suggests that the degree of isolation of these woods from surrounding potential source populations IS sufficient to prevent colonisation and the development of a metapopulation The mobility of some species, such as great-spotted woodpecker Dendrocopus major, may alter the scale at which turnover should be considered for that species (Haila et al 1993) Apparent turnover may result from pairs nesting in one wood rather than another withm the same home range However, most of the species in our woods, and certainly most of those responsible for the majonty of turnover, were thought to be more or less site faithful (Lack 1954) The results of a marking program to be reported elsewhere, have so far indicated that movements between woods of breeding adults of the species typical of our small woods are uncommon (Hinsley unpubl data) Other evidence suggests that breeding individuals may range more widely than is apparent from temtory size determined by song post location (e g Hanski and Haila 1988) Indeed, we have occasionally seen some of our marked birds in woods adjacent to their "own" However, such individuals do not sing and generally keep a low profile and withm the context of our study would probably be recorded as transient Such movements by small birds are also likely to be less common in fragmented than in continuous woodland because of the costs (energy and time) and nsks (predators such as sparrowhawks) involved in regularly crossing open ground Council for access to woodland, to T Sparks for statisucal advice and to two referees for comments improving the manuscnpt This work was initiated as part of the Joint Agncultural and Environment Program (JAEP)

Journal

Ecography – Wiley

Published: Mar 1, 1995

There are no references for this article.