This article was downloaded by: [Charman, Elisabeth] On: 10 February 2010 Access details: Access Details: [subscription number 919184766] Publisher Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 3741 Mortimer Street, London W1T 3JH, UK Bird Study Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t904369352 National-scale analyses of habitat associations of Marsh Tits Poecile palustris and Blue Tits Cyanistes caeruleus: two species with opposing population trends in Britain Jane Carpenter a; Jennifer Smart b; Arjun Amar b; Andrew Gosler a; Shelley Hinsley c; Elisabeth Charman b a Edward Grey Institute for Field Ornithology, Department of Zoology, University of Oxford, Oxford b Royal Society for the Protection of Birds, Sandy, Bedfordshire c Centre for Ecology and Hydrology, Huntingdon, Cambridgeshire First published on: 11 November 2009 To cite this Article Carpenter, Jane, Smart, Jennifer, Amar, Arjun, Gosler, Andrew, Hinsley, Shelley and Charman, Elisabeth(2010) 'National-scale analyses of habitat associations of Marsh Tits Poecile palustris and Blue Tits Cyanistes caeruleus: two species with opposing population trends in Britain', Bird Study, 57: 1, 31 — 43, First published on: 11 November 2009 (iFirst) To link to this Article: DOI: 10.1080/00063650903026108 URL: http://dx.doi.org/10.1080/00063650903026108 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material. Bird Study (2010) 57, 31–43 National-scale analyses of habitat associations of Marsh Tits Poecile palustris and Blue Tits Cyanistes caeruleus: two species with opposing population trends in Britain JANE CARPENTER1*, JENNIFER SMART2, ARJUN AMAR2, ANDREW GOSLER1, SHELLEY HINSLEY3 and ELISABETH CHARMAN2 1 Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Edward Grey Institute for Field Ornithology, Department of Zoology, University of Oxford, South Parks Road, Oxford, OX1 3PS, 2Royal Society for the Protection of Birds, The Lodge, Sandy, Bedfordshire, SG19 2DL and 3 Centre for Ecology and Hydrology, Monks Wood, Abbots Ripton, Huntingdon, Cambridgeshire, PE28 2LS Capsule Marsh Tits were strongly associated with both the amount and species diversity of woodland understorey; Blue Tits were associated with large trees and deadwood. Aims To gather quantitative information on the habitat requirements of Marsh Tits, in comparison with those of Blue Tits, across a large number of sites in England and Wales, and secondly to evaluate the range of habitat conditions likely to encourage the presence, and increase the abundance of, each species. Methods Counts of birds were made at each of 181 woods across England and Wales, and habitat data were collected from the same locations in each woodland. Marsh Tit and Blue Tit presence and abundance were related to habitat characteristics, interspecific competition and deer impact. Results Shrub cover and species diversity were important for the presence and abundance of Marsh Tits, across their geographical range in Britain. Blue Tits were associated with large trees and deadwood. Conclusion Our results support the hypothesis that changes in woodland management, leading to canopy closure and a decline in the understorey available, could have had an impact on Marsh Tits, and may have led to the observed population decline. These same changes were also consistent with population increase in Blue Tits. Since the 1920s, woodland habitat across Europe has been changing rapidly (Tucker & Evans 1997). Afforestation of open land, along with intensively managed plantations of conifers or non-native species, have left only fragmented patches of semi-natural forest (Tucker & Evans 1997). In Britain, the pattern is similar. Although the forested area increased from 5% at the turn of the century (Richards 2003, Hopkins & Kirby 2007) to 12% in 2007, over 60% of this increase was due to conifer plantation (Mason 2007). Furthermore, the age structure of trees in British woodlands is biased towards maturity with at least 70% of both conifer and broadleaved forest stands now entering a closed-canopy stage (Mason 2007). This change in broadleaved forest is probably due to the cessation of woodland management *Correspondence author. Email: mothgirl1@yahoo.co.uk © 2010 British Trust for Ornithology techniques such as coppicing (Amar et al. 2006, Hopkins & Kirby 2007), although changing timber markets could also be relevant (Fuller et al. 2007). There has been increasing concern about the health of woodland bird populations in Britain (Fuller et al. 2005, Amar et al. 2006, Hewson et al. 2007). National bird population monitoring schemes have reported declines in the populations of many woodland bird species over the last 40 years (Eaton et al. 2006, Gregory et al. 2002). The most recent revision of the list of Birds of Conservation Concern in the UK placed seven species on the Red List, and nine on the Amber List (Gregory et al. 2002). However, compared with the wealth of research into reasons for the decline of farmland birds in Britain (Brickle et al. 2000, Chamberlain et al. 2000, Vickery et al. 2004), there has been limited research into factors affecting woodland bird populations. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 32 J. Carpenter et al. The joint Royal Society for the Protection of Birds (RSPB) and British Trust for Ornithology (BTO) Repeat Woodland Bird Survey (RWBS) was partially designed to address this gap in knowledge, as well as to test whether population trends in woodland birds, as detected by the Common Bird Census, were valid. It focused first on the changes in bird populations in woodland habitats across 20 years and second on habitat change within woods and the link between population decline, habitat change and current conditions (Amar et al. 2006). Two hundred and fifty two woodland sites from an original census in the 1980s were revisited in the early 2000s. Nine out of 34 bird species showed large (<25%) population declines, but a further 11 showed a large population increase (Amar et al. 2006, Hewson et al. 2007). There was evidence that changes in woodland structure resulting from woodland maturation, a reduction in woodland management and, possibly increased deer browsing in some areas, could be important factors influencing the declines of some species. Basic ecological data for many woodland bird species are lacking (Amar et al. 2006), thus, the RWBS dataset was also used to determine the habitat requirements of several declining species, in parallel with those of some whose populations are increasing, to draw comparisons and gain insights into possible reasons for these contrasting trends (Smart et al. 2007). Here, we report on the findings for two closely related species; the declining Marsh Tit Poecile palustris and the increasing Blue Tit Cyanistes caeruleus. The Marsh Tit was one of three lowland woodland bird species to be ‘Red Listed’ during the last update of the Birds of Conservation Concern, due to a long-term population decline of over 50% in the last 25 years, detected by the national monitoring schemes (Gregory et al. 2002). This decline was also demonstrated by Perrins (2003) using long-term ringing data showing that the decline could be as much as 70% since the 1960s. However, recent studies suggest that the population may have stabilized at a historically low level, and may even have increased slightly in recent years (Amar et al. 2006, Eaton et al. 2006). There has been little research on Marsh Tits in Britain since the 1950s (Colquhoun & Morley 1943, Hartley 1953, Morley 1953, Gibb 1954), and hence there is little current understanding of the causes of decline. The early studies identified Marsh Tits as being more specialized ecologically than their dominant counterparts, Blue Tits and Great Tits Parus major, due to their apparent reliance on the understorey layer within mature woodlands, year-round territoriality and habit of storing food (Gibb 1954, Perrins 1979). © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 However, all of these early studies were focused on only one or two woodlands in Oxfordshire. Hinsley et al. (2007) provide recent evidence of the continued importance of the understorey to this species. Carpenter (2008) showed that although niche separation continues between Blue and Marsh Tits to some extent, there is also considerable overlap in the foraging behaviour of the two species, both in terms of actual foraging behaviour, and vertical and horizontal location in the woodland habitat. Furthermore, some evidence for competition between the two species was also found. This raises questions about the possible role of, first, habitat change, and, secondly, competition, in the species’ decline. British woodland habitat has changed in recent years (Amar et al. 2006, Mason 2007), attributed to canopy closure (Fuller et al. 2005) and overgrazing by deer (Gill & Fuller 2007). However, Broughton et al. (2006) found little difference in understorey cover between areas of a woodland occupied and unoccupied by Marsh Tits, and instead found evidence of the importance of canopy characteristics. They noted that understorey cover was very homogenous at their site, perhaps explaining this result, but the study still highlights the lack of knowledge of how Marsh Tits’ use of understorey is influenced by other factors. Secondly, recent evidence of a localized effect of competition by Blue Tits (Carpenter 2008) suggests this also needs further investigation. This is particularly pertinent (Perrins 2003) given the recent increase in the Blue Tit population (+33% in the last 25 years, Eaton et al. 2006) in contrast to the Marsh Tits’ decline. Only one published study (Hinsley et al. 2007) has examined habitat requirements of Marsh Tits in Britain in more than two woodlands, although only certain variables, again concentrated on the understorey, were included in the large-scale analysis. Thus, analyses of the extensive dataset collected for the RWBS (a total of 252 woods, of which we could use 181) provided a valuable opportunity to investigate the current habitat associations of Marsh Tits across the species’ range. Our study had two main aims: first, to gather quantitative information on the habitat requirements of Marsh Tits, in comparison with those of Blue Tits, across a large number of sites on a national scale, and, secondly, to evaluate the range of habitat conditions likely to encourage the presence, and increase the abundance, of each species. This provided a foundation upon which habitat management, and formal hypothesis testing of its effectiveness, could be based. Direct analysis of changes in bird abundance in relation to habitat change within the RWBS data set was carried out by Amar Habitat associations of two woodland birds et al. (2006) and indicated that, for the RWBS data, such an approach was not appropriate for Marsh Tits. METHODS Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Data collection A detailed account of the methods used for field data collection is given in Amar et al. (2006) and Smart et al. (2007). However, a summary is included here. Although the RWBS dataset consisted of sites censused by volunteers working either for the BTO (territory mapping, total number of sites available = 113) or the RSPB (point counts, total number of sites available = 252), the different surveys had different methods and could not be merged. Therefore, only the RSPB sites were used. Figure 1 shows the distribution and clustering of sites used. For the current study, we only used woodlands in England and Wales, to reflect the current distribution of Marsh Tits. This gave a total sample size of 181 woods. For the purpose of these analyses, Kent and Hertfordshire sites (n = 6) were joined with Buckinghamshire to create 33 a south-east England group of sites. For Marsh Tits, Gloucestershire and the Forest of Dean were grouped together, as were Suffolk and Northamptonshire, to increase locality-specific sample sizes or to remove zero marginals in occupancy analyses. Bird presence and abundance estimates were obtained using a point count method in 2003 and/or 2004. Point counts lasted 5 minutes and were carried out twice, first in April or the first week of May, then in the last 3 weeks of May or first half of June. There were at least 10 points per wood. Points were at least 100 m apart, and at least 50 m from the edge of the wood. The maximum count across the two visits (and across the 2 years if the site was surveyed in both years) gave the abundance estimate. Volunteers recorded whether birds were nearer or further than 25 m from the point at first detection. This was done to enable a density to be derived for each species per site, using the software program distance (Buckland et al. 2001). However, problems were identified in analysing our data with distance. This was due to the use of only two distance bands when recording species, which were determined from Figure 1. The location of all study woodlands in England and Wales showing the localities within which woodlands were clustered. For both species, Kent and Hertfordshire sites were joined with Buckinghamshire to form ‘south-east England’ sites. For Marsh Tits, Gloucestershire and the Forest of Dean were grouped, as were Suffolk and Northamptonshire. © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 34 J. Carpenter et al. historical methodologies (see Amar et al. 2006). This made the density estimates very sensitive to incorrect assignment of individual birds into either of the two bands, and we therefore concluded it would be far more robust to use the actual counts in subsequent analyses. Habitat data were collected at each point-count location in study sites, with this point forming the centre of a 25 m radius circle. Some data were collected at this scale, and some from four 5 m radius sub-plots centred 12.5 m from the centre of the plot in each of the four cardinal directions. For variables recorded at this sub-plot level, means of the four datasets were taken. Habitat data at the woodland scale were calculated as the mean of each variable across all point-count locations per wood. A table outlining each habitat variable, the level and unit of measurement and a description of how each habitat variable was collected is provided in Appendix 1. Certain non-habitat variables were also included in the analyses to test some suggested reasons for Marsh Tit decline (see Appendix 2). The composition of surrounding habitat within 3 km radius buffer circles centred on the central location of each site was calculated. This was done using the Centre for Ecology and Hydrology’s (CEH) Land Cover Map (LCM) 2000 within ArcGIS Version 9. The percentage composition of all habitat classes at LCM level 2 within these circles was calculated. The 15 habitat variables with the highest percentage around sites contributed 98% of the total area, and these variables were then grouped into eight broad habitat categories. Principal components analyses (PCA) were used to reduce the number of landscape variables entering the analyses. Principal Component (PC) 1 explained 28% of the variation, and described a gradient from an agricultural landscape to a non-agricultural landscape, whereas PC2 (explaining 17% of the variation) was a gradient from a wooded landscape to a non-wooded, grassier landscape (Amar et al. 2006). Data on spring weather conditions for each site were obtained from the UK Met Office Climate Impacts Programme (CIP). The 5-year average (1996–2000) was calculated for three weather variables for April and May: temperature, rainfall and the number of days where rainfall ≥ 1 mm. PCA was again used to reduce the number of climate variables entering the model. PC1 explained 76% of the variation, and described a gradient from the drier east to the wetter west (Amar et al. 2006). Statistical analysis A three-stage model selection process was used with both species, and for both models for Marsh Tits (see © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 later). First, the significance of each covariate was tested individually, and any which were not significant at the 10% level were removed from further analysis. At this stage, we also tested for quadratic relationships in nine variables (eight associated with tree and understorey structure plus altitude) thought most likely to show such relationships. In the second stage possible covariates were categorized into groups of similar variables (Table 1), these being: (a) those associated with large-scale variables; (b) field-layer characteristics; (c) understorey structure; (d) tree structure; (e) deadwood; (f) landscape; (g) deer impact; and (h) interspecific competitors (the latter for Marsh Tits only). For some covariates there were no similar variables, these, therefore, remained ungrouped and did not enter Stage 2. If more than one of the variables in a given group was significant at the univariate stage, they were entered into a multivariate backward stepwise model, and only those terms which remained significant at the 10% level were entered into the final model stage. In the third stage, those variables remaining from Stage 2, and any from Stage 1 which were not in a group, were entered into a final model. This model was run in a backward stepwise fashion again, removing the least significant term, until only those terms significant at the 5% level remained. These then formed the final models for the two species. We carried out this three-stage process to account for intercorrelations between predictors. We did not want to reduce the number of variables entering the model, as so little is known about the requirements of these bird species, and to do so could have inadvertently removed important variables. We are aware that this is something of a data-mining approach, but feel that as a first stage exploration of national habitat requirements this approach is justified. The results we present are therefore meant as a first step, and further research testing our initial findings more stringently should follow. We were interested, first, in the woodland-scale correlates of species presence. However, as Blue Tits were present in all of the woodlands studied, this analysis was only carried out for Marsh Tits. The probability of Marsh Tit presence was modelled using the binary logistic regression procedure (proc logistic) in sas 9.1. Goodness-of-fit was tested using the Hosmer– Lemeshow statistic (Hosmer & Lemeshow 1989), and a range of model performance statistics, as presented in the results, was examined. Secondly, in occupied woods only (for both Blue Tits, n = 181 and Marsh Tits, n = 114), we examined Habitat associations of two woodland birds 35 Table 1. A comparison of the results of modelling the habitat correlates of Marsh Tit presence and Marsh Tit and Blue Tit abundance in woodlands. Marsh Tit Groups Variables Large-scale Locality Weather PCA Bracken Bramble Herb Grass Moss Bare ground Leaf litter Cover 0.5–2 m Cover 2–4 m Cover 4–10 m Horizontal visibility Canopy cover Basal area dbh Max height Dead trees Dead limbs Fallen wood PC1 3 km PC2 3 km Deer PC1 Deer PC2 BT abundance GT abundance BT + GT abundance Dominant tree Lichen Ivy Shrub diversity Water features Altitude Size Tracks Drey density Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Field layer Understorey Tree size Deadwood Landscape GIS Deer Interspecific competition Ungrouped Blue Tit Presence Abundance Abundance Logistic GLM GLM n = 181 n = 114 n = 181 36.3 °°°° 0.02 ++++ 0.52 −− 0.89 ++ 0.19 + 0.53 −− 0.23 −−− 0.34 ++ 0.13 +++ 6.4 ++ 1.86 −−−− 8.0 +++ 3.16 °°° 0.00 + 10.51 °°°° 0.62 0.75 3.80 2.04 0.01 0.02 0.54 10.99 1.71 −− −−−− 0.05 1.65 + 0.36 − 0.90 ++++ 0.09 0.46 na na na ++ 5.05 0.07 6.6 − −− −− +++ ++ + −−−− −− 1.12 1.51 −− ++ 0.07 +++ 0.07 +++ 13.46 +++ 1.96 °° + ++ 1.91, 1.74, ns, ∩ 4.80, 5.08, ns, ∪ 0.55 − 4.87 1.03 +++ ++ 0.61 ++ 12.64 ++++ 0.18 +++ 0.94 −−−− 2.61 −−− 0.43 ++ 0.02 −−−− na na na 1.56 °°° −−−− ++++ 0.49 −− 2.59 0.05 + +++ 1.80 − 0.00 ++ 0.50, 0.42, ns, ∪ 0.52 + Note: Variable names in bold are those where the effect of the quadratic term was tested. Fully shaded grey cells are the variables retained in the final model stage, grey shaded values (not full cells) denote the variables retained after the within-group analysis and un-highlighted values the variables significant at a univariate stage. Effect sizes of the final model variables are shown; and of all other significant variables after being added into the final model one by one (presence analysis = Wald statistic, abundance analysis = F statistic). The number of symbols shows the level of significance (i.e. +P < 0.1; ++P < 0.05; +++P < 0.01; ++++P < 0.001) and the direction of the relationship (i.e. +, −, ∩ or ∪); °, categorical variable, hence no directional effect; na, variable not appropriate for the species/test; ns, not significant; GLM, generalized linear model; PCA, principal components analysis; PC, principal component; GIS, geographic information system; GT, Great Tit; BT, Blue Tit. the correlates of bird species abundance. A generalized linear model (glm: proc genmod) in sas 9.1 was used. The total number of individuals of each species counted in a wood was fitted as the response variable. A Poisson error structure was specified, with a logarithmic link and the natural logarithm of the number of points surveyed in each wood as an offset, to account for the likelihood of higher species counts in woods where more points were surveyed. The use of this offset, meant that we were effectively modelling the number of birds per point. The proportion of deviance (R2) explained by locality, the final model covariates, and © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 36 J. Carpenter et al. both locality and the covariates combined were examined, and are presented in the results. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Conditions required for habitat management The second aim of this study was to evaluate the range of habitat conditions which should encourage the presence, and increase the abundance, of each species. Therefore, we used the parameter estimates from the model outputs to calculate the range of habitat conditions that were likely to lead to a greater than average probability of wood occupancy and greater abundance. For each species, we completed four steps. First, we calculated the average occupancy and abundance of each bird species, and, secondly, using each covariate retained in final models or univariately significant at P < 0.01, we ran the univariate analysis to obtain the intercept and parameter estimates. Thirdly, we used these figures, along with the average occupancy and abundance, to solve for the value of the covariate x1 at which average occupancy or abundance occurred. We repeated this step but used the maximum occupancy or abundance to solve for x2 at which maximum occupancy or abundance occurred (details of the equations used are given in Smart et al. 2007). Finally, we also calculated the maximum and minimum values measured for each covariate and constrained the predicted range of habitat conditions (x1–x2) between these values. Some of the units of measurement for the habitat covariates were not ideal for translation into management prescriptions. We therefore converted these to more meaningful measures. For example, shrub diversity is calculated for the main analyses as the proportion of all shrub species present (n = 36); for the management prescriptions we converted this back to an actual number of shrub species, to give woodland managers a target number, by multiplying the given value by 36. Full details of all conversions are provided in Smart et al. (2007). concordant = 88.6, R2 = 0.54, Hosmer and Lemeshow goodness-of-fit = 0.77; Table 1). Marsh Tits were more likely to inhabit woods in the south and east of England, and at higher values of shrub diversity, canopy cover and understorey cover at 2–4 m (mean ± se: shrub diversity, occupied = 0.27 ± 0.01, unoccupied = 0.20 ± 0.01; canopy cover, occupied = 12.5 ± 0.17, unoccupied = 11.7 ± 0.23; cover 2–4 m, occupied = 26.5 ± 1.37, unoccupied = 18.1 ± 1.49; Fig. 2). Abundance of Marsh Tits and Blue Tits in occupied woods The abundance of Marsh Tits in woods (n = 114) was associated with 17 of the covariates in the univariate analysis (Table 1). Eight of these were entered into the final model stage. Locality and three covariates were retained in the final model; these were strongly associated with Marsh Tit abundance (P < 0.01). This model explained 30% of the variation in Marsh Tit abundance (locality only, R2 = 0.18; other covariates only, R2 = 0.22, Table 1). Abundance increased with decreasing horizontal visibility, increasing predominance of woodland in the landscape and increasing numbers of interspecific competitors (Fig. 3). The abundance of Blue Tits in woods (n = 181) was associated with 20 of the covariates in the univariate analysis (Table 1). Twelve of these were entered into the final model stage. Locality and two covariates were retained in the final model, which explained 27% of the variation in Blue Tit wood-abundance (locality only, R2 = 0.18; habitat covariates only, R2 = 0.13; Table 1). Abundance increased strongly with increasing diameter at breast height (dbh) of trees and increasing number of dead limbs (Fig. 4). Removing the two apparent outliers from the dead trees dataset (see Fig. 4) did not change the inclusion of this variable in the final model. Therefore, we left these two data points in the dataset. Conditions required for habitat management RESULTS Occupancy of woods by Marsh Tits One hundred and eighty one woods were included in the occupancy analysis, of which Marsh Tits occupied 114. For the first stage of the analysis, 22 of the 34 covariates tested had a univariate association with Marsh Tit occupancy (Table 1). Of these, 14 were entered into the final model stage. Three habitat covariates plus locality were retained in the final model (area under curve = 0.89, % © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 Our second aim was to provide woodland managers with criteria to manage woodlands for Marsh Tits. Table 2 shows the final model covariates for each species, and the range of conditions required to achieve mean to maximum occupancy and abundance. Table 2 shows that, for Marsh Tits, ensuring good understorey cover (and hence low horizontal visibility) and high shrub species diversity are the most important aims for habitat management. Conversely, for Blue Tits, there was no importance placed on managing the 37 Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Habitat associations of two woodland birds Figure 2. The influence of (a) shrub diversity; (b) vegetation cover 2–4 m; and (c) canopy cover on the probability of Marsh Tits occupying woods (final model: R2 = 0.54; locality, Wald χ27 = 36.3, P < 0.0001; shrub diversity, Wald χ21 = 6.6, P = 0.01; cover 2–4 m, Wald χ21 = 6.4, P = 0.01; canopy cover, Wald χ21 = 8.0, P = 0.05). Bars represent the number of woods from which Marsh Tits were absent (grey bars) or present (clear bars). Lines were fitted from the final model output (solid line) and from the final model minus the locality effect, if the habitat covariate was still significant (dashed line; P < 0.05). Each line was fitted after accounting for the parameter estimates of the other continuous explanatory variables in the model, assuming a mean value for each. For explanation of units see Appendix 1. Figure 3. The effect of (a) horizontal visibility; (b) landscape PCA 2; and (c) the abundance of interspecific competitors on Marsh Tit abundance within occupied woods (final model: R2 = 0.30; locality, F7,114 = 3.16, P = 0.004; horizontal visibility, F1,114 = 10.99, P = 0.001; landscape PCA2, F1,114 = 5.05, P = 0.03; interspecific competitor abundance, F1,114 = 13.46, P = 0.0004). Lines were fitted from the final model output (solid line) and from the final model minus the locality effect, if the habitat covariate was still significant (dashed line; P < 0.05). Each line was fitted after accounting for the parameter estimates of the other continuous explanatory variables in the model, assuming a mean value for each. For explanation of units see Appendix 1. © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 38 J. Carpenter et al. Figure 4. The effect of (a) maximum tree diameter at breast height; and (b) the number of dead limbs on trees on Blue Tit abundance within occupied woods (final model: R2 = 0.27; locality, F9,181 = 4.8, P < 0.0001; dbh, F1,181 = 5.8, P = 0.02, number of dead limbs, F1,181 = 11.9, P = 0.0007). Lines were fitted from the final model output (solid line) and from the final model minus the locality effect, if the habitat covariate was still significant (dashed line; P < 0.05). Each line was fitted after accounting for the parameter estimates of the other continuous explanatory variables in the model, assuming a mean value for each. For explanation of units see appendix 1. understorey layer, and a lack of management, increasing the availability of large mature trees and dead wood were found to favour high abundance of Blue Tits. DISCUSSION Understanding the habitat requirements The first aim of this study was to provide quantitative evidence of the habitat and other requirements of © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 Marsh Tits and Blue Tits across England and Wales. Although we would not expect wholly uniform birdhabitat effects across all localities, we were interested in detecting general national trends, which existed regardless of the locality concerned, to enable us further to understand the requirements of these species at the national scale. Therefore, although there are likely to be regional, or locality-specific, trends contained within the data, we have not focused on these here, and consider them to be beyond the scope of this study. Marsh Tit occupancy was best predicted by locality and three of the habitat covariates: cover at 2–4 m, canopy cover and shrub diversity. This final model explained 54% of the variation, suggesting that the model fitted the data well, and indicating that these understorey variables, and canopy cover, are important for the presence of Marsh Tits in woodland. Furthermore, three out of the four ‘understorey group’ variables tested were significant for Marsh Tit occupancy, two of which entered the final model stage. Only cover at 4–10 m in the understorey group showed no detectable relationship. This was strong evidence that a diverse understorey with good cover, up to 4 m, is important for Marsh Tit woodland occupancy. At occupied sites, Marsh Tit abundance was best predicted by locality, horizontal visibility, the amount of wooded habitat in the surrounding landscape, and the number of interspecific competitors. This model explained 30% of the variation in Marsh Tit abundance. There was a strong negative relationship between Marsh Tit abundance and horizontal visibility, i.e. high abundance occurring with low visibility, although the other understorey group covariates showed only weak relationships. Therefore, we conclude that the amount of understorey cover is important in predicting both occupancy and abundance of Marsh Tits in woods in England and Wales. Given recent and past studies on Marsh Tits, relationships with the understorey layer (Cramp & Perrins 1993, Hinsley et al. 2007), deadwood, (Cramp & Perrins 1993) woodland size (Hinsley et al. 1996), tree height (Broughton et al. 2006) and a wooded landscape (Cramp & Perrins 1993), could have been expected. We did find relationships with understorey cover and landscape PC2 (wooded landscape), and the relationship with canopy cover perhaps supports the relationship with tree height found by Broughton et al. (2006); only a weak negative relationship with dead wood was recorded, however. As Marsh Tits are hole-nesting birds and, in the UK, do not readily take to nest boxes, Habitat associations of two woodland birds 39 Table 2. Predicting the habitat conditions which are likely to provide average to maximum occupancy (Marsh Tit only) and abundance (Marsh Tit and Blue Tit) of the study species. Species/test Group Marsh Tit presence Understorey Tree size Other Marsh Tit abundance Understorey Landscape Blue Tit abundance Tree size Deadwood Variable Range Locality Cover 2–4 m (%) Canopy cover (%) Shrub diversity (no. spp) Locality Horizontal visibility (%) Wooded habitats (%) Locality dbh (cm) Dead limbs (No. ha−1) 0–66 0–100 0–20 30–96 0–66 24–127 0–528 Average Max − Wales, + East 22 76 8 − Wales, + South 65 29 + South and East 59 56 66 100 20 39 66 127 528 Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Note: Only those habitat variables retained in the final model (see text) are presented. Range = the range of the habitat variable present in the data. this result seemed surprising. However, it might have arisen because Marsh Tits readily use cavities within living trees, rather than those in dead wood, probably to reduce the likelihood of nest predation (Wesołowski 2002). Furthermore, if dead wood is not limiting, and hence there is no shortage of nest holes, then a relationship may not be detected. Marsh Tits have been shown to be sensitive to woodland size (Hinsley et al. 1996), but we did not detect a relationship in this analysis. However, this could be because most of the woodlands used here were relatively large (<10% were smaller than 20 ha, see Hinsley et al. [2007]). The relationship with shrub diversity was a novel finding in this study. Predation by grey squirrels, deer impact, and interspecific competition were included in the analyses, specifically because these factors have been proposed as potential causes of Marsh Tit decline (Fuller et al. 2005). However, the results obtained here provide little support for these hypotheses as major drivers of decline. Marsh Tit occupancy showed a positive relationship with squirrel drey density and both measures of deer activity. Marsh Tit abundance was positively related to drey density and interspecific competitor abundance. Furthermore, the positive relationship with interspecific competitors was retained into the final model, and was highly significant (P < 0.001). This result is consistent with that of Siriwardena (2006), who found significant positive relationships between Marsh Tits and their potential competitors. Lewis et al. (2007) also found that interspecific competitors did not appear to be a factor in the decline of the closely related Willow Tit Poecile monatana. These results may suggest that all three of these factors are unlikely to be important in Marsh Tit decline. However, Amar et al. (2006) discuss problems with the deer impact PCA analyses, suggesting that these results should be interpreted with caution. Furthermore, the presence of deer, squirrels and other tit species could all be expected, in general, to be positively associated with mature woodland and thus, given sufficient habitat quality, positive relationships with Marsh Tits in their prime habitat might be expected. Therefore, these results remain speculative and experimental work is required to investigate these aspects in more detail. As Blue Tits occupied all of the study woodlands, it was not possible to carry out a wood-occupancy analysis for this species. Blue Tit abundance in woodland was best predicted by locality and two habitat covariates, tree dbh and the number of dead limbs. This model explained 27% of the variation in Blue Tit abundance. Understorey group characteristics were not as important for Blue Tits as for Marsh Tits, with no or only weak effects for three of the four covariates. Cover up to 2 m entered into the final model stage, but was not retained. All three dead wood covariates showed positive relationships with Blue Tit abundance, and two of these entered the final model stage. This is interesting, given the unimportance of this dead wood group to Marsh Tits, and the fact that, unlike Marsh Tits, Blue Tits readily take to nest boxes. However, even fallen wood was positively related to Blue Tit abundance, suggesting that dead wood may be important as more than just a provider of nesting cavities for Blue Tits. It is also possible that abundance of dead wood acts as an indicator of mature woodland, but this would be expected to benefit Marsh Tits as well. Defining the habitat requirements of Blue Tits from the literature (Cramp & Perrins 1993) was more difficult than for Marsh Tits, because Blue Tits were more generalized in their requirements. © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 40 J. Carpenter et al. The predicted positive relationship between Blue Tit abundance and dead wood (Cramp & Perrins 1993) was borne out in our analysis, but many of the other relationships were not as predicted, or only weakly so. The other habitat covariate retained in the final model, tree dbh, was not predicted and suggests a possible importance of tree age for Blue Tit abundance. This could explain the observed result with dead wood, as older trees are more likely to produce dead wood. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Conditions required for habitat management Our second aim was to provide woodland managers with criteria to manage woodlands for the benefit of Marsh Tits. Conditions were also calculated for Blue Tits to allow us further to understand the opposing population trends of the two species. As expected, given its increasing population trend (Eaton et al. 2006), the values required for greater than average Blue Tit abundance are those which already exist in much of British woodland today, with plenty of mature trees, and little need for understorey layer. Much of the historical management of woodlands, such as coppicing, has now ceased, and many of these woodlands are no longer actively managed (Mason 2007), allowing Blue Tits to thrive. This same lack of management, allowing canopy closure, may have been detrimental to the Marsh Tit population, due to a reduction in available understorey vegetation. However, understorey cover from 2–10 m in the RSPB surveyed sites used in the RWBS, in fact, increased substantially between 1980 and 2003 (Amar et al. 2006). Interestingly, at these sites, the Marsh Tit population had actually increased by 27%, compared with a 27% decline in the BTO surveyed sites (Amar et al. 2006). This suggests that the Marsh Tit population increased at sites where understorey also increased, and provides further evidence for the potential importance of the understorey to this species. We have provided woodland managers with the criteria to begin to assess whether their woodland is suitable for management for the benefit of Marsh Tits. However, before a woodland manager embarks on such a programme, multiple factors need to be considered. For example, the woodland location, whether Marsh Tits were present historically, the impact on the bird and plant species community currently present, the historical use of the woodland, and, importantly, the problem of deer numbers. Despite the apparent lack of impact of deer found in our analyses, there is no doubt that deer have radically altered the nature of many of © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 our woods (Gill & Fuller 2007, Hopkins & Kirby 2007). In such woodlands, management to encourage a healthy and diverse understorey is unlikely to succeed without undertaking some level of deer control, and more detailed analyses of deer impacts are urgently required. The Marsh Tit population appears to have stabilized at a historically low level (Eaton et al. 2006). However, managing woodlands for their benefit, and hence encouraging a reversal of their population trend, is still to be encouraged, especially as other threats to their population, such as impacts of climate change, are still poorly understood. Two studies on the potential impact of climate change have produced contrasting results. Carpenter (2008) showed that by the 2080s, there will be little climate space still available to Marsh Tits in Britain, whereas Huntley et al. (2007) showed little impact on climate space in Britain by the late 21st century. Changes in woodland management have been implicated in the decline of other woodland bird species (Amar et al. 2006). Experimental manipulations of woodland habitat, at a woodland scale, to encourage a diverse understorey with good cover, are required to understand how woodland policy and management influence the relationships between woodland structure and the diversity and abundance of bird, and other wildlife, populations. ACKNOWLEDGEMENTS We are indebted to the RSPB and BTO volunteers who carried out the survey work. Without them, the dataset on which this study is based would not exist. We wish to thank Dr Ken Smith for assisting in the setting up of this work, and the RWBS project steering group for providing assistance and advice. Dr Rob Fuller, Dr Joseph Tobias and Dr Tomasz Wesołowski provided valuable comments on an earlier draft, which greatly improved the manuscript. The original RWBS survey was funded by Forestry Commission England and Wales, Natural England, Defra, RSPB, BTO and the Woodland Trust. The current work was funded by the RSPB, Natural England and by a Natural Environment Research Council postgraduate training award to JEC. REFERENCES Amar, A., Hewson, C.M., Thewlis, R.M., Smith, K.W., Fuller, R.J., Lindsell, J.A., Conway, G., Butler, S. & MacDonald, M. 2006. What’s Happening to our Woodland Birds? Long-Term Changes in the Populations of Woodland Birds. RSPB Research Report no. 19. RSPB, Sandy, UK. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Habitat associations of two woodland birds Brickle, N.W., Harper, D.G.C., Aebischer, N.J. & Cockayne, S.H. 2000. Effects of agricultural intensification on the breeding success of corn buntings Miliaria calandra. J. App. Ecol. 37: 742–755. Broughton, R.K., Hinsley, S.A., Bellamy, P.E., Hill, R.A. & Rothery, P. 2006. Marsh Tit Poecile palustris territories in a British broad-leaved wood. Ibis 148: 744–752. Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers, D.L. & Thomas L. 2001. Introduction to Distance Sampling: Estimating Abundance of Biological Populations. Oxford University Press, Oxford. Carpenter, J.E. 2008. An investigation of causes of population decline in the marsh tit Poecile palustris in Britain. PhD Thesis, University of Oxford. Chamberlain, D.E., Fuller, R.J., Bunce, R.G.H., Duckworth, J.C. & Shrubb, M. 2000. Changes in the abundance of farmland birds in relation to the timing of agricultural intensification in England and Wales. J. App. Ecol. 37: 771–788. Colquhoun, M.K. & Morley, A. 1943. Vertical zonation in woodland bird communities. J. Anim. Ecol. 12: 75–81. Cramp, S. & Perrins, C.M. (eds) 1993. The Birds of the Western Palearctic, Vol. 7. Oxford University Press, Oxford. Eaton, M.A., Ausden, M., Burton, N., Grice, P.V., Hearn, R.D., Hewson, C.M., Hilton, G.M., Noble, D.G., Ratcliffe, N. & Rehfisch, M.M. 2006. The State of the UK’s Birds 2005. RSPB, BTO, WWT, CCW, EN, EHS and SNH, Sandy, UK. Fuller, R.J., Noble, D.G., Smith, K.W. & Vanhinsbergh, D. 2005. Recent declines in populations of woodland birds in Britain: a review of possible causes. Br. Birds 98: 116–143. Fuller, R.J., Smith, K.W., Grice, P.V., Currie, F.A. & Quine, C.P. 2007. Habitat change and woodland birds in Britain: implications for management and future research. Ibis 149(suppl.): 261–268. Gibb, J.A. 1954. The feeding ecology of tits, with notes on the Treecreeper and Goldcrest. Ibis 96: 513–543. Gill, R.M.A. & Fuller, R.J. 2007. The effects of deer browsing on woodland structure and songbirds in lowland Britain. Ibis 149(suppl.): 119–127. Gregory, R.D., Wilkinson, N.I., Noble, D.G., Robinson, J.A., Brown A.F., Hughes, J., Procter, DA., Gibbons, D.W. & Galbraith, C.A. 2002. The population status of birds in the United Kingdom, Channel Islands and Isle of Man: an analysis of conservation concern 2002–2007. Br. Birds 95: 410–450. Hartley, P.H.T. 1953. An ecological study of the feeding habits of the English titmice. J. Anim. Ecol. 22: 261–288. 41 Hewson, C.M., Amar, A., Lindsell, J.A., Thewlis, R.M., Butler, S., Smith, K. & Fuller, R.J. 2007. Recent changes in bird populations in British broadleaved woodland. Ibis 149(suppl.): 14–28. Hinsley, S.A., Bellamy, P.E., Newton, I. & Sparks, T.H. 1996. Influences of population size and woodland area on bird species distributions in small woods. Oecologia 105: 100–106. Hinsley, S.A, Carpenter, J.E., Broughton, R.K., Bellamy, P.E., Rothery, P, Amar, A., Hewson, C.M. & Gosler, A.G. 2007. Habitat selection by Marsh Tits Poecile palustris in the UK. Ibis 149(suppl.): 224–233. Hopkins, J.J. & Kirby, K.J. 2007. Ecological change in British broadleaved woodland since 1947. Ibis 149: 29–40. Hosmer, D.W. & Lemeshow, S. 1989. Applied Logistic Regression. Wiley, New York. Huntley, B., Green, R., Collingham, Y. & Willis, S. 2007. A Climatic Atlas of European Breeding Birds. Lynx Edicions, Barcelona. Lewis, A.G.J., Amar, A., Cordi-Piec, D. & Thewlis, R.M. 2007. Factors influencing Willow Tit Poecile montanus site occupancy; a comparison of abandoned and occupied woods. Ibis 149(suppl.): 205–213. Mason, W.L. 2007. Changes in the management of British forests between 1945 and 2000 and possible future trends. Ibis 149: 41–52. Morley, A. 1953. Field observations on the biology of the Marsh Tit. Br. Birds 46: 233–8, 273–87, 332–46. Perrins, C.M. 1979. British Tits. Collins, London. Perrins, C.M. 2003. The status of the Marsh and Willow Tits in the UK. Br. Birds 96: 418–426. Richards, E.G. 2003. British Forestry in the Twentieth Century, Policy and Achievements. Koninklijke Brill, Leiden. Siriwardena, G.M. 2006. Avian nest predation, competition and the decline of British Marsh Tits Parus palustris. Ibis 148: 255–265. Smart, J., Taylor, E., Amar, A., Smith, K., Bierman, S., Carpenter, J., Grice, P., Currie, F. & Hewson, C. 2007. Habitat Associations of Woodland Birds: Implications for Woodland Management for Declining Species. RSPB Research Report No. 26. RSPB, Sandy, UK. Tucker, G.M. & Evans, M.I. 1997. Habitats for Birds in Europe: a Conservation Strategy for the Wider Environment. Birdlife International, Cambridge, UK. Vickery, J.A., Evans, A.D., Grice, P.V., Aebischer, N.J. & Brand-Hardy, R. (eds) 2004. Ecology and Conservation of Lowland Farmland Birds II: The Road to Recovery. Ibis 146(suppl.): 193–204. Wesołowski, T. 2002. Anti-predator adaptions in nesting Marsh Tits Parus palustris: the role of nest-site security. Ibis 144: 593–601. (MS received 12 December 2008; revised MS accepted 7 May 2009) © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 42 J. Carpenter et al. APPENDIX 1. HABITAT VARIABLES The location of different aspects of woodland habitat structure, the variable name, level and unit of measurement and description of how each variable was measured during habitat surveys of 252 UK woodlands in 2003 and 2004. Variable names in bold are those variables where we tested for a quadratic effect. Location Variable Level/unit Description Field layer All variables Sub-plot/% cover Understorey Cover 0.5–2 m Cover 2–4 m Cover 4–10 m Horizontal visibility Sub-plot/% cover Sub-plot/no. Canopy cover Sub-plot/no. Basal area Plot centre/no. tree stems dbh Max height Dead trees Dead limbs Plot centre/cm Plot centre/m Plot centre/no. Sub-plot/no. Ground wood Sub-plot/no. Dom tree Plot centre/category: ash, beech, birch or oak Lichen/ivy Shrub diversity Sub-plot/category Plot centre/index Water features Altitude Size Plot centre/presence Plot centre/m Wood-level only/ha Tracks Plot centre/category The % cover of each variable below 0.5 m was estimated across the sub-plot. The total % cover of vegetation of the 5 m sub-plot as if viewed from above, considering the vegetation in each height band in turn. Number of orange sections (max 12) at least 50% visible of an orange and black pole recorded. Pole 2.4 m long, sections 10 cm long. The number of 2 cm squares (max 16) in a 4 × 4 wire grid in which at least 50% of the square was occupied by canopy-level vegetation (min 10 m high) when viewed directly from below. The grid was held horizontally 60 cm above the observer using a marked stick with a plumb line. Using a standardised relascope to count the number of stems of each tree species that scored accordingly (Smart et al. 2007). Tree with the maximum diameter at breast height Tree with the maximum height. Number of dead trees. Number of dead limbs attached to trees at any height in the sub-plot. Number of pieces of dead wood on the ground > 10 cm diameter and 1m in length. Dominant tree species – proportion of oak, ash, beech and birch from the total number counted by the relascope. Species with the highest proportion equals the dominant species. Abundance scored as 0 = absent, 1 = present, 2 = frequent. Total number of shrub species divided by 36 (total number of shrub species recorded across all RWBS sites). Presence/absence wet features (bog, stream, flush or pond). Recorded from a GPS unit. Using the National Inventory of Woodland and Trees, the area of all polygons of contiguous (no gaps > 25m) non-coniferous woodland was calculated. Presence of tracks: 0 = none, 1 = single foot track, 2 = vehicle width track. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Tree structure Deadwood Other habitat © 2010 British Trust for Ornithology, Bird Study, 57, 31–43 Habitat associations of two woodland birds 43 APPENDIX 2. NON-HABITAT COVARIATES Additional non-habitat covariates used in analyses, with a brief description of how each variable was measured. See Amar et al. (2006) for further details of methods used for the predation and deer damage categories. Category Covariate Description Predation Squirrel density Counts of squirrel dreys along a 1000 m transect. Counts were analysed using DISTANCE software to provide an estimate of drey density per woodland. Deer damage Deer PC 1 Principal component axis 1 – deer activity. A high score indicates high levels of deer activity. Deer PC 2 Principal component axis 2 – deer activity. This separates sites with abundant browsed bramble from those with a high browse line. Downloaded By: [Charman, Elisabeth] At: 11:35 10 February 2010 Deer PCA based on stool counts, deer slots, deer pellets, browse line, browse height, browsed bramble, browsed stems, frayed stems and trackways per 100m. Interspecific competitors Blue Tit, Great Tit and combined Blue and Great Tit abundance Included in Marsh Tit wood-abundance analysis only. Abundance taken from RWBS dataset, and summed for combined species abundance. PCA, principle component analysis; RWBS, Repeat Woodland Bird Survey. © 2010 British Trust for Ornithology, Bird Study, 57, 31–43