Opusc. Zool. Budapest, 2020, 51(Supplementum 2): 69–86 _____________________________________________________________________________ Changes in surface active myriapod communities during the restoration of woodland to wood pasture: a long-term study H.J. READ1*, C.P. WHEATER2, M. ALBERTINI1 & M. WOOLNER1 1 Helen J. Read, Martin Albertini & Martin Woolner, Burnham Beeches, City of London, Farnham Common, Bucks. SL2 3TE. U.K. *Corresponding author, E-mail: [email protected] 2 C. Philip Wheater, Faculty of Science and Engineering, The Manchester Metropolitan University, Manchester M1 5GD. Abstract. The myriapod communities of an area of woodland restored to wood pasture through tree felling was com- pared with a control site in dense woodland, using pitfall trapping over a 28-year period. 14 species of millipede and 13 of centipede were recorded. Numbers of individual species varied over time even within the control plot. There were significantly more centipede individuals and species and more millipede species in the control plot than the res- toration area. NMDS ordination indicated three groupings for the millipedes in the restored area with the community changing following the initial restoration to a tight cluster but then further movement in later years. In contrast the control community showed little variation. 2015 showed the greatest change from other years, probably due to high abundance of two species of polydesmid millipede that year, in both plots. The results are discussed in comparison with other studies looking at the impact of the edge effect of forests. The community of the restored area has perhaps not changed as much as expected, which may indicate a high degree of resilience, although it may be acting more like a forest clearing than an edge. Keywords. Millipedes, centipedes, habitat restoration, beech trees, forest edge. largely open environment. The soils are acidic INTRODUCTION with ground vegetation being similar to heath- land, with Calluna vulgaris (L.) (ling/common B urnham Beeches is a 383 ha nature reserve heather) and grasses typical of acid soils. comprising beech dominated woodland (Fagus sylvatica L.) 40 km west of London. It is In recent years, the idea that most of Europe, part of a Natura 2000 site (i.e. a site of European prior to human influence, was climax woodland nature conservation importance), for Atlantic with more or less continuous tree cover has been acidophilous beech forests (Natura 2000 code challenged. Recent research has indicated that it 9120). The main reason for the designation is its was more likely to have been dominated by a population of 400–500 year-old beech trees that type of savannah with large herbivores driving were for centuries cut as pollards, producing a the development of scrub and woodland and crop of wood every 15–20 years (Le Sueur maintaining a greater proportion of open areas 1931). The trees were set in wood pasture with scattered trees (Vera 2000). In medieval grazed by domestic livestock, probably a mix- times, southern England had abundant areas of ture of cattle, equines and pigs. The southern wood pasture but most were abandoned or de- 220 ha of the nature reserve was common land stroyed in the early 20th century through loss to which, although owned by the Lord of the agriculture or building development, or through Manor, gave local people rights to graze their lack of grazing. These medieval wood pastures, livestock and cut the trees for fuel wood. Old although heavily managed, were perhaps argua- descriptions and photographs indicate that the bly a more accurate reflection of the natural majority of the area was ‘savannah-like’ with vegetation of the region than was previously scattered trees and some patches of scrub in a thought. _____________________________________________________________ urn:lsid:zoobank.org:pub:1AD83119-2463-4AE7-B269-E3B1B80922F7 published: 07 August 2020 HU ISSN 2063-1588 (online), HU ISSN 0237-5419 (print) http://dx.doi.org/10.18348/opzool.2020.S2.69 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ At one time almost all the beech and oak wood) of invertebrate was examined by Alexan- (Quercus sp.) trees within Burnham Beeches der et al. (2010) however little work has been were managed as pollards, cut at regular inter- published on the effects of similar management vals to produce a crop of wood which was prob- on surface-active invertebrates. Myriapods can ably largely used by local people for domestic provide an interesting insight to the changes fuel. Repeated cutting causes the trees to de- taking place as a shaded woodland floor with velop a knobbly area, around the point of cut- abundant leaf litter, is replaced by a grassy ting, developing decay pockets and water pools sward with a much greater exposure to light. which provide micro habitats for many saprox- Millipedes being detritivores might be consid- ylic invertebrates. Cutting also enables the trees ered more likely to be impacted than the more to live longer than uncut trees so the older gen- predatory groups of centipedes, for example. eration of pollards are around 500 years old. Over 60 red listed species have been recorded There are other studies that have docu- (Read 2010) on the reserve, almost all associ- mented myriapod communities in similar envi- ated with these old trees. Tree management ronments, including disturbance effects on for- through pollarding ceased at Burnham Beeches ests due to tree clearance (Smith et al. 2017 and around 200 years ago; grazing also declined so Stašiov & Svitok 2014) the latter in beech for- that by the 1920s this merely involved a few ests, studies of gradients between forested and donkeys. In 1951 the nature conservation value open areas (Bogyó et al. 2015, De Smedt et al. of the area was recognised and the area was 2016) and studies on the impact of forest frag- designated as being important for nature conser- mentation (Riutta et al. 2012). None of these vation in a UK context. However, by the 1980s situations are quite equivalent to wood pasture the old trees had become swamped by younger but are similar. secondary woodland, dominated by Betula spp. (birch) and Ilex aquifolium (L.) (holly), resulting The current study reports on the results of from cessation of grazing, and were dying part of a larger project looking at changes in the through lack of light. In addition, they had be- ground active invertebrates during restoration of come very top heavy because of the lapse in wood pasture over a period of nearly 30 years. their cutting cycle, resulting in trees splitting In general, most studies cover relatively short apart as the larger branches became too heavy periods of time and typically use different sites for the ageing trunk beneath. Since then, sub- at different stages of succession as a substitute stantial restoration work has been carried out to for following long term changes (e.g. Bogyó et attempt to rescue the trees and to restore the al. 2015)- The work of Tajovsky (2000), and wood pasture in which they stand. This has in- Tajovsky et al. (2017) is notable for the long cluded tree clearance and the re-instatement of time scales of recording myriapod communities, grazing using cattle, ponies and, in some years, even though not always continuous, one such pigs in the autumn. Wood pasture is increas- study covered a time scale of 25 years (Ta- ingly being recognised as a distinct habitat in jovsky et al. 2017). the UK and restoration is being carried out in a range of locations. However, this restoration has only recently started and documenting changes MATERIALS AND METHODS taking place is valuable to both assess the im- portance of the habitat and the implications of Two areas were sampled. The restoration restoration. Substantial work has been carried area was actively managed for nearly 30 years out on the trees at Burnham Beeches (Read et to restore wood pasture (through tree felling, re- al. 2010) and the preliminary impacts on ground instatement of grazing and periodic control of vegetation and some ground active spiders has bracken and bramble). The tree clearance was been reported (Read 2000). The impact of carried out in two phases, winter 1991/92 and felling younger trees to give more light to an- March 1994 during which the number of mature cient ones in terms of tree response and saprox- trees in the vegetation plot was reduced from 19 ylic species (those associated with decaying to 6 and shrub layer holly from 20% to a single 70 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ bush. A control site nearby was also surveyed Trapping usually took place between April where no restoration had taken place and which and October/November but for operational rea- remained densely vegetated with beech/holly sons the trapping periods varied slightly be- woodland throughout the study (although at tween years, with two years when it started later several times work in the vicinity may have (in early June in 1992 and in early May in 1996) made minor changes to the light levels). and a further two years when it finished earlier (late September in 1993 and 1994). To include Pitfall traps (plastic vending machine cups) as much of the data as possible for analysis, the 8 cm in diameter and 10 cm deep were installed, longest period of time for each year was used containing 50 ml of 4% formalin and a few where the two trapping grids (i.e. restoration and drops of detergent, with wooden rain covers control areas) were in agreement. As the number held about 3 cm above them on large nails. The of days varied between years the analysis was quantity of fluid was sufficient to ensure cap- carried out on numbers of myriapods per 100 tures remained in fluid except on a few occa- trapping days. sions when the number of wood ants was very Vegetation was also monitored annually high. Ten traps were installed in each site in a around the pitfall traps. For the experimental 5 x 2 formation with traps 2 m apart, a format area the plot was 30 x 30 m in size and was lo- regularly used in the UK (Read 1987). Traps cated just to one side of the traps. For the con- were emptied and re-set every two weeks during trol, due to the shape of the woodland block, it the season taking care to ensure that the lip of was 20 x 45 m with the pitfall traps central to the cups were flush with the ground surface. the plot. Each year 25 random quadrats, 0.25 Traps were first installed in the restored area in m2, were thrown (positions determined using 1990 and in the control (unrestored area) in random number tables) and percentage cover 1992 and run each year. Catches were sorted estimated for all species. A mean percentage and the Myriapoda were identified to species cover figure was calculated for the whole plot level where possible and adults distinguished per year and multiplied by the frequency to give from immatures. Data presented here are for a value that was arcsine transformed before or- millipedes and centipedes for each year up to dination since the data were not normally dis- and including 2017. Although pitfall trapping tributed. has been used in previous studies (for example Stašiov & Svitok 2014), its validity as a method A notable feature of this part of Burnham of comparing myriapod communities has been Beeches is the high density of wood ant (For- questioned and there are clear short comings mica rufa L.) nests. This species forages on the (see for example Gerlach et al. 2009, Tuf 2015) ground and also in the tree canopies, which it but they are still regularly used in quantitative reaches by walking along the ground using trails studies for example De Smedt et al. (2019) and that are variable in location and extent. Thus, in the UK Environmental Change Network this species was found in high numbers in the (2020). Indeed, despite the considerable debate pitfall traps at certain times and is generally over the years regarding pitfall trapping as a very abundant on the surface across both areas suitable technique for sampling surface active where the traps were located. As wood ants are invertebrates, the technique remains reasonably predatory and opportune generalists, their pres- popular because it is relatively easy (and cheap) ence can have profound effects on the popula- to operate and usually provides statistically via- tions of other invertebrates (e.g. Fowler & ble results (Wheater et al. 2020). Myriapoda MacGarvin 1985 and Punttila et al. 2004), in- have long been sampled using pitfall traps (e.g. cluding surface active species (e.g. Reznikova & Van der Drift 1963) and the method was used Dorosheva 2004). The impact is usually nega- here because of the need to trap in a way that tive, but some species are positively associated drew least attention to members of the public with wood ant abundance (for example the and was feasible to run by volunteers, without staphylinid beetle Pella humeralis (Graven- taking up too much time and/or needing specific horst)) and on others there may be no apparent expertise to deploy. impact. 71 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ Table 1. Species of millipede found in the two sites. Restored site Control site mean ± SE mean ± SE per year Millipedes per year (percentage of (percentage of years years occurring: n = 26) occurring: n = 28) Glomeris marginata (Villers, 1789) 6.8 + 1.13 (82.1%) 19.9 + 5.80 (96.2%) Nanogona polydesmoides (Leach, 1814) 0.3 ± 0.17 (21.4%) 0.1 ± 0.04 (7.7%) Chordeuma proximum Ribaut, 1913 1.2 ± 0.38 (57.1%) 6.8 ± 1.67 (96.2%) Nemasoma varicorne C.L. Koch, 1847 0.02 ± 0.02(3.6%) 0 (0%) Proteroiulus fuscus (Am Stein, 1857) 0.7 ± 0.19 (57.1%) 1.0 ± 0.22 (57.7%) Tachypodoiulus niger (Leach, 1814) 8.2 ± 0.09 (100%) 11.7 ± 1.58 (100%) Cylindroiulus britannicus (Verhoeff, 1891) 4.0 ± 1.63 (35.7%) 0.9 ± 0.36 (26.9%) Cylindroiulus caeruleocinctus (Wood, 1864) 0.02 ± 0.02(3.6%) 0.02 ± 0.021 (3.8%) Cylindroiulus parisiorum (Brolemann & Verhoeff, 1896) 0.07 ± 0.00 (7.1%) 0.02 ± 0.02 (3.8%) Cylindroiulus punctatus (Leach, 1815) 5.1 ± 0.69 (100%) 4.1 ± 0.57 (100%) Ophyiulus pilosus (Newport, 1842) 0.07 ± 0.03 (14.3%) 0.1 ± 0.05 (15.4%) Polydesmus angustus Latzel, 1884 124.7 ± 20.00 (100%) 90.4 ± 20.08 (100%) Polydesmus denticulatus C.L. Koch, 1847 33.4 ± 8.20 (89.3%) 94.5 ± 22.72 (100%) Brachydesmus superus Latzel, 1884 0.07 ± 0.05 (7.1%) 0.3 ± 0.15 (19.2%) Analysis comprised repeated measures ANOVA (by year) using Tukey-Kramer multiple RESULTS comparison tests (Wheater & Cook 2000), using StatView v5.0, to explore potential differences A total of 14 species of millipedes and 11 between the sites on the basis of the numbers of species of centipedes were found in the wood individuals per 100 days trapping, the number of pasture restoration area compared to 13 species species, the proportion of immature animals com- of millipedes and 12 species of centipedes in the pared to the number of adults (millipedes), and control site (Tabs. 1–2). Figures 1–9 show the various measures of species diversity (Shannon H’, abundance of selected species which were Simpson 1-D, Berger-Parker 1/d, and Evenness J’ caught in sufficient numbers to illustrate pat- based on the Shannon index – Wheater & Cook terns of variation over time. Most of the species 2015). The communities of millipedes and centi- found were common British species with the pedes were ordinated separately using NMDS, most unusual in a UK context being the milli- with the Gower distance measure (Wheater et al. pedes Chordeuma proximum and Cylindroiulus 2020), using PAST v3 (Hammer et al. 2001). Spe- parisorum and the centipedes Lithobius muticus cies turnover was measured as change in ordina- and L. macilentus. Chordeuma proximum has a tional space between years, calculated as differ- strong southwest distribution in the UK and is ences in ordinational space across the first three an Atlantic species strongly associated with major ordination axes using an extension of the Py- woodlands and may be associated with acid thagorean theorem. Correlations between the axes sandy soils (Lee 2006). The relatively high of major variation for myriapods and vegetation numbers caught, including juveniles, especially variables was completed using FCStats v2.1a in late springs, indicate it is well established at (Wheater & Cook 2015, Wheater et al. 2020). Burnham Beeches. Cylindroiulus parisorum in 72 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ contrast is a predominately south eastern species west (Barber pers. comm.). The majority of the and rather less well recorded in the UK. Alt- records are for woodland where it also appears hough there is some suggestion of an association to be most abundant (Barber & Keay 1988), it is with synanthropic habitats in continental Eu- only known from females in the British Isles rope, where it is very sparsely distributed (Kime and is the only British Lithobius apparently & Enghoff 2017), in the UK this species has showing parthenogenesis (Barber pers. comm.). been associated with decaying wood and loose There were just 2 individuals caught during this bark (Lee 2006) so the finding in a nature re- project, both in the control plot. Lithobius muti- serve known for its decaying wood habitat is not cus has a strong south eastern distribution in the unexpected. Lithobius macilentus has been UK and is a characteristic scrub and woodland found across the UK but the records are very species. It was regularly caught in the traps, es- patchy and it has not been found in the south pecially in the restoration area. Table 2. Species of centipede found in the two sites. Restored site mean ± SE Control site mean ± SE Centipedes per year (percentage of per year (percentage of years occurring: n = 28) years occurring: n = 26) Strigamia acuminata (Leach, 1814) 0.2 ± 0.07 (17.9%) 1.6 ± 0.33 (84.6%) Strigamia crassipes (C.L. Koch, 1835) 0.7 ± 0.27 (46.4%) 1.9 ± 0.46 (61.5%) Geophilus easoni Arthur et al., 2001 1.0 ± 0.25 (67.9%) 1.9 ±0.42 (84.6%) Geophilus flavus (De Geer, 1778) 0.1 ± 0.04 (21.4%) 0 (0%) Geophilus insculptus Attems, 1895 0 (0%) 0.03 ± 0.03 (3.8%) Geophiulus truncorum (Bergsoe & Meinert, 1886) 0.1 ± 0.05 (14.3%) 0.3 ±0.09 (34.6%) Cryptops hortensis Donovan, 1810 0.2 ± 0.06 (25.0%) 0.4 ± 0.11 (38.5%) Lithobius forficatus (Linnaeus, 1758) 8.0 ± 1.06 (100%) 15.0 ± 1.02 (100%) Lithobius macilentus L. Koch, 1862 0 (0%) 0.1 ± 0.05 (11.5%) Lithobius muticus C.L. Koch, 1847 8.0 ± 0.89 (100%) 1.2 ± 0.31 (61.5%) Lithobius variegatus Leach, 1813 2.9 ± 0.90 (71.4%) 4.8 ± 0.88 (96.2%) Lithobius crassipes L. Koch, 1862 0.3 ± 0.11 (28.6%) 0.6 ± 0.28 (26.9%) Lithobius microps Meinert, 1868 0.04 ± 0.03 (7.1%) 0.03 ± 0.03 (3.8%) The numbers of individuals in each species high). Chordeuma proximum shows a very varied considerably with some only being found spikey pattern and in some years it is much twice but other species being found in their more abundant than others. The two larger pol- hundreds. Numbers of Cylindroiulus punctatus ydesmid species were caught in very high num- declined over time in both plots whereas Cylin- bers in 2015. Generally, Polydesmus denticula- droiulus britannicus and Glomeris marginata tus was relatively more abundant in the control increased over time (for C. britannicus from plot and P. angustus in the restoration plot but complete absence in the early years and Glom- both showed a spike in 2015 which was espe- eris marginata from very low numbers to quite cially pronounced in the control plot. 73 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ Centipede species were also variable with abundant in the restoration plot in the early some predominantly found in the control plot years but numbers declined over time while they (like Strigamia acuminata), others (like increased in the control plot. Generally, there Lithobius muticus) more abundant in the resto- were more species in the control plots than the ration plot. Lithobius variegatus was more restoration areas. Figures 1–6. Numbers of individuals in various myriapod species in restoration and control plots over time. 1 = Cylindroiulus punctatus; 2 = Cylindroiulus britannicus; 3 = Glomeris marginata; 4 = Chordeuma proximum; 5 = Polydesmus denticulatus; 6 = Polydesmus angustus. 74 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ Figures 7–9. Numbers of individuals in various myriapod species in restoration and control plots over time. 7 = Strigamia acuminata; 8 = Lithobius muticus; 9 = : Lithobius variegatus. Table 3. Analysis of variance for millipede numbers and diversity (with year as a repeated measure). Millipedes F P Comments 1,25 Numbers of individuals per 100 1.006 0.3254 No significant difference between sites trapping days Number of species 4.475 0.0445 Control significantly higher than restored site Proportion of immatures to matures 7.933 0.0093 Control significantly higher than restored site Shannon index (H’) 0.393 0.5362 No significant difference between sites Simpson index (1-D) 13.935 0.0010 Restored site significantly higher than control site Berger-Parker index (1/d) 8.014 0.0090 Control significantly higher than restored site Evenness (J’) 0.035 0.8535 No significant difference between sites Comparison between the two sites (restored of the number of millipedes per 100 trapping vs. control) showed that the numbers of individ- days in the two sites. Similar patterns can be uals per 100 days trapped were not significant seen at the two sites over time except for a more between the two areas for millipedes (Tab. 3) dramatic increase in 2015 for the control site. but showed significantly more centipedes in the Figure 11 clearly shows that the number of cen- control area than in the restored site (Tab. 4). tipedes per 100 trapping days is consistently Figure 10 illustrates the patterns found in terms lower in the restored site, the restored site has 75 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ two major peaks (one at the beginning of the significantly more species than the restored site survey and one part way through), whilst the (Tabs. 3–4). The proportion of immatures to control is reasonably stable with an increase at adults was significantly higher in the control site the end. For both classes, the control area had for millipedes, but not for centipedes. Table 4. Analysis of variance for centipede numbers and diversity (with year as a repeated measure). Centipedes F P Comments 1,25 Numbers of individuals per 100 6.779 0.0153 Control significantly higher than restored site trapping days Number of species 9.441 0.0051 Control significantly higher than restored site Proportion of immatures to matures 2.093 0.1604 No significant difference between sites Shannon index (H’) 0.384 0.5411 No significant difference between sites Simpson index (1-D) 0.120 0.7324 No significant difference between sites Berger-Parker index (1/d) 2.391 0.1346 No significant difference between sites Evenness (J’) 0.074 0.7882 No significant difference between sites Figures 10–11. Numbers of myriapods per 100 trapping days over time. 10 = millipedes; 11 = centipedes. Figures 12–13. Indices over time for millipedes. 12 = Simpsons diversity index; 13 = Berger Parker Index. 76 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ No significant differences were found be- The NMDS biplots (Fig. 14) for millipedes tween the sites in terms of any of the diversity show some separation along coordinate 1 (the measures for centipedes. However, for milli- axis of major variation) with the communities in pedes, Simpson’s index was significantly lower the restored site for some early years lying to the in the restored site, whilst Berger-Parker’s index left of the biplot, close to a large (fairly tight) was higher in the control site. Figure 12 shows cluster comprising the bulk of the data and the that the Simpson index for millipedes has re- later years (2015–2017) being somewhat spread mained more or less consistent over time for the out to the right. This indicates that perhaps some control site, whereas for the restored site it has changes did take place early on after restoration declined over time, becoming less diverse. Since began, there was then a period of consolidation this index is more influenced by the presence of and then some larger changes occurred much common species, this implies that there are later in the process. Although a similar spread of fewer common species in the restored site over data can be seen on the biplot for the control site time. The Berger-Parker index was also signifi- (with change occurring from bottom left to top cantly different between sites and showed a right – Fig. 15), the data for the first 23 years are clear increase over time for the restored site generally much more tightly clustered, implying (Fig. 13). This suggests that species dominance a more consistent community structure over has decreased over time, possibly related to the time. For the centipedes no obvious patterns overall decline of several species including C. emerged for either the restoration area or the punctatus (Fig. 1) and P. angustus in the later control and similar groupings were found in the few years (Fig. 6). plots for the restoration and the control. Figure 14. NMDS biplot of millipede communities over time for the restored site. Showing a general trend from early to later years along coordinate 1 (ellipses indicate groupings of early to later years). 77 Read et al.: Myriapod communities during the restoration of woodland to wood pasture _____________________________________________________________________________ Figure 15. NMDS biplot of millipede communities over time for the control site. Showing a fairly tight grouping of years and three possible outliers (arrowed). Figure 16. Degree of change in multidimensional space over time for millipedes. Indicating amount of community change. Using change in multidimensional space as rol site (repeated measures ANOVA F = 1,24 a measure of change in community structure 11.392, P = 0.0029). For centipedes the amount between years, there were no differences be- of change between years is similar, although the tween sites for either millipedes or centipedes. patterns differ across the years with no discerni- However, it can be seen for millipedes that the ble pattern. major change at both sites was in 2015 and 2016 (Fig. 16). If these dates are omitted from the The myriapod communities were then com- analysis, the restored site can be seen to have a pared with the vegetation. Spearman Rank Cor- higher degree of change over time than the cont- relations between the coordinate 1 scores from 78