Introduction

Natural woodlands in the United Kingdom represent a natural habitat for a wide range of species that have evolved in that habitat. But plantation woodlands require management for economic purposes. Such management may also create positive or negative impacts on a wide range of species through creation of clear space or thinning of tree canopies (Carnus et al. 2006). Within the U.K. during the last decades there has been an initiative for habitat restoration to re-establish native broadleaves in Plantations on Ancient Woodland Sites (PAWS) (Pryor 2003). This involves the removal of conifers to gradually restore sites to their semi-natural state (Harmer 2006). PAWS are twentieth century plantations (frequently conifer) on what were once Ancient Semi Natural Broadleaved Woodlands, dating back to 1600 or earlier and cover ca. 165000 ha in England and Wales. Conifer Plantations on Ancient Woodland Sites present particular challenges for the design and management of felling where legally protected species occur. However, the appropriate management approach is untested and poorly understood. Therefore, there is a need to develop suitable scientific knowledge on the influence of management practices on woodland species, especially those of a protected status.

The hazel dormouse (Muscardinus avellanarins L.) is a woodland species native to Britain and Europe that is commonly regarded as an indicator of woodland health due to its apparent requirement for high species diversity and complex habitat structure (Morris 2003, Bright et al. 2006). It has become extinct over half its range in England and Wales during the last half century (Hurrell & McIntosh 1984) and may still be declining (Goodwin et al. 2017). It became a legally protected species in Europe under the EU Habitats Directive, confirmed in Britain by the Wildlife and Countryside Act, 1981 (Bright et al. 2006). The hazel dormouse is a sedentary and non-migratory hibernator (Juškaitis & Büchner 2013). It is a selective feeder and when possible feeds on energy-rich material such as buds, flowers, insects and fruits as seasonally available (Richards et al. 1984, Morris 2004). Hazel dormice are typically found in deciduous broadleaved woodland, hazel coppice and dense species-rich hedgerows in Britain (Bright et al. 2006). However, there has been increasing evidence (Sanderson et al. 2004, Trout et al. 2012b, Juškaitis 2014) of dormice in conifer woodlands, especially in PAWS although the dormouse population density is usually lower than in broadleaf woods – ca. 2/ha vs. ca. 4–5/ha (Juškaitis & Büchner 2013). Thus, knowledge needs to be gained on the effect of varying management practices at these sites and their impact on dormice abundance, overwinter survival and future population stability. Previous studies based on survey data and observations on their arboreal behaviour suggest that dormice thrive best where the woodland contains dense and diverse low-growing shrubs (Bright & Morris 1994). Radiotracking has shown they do however make nests in conifers (Trout et al. 2012b). In autumn, this species gains weight from available arboreal food sources before descending to the ground for hibernation (Juškaitis & Büchner 2013). They hibernate in small nests built on the ground for about six months until the following spring when survivors become active and resume their arboreal lifestyle (Bright & Morris 1994). Within a sample of 20 managed conifer plantations Trout et. al. (2012a) have shown that dormouse abundance in nest boxes increases as the shrub layer index increases; and Juškaitis & Siožinytė (2008), Juškaitis & Büchner (2013) indicate that high shrub density is deemed optimal habitat. Previous vegetation analysis in a conifer study site showed that felling small groups of trees or larger clear fells appears to result in more regenerating vegetation than standard overall thinning (Trout et al. 2012b). The presence of this protected species has implications for management policies aimed at the restoration of coniferous PAWS to native broadleaved woodland (Thompson et al. 2003). Local extinctions have often resulted from inappropriate land management in the past (Bright & Morris 2002) as reduced population density from large scale forest operations may result in isolated populations falling below the point of viability. Once local extinction occurs, it is quite likely in the UK's fragmented countryside to be permanent (Bright & Morris 1996). Developing pragmatic woodland management methods that are not greatly or permanently disruptive to the local hazel dormouse population is thus very desirable when designing best practice guidelines for conifer management without needing a Licence; complemented further by enhancing the local landscape connectivity to more easily enable dormouse movements between woods (Bright et al. 2006).

This study examines the effect on the local dormouse population at one woodland site in the U.K., of repeating different experimental conifer removal patterns for PAWS restoration, extending beyond a decade. There are various different practices used to remove conifers from woodland for PAWS restoration, from repeated periodic traditional thinning of an area to a single complete clear-fell (Thompson et al. 2003). We examined the impact on dormice numbers of different PAWS restoration techniques on four adjacent Experimental Study Areas (ESA) in one conifer forest compartment with repeated removal operations. Different restoration options are likely to alter the structure of the habitat and consequently may affect dormouse populations, in addition to the thinning and log removal operations themselves disturbing or killing dormice. The objective ofthis experiment was to determine way(s) for forest managers to minimise the short and medium-term adverse effects on the legally protected hazel dormouse during PAWS restoration operations within an economic framework. It was not aimed specifically to enhance the dormice population (as conservation groups might undertake where economic considerations are secondary). Previous experience at this poor sandy soil site indicated that it took at least three years for regrowth to become thick enough for dormice to utilise (Trout et al. 2012b). We therefore predict that at this site the population density indices of dormice will be differentially affected by the various restoration procedures, although the actual numbers of individuals will be small in this suboptimal habitat. We also investigated the immediate impact of the tree removal action on dormouse survival at the whole site level. The study sought to advance how we might manage conifer plantations and improve our understanding of the effects that different management regimes have on the protected hazel dormouse – perhaps as a model species – so that we can manage our woodlands sustainably both for economic and biodiversity purposes.

Material and Methods

Management treatments

There were two forest operational constants in the experiment. Firstly, tree felling and log removal from ESAs was timed between the second half of October and late November to minimise impacts on dormice by avoiding the main breeding season and, as far as practicable, before hibernation to allow the majority of animals to actively escape the felling works. Secondly, each felling resulted in only ca. one third of the trees (not necessarily 1/3 of the area) in an ESA being actually cut and removed at a time. This was intended to provide a significant area of undisturbed refuge as described by the Forestry Commission guidance. A peripheral wildlife corridor round the site was also left uncut to allow dormice to travel easily to find alternative habitat after any felling operation had created (temporary) habitat fragmentation (Perault & Lomolino 2000). Trees were removed using one of four treatment patterns. Fig. 1 indicates the areas used within the forest.

Fig. 1.

The Experimental Study Areas (ESA) in the Ribbesford Forest edged by the hard forest road (dashed line). The patterns of small group felling in ESA 1 and ESA 2, large felling plots in ESA 4 and the racks with overall thinning of ESA 3 can be clearly seen following the 2003 operations.

Table 1.

Details of operational methods on each Experimental Site Area (ESA).

Racks (a complete line of trees felled and cleared for machinery access and log removal) were located in every seventh row of the plantation. Broadleaf trees not in a “rack” were left whenever possible. All log piles were placed close to racks and log removal was along the racks using a wheeled forwarder machine with a hydraulic loading grab. The remaining 4 ha had been more recently planted (1995) with conifers and was not thinned.

The treatment applied in 2003 in each ESA was repeated in the same ESA in 2009 and 2015. Tree numbers equivalent to approximately 1/3^rd of each ESA were removed in 2003 in the patterns described in Table 1, felling into the racks or group fell areas where practicable. In 2009 ca. 33 % of the remaining trees were removed and a third of those remaining were removed in 2015, leaving 30 % of the original total conifer trees untouched after 2015. Only in ESA 4 were there significant blocks of trees. In ESA 3 the trees were just more thinly but evenly spaced but the falling trees impacted that entire ESA. A fuller description and methodology are detailed in Trout et al. (2012b). Before each conifer removal operation in autumn all nest boxes in the ESAs were removed, to be replaced afterwards during that same winter hibernation period, so the nest boxes were available in the spring following the operations. The base of each tree with a nest box had been previously marked with paint to aid replacement in the same location. Where a marked tree had been felled, the box was replaced on the nearest remaining tree to the stump with the paint. Within the clearings created in small or large group fell areas, wooden posts were substituted in place of trees unless and until a broadleaf tree was available. In each ESA, the operators were asked to fell trees into the planned cleared patches or the racks and not routinely into the unfelled matrix. This was not possible for ESA 3 where all the matrix was affected by the felling in every row of trees. All boxes were inspected for evidence of dormice (animal/s or nest) once a month from May to October inclusive. Each dormouse of 7 g or larger found since 2002 had a Passive Integrated Transponder (PIT) tag 8-9 mm long × 1.4 mm diameter inserted as a method of identifying each individual. PIT tagging was conducted under a European Protected Species Science Research Licence from Natural England. Each survey involved the recording of every individual, its age-class, weight, gender and any suckling young. Capture records were collated by month and consolidated into yearly bins to allow the number of captures to be calculated. Nest-based juveniles (defined here as individuals from birth but < 7 g and thus not individually marked nor had they undertaken hibernation) were aggregated to provide an annual site total. Overall dormouse population indices for the entire site were derived by summing the total dormouse captures each year. The PIT tag recapture data enabled the number of individual dormice > 7 g known to be alive at the site and in each ESA to be calculated. The overall overwinter survival of adults and of juveniles was measured by collating the recapture data for each marked individual.

We analysed the difference in overall dormouse captures and individuals in the year following a felling treatment compared to other years using a GLM. Each year was entered as a categorical predictor in the model to determine if there was an immediate effect of tree felling on local adult and on juvenile dormouse abundance in nest boxes. GLMs with poisson errors were also used to analyse the trend in dormouse numbers for each ESA over time (with year entered as a continuous explanatory variable). The same method was applied to examine the sitelevel population trend of juveniles from 2011 to 2016. In models where over-dispersion was detected, standard errors were corrected using a quasi-GLM model (Ver Hoef & Boveng 2007). To account for the slight differences in size of each ESA, the numbers of known dormice were normalised into individuals per three hectares. They were analysed using a one-way Analysis of Variance (ANOVA), with ESA entered as a categorical predictor, to determine differences in population density between each ESA. Due to a skew in the data distribution a square-root transformation was performed before conducting the ANOVA. All statistical analyses were conducted using R version 3.4 (R Core Team 2017) using the package “Mass”. Functions from the packages “AER” and “DHARMa” were used to test for overdispersion.

Survival estimates and population modelling

Because of the small numbers of individuals the data were collated to give one annual adult and one juvenile overwinter survival figure for the site. Survival probabilities, abundances and population growth rates were estimated by developing an Integrated Population Model (IPM) (Abadi et al. 2010, Kéry & Schaub 2012). Our IPM used a stagestructured formulation comprising a state space model using annual survey counts, a fecundity model using the number of litters and a capture -mark-recapture (CMR) model using the PIT tagging data in yearly bins using methods described in Harris et al. (2015). We treated our datasets as independent, but since the survey methods used were designed to obtain data on every individual in the population over time, some individuals may be included in more than one data type. However, because the survey methods are exhaustive and previous studies have demonstrated only minor impacts on the accuracy of parameter estimation from data sets (Abadi et al. 2010) we assumed independence within this data set. Goodness-of-fit tests on our CMR models suggested a good fit to these data (Cooch & White 2014). The analyses were run using JAGS version 3.4.0 (Plummer 2003) called from R version 3.4.2 (R Core Team 2017) with package “R2jags” (Su & Yajima 2012). To assess convergence, three independent chains with different starting values of 100000 MCMC iterations, with a burn-in of 50000 iteration thinning every 100^th observation. We confirmed model convergence using the Gelman-Rubin statistic (Gelman & Rubin 1992). Overwinter survival rates were compared between years following thinning and other years using a t-test on the angular transformed percentages. Results should be treated with caution due to the very small number of data sets for a year following thinning.

Fig. 2.

Total dormouse captures recorded for each year in the forest compartment; individuals with a microchip and numbers of dependent nestlings too small to microchip (< 7 g). Grey vertical bars indicate autumn conifer removal operations.

Fig. 3.

Individual dormice > 7 g handled in each experimental area (ESA). Grey vertical bars indicate autumn conifer removal operations.

Fig. 4.

Annual overwinter survival rate estimates for marked adults and for marked juveniles (> 7 g) calculated from the I PM from 2002/03 to 2016/16. Error bars show the standard error of the survival rate estimates. The median survival values for adults and for juveniles are indicated by the dotted horizontal lines. Grey vertical bars indicate autumn conifer removal operations.

Results

Microchipped dormice numbers fluctuated from year to year, captures on the site varied from 22 to 115 (Fig. 2). The numbers of individual animals identified by the PIT tag also varied but less so, from 22 to 55. On average there was a flat non-significant trend across the period with a mean of 69.5 captures of tagged dormice (slope = 0.017, SE estimate = 0.026, t = 0.669, P = 0.516, SD = 26.7, Var = 713.2) comprising a mean of 36 individuals (slope = 0.013, SE estimate = 0.020, t = 0.646, P = 0.531, SD = 10.5, Var = 109.7). Of 341 animals microchipped, one reached an estimated five years old (63 months), seven lived to four years and 17 to three years. Fifty-two animals were captured five or more times, the maximum being 17 times. Half of all the animals (50.15 %) microchipped from 2002 to 2016 were caught on more than one occasion. A small number of marked individuals were known to move across adjacent ESA boundaries (16/227 relating to 445 captures = 3 %) mostly juveniles dispersing from the nest.

The year following the 2003 and 2009 forest thinning operations the numbers of dormouse captures in nest boxes, as well as the number of individuals seen, increased markedly but then stabilised the following year (Fig. 2). A quasi-GLM (to correct for overdispersion) shows a small significant increase in captures the year after felling (t = 2.539, P = 0.026) and also a significant increase in the number of individuals known to be alive (t = 3.01, P = 0.011). The temporary increase after the 2009 felling operations was greater than after felling in 2003. After the third thinning, in 2015, the total captures and numbers of individuals changed much less strongly but these are minimum counts as some young born in 2015 and some longerlived individuals may still have been alive on site but not recaptured in 2016. The results do not give any indication of the numbers, if any, of individuals present but never captured.

The total number of small juveniles < 7 g recorded in nest boxes across the whole site varied from 5-50 per year with alternate years of small and large numbers until 2011 (Fig. 2). Their numbers increased after each thinning operation. Juveniles of less than 7 g were not microchipped and many young nestlings had left the maternal nest before the next inspection a month later. A GLM with poisson errors shows a significant rising trend in the numbers of juveniles recorded over recent years 2011-2016 (slope = 0.414, SE estimate = 0.059, z = 7.077, P < 0.001).

The number of individual dormice captured in each of the four ESAs over the period is shown in Fig. 3. Only in one year (2011 for ESA 3) were no dormice reported, so populations remained in all ESAs, even if at a low level, despite the repeated felling operations. There was no significant difference overall in dormice numbers between the different experimental areas (df = 1.11, P = 0.354). Although ESA 3 had 10-15 individuals per year present early in the experiment, there was a significant declining overall trend in this treatment area (slope = –0.151, SE estimate = 0.043, t = –3.533, P = 0.004, R² = 0.5) and only 2-4 individuals seen per year in the last five years. Conversely ESA 4 had an overall increasing trend (slope = 0.086, SE estimate = 0.037, t = 2.306, P = 0.04, R² = 0.34). There was no significant change in dormice numbers over time in ESA 1 (slope = –0.075, SE estimate = 0.051, t = –1.462, P = 0.169) or ESA 2 (slope = 5.5e-04, SE estimate = 3.77e-02, t = 0.015, P = 0.989).

The mean adult overwinter survival was estimated at ca. 60 %. We recorded 160 nestlings too small to microchip or already dead in the nest. The survival (the number of individuals recorded in a given year who were found again after hibernation in subsequent year/s) of 7 g+ chipped juveniles was consistently lower than that of adults; mean 25 % (Fig. 4). The immediate impact of the autumn thinning (i.e. 2003, 2009) on dormouse survival in that winter was not significant for either adults (t = –0.52, df = 12, P = 0.615) or juveniles (t = 0.53, df = 12, P = 0.603). Adult survival over the next winter following felling (i.e. 2004, 2010) was not significantly different (t = 1.36 with 12 df, P = 0.199) compared to other years, but survival was significantly lower for juveniles born the season after felling (t = 2.36 with 12 df, P = 0.036).

Discussion

This is the first time that a long term study has been used to determine the impact of different repeated felling operations within a conifer plantation on a wild hazel dormouse population. The important implication so far is that the dormouse population was sustained after repeated disturbance by forest operations at this site. Our first hypothesis of different harvesting treatments resulting in significantly differential impacts on dormouse abundance was not validated within the first decade, although important differences have since become apparent.

The numbers handled varied annually but the data and analyses show that on this site the overall capture index increased significantly the year following felling operations in 2003 and 2009, but not after 2015. However, the number of microchipped individuals known to be alive on the site provided a more accurate and more stable estimate of the population present. However, the actual numbers were small because of the poor habitat that 30-45 year old conifers provide and the limited size of the ESAs. Numbers found also increased significantly the year after felling operations but stabilised subsequently. Perhaps the sudden forest disturbances reduced the availability of natural refuges and created a greater need for dormice to enter nest boxes in the following season, in comparison to other years. This would also explain the simultaneous increase we observed in the number of small young in nests. It would appear that any reduction in the food or refuge the year following the late autumn removal of trees and crushing of some understory cover did not reduce adult numbers. We conjecture that this may have impacted juvenile survival of the following cohort as they did not have such wide experience of locating seasonal food or good refuges or were outcompeted by the resident adults. The tree felling took place before hibernation. Previous research from the same site (Trout et al. 2012b) has indicated a very high incidence (ca. 90 %) of damage to or destruction of grids of artificial (crushable plastic ball) hibernation nests where trees were felled. This was independent of the use of a mechanical harvester or hand felling but related to the actual felling itself. For the group felling treatments 1, 2 and 4 the proportion of the ESA affected was ca. 1/3 but in the overall traditional rack and thin ESA the result was ca. 90 % damage or loss across that whole area. Each ESA was adjacent to at least one other and a small number of microchipped animals (ca. 3 %) moved from one ESA to an adjacent one. There were no records of any dormice using two ESAs simultaneously. When an animal moved from one ESA to another, it apparently did so as either a permanent change to its home range or involved juveniles dispersing from their natal nest.

We know that juveniles crossed ESA boundaries as they dispersed from their natal nest (Trout et al. 2017) on the same site. However we have no figure for those that crossed the hard forest road and out of the area. They were assumed to be dead. Shrub regeneration was slow on this site, taking three years or more and all except the very oldest marked individuals would have died before any clear-felled areas had re vegetated significantly to become attractive. The whole shrub layer of the overall thinning treatment in ESA 3 hardly revegetated at all since the pine canopy quickly regrew and all the vegetation was crushed again as racks were reused in later thinning operations. The marking of individuals with PIT tags enabled more useful estimates of abundance using individuals known to be alive than using the total captures as the index on each ESA, since it is the number of dormice locally that is important in practice. However, as with most vertebrate studies, there is still no knowledge of the proportion of animals present that never enter nest boxes or moved off site permanently. Vogel et al. (2012) found that eight nights of trapping in Switzerland over two months obtained 2.75 times more dormice than eight nest box inspections. Trout et al. (2012a) found on this U.K. site a similar differential of ca. 2.5 times higher numbers of marked individuals “known to be alive” vs. the maximum count of dormice found in nest boxes during any monthly check.

In this study, survival was measured by assessing the proportion of microchipped animals pre-hibernation which were recaptured in the following or a subsequent year. Adult survival changed little through the whole period, indicating a degree of stability that overcame events. However, juvenile (those marked at 7 g+) survival overwinter varied widely from year to year. Interestingly, neither adult nor juvenile survival over the winter following the autumn conifer removal operations was significantly affected, but there was a significant decline in juvenile overwinter survival for those born the year following a forestry operation (even though the actual numbers of young < 7 g found in nest boxes pre-hibernation rose sharply). We assume adverse effects of the forestry operations resulted in this increased mortality overwinter for the next juvenile cohort. We surmise the loss of conifers and crushing of the shrub layer impacted food supply in the next summer/autumn and this might have led to greater juvenile mortality during their initial dispersal. Conversely, it is known from faecal analysis that they eat conifer flowers and there are many insects available, including aphids, on conifer bark and leaves (Juškaitis 2014). The older dormice are likely to have established home ranges whereas juveniles may need to disperse further to establish their own home range somewhere else, but this presumably carries a higher risk of death. We cannot fully explain why the numbers of small juveniles varied so much from year to year, but density dependent factors may be occurring as suggested by Combe et al. (unpublished). In some litters the individuals were too small to microchip during the nest box inspection and had dispersed before the next monthly visit and were not captured, so they cannot be included in the analysis unless captured later that year as juveniles and marked. In low density populations such as here, failure to microchip one entire litter (of up to six individuals) could cause variations in the apparent long term survival trends for the site since only a portion may be captured and marked subsequently. Understanding juvenile dispersal and survival more fully is important to inform government forestry policy and best practice, particularly when looking at differences in survival between early and late born young. Tree thinning currently takes place in late autumn so later born young will have very little experience of such disturbance or where to find alternative food. Increasing the frequency of nest box inspections and reducing the eligible weight for microchipping by 1 gm would greatly improve the quantity and quality of data collected.

Our results indicate that the standard thinning of the entire area of ESA 3 showed the most pronounced decline in dormouse abundance over the later part of this study, leading to very few recorded individuals. More marked dormice had been recorded moving away from this study area in the early years (2004-2009) than any other (Trout et al. 2012a). It also had weak ground and shrub vegetation. The canopy re-closed quickly after thinning and soon reduced the light reaching the forest floor, repeating the poor regeneration of the shrub layer and thus perpetuating a poor habitat for dormice (Thompson et al. 2003, Allen 2008). Moreover, at ESA 3 the whole area was repeatedly degraded by the falling trees in thinning operations crushing all the vegetation with no parts left as undisturbed refuges. These reasons probably explain the declining abundance in ESA 3 and why the standard thinning operation appears to be the least favourable treatment for this species within the area thinned. However this treatment type is avery common silvicultural practice in the U.K. because it promotes improved incremental tree growth by encouraging crown growth and thus improving timber yields (Peterken 2001). Managers need to consider ways to mitigate the harm apparently caused by this particular option, perhaps by ensuring only a portion (e.g. 2/3) of that forest compartment is thinned in this manner in the same year or ensuring that some other suitable habitat remains available and connected nearby. Management decisions relate to future timber quantity or value (e.g. used for chipping, firewood or sawlogs) versus a larger stem diameter for fewer trees.

On the other hand, the forestry treatments in ESA 1, 2 and 4, where only 1/3 of the trees were felled in a pattern, produced significant gaps in the canopy resulting in patches with more light reaching the forest floor in the long term and a greater regeneration of shrub and sub-canopy vegetation and increased ground and shrub vegetation density, as preferred by dormice (Thomas et al. 1999). Trout et al. (2012a) already showed that ESA 4 at this site produced the highest density of shrub vegetation, whilst ESA 1 and 2 also provided patches of good regeneration in the shrub layer. The undisturbed two-thirds of the area was used by dormice and the regeneration patches were also used after several years. However it is suggested that those areas of conifer remaining completely unfelled may be now less suitable for dormice if the Corsican pine continues to grow at its original planted density and completely degrades the shrub layer by shading (as predicted by Juškaitis & Šiožinytė 2008). However food is available on the pine bark and canopy of old trees, in the form of both insects and pine flowers, including pollen.

ESA 1 apparently lost most of its dormice as a result of the excessively directional first felling but did continue to support a few. Unlike the other three study areas, the hand felling of the trees in 2003 had been unfortunately directed towards and into the refuge strip and the hard forest track bounding three edges of this treatment area and not towards other refuge areas. This is likely to have driven dormice permanently out of that area and over the track. It was the only treatment area with a subsequent overall net inflow of marked dormice (Trout et al. 2012b). ESA 4 appeared to be the best potential future area for dormice as relatively large blocks of future suitable dense shrub vegetation were periodically created. After several years larger numbers of dormice were found. This system mirrors the hazel coppice system of patch felling. This must, however, be balanced against the future economics of a poor silvicultural practice where some conifer areas are left completely unthinned for a further 5 and 10 years.

The U.K. Forestry Commission PAWS restoration policy encourages active management to gradually restore sites to their semi-natural state (Harmer 2006). We suggest a desirable and workable option indicated by this study is a phased patch-felling operation over perhaps 3-4 occasions and 5-7 years apart. This would result in temporal and spatial diversity of woodland structure and would promote the growth of ground and shrub flora. The size of any patch can be determined locally whilst not affecting a major proportion of the contiguous woodland (e.g. having crossing points across rides or along hedges) including keeping a peripheral strip for displaced dormice to move along. However, thinning or clear felling a large proportion or all of any contiguous area used by dormice in a single operation would appear to be very damaging, especially in small or isolated woods (Bright et al. 2006). In contrast, after an unmanaged planted conifer site develops a closed canopy, it becomes extremely uniform in structure and the ground vegetation and shrub layer become very sparse or non-existent about a decade later. Unless forest management takes place, the canopy shading leads to increasingly unsuitable conditions for dormice until it eventually starts to break up as conifers self-thin and fall, and a vigorous ground vegetation starts to re-establish in a haphazard way (Mason 2007). The study areas at this site comprising treatments that created racks with many small group fells or fewer larger patches would appear to be the optimum for dormice whilst also enabling reasonable silvicultural development due to standard thinning of one row either side of racks, felling the trees into the rack. Using an eight row system instead of seven in an ESA 3 situation, repeated thinning would enable some shrub layer to regenerate a little as the two adjacent racks would only be damaged during alternate thinning operations. We retained uncut peripheral wildlife corridors to allow dormice to travel easily to find alternative habitat after felling operations had created temporary habitat fragmentation (Perault & Lomolino 2000) allowing subsequent repopulation from around the periphery to occur (Bright et al. 2006).

Conclusion

In this small scale un-replicated field experiment on four plots with only small numbers of individuals present, dormice populations were retained in each thinned area of this 24-35 year old conifer plantation. Significantly more dormice captures were made in the year after felling operations have taken place (including more small juveniles in next boxes) although our annual abundance index using marked individuals remained relatively unchanged. Overall adult survival from year to year was constant and significantly higher than juvenile survival. Juvenile survival from those born in the year after felling operations was also significantly lower, although whether this is due to mortality or dispersal or density dependent factors is unclear. There was no overall statistical difference in the population density between experimental areas, although the traditional overall thinning treatment resulted in a significant decreasing abundance over the course of this study and large patchwork felling an increase in abundance. Dormice are most vulnerable to felling operations when hibernating, but autumn nests (especially maternity nests) are also at risk. The standard thin treatment ESA 3 causes most disturbance, with trees falling across the entire area, leaving virtually no undamaged refuge areas and produces the least favourable habitat, making this forestry treatment the least desirable option for dormice if undertaken over a large area and especially during the hibernation period. The results of this study are positive in that forest managers do have a number of harvesting options for economic PAWS restoration and conifer management whilst retaining dormouse populations in compliance with protective legislation relating to the “local population” by using a patchwork approach across entire compartments. Further positive enhancements can be applied by following conservation best practice, such as retaining perimeter undisturbed corridors and intra-wood connectivity, a directional felling pattern for clear fells encouraging movement towards suitable habitat and by avoiding the clearance or overall thinning of large proportions of dormouseoccupied woodland in one operation. If best practices are followed, other economic, silvicultural and conservation objectives can be included as part of the management regime.

This study is not yet fully complete and further monitoring following the final thin will indicate the longer term effects of the PAWS restoration treatments on this dormouse population. Moreover, it is important to replicate this study, or similar conifer felling methods, in other U.K. forests to improve confidence in the application of these results. The implications for dormice drawn from this study should be treated with some care as what applies to this particular site may not apply to all conifer woodlands. Other intersite and/or intra-site variables such as soil type, wind effects, tree growth rates, local shrub species and their growth rates may also be significant factors in decision making. Studies such as this are extremely important for informing government policy regarding best practice woodland management for economic PAWS restoration in the presence of dormice. Management to deliberately encourage dormice in conservation woodlands can take a different approach.