Introduction

Many animals, both invertebrate and vertebrate, build nests (e.g. stingless bees Trigona spinipes [1]; grass-cutting ants Atta vollenweideri [2]; chimpanzees Pan troglodytes [3]; great tits Parus major [4]). Selecting a suitable nest-site is important as it provides shelter from predators or adverse weather conditions, and increases fitness and survival of young [5–7]. Most nest-building birds, for example, invest considerable time and energy choosing their nest-site because certain sites greatly influence reproductive success [8] and the same is true for large mammals (e.g. badgers Meles meles [9]), and for many small mammal species (e.g. [5,10,11]). Knowledge of nest-site requirements is essential for the conservation of rare or specialist species [7,12], especially where nest-site availability limits population sizes, as has been observed in a variety of arboreal mammals (e.g. grey mouse lemurs Microcebus murinus [13]; northern flying squirrels Glaucomys sabrinus [14]; greater gliders Petauroides volans [15]; common dormice Muscardinus avellanarius [16]).

In the UK, a lack of appropriate woodland management and habitat fragmentation has resulted in the reduction of suitable habitat for dormice, at the edge of their range, leading to extirpations [17]. As a result, and despite legal protection, dormouse distribution has reduced by more than half since the 19^th century, and the species is now of conservation concern in the UK [18]. Dormouse nesting ecology is difficult to study because dormice are cryptic, nocturnal and arboreal; their natural nests are difficult to locate as they are usually concealed in thick foliage or in tree cavities, and may be as high as 15 m in the canopy [17,19]. This makes studies relying on natural nest-sites logistically challenging, or even misleading, because of the high risk of not finding nests [20]. Nestbox occupation data provide an opportunity to estimate relative abundance and distribution of dormice with minimal labour [21]. Dormice are found in nestboxes from mid-May to October, and are known to use them across their range, thereby allowing the comparison of findings across similar studies [22]. Nestboxes also benefit dormouse conservation. Bright and Morris [20] conducted a radio-tracking survey and found that artificial nestboxes were by far the most frequently used nest-sites compared to natural nests. They argued that, where nestboxes are present, almost the whole population would use them, and providing nestboxes appeared to double the number of dormice present in an area [20]. Some cavity-nesting bird species such as blue and great tits are known to use artificial nestboxes almost exclusively when they are available, and numerous studies have benefited from the study of these species in nestboxes [23]. As dormice also readily breed in nestboxes [24], this also allows the study of their breeding ecology. Both male and female dormice use nestboxes, and they can be found either singly or in groups of two or more (e.g. male-female breeding pairs, groups of juveniles, mothers with litters) and this fluctuates depending on the time of year. Dormice can have several litters per year, although exact numbers of litters and young per litter differ across their range [22] (note that two litters per year were commonly found in some nestboxes at the present study site; one in early summer and one in the autumn). Any findings that relate habitat features to nestbox preference or breeding success in nestboxes could therefore easily be used in an applied sense (e.g. changing nestbox location) and may have more immediate conservation implications than findings relating to habitat features in natural nest-sites (because these cannot be moved), although factors influencing the selection of natural and artificial sites may not be identical.

There is a growing focus on long-term volunteer-collected datasets in ecology [25,26] because volunteer-run programmes provide large quantities of data at minimal cost [27,28]. Deploying a team of volunteers can also save substantial amounts of time compared to using professional ecologists [27]. In the UK, many conservation organisations rely heavily on volunteers to collect data (e.g. the British Trust for Ornithology BTO, the Royal Society for the Protection of Birds RSPB, the People’s Trust for Endangered Species PTES, the Mammal Society, the Marine Conservation Society, the Wildlife Trusts and the Bat Conservation Trust), however, volunteer-collected data are often questioned because they lack the rigour and precision of scientific studies (e.g. [29]).

The dormouse is a popular and charismatic species in the UK. Currently, over 1,000 volunteers participate in the National Dormouse Monitoring Programme (NDMP) run jointly by the PTES and Natural England. These volunteers have been submitting records since 1988, and in 2011, there were 305 sites involved in the scheme (with some annual variation – S. Sharafi, PTES, pers. comm.). Volunteers are required to check nestboxes at a site at least twice a year (May/June and Sept/Oct) to monitor evidence of dormouse occupation. The records are analysed by the PTES to estimate national trends in dormouse numbers and distribution.

Understanding breeding dormouse population nestbox requirements is crucial if nestboxes are to be maximally effective for conservation. Using long-term (18-year) volunteer-collected data collected as part of the NDMP, this study: (1) tests whether dormice actively choose (rather than randomly occupy) nestboxes; (2) examines some of the biotic and abiotic factors responsible for this selection; and (3) provides recommendations on using large volunteer datasets, discussing the attributes and limitations such datasets present.

Methods

Results

Occupation of nestboxes was relatively low, with an average of 7.3% of boxes occupied in any given year (S.D. = 3.3, minimum 2%, maximum 13%). Occupation of nestboxes was not random (Z = 5.07, n = 97, p < 0.01), indicating active nestbox selection. The final stepwise-reduced model was highly significant (F_8,88 = 5.68, p < 0.01) and the suite of habitat variables entered explained 28% of variability in dormouse occupancy (adjusted r² = 0.28) (Table 2). It is important to note that the stepwise approach creates a best-fit model of numerous predictor variables in a multivariate framework, balancing model explanatory power and parsimony. Overall, this model is highly significant, and all explanatory variables in the model are important in achieving the overall significance and R² value, and warrant further discussion. Not all explanatory variables are independently significant in this final model (Table 2) since many of these are important in association with other variables (i.e. there is no simple univariate relationship). The most important factor determining occupancy was the distance from the wood edge. This was a positive correlation, indicating that dormice preferred nestboxes towards the centre of the wood. There was a negative relationship between occupancy and the circumference of the nestbox tree; smaller trees were associated more strongly with nestbox use than larger trees. There was also a negative relationship with the presence of ferns. Presence of oak and canopy clumpiness, as well as the number of trees and the presence of hawthorn were also important factors in the final best-fit model.

The stepwise regression also showed that there was a positive relationship between dormouse occupancy and bird nests (Table 2), indicating: (1) no evidence of competition in the study population and (2) that nestboxes were selected on the basis of similar, or at least closely correlated, variables. When bird nests and dormouse occupancy were further examined for all nestboxes within years, a relatively strong significant negative correlation was found (r_s = -0.56; n = 18; p = 0.016), implying potential competition or mutual exclusion on a yearly basis (Figure 1). Comparison of the observed percentage of occupation of each nestbox in a given year by birds, dormice or both showed no significant difference to calculated expected values where dormouse and bird occupation were calculated independently of one another (χ² = 0.01; df = 3; p > 0.99), hence, there was no evidence of competition between birds and dormice at this site.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Percentage of nestboxes containing bird nests and dormice (individuals rather than nests) between 1994 and 2011.

https://doi.org/10.1371/journal.pone.0067986.g001

In the stepwise binary logistic regression, five factors were found to be significant and these explained 28% of dormouse nestbox selection in total (estimated R² = 0.28, Wald = 0.83, p < 0.01). Distance from the edge of the wood remained the most significant explanatory variable (p < 0.01), followed by canopy clumpiness (p = 0.03). Ferns, honeysuckle and sycamore were also important in the final best-fit model (Table 3).

Discussion

This study demonstrates three main points. Firstly, dormice actively select nestboxes, a point often overlooked or impossible to test in habitat selection studies (e.g. [36,38]). Secondly, a suite of habitat factors can explain a considerable degree of this nestbox selection, which could inform the placement of nestboxes for the purpose of dormouse research and conservation. Thirdly, volunteers can collect useful data on dormouse nestbox occupation.

Several of the factors included as candidate variables in the model influenced dormouse nestbox selection, together explaining 28% of variability in occupancy. The most influential factor was the distance from the edge of the wood, which may be due to edge effects (such as increased predation or competition [39]), although some nestboxes were occupied despite being close to the wood edge and the presence of potential dormouse predators was not recorded in this study (for example, corvid birds are potential predators of dormice in edge habitats – see 40. As Midger Wood is a small wood (9 ha), it is not possible to determine at what point distance to the edge of the wood would cease to be important, for example, in a much larger wood. Additionally, the shape of the woodland might affect the importance of the distance to edge variable on dormouse nestbox selection, since this affects the edge:interior ratio. The influence of edge effects on nest-site selection has been studied mainly in avian populations [41] and there are currently no studies on its effect on dormice, although edge effect influences in smaller woods have been proposed [42]. Contradictory results show that dormice readily occupied nest tubes on the fringe of dense scrub in Dorset (S. Eden, pers. comm.), possibly because these were less favoured by competing small mammals (note: nest tubes consist of a length of corrugated plastic tubing and square in section containing a sliding wooden tray [43]). There may be very different selection pressures influencing populations across different habitats.

Since dormice selected nestboxes on thinner trees in this study, it may be that larger trees supported more natural nest-sites such as cavities or dense foliage in the canopy. There is much contention as to whether dormice prefer to use nestboxes or natural nest-sites and factors vary greatly in different habitats. In young woodlands, hedgerows and scrub, dormice may favour unenclosed natural nest-sites (e.g. woven into bramble) over tree hollows or artificial nest-sites [16,44], although in diverse, low-growing woodlands, radio-tracked dormice preferred nestboxes over natural nest-sites [20]. In coppice-with-standards woodland, radio-tracked dormice spent the majority of time either in nestboxes (34% of dormouse tracking days) or in natural tree hollows (41%) and far less time in natural nests in bramble (8%) [19]. Juškaitis [22] found that dormouse nestbox occupation was negatively, but weakly, correlated with tree crown density; the positive relationship with canopy clumpiness found in this study might be due to similar reasons, as gaps in the canopy would mean fewer arboreal routes, which may cause dormice to descend to nestboxes. High canopy clumpiness meant that there were areas of dense cover but also large gaps that let through direct sunlight, which would benefit the plant species that dormice use for food and nest material. It is still unclear how selection for natural nest-sites interacts with nestbox selection mechanisms, and this would be an interesting area for further investigation. Note that studies into natural nest-sites in woodland are facilitated by radio-tracking, and this is unlikely to be feasible using NDMP volunteers due to the legislation surrounding fitting radio-tracking devices to dormice, and the prohibitive costs involved. Nestboxes therefore remain a more practical way of studying dormice with the help of volunteers.

The presence of certain plant species influenced dormouse nestbox selection: dormice were positively correlated with oak and honeysuckle, and negatively correlated with ferns, sycamore and hawthorn. Food sources influence nestbox selection, as dormice rarely travel further than 100 m from their nests but require a diversity of food sources to ensure that food is available continuously throughout the active season [12,20]. Honeysuckle and oak are important food sources [12,18,45,46], with honeysuckle also forming an important component of dormouse nests in Midger Wood [30]. It is therefore unsurprising that these plant species are important explanatory variables in the final models. The presence of ferns is characteristic of dark and damp areas [47], which may be avoided by dormice. The negative relationship with sycamore is unlikely to be biologically meaningful as this species was only present near three nestboxes (these never contained dormice, and this is the reason for its statistical inclusion in the stepwise regression).

The lack of competition between dormice and other nestbox inhabitants was of particular interest in this study because competition for nestboxes occurs in other studies (e.g. [22,48]). Although the lack of competition between birds and dormice agreed with the findings of Morris et al. [24]), years in which dormice occupied more nestboxes generally coincided with years in which birds occupied fewer nestboxes, implying that larger scale effects such as population fluctuations might influence nestbox occupancy. The amount of volunteer-collected data available on birds and dormice might provide an opportunity to investigate this relationship further.

The remaining variability in our study might be explained by chance, variables that were not measured as part of this study (e.g. climate, predators, parasites, pathogens etc), or by dormouse learning and previous experience. Indeed, Marsh and Morris [49] found that boxes favoured by dormice in one year tended to be reselected by them in the following year; however, since individuals were not individually marked for identification at the study site, it was not possible to investigate this. Furthermore, since our study only investigated one small woodland in the UK, it is possible that the results may be site and size specific, and further exploration would be needed to elucidate the generality of the results. The temporal span of the dataset was 18 years, and some of the explanatory factors may have changed over this time despite consistent management by GWT. The influence of parasites and predators on dormouse nestbox selection would be an interesting topic for future study, but as this would require annual records of the relevant variables, it was not possible to examine this here using a historical dataset.

Using volunteer-collected data has both advantages and disadvantages. Alongside the usual benefits of saving time and money compared to recruiting professionals [e.g. 27, 28], a key advantage of this volunteer-collected dataset was its longevity; this can also be an important attribute of useful volunteer-collected data [50]. Additionally, volunteers surveyed the nestboxes monthly from April to November, the highest recommended number of nestbox checks in a year [51]. As a result of this, the dataset was large (>30,000 data points: 97 boxes * 6 months * 18 years * 3 species – dormice, small mammals and birds), which reduced the chance of a type II error.

Volunteer-collected data also has indirect benefits. For example, volunteers also monitored any issues at the site (e.g. fallen trees across footpaths) and reported these back to GWT, thus facilitating the overall management of the site. Most of the regular volunteers at Midger Wood were members of GWT, providing financial support through their memberships and therefore contributing to the cost of managing the site as well as collecting data. Newman et al. [27] found that at least 30% of their volunteers joined conservation organisations after they had volunteered on their project. Meaningful interactions with the natural world also have the potential to enhance human wellbeing and quality of life [52,53] and volunteers who participated in mammal surveying projects gained fulfilment and knowledge [27]. When asked, volunteers at Midger Wood stated that they gained enjoyment from monitoring the nestboxes and some had been participating in dormouse surveys at the site for 18 years.

The volunteer dataset did, however, present some analytical challenges. Although nestbox occupancy data were collected regularly and followed the majority of NDMP guidelines, volunteers did not collect any habitat data, despite habitat data being requested at 5-year intervals for the NDMP. These habitat data would have proved extremely useful in the present study. Habitat characteristics were measured by the authors at the end of an 18-year period, and we were aware that some of these would have changed during this time. A careful analysis and a consideration of variables that may have changed resulted in useful trends being identified. Some nestbox records were difficult to interpret and, if they could not be confirmed, they had to be discarded from the dataset (<5% of the records). There was a certain degree of variability in the records that made computerising the dataset time-consuming (e.g. a record of “*DORMOUSE*” was described as an unoccupied dormouse nest at the bottom of the recording form, not the presence of a dormouse as suggested). Exact records (i.e. how nestbox contents were recorded on the form) were variable between different people despite using the same data recording forms, and this issue increased with the fluctuating number of volunteers.

Conclusions and Recommendations

This study has developed work by previous researchers and has furthered understanding of dormouse nestbox selection. It indicates that dormice select nestboxes based on a combination of factors. While views on the importance of nestboxes for dormouse conservation differ, many, but not all of the results of this study are likely to be relevant for natural nest selection too. Large scale features, such as distance to the edge of the wood, or combinations of plant species in the nearby vicinity are likely to apply equally to natural nests and nestboxes. Some localised factors may differ, as nestboxes provide shelter that may be absent on thin trees with low structural complexity, which would prevent dormice from building natural nests on these trees. Nevertheless, these results are important in informing conservation management decisions where nestboxes are used, and, in combination with other studies, in understanding the broad principles of dormouse habitat selection in any woodland.

Monitoring dormice using volunteers can provide an adequate quantity of analysable data, and useful information can be extracted from data that might usually be considered less reliable compared to rigourous scientific data, as shown in other studies (e.g. [50,54]). Volunteer schemes with large historical datasets are irreplaceable and invaluable as they can produce important ecological information and can help identify important sites and management strategies [28,54]. NDMP records vary greatly in quantity and quality between sites and years (S. Sharafi, PTES, pers. comm.), so it would be useful to determine the reasons behind this variation in order to uncover ways in which to reduce it, thus improving the national database. Volunteers should be informed of the importance of completing forms consistently and of collecting regular habitat data, and guidance on this matter should be given to the leaders of monitoring groups. Volunteer schemes would undoubtedly benefit from scientific input to improve data collection, thereby facilitating scientific study of those data and allowing the results to be of maximum usefulness for applied ecology and conservation.

Acknowledgments

Many thanks to Dora Clarke for providing 18 years worth of dormouse records, to the volunteers who monitor the nestboxes, and to Chris Newman and an anonymous reviewer for useful comments on the manuscript.