Influences of adjacent suburbia, fire regimes and vegetation on the mammals of a peri-urban reserve

Urban development is thought to negatively affect most native mammals. Here we assess whether adjacent suburbia, fire regime and vegetation influence the activity of mammal species in a reserve in Hobart, Tasmania, Australia. We used multiple regression to analyse the relationship between distance from houses, fire history, vegetation structure and floristics, and mammal activity, derived from camera visits and signs. Animal diggings and visits by long-nosed potoroos, southern brown bandicoots and cats were greater close to houses, while visits by red-necked wallabies, rufous-bellied pademelons, brushtail possums and short-beaked echidnas were not. The structure of the vegetation, particularly related to shelter, was important in many models. Many models also included floristic ordination axes that reflected drainage, fire regime and distance from high-density housing. Positive relationships between the introduced predatory cat and some of its potential native prey animals may reflect the influences of close proximity to suburbia, such as elevated domestic mesopredator populations, and elevated resource availability in domestic gardens. Our results suggest that urban areas can have a valuable role in nature conservation, despite, or because of, their effects on remnant native ecosystems.


Introduction
Increasing urbanisation is one of the leading causes of global biodiversity loss and species extinction (Czech et al. 2000;Marzluff 2001;McKinney 2006). It can impact species through a range of processes including: loss, degradation and fragmentation of natural habitats; increases in exotic species; noise and light pollution; and altered hydrology. Impacts vary depending on the intensity of urbanisation (McKinney 2002(McKinney , 2008, as in Melbourne, Victoria, Australia where suburbs (built up residential areas outside the Central Business District) with treed gardens have very different (summarised in McKinney 2002) have reported higher species richness in some suburban areas than in more natural areas or more intensively modified parts of the city, a pattern consistent with the intermediate disturbance hypothesis. For example, Brady et al. (2011) found that native mammal species richness peaked at moderate levels of development intensity. Suburbs in a matrix of largely natural habitat may provide environmental heterogeneity, which is further enhanced by productive suburban gardens (Falk 1976;McKinney 2002). However, other studies have reported reduced species diversity in suburban and exurban areas (Marzluff 2001;Daniels and Kirkpatrick 2012). Globally most urban conservation studies have focussed on plants, invertebrate and birds (Shwartz et al. 2014). More studies on the impacts of urbanisation and urban remnant management on mammals are required. We know that urbanisation can cause a decrease in mammal species richness and abundance (Tait et al. 2005;van der Ree and McCarthy 2005;Ordenana et al. 2010;Daniels and Kirkpatrick 2012;Luck et al. 2013) as well as have adverse physiological consequences (Brearley et al. 2012). While urbanisation generally has adverse impacts on mammal populations, some species are more abundant in exurban environments than in similar native habitats (Ordenana et al. 2010;Daniels and Kirkpatrick 2012;Maclagan et al. 2018).
With urbanisation increasing globally there is a need to better understand factors that could help maintain and promote retention of native species or mitigate against negative factors. Habitat structure, vegetation cover, vegetation extent, urban intensity/housing density, native vegetation proximity/urban proximity, functional connectivity of habitat fragments, road density and access by domestic dogs have all been found to be important for determining presence and abundance/activity of mammal species in urban and peri-urban (on the margin of built up areas) environments (Baker and Harris 2007;FitzGibbon et al. 2007;Garden et al. 2007;Ordenana et al. 2010;Luck et al. 2013;Caryl et al. 2016;Bryant et al. 2017;Van Helden et al. 2020;Zungu et al. 2020). Where large remnants of native vegetation occur adjacent to, or enclosed by, urban development, management of potential threats to biodiversity such as fire and introduced plants can be difficult if maintenance of native mammal biodiversity is a primary aim. Planned fires to protect nearby houses from fires lit by arsonists can deleteriously affect habitats for mammals if undertaken too frequently or infrequently or if not enough unburnt habitat is retained.
The present paper provides a Tasmanian case study of the effects of vegetation, fire history and proximity to houses on the activity of the more common mammals of a peri-urban reserve. Our work is of particular interest in that Tasmania is one of the few places where the suite of rticial-weight-range mammals, susceptible to extinction through European fox (Vulpes vulpes) predation, survive.

Study area
The study was conducted in a small conservation reserve (Peter Murrell State Reserve and Conservation Area; 277 ha; 43°00' S, 147°18' E) comprising eucalypt forest, eucalypt woodland (mainly black peppermint Eucalyptus amygdalina) and heathland that is isolated from other areas of native vegetation by development, including adjacent residential suburbs (Fig. 1). The study area is described in detail by Kirkpatrick (1999) and Driessen et al. (2010).

Experimental design
This study used a design previously established to investigate the response of mammals to planned burns, comprising four fire-management blocks (each 17-20 ha) (Driessen and Fig. 1 The locations of transects and houses in the study area. Vegetation, scat and digging surveys were conducted at all stations (yellow triangles and red circles) and cameras surveys were conducted at four stations per transect (red circles). Red line indicates boundary of the reserved land Jarman 2010Jarman , 2014Driessen et al. 2021). The transects were located to adequately cover the area of the fire-management blocks, with spacing taking into account the home ranges of the medium-sized and small-sized mammals that were the target of the fire study. Two transects, 100 m apart, were established in each block and each transect consisted of 10 stations (marked by a steel stake and used for vegetation and fauna surveys) each ~ 50 m apart ( Fig. 1). At the time of our surveys in 2018, block AB had been subject to a moderate intensity (500-3000 kWm − 1 ) planned burn in 2010, and block IJ and the eastern 60% of block CD had been subject to a moderate intensity planned burn in December 2013. Most of the reserve (94%) was burnt by an intense fire in January 1988 (30 years prior to our study).

Data collection
Camera-trapping took place in seven-day sessions at four stations per transect; every second station from the east or south except for the last station (total 32 stations, see Fig. 1); one session in autumn and one in spring 2018. We used the same camera stations in each seven-day session. KeepGuard KG 680 V cameras were attached to small trees and angled down towards a PVC bait container (peanut butter, rolled oats and sesame oil) and pegged into the ground. Heights of cameras above ground (0.5-1 m) and distances to the bait stations (1.2-2 m) were as standardized as natural variations in tree-trunk shape and ground micro-topography allowed. By aiming the cameras towards the bait container we minimised possible variation in detection due to variation in camera height and distance between camera and bait station (McIntyre et al. 2020). Cameras, with a minimum triggering interval of 30 s and high-level sensitivity, took single-frame images, with date and time recorded on each. The cameras use infrared illumination at night and have a 50° field of view and a trigger time of 1 s. KeepGuard KG 680 V cameras have been shown to perform as well as or better than some other commonly used cameras (e.g., Reconyx H600) (Driessen et al. 2017) and no performance issues occurred with the cameras during their deployment for this study. All images were assessed for the presence of an animal and in most images the species was identified. Our metric for mammal activity was the number of visits by a species rather than the number of images (Jarman and Driessen 2019). We defined a visit as one or more images of a species with no interval between animal images greater than 5 min. Our choice of 5 min as the maximum allowable within-visit interval was based on our analyses of inter-image durations in large samples of timed images of long-nosed potoroos (Potorous tridactylus) (Jarman and Driessen 2019) and was appropriate for other species recorded in the study area (P. Jarman and M. Driessen, unpublished data).
In autumn 2018, all observed vascular plant taxa were recorded in 80 10-by-10 m plots centred on each of the 80 stations. The tape that defined the borders of each plot was used as an intercept line. The outline intercept cover was measured to the nearest cm for shrubs, graminoids (all monocotyledonous plants), ferns, herbs, bryophytes, lichen, bare ground (clear to the sky), rock and litter. A Bitterlich wedge count from the middle of the quadrat was undertaken by species to determine basal area. Slope (degrees), aspect (degrees) and topographic position (lower, upper, ridge and valley) were also recorded. The distance in metres to the nearest house and the nearest high-density housing (250-1000 houses/km 2 ) was estimated using Google maps. Scats of macropods (red-necked wallaby (Notamacropus rufogriseus) and rufous-bellied pademelon (Thylogale billardierii)), and possums (brushtail (Trichosurus vulpecula) and ring-tailed (Pseudocheirus peregrinus)) were counted for the whole plot of all stations, as were all marsupial and monotreme diggings.
In autumn 2018, an index of horizontal vegetation density was recorded for all stations using a 100-cm-tall by 50-cmwide cover board marked with 200 5-by-5-cm squares. The cover board was placed on the ground 5 m to the north, and then to the south, of each station and the numbers of squares obscured by vegetation between 0 and 50 cm and between 50 and 100 cm above the ground were recorded by an observer standing at the station. The average of the north and south scores provided an index for the station of vegetation density between 0 and 50 cm and between 50 and 100 cm.
Fire history data since 1988 was obtained from the Peter Murrell State Reserve and Conservation Area Fire Management Plan (PWS 2006) as well as a fire mapping database maintained by the Tasmanian Parks and Wildlife Service. From this information the times since last fire in 2018 and number of fires between 1988 and 2018 were determined for each station.
Variables used in the analyses are summarised in Table 1 together with minimum, maximum and mean values (± 95% confidence intervals) for the 80 stations.

Statistical analysis
All analyses were undertaken in Minitab18, except for ordination of the vegetation data.
Non-metric global multidimensional scaling ordination in DECODA (Minchin 1990) with the default options was used to reduce the plant species presence/absence data to four variables. The four-dimensional ordination solutions had a minimum stress of 13.4%. The first axis discriminated assemblages of dry heathy vegetation, characterised by such species as Bossiaea cinerea, Allocasuarina monilifera and species of poorly-drained ground included Gymnoschoenus sphaerocephalus, Leptocarpus tenax and Restio monocephalus ( Table 2). The scores on this axis were also predicted by slope, with the positive values associated with species Epacris impressa, from other types of assemblages ( Table 2). The heathy extreme was on steeper slopes (Table 3). The second axis discriminated species assemblages characteristic of heaths on poorly-drained ground from the rest. The Cover on cover board 50-100 cm above the ground at each station 21.7 ± 5.2 0 93 Vegetation density (0-50 cm) Cover on cover board 0-50 cm above the ground at each station 63.5 ± 5.5 5 100 Distance to house Distance (m) of station from nearest house on any lot size 533. Cat (Felis catus) and short-beaked echidna (Tachyglossus aculeatus) camera visits were too infrequent to treat as continuous variables, so were treated as presence/absence. One-way ANOVA was used to determine whether animal observations and some environmental variables varied by the time since the last fire, the presence/absence of cats and the presence/absence of echidnas. Pearson product moment correlation coefficient was used to determine the significance of linear relationships between continuous variables.

Overview
A total of 1,795 visits to camera by mammals was recorded comprising 8 native species and 3 non-native species. The native species were rufous-bellied pademelon (707 visits), long-nosed potoroo (657), brushtail possum (218), rednecked wallaby (115), southern brown bandicoot (Isoodon obesulus) (82), short-beaked echidna (12), ringtail possum (3) and Tasmanian devil (Sarcophilus harrisii) (1). The nonnative species were house mouse (Mus musculus) (72), cat (10) and dog (Canis familiaris) (2). of the poorly-drained ground associated with gentler slopes ( Table 3). The scores on axis 3 were strongly correlated with graminoid cover, with high cover of Gahnia radula and Hypolaena fastigiata at the positive end of the axis. (Table 2). The scores on this axis were also correlated with the two distance from houses variables ( Table 3). The scores on the fourth axis discriminated recently burned vegetation from less recently burned vegetation, with strong correlations with time since last fire and number of fires, as well as distance from high-density housing ( Table 3).
Multiple regression analysis, guided by the 'Best Subsets' routine in Minitab, was undertaken with each of the taxa, and indicators of taxa, for which quantitative data were available as dependent variables (Table 1). The maximum number of predictor variables was determined by the sample size, with at least 10 samples per variable being used to avoid overfitting. The selected best model was that which had the highest adjusted r 2 and significant (P < 0.05) slopes on each variable. Where possible, our analyses remove collinearities from predictive equations. However, the precision of measurement can affect the relative strength of explanatory variables, thereby influencing inclusion, so we have been careful to assess the logic of the variables included in our models.

Fire regime
The number of years since the last fire had positive coefficients in the models for southern brown bandicoot visits and total diggings density, and negative coefficients in the models for red-necked wallaby visits and macropod scat density (Table 4). There was a positive coefficient for number of fires in the southern brown bandicoot visit model ( Table 4).

The role of vegetation and topography
The total basal area of trees had a positive coefficient in the model for possum scat density ( Table 4). The basal area of the most common tree, Eucalyptus amygdalina, had positive coefficients in the models for the southern brown bandicoot visits and total diggings density and a negative coefficient in the model for macropod scat density ( Table 4).
Vegetation density at 0-50 cm had a negative coefficient in the total diggings density model, while vegetation density at 50-100 cm had positive coefficients in the models for the red-necked wallaby and rufous-bellied pademelon visits. Fern cover had negative coefficients in the red-necked wallaby and brushtail possum visits models, graminoid cover had a positive coefficient in the macropod scat density model and shrub cover had a positive coefficient in the long-nosed potoroo visits model. The scores on axis 2 of the floristic ordination had a negative coefficient in the longnosed potoroo visits model, indicating an association with plant communities of well-drained ground ( Table 3). The scores on axis 4 of the ordination had negative coefficients in the models for brushtail possum visits and total diggings density (Table 4), indicating their preference for communities of recently burned ground ( Table 3). The model for total diggings also included a positive coefficient for slope (Table 4).
There were sufficient data to develop multiple regression models for the long-nosed potoroo, southern brown bandicoot, red-necked wallaby, rufous-bellied pademelon, and brushtail possum visits, macropod scat, possum scat and total diggings density. The best models had R 2 values varying from 6.1% for possum scat density to 73.5% for southern brown bandicoot visits (Table 4). The data for cats and short-beaked echidnas could not be modelled beyond one predictor because of the low numbers of visits. There were no significant relationships for the house mouse.

Distance to houses
The distance to high-density housing had negative coefficients in the models for southern brown bandicoot visits, long-nosed potoroo visits and total diggings density (Table 4). Cat visits occurred closer to high-density brown bandicoots (F = 29.3 1,30 , P < 0.001), but negatively associated with macropod scat density (F = 4.4 1,30 , P = 0.045). Both southern brown bandicoots and cats decrease rapidly in visits with distance from high-density housing (Fig. 2).

Discussion
Within a remnant, native vegetation reserve, we investigated the relationship between adjacent suburban environment and the activity of native and introduced mammals taking into account recent fire regimes and habitat variation.

Suburban influences
The human threat hypothesis suggests that population performance declines with increasing human dominance (Lawson et al. 2008). Thus, we might have expected decreased mammal activity with increased proximity to suburbia. However, the assumption that human-modified or "novel" habitats are poorer in quality to remnant habitats has been questioned (Maclagan et al. 2018;Soanes and Lentini 2019), and for some species abundance is greater in novel habitats than in remnant habitats (Daniels and Kirkpatrick 2012; Maclagan et al. 2018). We found no evidence of decreased mammal activity with increasing proximity to houses, but we did find that the activity of the southern brown bandicoot, the long-nosed potoroo and the cat

Fire regime influences
Fire regimes can be an important influence on mammal community assemblages in many ecosystems around the world (e.g., Whelan et al. 2002;Fontaine and Kennedy 2012). For protected areas adjacent to suburban areas, fire management is a critical issue for land managers because of the risk of bushfire to human life and property and this can sometimes be at odds with management for conservation (Driscoll et al. 2016;Ryan et al. 2020). Within our study area, fire is primarily used as a tool to manage risk to human life and property and therefore has potential to impact on the persistence of mammal species that occur therein. Time since last fire featured in the models associated with southern brown bandicoot and red-necked wallaby activity (visits, diggings and scats). The negative effect of recent fire on southern brown bandicoot visits and total diggings density is consistent with that found for diggings of mainland southern brown bandicoots (Claridge and Barry 2000) and may reflect a lack of shelter and food resources. However, other studies have found limited effect on their activity following fire (Shan et al. 2006;Hope 2012;Driessen et al. 2021). The presence of cats at our study site may have played a role in our study as Arthur et al. (2012) suggested that cat predation may retard southern brown bandicoot recovery after fire. The positive association of southern brown bandicoots with areas of the reserve that had been burnt several times suggests that burning may be a positive influence for this species over the longer term; but this requires further investigation. In contrast to southern brown bandicoots, red-necked wallaby activity (visits and macropod scats) was greater in recently burnt areas of the reserve, consistent with two previous studies (Southwell and Jarman 1987;Driessen et al. 2021), and most likely reflecting a tendency of these macropods to feed on fresh growth in recently burnt areas (Southwell and Jarman 1987;Kirkpatrick et al. 2011Kirkpatrick et al. , 2016Styger et al. 2011). However, Dixon et al. (2019 found that red-necked wallabies occurred more frequently on long-unburnt than more recently burnt subalpine forests and woodlands. While fire regime variables were in the models for total diggings density, macropod scat density, red-necked wallaby visits and southern brown bandicoot visits, the effects of fire regime on vegetation structure is also likely to be important for those species with shrub cover or vegetation density in their models (Monamy and Fox 2000). The pademelon and potoroo fit this description. The opposite responses to time since fire shown by red-necked wallaby and brushtail possums is striking. Such differences between species in response to fire regime and vegetation have been widely observed in non-urban areas (Catling et al. 2001;Lindenmayer et al. 2008;Arthur et al. 2012).
increased closer to high-density housing. For the southern brown bandicoot this is consistent with other studies on this species (Daniels and Kirkpatrick 2012;Bryant et al. 2017;Maclagan et al. 2018), whereas the relationships between long-nosed potoroos and urban environments have not been previously studied. Not surprisingly, cats were associated strongly with high-density housing; the surprise was their positive associations with long-nosed potoroos and southern brown bandicoots. Both native mammals were more widespread than the cat in the study area, but the numbers of visits per site were higher where cats occurred, which was higher close to high-density housing than otherwise. Given that the direct effects of the cat on such mammals is almost certain to be negative, through the disease toxoplasmosis and direct predation, alternative explanations for the association between these species are: (1) that the habitat for these mammals in the reserve is better close to high-density housing than elsewhere and the cats are domestic, fed by the householders; (2) that the habitat is better near the high-density housing and the cats are attracted by the high density of prey; (3) that the small mammals have elevated populations close to high-density housing because they utilise domestic gardens as well as the ecosystems in the reserve, as do the domestic cats. The first two hypotheses are not consistent with the strong effect of proximity to high-density housing in the same multiple regression models as other environmental variables. This variable, thus, has an effect independent of the other variables included in the models, and is more explanatory than them. The third hypothesis is consistent with observations of bandicoots occurring in exurban gardens (Daniels and Kirkpatrick 2012) and with the high productivity associated with the elevated fertility and moisture availability of suburban gardens (Falk 1976;Bidwell et al. 2006;Bell et al. 2011).
The resources created by gardeners may have a fecundity and survival effect that outweighs mortality from high numbers of domesticated mesopredators. Gardeners also provide a complex environment within which predation may be less effective than in some native vegetation. Thus, high mesopredator abundances and high mammal abundances are likely to be independently determined by the actions of householders in maintaining domestic animals and creating and maintaining gardens rich in food resources for native animals. The models suggest that elevated populations of mammals may also partially depend on dense sheltering growth in the native vegetation near the gardens, indicating the desirability, for conservation of native mammals, of a low fire frequency (15-25-year intervals). Mosaic burning is desirable as several native species appear to benefit from recent burning.
MacLennan 1984), explaining the negative association with fern cover in their best model and their positive association with macropod scat density. The model for macropod scat density may reflect the importance of graminoids as a food source for macropods in the area and a preference for open vegetation for feeding (McArthur et al. 2000), including recently burned areas as previously mentioned. However, the models for red-necked wallaby visits, as well as rufousbellied pademelon visits, included a positive influence of vegetation density 50-100 cm above the ground, seemingly contradicting the preference for openness implied by the macropod scat density model. However, this relationship reflects the well-known tendency of both wallabies and pademelons to shelter in dense vegetation during daylight hours (Le Mar and McArthur 2005;While and McArthur 2005). The avoidance of dense fern cover indicated by the red-necked wallaby visit model could reflect the association of Pteridium esculentum with low graminoid cover.

Caveats
While the reserve supports many mammal species, some of which appear to benefit from the adjacent suburb, there are several species that may already have been disadvantaged by the encroachment of suburbia, its associated disturbances and the increasing isolation of the reserve. Tasmanian devils, eastern quolls (Dasyurus viverrinus), eastern bettongs (Bettongia gaimardi) and the platypus (Ornithorhynchus anatinus) have been recorded from the area in the past but are uncommon or absent now (Driessen and Jarman 2010; Driessen et al. 2021).

Conclusions
We conclude that the effects of adjacent suburbs on native mammals in the reserve were positive or neutral, depending on taxon. The reserve is isolated from other areas of native vegetation by suburbia and farmland, so could theoretically be subject to loss of species that require larger areas of habitat to maintain viable populations, accounting for the apparent absence of species found in the areas in the past. Within the reserve, fire and vegetation structural variables were most associated with variation in mammal visits and signs, reinforcing the importance of continuing to maintain a diversity of fire ages in different vegetation types.
Brushtail possum visits were positively associated with recent fire (axis 4), which is consistent with an increase in activity in forest and woodlands following cool burns (Driessen et al. 1991(Driessen et al. , 2021 and wildfire (Lindenmayer et al. 2008) but not with abundance after wildfire in the study by Catling et al. (2001).
In peri-urban areas fire management to reduce fuels is a priority for fire managers and is complex and challenging. Bioturbation by wildlife as a result of foraging activities can assist with reducing fuel loads (Ryan et al. 2020). Management practices should facilitate maintenance of these populations and protection of life and property. Scale of fire is also important. Here, we investigated low-moderate severity fires, which aim to mitigate against larger-scale and or intense fires to reduce impacts on adjacent human life and property and to avoid adverse biodiversity outcomes. For example, Ramalho et al. (2018) used population viability analysis models based on field collected data to demonstrate that isolated metapopulations of the southern brown bandicoot were predicted to be highly sensitive to fire, with fires that burned large areas having greater negative impacts on population abundance than small fires. Fire management that limits the size of unplanned fires and promotes a mosaic of vegetation within and between remnant habitats will help protect human life and property as well as biodiversity. However, appropriate fire mosaics in peri-urban environments and the appropriate means for achieving them require further investigation.

Vegetation influences
In addition to proximity to houses and fire regime, variables associated with vegetation were also important in determining variation in mammal species activity within our habitat remnant. The positive effect of Eucalyptus amygdalina basal area in the model for the southern brown bandicoot visits is consistent with the year-round importance of fungi in their diet (Quin 1985;Mallick et al. 1998) and the importance of trees in supporting soil fungi (Johnson 1994;McMullan-Fisher et al. 2010). However, no such relationship was recorded for fungivorous long-nosed potoroos, which occurred widely throughout the reserve. The association of the long-nosed potoroo with high shrub cover and good soil drainage in its best model is consistent with previous descriptions of its habitat (Seebeck et al. 1989;Bennett 1993;Norton et al. 2010), although, at the local scale, Bennett (1993) also found that the long-nosed potoroo was not strongly correlated with any structural feature of the vegetation. It was not surprising that possum scat density increased as total basal area of trees increased, given the importance of trees for both ringtail and brushtail possums. Brushtail possums feed on the ground to some extent (Kerle 1984;Bennett AF (1993)  Water and Environment. This study was partly funded by the Tasmanian Government.
Author contribution Jamie Kirkpatrick designed the vegetation, diggings and scat data collection, collected the vegetation data and undertook most of the analysis and writing up. Michael Driessen was the major designer of the larger project, collected much of the mammal and environmental data and contributed substantially to the writing of the paper. Peter Jarman contributed to the design of the project, collected much of the mammal data and contributed to the writing of the paper. Lauren Jakob collected the diggings and scat data and contributed to the writing of the paper.

Funding Open Access funding enabled and organized by CAUL and its Member Institutions
Data Availability Data and material are available from the authors upon written request.
Code Availability Not applicable.

Conflicts of interest/Competing interests
The authors declare that they have no conflict of interest.
Ethics approval This study followed the Standard Operating Procedures for Camera-trapping of Wild Tasmanian Mammals 2013 by the Department of Primary Industries, Parks, Water and Environment.

Consent to participate Not applicable.
Consent for publication was obtained from the Tasmanian Government. All authors consent for the publication.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.