Moreover, mammals are one of the most important taxa concerning biological invasions on islands (Courchamp et al. 2003), either with a domesticated or wild-origin.
Detecting changes in mammalian biodiversity require data to be collected systematically in space and time, which can be both challenging and time consuming. Mammalian biodiversity assessments are a two-stage process, beginning with an inventory of the taxa of concern at a particular ‘site’ followed by periodic monitoring to detect changes at that ‘site’, including natural (e.g. impact of seasonality) or human-mediated (e.g. forest fire) changes. The inventory /monitoring planning and techniques used for the assessment will, therefore, differ, depending on the taxa of concern (physical characteristics and mode of life) and the size (from local to landscape levels) and characteristics (homogeneous vs heterogeneous) of the ‘site’.
Based on the previous considerations bellow we propose four distinct protocols to survey bats, small mammals, carnivores and large herbivores in forested habitats on islands.

PROTOCOL 7 – BAT INVENTORY / MONITORING
Taxa – Bats
Experimental design – Subsampling within the main 50 m × 50 m plots. However, the use of small plots (e.g., 50 m × 50 m plots) is rarely suitable for monitoring flying mammals, i.e. bat population trends. Instead, monitoring should focus on finding and estimating population areas (such as large roosts in trees or caves), and in documenting their presence across measured gradients using mist nets, bat detectors, and harp traps.

Frequency – The monitoring protocol as described in the main text is adequate for sampling every year.

Description
Most countries strictly regulate the capture and handling of bats and other mammals for research; Sikes et al. (2016) provide guidelines that are widely followed. See Kunz (1988) for a broad overview of methods for studies of bats.
Bats are often diverse and abundant on islands in the wet tropics. Their highly variable body size and roosting habits require a diverse set of monitoring techniques. The large flying foxes (also called giant fruit bats) of the Old World tropics often roost in large colonies (from tens to tens of thousands individuals) in the tops of large trees. These can be monitored readily by observation with binoculars, using standardized procedures to estimate the number of individuals (Kunz 1988; Kunz et al. 1996). Because individuals often live for more than a decade, long-term monitoring is essential to determine population trends. Many bats roost in caves. The number of individuals and species may vary greatly depending on the characteristics of the individual cave (length, height, humidity, temperature, and type of disturbance), so it is necessary to monitor multiple caves to determine overall bat population trends. Use of flashlights and binoculars may allow estimates of species abundances, but care should be taken to minimize disturbance, especially during the season when females gather into maternity roosts. Many species within a given genus or family may be similar in appearance, so it may be necessary to capture, photograph, and measure some individuals to determine species identity. Captures may be made with small nets (“butterfly nets”) inside the cave, or with mist nets at cave entrances. Use of mist nets in this manner must be done with great care, since the number of bats emerging may increase dramatically at dusk, overwhelming a mist net. Care should be taken to handle bats carefully to avoid harming the bats, and to avoid transmission of any disease-causing agents to the handlers (Sikes et al. 2016). Mark-and-recapture methods often add greatly to monitoring programs (Kunz 1988; Jones et al. 1996; Kunz et al. 1996).
Because caves are often the location of hunting activities, monitoring should include keeping records of evidence of disturbance. Such evidence may include signs of smoke from fires set within the caves (to asphyxiate the bats), remains of old nets (often
fish-nets), and the presence of darkened areas where bats formerly roosted but are unoccupied at the time of monitoring (Kunz 1988; Kunz et al. 1996).
Many species of bats roost in inconspicuous places, such as hollow trees, within vegetation, in crevices and cracks in large rocks, etc. These populations can be monitored in several ways. Bats of small body size (100 g or less) that eat fruit (Pteropodidae in the Old World, Phylostomatidae in the New World) are relatively easy to capture with mist nets, and often constitute communities rich in species and individuals. Bats that consume insects are more difficult to capture because of their active use of highly sensitive echolocation systems (“sonar”); mist nets and harp traps can be successful, but, for most monitoring projects, the extensive and specialized effort may not be productive. In places where a “dictionary” of the calls produced by the bats has been compiled, use of ultrasonic detectors (“bat detectors”) may readily allow the presence of individual bat species to be documented.
Bats typically utilize virtually all habitats in lowland tropical areas, from rural to old-growth forest. Monitoring programs should carefully document the habitat in each sampling area along disturbance gradients, since the various species of bats typically forage in a restricted range of specific habitat types. Bats typically are most diverse in lowland habitats (below about 1000 m), and decline in diversity and abundance with increasing elevation. However, the percent endemism among bats often increases with elevation, so that incorporating an elevational component t is often important for monitoring purposes.

PROTOCOL 8 – NON-FLYING SMALL MAMMALS INVENTORY/ MONITORING
Taxa – Non-flying small mammals
Experimental design – Within a relatively small island it may be possible to cover the entire area with a regular or random distribution of sampling points. On larger islands a spatial sampling should be implemented by selecting sub-sampling units that allow an ‘a-posteriori’ overall estimate.
Small non-flying mammals often have individual home ranges that exceed one ha, so sampling areas for them will often need to encompass 10 ha or more (Jones et al. 1996; Wilson et al. 1996).

Frequency – For inventory purposes at least one sampling event in the pre-breeding and post-breeding seasons to account for life-cycle induced variations in behaviour and abundance. For monitoring purposes a minimum interval of five years is adequate to assess community changes and/or population trends.

Description
Non-flying small mammals, generally defined as species weighing one kilogram or less, comprise the largest number of mammal species world-wide. The ecologically and morphologically highly diverse rodents are often the most species-rich and abundant native small mammals on islands, but small insectivores (such as shrews and tenrecs) may also be diverse and abundant, and marsupials are present on many islands in the vicinity of South America and Australia. Other groups of small mammals tend to be locally distributed, but may be abundant (e.g., the “wood shrews” of SE Asia that are non-spiny relatives of hedgehogs).
Some small mammals that are active during the day, such as squirrels and small primates, can be observed using binoculars, and other species (such as small arboreal rodents) can be seen at night with flashlights (torches); establishing standardized transect trails can document their abundance. Such trails should be walked at some standard rate of speed, and distances carefully documented, so that the number of observations per hour and per kilometer can be calculated (Rudran et al. 1996).
Many species of small mammals, perhaps most, are difficult to detect even when they are abundant. Some of these can be captured in cage traps. For such species, lines of traps can be laid out either in long lines or in grids, so that either the number of captures per trap-night (for trap lines), or the density per hectare (for grids), can be calculated (Wilson et al. 1996). In such studies, it is often necessary to use several types of bait, since different species consume different foods. For example, traps using a “seed-type” bait (roasted coconut and peanut butter are often successful) can be used in conjunction with traps baited with live earthworms (attached with safety pins to the trap) or fresh fruit (such as banana slices) to attract species that do not eat seeds (e.g., Rickart et al. 2011, 2016).
Small mammals usually show dramatic variation in diversity and abundance along elevational gradients; they typically are least diverse and abundant in the lowlands, and
increase in both respects with increasing elevation. When cloud forest (= mossy forest) is present, maximum diversity of small mammals often occurs in the lower portions of that habitat. At the highest elevations on a given mountain, there is often a decline in both diversity and abundance, but endemism is often high at higher elevations. It is therefore important that monitoring protocols include sampling areas along the entire elevational gradient to reflect all of the potential variation that may be present. Additionally, most species show specific responses to types and extent of habitat degradation, and so it is often necessary to include disturbance gradients at multiple elevations to determine the full extent of changes that are taking place (Rickart et al. 2011; 2016).

PROTOCOL 9 – Carnivores inventory / monitoring
Taxa – Small/Medium-sized carnivores
Experimental design – Within a relatively small island it may be possible to cover the entire area with a regular or random distribution of sampling points. On larger islands a spatial sampling should be implemented by selecting sub-sampling units that allow an ‘a-posteriori’ overall estimate. If forest cover is heterogeneous (i.e., different forest types), sampling site selection should involve stratification to allow estimates per stratum (forest type). To maximize species detection (including rare species) the combined use of sign surveys, camera-trapping stations, and tracking-plates is advisable. However, although signs of presence are a good indicator of carnivore occurrence, there is evidence that species assignments based on scats can have a high error rate, supporting the need for the complementary use of molecular markers.

Frequency – For inventory purposes at least one sampling event should be undertaken in the pre-breeding and post-breeding seasons to account for life-cycle induced variations in behaviour and abundance. For monitoring purposes a minimum interval of five years is adequate to assess community changes and/or population trends.

Description
Oceanic islands faunas are characterised by a general absence of native mammalian predators; when naturally present they are normally limited to continental-shelf islands. This is the case of the Kodiak bear, a subspecies of the brown bear (Ursus arctos middendorffi), which is native to the islands of the Kodiak Archipelago (SW Alaska)
(Van Daele 2003). Most carnivore predators nowadays inhabiting oceanic islands are small/medium sized mammals that result from human-mediated intentional introductions either as domestic pets (cats Felis catus and dogs Canis familiaris) or to hunt or control rabbits (e.g., the ferret Mustela furo on Madeira islands – Mathias 1988), commensal rodents (e.g., least weasel Mustela nivalis in Mediterranean and Atlantic islands – Rodrigues et al. 2017) or snakes (e.g. the small Indian mongoose in Caribbean and Pacific islands – Hays & Conant 2007). These invasive species are a great threat to island biodiversity, and for example, domestic cat feral populations are nowadays found on most major island groups, from subantarctic to arid regions, and are known to be the direct cause of severe reductions or extinctions of numerous populations of insular vertebrate species (Courchamp et al. 2003).
Ferrets, weasels, mongooses and feral cats vary in size and habits, and, as the case for a great majority of carnivores, are elusive and highly mobile animals, have a solitary life-style, live in low-density populations, occupy vast home ranges, and are often nocturnal. These characteristics translate into low opportunities to observe individuals, expected low detection rates, substantial field effort, and the need for a combination of transect-based and station-based noninvasive sampling techniques to maximize species detection.
Sign surveys are among the most widely used noninvasive field techniques to survey carnivores, based on the assumption that each species produces distinctive tracks and scats. These are conducted along transect trails in adequate soil conditions that can be standardized by time or length of search, to allow relative comparison between sampling sites/events on the basis of the number of tracks/scats per time/space unit (Wilson & Delahay 2001). However, although signs of presence are recognized as good indicators of carnivore presence, increasingly evidence on species assignment shows frequent misinterpretation on the basis of scats size, shape and odour (Monterroso et al. 2013). These findings support the need to use molecular markers to determine species identifications, particularly when diverse communities of similar-sized or phylogenetically related species are present, such as the case of insular ferrets and weasels. Noninvasive genetic sampling of carnivores is a tool that is gaining importance in monitoring, as scats are relatively easy to find, provide large samples, and allow species, gender and individual identification, giving good estimates of population abundance and structure (Kelly et al.
2012). However, being costly, it should mostly focus on species with potentially confounding scats (Fernandes et al. 2008).
Trapping via digital infrared remotely triggered cameras is a non-invasive station-based technique that is becoming a mainstream tool in carnivore studies (Rowcliffe & Carbone 2008) and an increasing body of literature is available about camera trapping techniques (models, trap placement, baits/lures, etc – e.g. McCallum 2013, Ferreras et al. 2017) and survey design requirements to meet statistical assumptions allowing robust occupancy and abundance analyses (Boitani & Powell 2012). For inventory purposes, camera trapping is probably the most efficient method for cryptic animals such as carnivores, but for abundance estimates it relies on individual recognition through coat colour patterns. Although this is not the case for weasels, mongooses, and most ferrets (depending on the variability of facial markings), which cannot be individually recognized, it is an appropriate and a widely used tool to estimate feral cat densities (Bengsen et al. 2012), and may also apply to dogs, which together probably have the greatest impact on island biodiversity (Medina et al. 2011).
For small-sized species, in particular those such as weasels, which avoid moving in open areas, camera-trapping is not the appropriate method as very low detection rates are expected, and track plates are the best option for inventorying purposes in forested habitats. Also a station-based approach, track-plates are carbon-blackened aluminium sheets that are partially covered with a white contact paper that captures animal footprints, thus allowing species identification. These devices may be baited or lured at the back (to increase attractiveness) and, in high cover habitats should be enclosed in a wooden box or a half plastic tube to avoid plant debris obscuring or destroying footprints (for details see Zielinski & Thomas 1995). Track-plates have proved to be efficient in detecting 11 species of small/medium sized carnivores, including weasels and feral cats (Loukmas et al. 2003). An additional advantage is that track plates can capture fine detail in the footprints, recording rows of dots corresponding to tiny papillae on the animal’s metacarpal pad, such as possessed by e.g. several mustelids (e.g., Herzog et al. 2007).
If the selection of appropriate techniques to survey island forest carnivores is relatively easy, the same does not apply to which field protocol should be implemented, as none can give an optimum number of sampling units, nor a sampling frequency or distance between sampling units to be implemented in a given site, because no single
numbers can be universally effective for any species and in all situations, especially if no information is available about the average home range of the target species. For carnivore surveys, however, too few or too short sampling sessions may fail to detect animals, and too close sampling units may violate the spatial independence assumption and produce biased occupancy patterns or abundance estimates due to spatial autocorrelation. A trade-off is therefore needed taking into account size of the island, species and local contexts, and logistic and financial constraints. The rule, however, is to have as many sampling sites as possible, and some hints can be given as to how best to conduct multi-species assessments: i) use a combination of techniques that maximises detection of carnivore taxa potentially occurring in the island, ii) select a sampling unit large enough to include several of the entire home ranges of the smallest species, iii) place sampling stations at sites where detections are most likely to occur, iv) install more than one transect/device per sampling unit to increase detection probability and/or avoid failure due to unpredictable events (e.g., extreme weather events, camera mechanical failure, vandalism), and v) replicate sampling events according to taxa life-history (e.g., pre and post-breeding seasons), resource seasonality or disturbance drivers (e.g., forest production cycles).
For a comprehensive and more detailed review of challenges, planning issues and methodological/analytical concerns when surveying carnivore populations, see e.g., Boitani et al. (2012).

PROTOCOL 10 – Herbivores inventory / monitoring
Taxa – Medium-sized and large herbivores
Experimental design – Similarly to what has been described for the carnivores sampling, depending on the island size, a regular or random distribution of sampling points (small islands) or a spatial sub-sampling (large islands), involving or not involving stratification (depending on the degree of forest heterogeneity), should be used. To maximize species detection (including rare species) the combined use of direct observations, sign surveys, and, eventually camera-trapping stations, is advisable. Depending on the diversity and similarity of species known to inhabit the focal islands, and its detectability (related with species size, forest closeness and understory thickness), signs of presence may be the only
indicator of herbivore presence, and the complementary use of molecular markers may be needed.

Frequency – For inventory purposes at least one sampling event in the high abundance peak (post –breeding), while for monitoring purposes a minimum interval of five years is adequate to assess community changes and/or population trends.

Description
Medium-sized and large herbivores are species weighing more than one kilogram including lagomorphs (rabbits and hares) and wild and domestic ungulates, found in many islands. Line transects, using either direct sightings or indirect signs (tracks and/or faecal pellets) provide the traditional survey method for these taxa. However, this method may be inefficient in dense vegetation, and is difficult to standardize among heterogeneous, fragmented areas where the small size of patches restricts the length of transects (Espartosa et al. 2011). This being the case, in forested habitats a combination of line transects with camera trapping (particularly for rare and nocturnal species) comprises the most appropriate field protocol.
Faecal and track sampling may be used to estimate species site occupation and, if regularly counted along pre-defined transects, may indicate a relative index of abundance, provided that in the first instance the transect is cleared of old signs. Observations on the rates of accumulation of faecal deposits in sample areas of ground may provide estimations of population density if average defecation frequencies per species are known (e.g., Mitchell et al. 1985).
Rabbit-census methods have been the focus of several studies that compared different field protocols based either on systematic spotlight counts conducted at night or at dawn and dusk (e.g., Poole et al. 2003) or pellet-counts along foot transects. According to Fernandez-de-Simon et al. (2011), a standard methodology based on cleared-plot pellets counts should be used to monitor rabbit abundance on a large scale.
As described in the carnivores protocol, selecting the appropriate survey technique(s) is a comparatively easy task, but the same does not apply to the field protocol, with number of sampling units, length of transects, and sampling frequency depending on island size, composition and configuration of forested habitats, and herbivores community composition. See carnivore protocol description for hints regarding field surveys.

References
Boitani L, Powell RA (2012) Carnivore ecology and conservation: a handbook of techniques. Oxford University Press, UK.
Boitani L, Ciucci P, Mortelliti A (2012) 2. Designing carnivore surveys. In: Boitani L, Powell R (eds.). Carnivore ecology and management: a handbook of techniques.Oxford University Press, Oxford, United Kingdom. pp. 8-29.
Courchamp F, Chapuis JL, Pascal, M (2003) Mammal invaders on islands: impact, control and control impact. Biol Rev 78:347-383. doi:10.1017/S1464793102006061
Espartosa KD, Pinotti BT, Pardini, R (2011) Performance of camera trapping and track counts for surveying large mammals in rainforest remnants. Biod Conserv 20:2815-2829. doi:10.1007/s10531-011-0110-4
Fernandes CA, Ginja C, Pereira I, Tenreiro R, Bruford MW, Santos-Reis M (2008) Species-specific mitochondrial DNA markers for identification of non-invasive samples from sympatric carnivores in the Iberian Peninsula. Conserv Genet 9:681-690. doi:10.1007/s10592-007-9364-5
Fernandez-de-Simon J, Díaz-Ruiz F, Cirilli F, Tortosa FS, Villafuerte R, Delibes-Mateos M, Ferreras P (2011) Towards a standardized index of European rabbit abundance in Iberian Mediterranean habitats. Eur J Wildl Res 57:1091-1100. doi:10.1007/s10344-011-0524-z
Hays WS, Conant S (2007) Biology and impacts of Pacific Island invasive species. 1. A worldwide review of effects of the small Indian mongoose, Herpestes javanicus (Carnivora: Herpestidae). Pac Sci 61:3-16.
Herzog CJ, Kays RW, Ray JC, Gompper ME, Zielinski WJ, Higgins R, Tymeson M (2007) Using patterns in track-plate footprints to identify individual fishers. J Wildl Manage 71:955-963. doi:10.2193/2006-408
Jones C, McShea WJ, Conroy MJ, Kunz, TH (1996) Capturing mammals. In: Wilson DE, Cole FR, Nichols JD, Rudran, R, Foster, MS (eds.). Measuring and Monitoring Biological Diversity, Standard Methods for Mammals. Smithsonian Institution Press, Washington, pp. 115-156.
Justiniano R, Schenk JJ, Balete DS, Rickart EA, Esselstyn,JA, Heaney LR, Steppan SJ (2015) Testing diversification models of endemic Philippine forest mice (Apomys) with nuclear phylogenies across elevational gradients reveal repeated colonization of isolated mountain ranges. J Biogeo 42:51-64. doi:10.1111/jbi.12401
Kelly MJ, Betsch J, Wultsch C, Mesa B, Mills LS (2012) 4. Noninvasive sampling for carnivores. In: Boitani, L, Powell RA (eds.). Carnivore ecology and conservation: a handbook of techniques. Oxford University Press, New York. pp. 47-69
Kunz TH (1988) Ecological and behavioral methods for the study of bats. Smithsonian Institution Press, Washington.
Kunz TH, Thomas,DW, Richards GC, Tiedemann CR, Pierson ED, Racey, PA (1996) Observational techniques for bats. In: Wilson DE, Cole FR, Nichols JD, Rudran R, Foster MS eds.). Measuring and monitoring biological diversity, Standard Methods for Mammals. Smithsonian Institution Press, Washington, pp. 105-114.
Lawlor TE (1986) Comparative biogeography of mammals on islands. Biol J Linnean Soc 28:99-125. doi:10.1111/j.1095-8312.1986.tb01751.x
Loukmas JJ, Mayack DT, Richmond, ME (2003) Track plate enclosures: box designs affecting attractiveness to riparian mammals. Am Midl Nat 149:219-224.
Mathias ML (1988) An annotated list of the mammals recorded from the Madeira islands. Bol Museu Mun Funchal 40:111-137.
McCallum J (2013) Changing use of camera traps in mammalian field research: habitats, taxa and study types. Mammal Rev 43:196–206. doi:10.1111/j.1365-2907.2012.00216.x
Medina FM, Bonnaud E, Vidal E, Tershy BR, Zavaleta ES, Josh Donlan,C, Nogales M (2011) A global review of the impacts of invasive cats on island endangered vertebrates. Glob Change Biol 17:3503-3510. doi:10.1111/j.1365-2486.2011.02464.x
Mitchell B, Rowe JJ, Ratcliffe P, Hinge M (1985) Defecation frequency in roe deer (Capreolus capreolus) in relation to the accumulation rates of faecal deposits. J Zool 207:1-7. doi:10.1111/j.1469-7998.1985.tb04910.x
Monterroso P, Castro D, Silva TL, Ferreras P, Godinho R, Alves PC (2013) Factors affecting the (in) accuracy of mammalian mesocarnivore scat identification in South-western Europe. J Zool 289:243-250. doi:10.1111/jzo.12000
Poole DW, Cowan DP, Smith GC (2003) Developing a census method based on sight counts to estimate rabbit (Oryctolagus cuniculus) numbers. Wildl Res 30:487-493. doi:10.1071/WR0201
Rickart EA, Heaney LR, Balete DS, Tabaranza Jr,BR (2011) Small mammal diversity along an elevational gradient in northern Luzon, Philippines. Mamm Biol 76:12-21. doi:10.1016/j.mambio.2010.01.006
Rickart EA, Balete DS, Alviola PA, Veluz MJ, Heaney LR (2016) The mammals of Mt. Amuyao: a richly endemic fauna in the Central Cordillera of northern Luzon Island, Philippines. Mammalia 80:579-592. doi:10.1515/mammalia-2015-0132
Rodrigues M, Bo, AR, Schembri PJ, de Lima RF, Lymberakis P, Parpal L, Cento M, Ruette S, Ozkurt SO, Santos-Reis M, Merilä J, Fernandes, C (2017) Origin and introduction history of the least weasel (Mustela nivalis) on Mediterranean and Atlantic islands inferred from genetic data. Biol Invasi 19, 399–421. doi: 10.1007/s10530-016-1287-y
Rowcliffe JM, Carbone C (2008) Surveys using camera traps: are we looking to a brighter future? Anim Conserv 11:185-186. doi:10.1111/j.1469-1795.2008.00180.x
Rudran R, Kunz TH, Southwell C, Jarman P, Smith AP (1996) Observational techniques for nonvolant mammals. In: Wilson DE, Cole FR, Nichols JD, Rudran R, Foster MS (eds.). Measuring and monitoring biological diversity, Standard Methods for Mammals. Smithsonian Institution Press, Washington, pp. 81-104.
Sikes RS and the Animal Care and Use Committee of the American Society of Mammalogists (2016) Guidelines of the American Society of Mammalogists for the use of wild mammals in research and education. J Mammal 97:663-688. doi:10.1093/jmammal/gyw078
Van Daele LJ (2003) The history of bears on the Kodiak Archipelago. Alaska Natural History Association. Anchorage, Alaska, U.S.
Wilson DE, Cole FR, Nichols JD, Rudran R, Foster MS (eds.) (1996) Measuring and Monitoring biological diversity, Standard Methods for Mammals. Smithsonian Institution Press, Washington.
Wilson GJ, Delahay, RJ (2001) A review of methods to estimate the abundance of terrestrial carnivores using field signs and observation. Wildl Res 28:151-164. doi:10.1071/WR00033
Zielinski WJ, Thomas E (1995) American marten, fisher, lynx, and wolverine: survey methods for their detection. Gen. Tech. Rep., PSW-GTR-157. Forest Service, U.S. Department of Agriculture, 163 p.