Spatial and temporal variation in the diet of introduced sambar deer (Cervus unicolor) in an alpine landscape

Study species

Sambar deer are a large-bodied ungulate species native to the tropical and temperate ecosystems of India, southern China and south-eastern Asia (Leslie 2011), and were first introduced into south-eastern Australia in the mid-to-late 1800s (Forsyth et al. 2015). A cautious and secretive species, sambar deer are generally solitary (Harrison 1998) and prefer densely forested ecosystems, with a tendency to emerge from these areas at night for foraging (Eisenberg and Lockhart 1972). In 1941, the first reports of sambar deer in the Victorian alpine bioregion were documented at Mount Howitt, and as of 2015, sambar deer were estimated to have established wild, self-sustaining breeding populations covering approximately 29% of Victoria’s land area (Forsyth et al. 2015). Sambar deer are listed under the Flora and Fauna Guarantee Act 1988 as a ‘potentially threatening process’ to biodiversity in the state of Victoria yet are also protected as wildlife under the Victorian Wildlife Act 1975, due to their value as game species (Davis et al. 2016).

Study area

The study was conducted on the Bogong High Plains (BHP), within the Alpine National Park, approximately 240 km north-east of Melbourne, Victoria, Australia (36°50′S, 147°20′E). The BHP occupy an area of approximately 120 km² and comprise vegetation communities ranging from woodlands, open and closed-heathlands, tussock grasslands, wetlands and snowpatch herbfields (McDougall 1982). The area experiences mean annual minimum and maximum temperatures of 2.7°C and 9.5°C respectively, and a mean annual precipitation of 2430 mm, much of which falls as snow in winter (June–August) (Lawrence 1995; Giljohann et al. 2011). The BHP are snow-covered from June to September, and the potential for ground frost is highest during the months of April, May and October. Sambar deer are the only species of deer observed on the BHP, inhabiting the area following snowmelt in late spring and remaining until the beginning of winter, when they descend to lower elevations (Downes 1983; Watter et al. 2020; Comte et al. 2022).

Two ecologically distinct sites on the BHP were selected for sampling and are separated by <10 km. The alpine Mt Nelse study site (36°50′06″S, 147°20′21″E) is 1884 m above sea level (a.s.l) and experiences long periods of late-lying snow following winter (Wahren et al. 2001). The sub-alpine Basalt Hill study site (36°53′24″S, 147°18′42″E) is 1640 m a.s.l and has been described as discordant with the rest of the BHP due to differing underlying geomorphology and plant species diversity (McDougall 1982; Wahren et al. 2001). The Basalt Hill location has a pronounced heterogenous landscape, comprising a mixture of sub-alpine vegetation communities, all in close proximity. At each of the study sites, collection area for faecal pellet samples was <5 km².

Within each site, we concentrated on collecting samples from three different vegetation types: snowpatch herbfields, grasslands and wetlands. Snowpatch herbfields are one of the rarest vegetation types in Australia (Wahren et al. 2001; Williams et al. 2015) and are typically dominated by low growing forbs, sedges and grasses due to the shortened growing season at high elevation (Burrows 1977; Wahren et al. 2001; Venn and Morgan 2007; Williams et al. 2015). Wetland ecosystems include the Alpine Sphagnum Bogs and Associated Fens ecological community, currently listed as endangered under the Commonwealth Environment Protection and Biodiversity Conservation Act 1999. These communities are of high importance hydrologically and are particularly sensitive to grazing and trampling by ungulate species (McDougall 1982; Williams and Costin 1994). Grasslands are widespread and diverse across the alpine landscape, but high levels of deterioration within these communities has been observed due to previous grazing practices, which has resulted in the establishment of a range of invasive plants (Carr and Turner 1959; McDougall et al. 2005).

Sample collection

To assess spatial and temporal variation in sambar deer diet, faecal pellets were collected monthly (January to March), from within each of the selected vegetation types at the two study sites. Because sambar deer expel many faecal pellets in one defecation, we define a faecal pellet group as a collection of pellets that varies in the total number of individual units yet is produced by only one sambar deer. In total, 90 faecal pellet group samples were collected (n = 45 at each of the study sites, comprising 15 samples from each month across the three vegetation communities). Grasslands have a higher percentage of cover compared with wetlands on the Bogong High Plains (McDougall 2003), and snowpatch habitats are much smaller in area, occupying less than 1% of the overall alpine landscape (Wahren et al. 2001). We therefore stratified the sample collection across each month and within each site, by collecting six groups of faecal pellets from grasslands, five groups from wetlands and four groups from snowpatch herbfields. We opportunistically surveyed the same vegetation types within each site monthly and only collected fresh samples, described as those that were still brown and exhibited a wet shine with no outer decay (Mayze and Moore 1990). Pellet groups that exhibited clear signs of contamination from non-food items, such as leaf or plant tissue, were ignored. Each faecal pellet group sample consisted of approximately 10–12 faecal pellets, which were stored in a plastic zip-lock bag and placed in −30°C freezers until processing.

Plant reference library

We collected tissue samples of plant species that could be consumed by sambar deer in each of the study sites yet lacked available reference sequences on the National Centre for Biotechnology Information (NCBI) GenBank nucleotide database. We used these tissue samples to generate a plant DNA reference library. Approximately 100 mg of tissue was collected from each of the selected plant species (n = 63), and a DNeasy Plant Mini Kit (Qiagen) was used to perform DNA extractions. Plant DNA extractions were undertaken in a Polymerase Chain Reaction (PCR)-free laboratory separate to the faecal pellet extractions to prevent contamination.

Faecal DNA extraction and PCR amplification

Each faecal pellet group sample was homogenised and subsampled three times, resulting in a total of 270 samples for DNA extraction. DNA extractions were performed with ~100 mg of faecal sample using a QIAmp Fast DNA Stool Mini Kit (Qiagen) following manufacturers’ protocols and were eluted in a final volume of 200 μL. Throughout the process, negative extractions containing no faecal material were performed to monitor potential contamination.

Each DNA extraction (faecal and plant) was amplified using the ITS2 gene region of nuclear ribosomal DNA, using the established primer set ITS2-S2F and ITS4 (White et al. 1990; Chen et al. 2010). The robustness of the ITS2 gene region for analysing plant material that has undergone digestion and excretion has been highlighted by Moorhouse-Gann et al. (2018). Amplification of the ITS2 region was carried out in a final volume of 15 μL, using 2 μL of DNA extract as a template, 6.25 μL of MyTaq™ Red Mix (Bioline), 0.5 μL of each forward and reverse primer and 5.75 μL of H₂O. The mixture was denatured at 94°C for 5 min, followed by 35 cycles of 30 s at 94°C, 30 s at 50°C and 45 s at 72°C. A final extension of 72°C for 10 min completed the PCR amplification stage. In cases where samples failed to amplify, DNA samples were diluted at 1/10 and 1/100 rates with H₂O until amplification was successful. PCRs included both negative extraction and H₂O controls to test for sample contamination.

Illumina adaptors and indexing primers were attached following the first round of PCRs, and these reactions comprised 6 μL of H₂O, 6.25 μL of MyTaq™ Red Mix (Bioline), 0.75 μL of the index primer pair (Illumina) and 2 μL of DNA PCR template for a 15 μL reaction. Conditions for this PCR consisted of an initial denaturation period of 95°C for 3 min, followed by 10 cycles of denaturation at 95°C for 20 s, annealing at 60°C for 15 s and elongation at 72°C for 30 s, with a final extension step at 72°C for 3 min. PCR products were cleaned using an SPRI magnetic bead mix to remove non-specific DNA fragments. Preparation for sequencing was undertaken following the Illumina MiSeq protocol, using a MiSeq Reagent v2 (2 × 250 bp) sequencing kit.

Sequence analysis

The sequence analysis tool USEARCH v11.0.667 (Edgar 2010) was used to generate a sequence list of Operational Taxonomic Units (OTUs), which were then analysed to investigate the diet of sambar deer. The fastx_truncate function was used to remove the primers, and sequences were then filtered based on length using the fastq_filter function, to remove sequences <200 bp. Contaminants and singleton sequences were removed using the fastx_uniques function, and the -unoise3 function was used to generate OTUs. Any OTUs with fewer than three reads in a sample were removed from the dataset for analysis, and any OTU present in negative extractions and H₂O controls were also discarded. After the completion of filtering, OTUs that were derived from each of three extraction subsamples were merged back into one single sample.

The final list of OTUs were first compared with our plant species reference sequence database and, if no match was found, compared with the National Centre for Biotechnology Information (NCBI) GenBank nucleotide database, using BLAST (Basic Local Alignment Search Tool; Altschul et al. 1990) with default parameters. We assigned all sequences to the lowest taxonomic level possible and considered correct assignment to species when a ≥97% match of a reference sequence to the full length of the OTU sequence was observed, following the convention that ≥97% similarity corresponds to species level (Schloss and Handelsman 2005). Likewise, ≥95% similarity was required for assignment to genus level and ≥90% for family level (Elbrecht et al. 2017). The Atlas of Living Australia (www.ala.org.au) and Royal Botanic Gardens Victoria VicFlora (www.vicflora.rbg.vic.gov.au) databases were used to ensure that plant species, genera or families identified from the OTUs had been previously recorded within or surrounding the study sites.

Statistical analysis

To determine whether the diet of sambar deer on the BHP had been sufficiently sampled, the cumulative richness of all detected OTUs, genera and families in each study site was plotted against the number of faecal samples using the ‘vegan’ package in R (R Development Core Team 2010; Oksanen et al. 2013). Differences in the number of OTUs sequenced each month were compared using a non-parametric Kruskall-Wallis H Test (Kruskal 1952).

The lowest taxonomic resolution observed for a majority of the detected OTUs was at the genus level (≥95% similarity), so much of the dietary preferencing analysis was performed at this taxonomic rank. All identified OTUs were grouped at the genus level and the frequency of occurrence was plotted for each detected genus. In accordance with Nawaz et al. (2019), we considered any genera that were present in ≥50% of all samples overall to be frequent plant dietary items, any genera that were observed in ≥10% of samples were regular dietary items and those with <10% occurrence were rare dietary items. To investigate dietary preferences, we determined if the genera detected across all samples occurred in sambar diet more or less frequently relative to their presence in the BHP plant records. Observation data were obtained from the online Atlas of Living Australia database by extracting all plant observation records within a 10-km radius from a location at the midpoint between both study sites. This midpoint included all genera available across Mt Nelse and Basalt Hill. A genus was judged as ‘preferred’ if its proportional representation across faecal samples within the study site was greater than its proportional representation in the environment as judged by ALA records (total records for target genera/total records for all genera).

To assess the feeding behaviour of sambar deer, we labelled each identified plant OTU as one of three plant growth forms (Forb, Woody, Graminoid), using information from the online Royal Botanic Gardens Victoria VicFlora database. The combined sequencing read depth of each growth form (calculated by performing the sum of all sequence reads for each OTU and grouping by growth form) across the 3 months and within each site was compared using a multivariate analysis of variance (MANOVA; O’Brien and Kaiser 1985), calculated using a Pillai trace statistic in the ‘car’ package in R (Olson 1976; Fox et al. 2007). The number of sequence reads for multiple dietary items can indicate the relative amounts of each of these items contained within a sample; therefore, sequence read depth data is used here to provide a proxy measure of overall plant biomass from each of the three plant growth forms consumed by sambar deer (Kowalczyk et al. 2011; Elbrecht and Leese 2015; Thomas et al. 2016; Deagle et al. 2019). Although the use of sequence read depth data for dietary analysis can be considered controversial due to a number of potential biases (Deagle et al. 2019), we cross-validated the results with the same analysis using cumulative OTU richness in addition to sequence read depth.

To examine the plant species composition of faecal samples among both sites and vegetation communities across different months, we performed a non-metric multi-dimensional scaling (nMDS) ordination using Bray-Curtis distance measures on presence/absence data in ‘vegan’ (Bray and Curtis 1957) and generated the ordination plot using the ‘phyloseq’ package in R (McMurdie and Holmes 2013). Stress values were used to evaluate the goodness-of-fit of the nMDS ordination, with values >0.3 considered to signify unsatisfactory representation of the data (Clarke and Warwick 2001). Analysis of similarities (ANOSIM) using ‘vegan’ was performed to test for significant differences in faecal pellet composition between sites (Mt Nelse or Basalt Hill), months (January, February or March) and vegetation communities (Snowpatch herbfield, grassland or wetland). The ‘indicspecies’ package in R was used to identify plants that were significantly associated with month, site or vegetation community, using the ‘multipatt’ function (De Cáceres 2013). The statistical significance of these associations was determined using a permutation test, with 9999 permutations.

Diet and dietary preferences

The plant genera detected in the diet of sambar deer were not evenly represented – the frequency of occurrence across faecal samples ranged from 1 to 81%. Eleven plant genera were represented in only one faecal sample, and approximately 60% of all genera were found in 10 or fewer samples. The combined regular and frequent plant diet (≥10% occurrence) of sambar deer comprised 34 genera, of which 20 were forbs, 11 were shrubs and three were graminoids (Fig. 2). Forbs dominated the diet, both in terms of number of taxa detected and frequency of occurrence across faecal samples. Only three genera were noted as comprising the frequently consumed portion of the diet (≥50% occurrence): Acaena, Eucalyptus and the introduced Acetosella.

Fig. 2.

Frequency of occurrence for each genus detected in the diet of sambar deer across the two study sites on the BHP. The growth form for each genus is represented by one of three different colours. Dashed lines separate frequent, regular and rare dietary items.

A number of genera were detected more often from faecal samples than would be expected based on plant observation records. These genera were therefore considered to be over-represented in the diet and included the native forbs Acaena, Gonocarpus, Galium and Epilobium, the native woody genus Eucalyptus and the introduced forbs Acetosella and Hypochaeris (Fig. 3). Conversely, several genera were under-represented in the diet, in comparison with their presence in plant observation records, including the most commonly observed plant genus within the entire study area (Poa) and two other graminoids (Carex, Luzula).

Fig. 3.

Relative occurrence of plant genera in 90 faecal samples and number of observations within the study area based on observations from the Atlas of Living Australia online database. The black dashed line indicates the equal detection of plants in faecal samples relative to observation counts in the landscape. Points above the line indicate higher occurrence of genera in the faecal samples than in the landscape, and points below the line indicate lower occurrence in the samples than would be predicted from recorded observations. Labels are included for the genera that were recorded in more than 400 observations and/or were detected in ≥37.5% of samples.

Of the OTUs that were identified to species level, the most frequently detected across all faecal samples were Acaena novae-zelandiae, Eucalyptus pauciflora, *Acetosella vulgaris, Poa costiniana, *Hypochaeris radicata and Gonocarpus montanus (asterisk here and throughout denotes non-native species). These species, as well as Coprosma hirtella, Oreomyhrris eriopoda, Scleranthus biflorus, Phebalium squamulosum subsp. ozothamnoides and *Salix cinerea, were also abundant in terms of overall sequencing read depth (each with a total read depth >5000).

Nine species that were detected in the diet are currently considered rare in the state of Victoria, (although all but one (Tasmannia vickeriana) are present in reasonable numbers in the study area). Four of these species were detected in few samples, or had low sequence read depths, suggesting that they comprise a minor component of the diet. These species were Tasmannia vickeriana (five samples, average read depth <70), Carpha nivicola (one sample, average read depth <40), Psychrophila introloba (one sample, average read depth <10) and Celmisia sericophylla (one samples, average read depth <5), which is currently listed as critically endangered under the Flora and Fauna Guarantee Act 1988. In comparison, five rare species were either detected in an increased number of samples or had high sequence read depths, suggesting that these species appear more regularly in the diet than would be expected. These species were Phebalium squamulosum subsp. ozothamnoides (29 samples, average read depth <300), Carex canescens (15 samples, average read depth <15), Ranunculus eichlerianus (10 samples, average read depth <15), Bossiaea foliosa (seven samples, average read depth <350) and Pimelea axiflora subsp. alpina (four samples, average read depth <200).

Temporal variation

There was no difference in the mean number of OTUs across the two study areas in each of the 3 months (χ²(2) = 0.828, d.f. = 2, P = 0.661), suggesting that the number of dietary items detected in the diet of sambar deer was similar throughout the study period. However, distinct monthly variation was observed in the sequence read depth of plant growth forms in Basalt Hill samples (Pillai’s trace = 0.411, F_6,82 = 3.54, P < 0.01) but not in Mt Nelse (Pillai’s trace = 0.226, F_6,82 = 1.743, P = 0.121; Fig. 4). These results were supported by similar analysis using cumulative OTU richness in addition to sequence read depth (Basalt Hill samples; Pillai’s trace = 0.541, F_6,82 = 5.064, P < 0.01, and Mt Nelse samples; Pillai’s trace = 0.254, F_6,82 = 1.993, P = 0.076). The sequence read depth of woody dietary items increased significantly in Basalt Hill samples in February (F_2,42 = 5.62, P < 0.01), yet the contribution of forbs remained high across all months in the Mt Nelse samples. Additionally, we observed no difference in the sequence read depth of graminoids across the 3-month study period in Basalt Hill (F_2,42 = 1.339, P = 0.271) or Mt Nelse (F_2,42 = 2.685, P = 0.079).

Fig. 4.

Monthly variation in the diet of sambar deer in terms of sequence read depth of respective growth forms.

ANOSIM analysis comparing the plant species composition of faecal pellets across the 3-month sampling period resulted in no significant differences among the Mt Nelse samples (R = −0.015, P = 0.636); however, there was significant variation observed in Basalt Hill samples (R = 0.249, P = 0.001; Supplementary Material). From the indicator taxa analysis of Basalt Hill samples, the woody genus Asterolasia was more associated with February samples (statistic = 0.527, P = 0.033), and the forb genus Leptinella was more associated with March samples (statistic = 0.527, P = 0.035). The genera Eucalyptus (statistic = 0.858, P < 0.01), Gonocarpus (statistic = 0.778, P < 0.01), Phebalium (statistic = 0.761, P = 0.01) and Veronica (statistic = 0.632, P = 0.01) were significantly associated with the February and March samples, but not January. We also observed temporal variation in the plant species composition of faecal samples collected from wetland sites at Mt Nelse (R = 0.264, P = 0.022) and from grassland sites at Basalt Hill (R = 0.259, P = 0.012). Indicator taxa analysis suggests that the genera contributing to this temporal variation in wetlands at Mt Nelse were Oreomyhrris (statistic = 0.8, P = 0.049) and Scleranthus (statistic = 0.8, P = 0.049), and that the genera contributing to temporal variation in grasslands at Basalt Hill were Eucalyptus (statistic = 0.907, P = 0.019), Gonocarpus (statistic = 0.876, P < 0.01), Galium (statistic = 0.828, P = 0.023), Phebalium (statistic = 0.828, P = 0.025) and Veronica (statistic = 0.707, P = 0.044).

Spatial variation

A significant difference was observed between the plant species composition of samples from Mt Nelse and Basalt Hill (R = 0.242, P = 0.01; Fig. 5), suggesting differences in sambar deer diet between the two sites. Twelve genera were significantly more associated with samples collected from Basalt Hill: Rubus (statistic = 0.760, P < 0.01), Poa (statistic = 0.713, P < 0.01), Phebalium (statistic = 0.672, P < 0.01), Galium (statistic = 0.659, P < 0.01), Stellaria (statistic = 0.581, P < 0.01), Coprosma (statistic = 0.562, P < 0.01), Trifolium (statistic = 0.516, P < 0.01), Veronica (statistic = 0.516, P < 0.01), Agrostis (statistic = 0.494, P < 0.01), Geranium (statistic = 0.447, P < 0.01), Tasmannia (statistic = 0.447, P < 0.01) and Cerastium (statistic = 0.422, P < 0.01). Seven of the genera that were significantly associated with samples collected from Basalt Hill were only detected at this site (Agrostis, Cerastium, Geranium, Rubus, Tasmannia, Trifolium and Veronica). The remaining five genera (Coprosma, Galium, Phebalium, Poa and Stellaria) were also detected in samples from Mt Nelse, but in fewer samples compared with Basalt Hill. Although several genera were only detected in samples collected from Mt Nelse (Acrothamnus, Caltha and Lolium), each of these genera were detected in three or fewer samples, so were not significantly associated with the Mt Nelse site.

Fig. 5.

Non-metric multi-dimensional scaling (nMDS) comparison of plant species composition of sambar deer faecal pellet samples collected from Basalt Hill and Mt Nelse, using Bray–Curtis dissimilarity based on presence/absence data (stress = 0.19).

Differences were observed between sites for direct comparisons of snowpatch herbfields (R = 0.33, P = 0.001), grasslands (R = 0.237, P = 0.001) and wetlands (R = 0.212, P = 0.011). However, comparisons between vegetation communities within individual study sites resulted in non-significant results for Mt Nelse (R = 0.015, P = 0.619) and Basalt Hill (R = 0.052, P = 0.124). Additionally, comparisons among vegetation communities using all samples resulted in no difference (R = 0.009, P = 0.294), suggesting that within and among each study site, sambar deer diet contains a mixture of items from multiple vegetation communities.

Overall diet

Although sambar deer have been described as grazers, browsers or intermediate feeders across their range (Varman and Sukumar 1993; Semiadi et al. 1995; Padmalal et al. 2003; Forsyth and Davis 2011), our study demonstrates that during the main flowering period in alpine environments, sambar deer exhibit an intermediate feeder diet mostly comprising forbs and shrubs, and they appear to adjust the composition of dietary items to adapt to food quality, quantity and availability. Forbs were the dominant dietary items, and several genera were frequently consumed by sambar deer (i.e. Acaena, Eucalyptus and *Acetosella). Other dietary studies of introduced deer in south-eastern Australia have reported similar dominance of forbs and shrubs in the diet of hog deer (Axis porcinus), red deer (Cervus elaphus) and sambar deer in a variety of environments ranging from coastal grassy woodlands to wet forests (Davis et al. 2008; Forsyth and Davis 2011; Roberts et al. 2015). In alpine environments, it is hypothesised that ungulates display opportunistic feeding behaviours due to the shortened growing season (Schaller and Gu 1994). Our results support this − the high contribution of forbs to the overall diet of sambar deer is likely a function of forb abundance in the alpine landscape (McDougall 1982). This may suggest that sambar deer are eating in response to what is locally available and therefore exhibiting feeding behaviours modified to the environments they inhabit (Hofmann and Stewart 1972). However, equally plausible is that the high contribution of forbs in the diet is also explained from a nutritional perspective, because forbs are of high nutritional quality (high metabolisable energy) and have the additional benefit of retaining their quality throughout the summer period while the quality of other plant growth forms decline (Holechek 1984; Shrestha et al. 2005; Goldberg et al. 2020).

Graminoids were not a main dietary item of sambar deer at either study site, or across the summer season. However, in Mt Nelse samples, graminoid abundance did increase towards the end of the summer season as the abundance of shrubs in diets declined. Stafford (1997) noted a dominance of grasses in the diet of sambar deer in New Zealand, a result that has been observed in other study systems (Kelton and Skipworth 1987; Padmalal et al. 2003). Although grasses, sedges and rushes represent an easily palatable and accessible supply of energy, (Jones and Wilson 1987; Frase and Armitage 1989), on the BHP, the intake of graminoids may be observed as only a minor component in the diet of sambar deer due to the availability of other, more nutritious growth forms (Goldberg et al. 2020). Therefore, it is likely that graminoids are only utilised by sambar deer to maintain energy intake and nutrient requirements when other growth forms are limited (Belovsky 1986; Frase and Armitage 1989). Further research on the diet of sambar deer outside of the summer growing season on the BHP could detail the prevalence of these growth forms during the colder months.

We detected several genera more often in faecal samples than would be expected based on their presence in the study area, including Acaena, Eucalyptus, *Acetosella, Gonocarpus, Galium, Epilobium and *Hypochaeris. The targeting of these particular plant genera suggests that sambar deer are not strictly opportunistic feeders in the Alpine National Park and are exhibiting some selective feeding behaviours towards certain plant genera. Selective feeding has the potential to alter the abundance of favoured dietary items within the landscape and is of particular concern when these items are rare or vulnerable (Kirby 2001; Keith and Pellow 2005). Currently, none of the genera detected as over-represented in the diet are considered threatened on the BHP, but their representation may help to understand movement patterns of sambar deer in alpine environments because a number of the over-represented dietary items are more common in the sub-alpine habitats, rather than the alpine sites sampled. In sub-alpine environments, sambar deer display strong crepuscular activity patterns because snow-gum woodlands and closed-heathland environments at lower elevations provide ideal habitat (i.e. cover) for the species during daylight hours (Leslie 2011; Davies et al. 2020). Considering these environments accommodate a majority of the genera observed as over-represented in the diet (McDougall 1982; Williams and Ashton 1987), it is probable that the high frequency of these dietary items is due to the movement of sambar between alpine and sub-alpine vegetation zones during active hours.

Considering the diverse range of plant material eaten by sambar deer in this study, the composition of individual faecal samples would be heavily dependent on the movement, gut passage retention times and foraging patterns of sambar deer prior to defecation (Mautz and Petrides 1971; Milne et al. 1978; Holand 1994; Gill and Beardall 2001; Pellerin et al. 2016). Therefore, as suggested by the presence of sub-alpine plant species in alpine feacal samples, the plant composition of faecal samples is unlikely to be a direct reflection of the specific vegetation community from where it was collected. Our results support the notion that sambar deer are using multiple vegetation communities for foraging, resting and defecating, and could have negative impacts through seed-dispersal of plant species into new communities and trampling of sensitive communities while browsing (Pakeman et al. 2002; Ismail and Jiwan 2015).

We did, however, observe differences between the plant species composition of faecal pellets collected from Mt Nelse and Basalt Hill, which not only suggests differences in the dietary items consumed by sambar deer that utilise these two sites, but may also suggest that sambar deer individuals are unlikely to be moving between these two areas. Mt Nelse and Basalt Hill are less than 10 km apart, which is well within home range estimates of sambar deer males and on the upper limit for females (Sankar 1994; Chundawat et al. 2007). However, despite this relatively short distance, the largely open, treeless nature of the alpine habitat between Mt Nelse and Basalt Hill may present too much of a risk for sambar deer; they are a naturally nervous and cautious species (Leslie 2011). The clear separation in dietary composition between sites suggests that localised feeding and movement behaviours are occurring on the Bogong High Plains. With the exception of sambar deer at high elevations descending to lower environments during winter, the species is typically considered non-migratory across much of its range and has a tendency to remain sedentary within locations (Schaller 1967; Leslie 2011). Further investigation utilising population level genetic techniques would provide better assessment of the movement and dispersal patterns of sambar deer within alpine landscapes and, most importantly, would aid in more effective management of current populations.

Species of conservation importance

Although many of the species detected in the diet of sambar deer are considered widespread in the environment, we detected several species of conservation significance. Rare wetland species such as Broad-leaf Flower rush (Carpha nivicola) and Alpine Marsh-marigold (Psychrophila introloba), and snowpatch species including Silky Snow-daisy (Celmisia sericophylla), which is currently listed as critically endangered under the Flora and Fauna Guarantee Act 1988, were each observed in a small number of samples. Although it appears that sambar deer are not selectively targeting these particular dietary items, the extent of browsing may only need to be minimal to have detrimental impacts on rare species (Davis et al. 2008). Therefore, it is recommended that routine monitoring of the abundance and distribution of rare plant species on the BHP is undertaken to ensure any negative impacts from sambar deer habitation is managed promptly.

We also detected Hawkweed (*Pilosella spp.) and Grey Sallow Willow (*Salix cinerea), both of which are invasive and pose serious threats to native vegetation, particularly snowpatch herbfields and wetlands (Morgan 2000; Karrenberg et al. 2002; Williams et al. 2008; Giljohann et al. 2011; Moore and Thomas 2011; Primrose et al. 2016). Although control programs exist to restrict and eradicate both species, the presence of each within the diet highlights that these species are both available in the landscape for consumption by sambar deer. Additionally, it remains possible that following consumption, sambar deer may facilitate the spread of these species through the processes of endo- or epizoochory (Albert et al. 2015).

It is possible that rare species could be widespread in the diet of sambar deer, but the DNA metabarcoding approach applied in this study may have failed to detect these dietary items due to low amounts of DNA compared with more common plant dietary item DNA (Deagle et al. 2019). Therefore, to better understand the extent of rare items in the diet, whether of conservation significance or invasive potential, a more refined method may be required. A targeted approach using species-specific primers in conjunction with real time quantitative PCR (qPCR) would be feasible for determining the presence of a single genus or species of interest in the diet of sambar deer (Harper et al. 2018) and, most importantly, could aid in tracking the distribution of rare dietary items through sambar deer faecal pellets.

Management implications

In the Alpine National Park, a long history of cattle grazing practices was discontinued in 2005 after their detrimental impacts on native vegetation, soils and water quality were highlighted (Carr 1977; Costin 1977; Van Rees 1982; Wahren et al. 1994; Williams et al. 2006). Since this time, the recovery of impacted plant species and communities has been slow (Driscoll et al. 2019). After cattle removal, sambar deer have become the primary large herbivore observed on the BHP, and our results highlight clear concerns for the management of the alpine plant species and vegetation communities in the presence of sambar deer. Although sambar deer displayed largely opportunistic feeding behaviours in this study, the preferencing of several genera highlights the ability of the species at times to be more selective with their dietary choices (Hanley 1982). The detection of rare, threatened native species and exotic weeds in the diet is of concern and should be carefully evaluated, with monitoring over a longer time period recommended where sambar deer coexist alongside plant species or communities of conservation significance.

The detection of distinct spatial and temporal variation in the diet of sambar deer within high elevation sites of the Alpine National Park provides evidence that individual deer have localised home ranges, which further emphasises the need for land managers to evaluate both the presence and potential impacts of sambar deer at small spatial scales. Further research on the population genetics of sambar deer individuals on the BHP is important, because the utilisation of advancing DNA analysis techniques for estimating population structure and dispersal characteristics would better detail the movement patterns of sambar deer within each of the study sites and, possibly, the broader Alpine National Park (Hampton et al. 2004; Brinkman and Hundertmark 2009; Davies et al. 2020).