Use of carrion fly iDNA metabarcoding to monitor invasive and native mammals

Severely fragmented habitats increase the risk of extirpation of native mammal populations through isolation, increased edge effects, and predation. Therefore, monitoring the movement of mammal populations through anthropogenically altered landscapes can inform conservation. We used metabarcoding of invertebrate‐derived DNA (iDNA) from carrion flies (Calliphoridae and Sarcophagidae) to track mammal populations in the wheat belt of southwestern Australia, where widespread clearing for agriculture has removed most of the native perennial vegetation and replaced it with an agricultural system. We investigated whether the localization of the iDNA signal reflected the predicted distribution of 4 native species—echidna (Tachyglossus aculeatus), numbat (Myrmecobius fasciatus), woylie (Bettongia penicillata), and chuditch (Dasyurus geoffroii)—and 2 non‐native, invasive mammal species—fox (Vulpes vulpes) and feral cat (Felis catus). We collected bulk iDNA samples (n = 150 samples from 3428 carrion flies) at 3 time points from 3 conservation reserves and 35 road edges between them. We detected 14 of the 40 mammal species known from the region, including our target species. Most detections of target taxa were in conservation reserves. There were a few detections from road edges. We detected foxes and feral cats throughout the study area, including all conservation reserves. There was a significant difference between the diversity (F3, 98 = 5.91, p < 0.001) and composition (F3, 43 = 1.72, p < 0.01) of taxa detections on road edges and conservation reserves. Conservation reserves hosted more native biodiversity than road edges. Our results suggest that the signals from iDNA reflect the known distribution of target mammals in this region. The development of iDNA methods shows promise for future noninvasive monitoring of mammals. With further development, iDNA metabarcoding could inform decision‐making related to conservation of endangered taxa, invasive species management, and impacts of habitat fragmentation.


INTRODUCTION
The accelerated loss of vertebrate species across the globe heralds the beginning of the Earth's sixth mass extinction event (Ceballos et al., 2017;Grooten & Almond, 2018;Tilman et al., 2017).Extinctions and decline of ecosystem functionality have been attributed mainly to human activities causing habitat loss (Brodie et al., 2021).Some regions face more extinction threats than others.For example, in Australia, the decline of mammal species represents about one third of all mammal species that have gone extinct in the last 500 years globally (Baillie & Groombridge, 1996).Since European colonization, 22 native Australian mammal species have been driven to extinction (Burbidge & McKenzie, 2009;McKenzie et al., 2007;Woinarski et al., 2011), comprising 6% of Australia's marsupial species (Baillie & Groombridge, 1996).Another 8 species of mammals now only largely exist as populations on a small number of islands or as translocated populations win intensively managed fenced reserves (Burbidge & McKenzie, 2009).
Mammal declines in Australia are attributed primarily to ecological disturbances from human activities: habitat clearing and fragmentation for agricultural use, introduction of domestic animals, changes in burning regimes, hunting and culling of native fauna, and increased predation from introduced predators (Short & Smith, 1994;Woinarski et al., 2019).Attempts have been made to curb species decline and protect existing populations of native animals by creating conservation reserves.However, the establishment of conservation reserves alone is insufficient to maintain biodiversity in an area (Woinarski et al., 2011).Their effectiveness is determined by appropriately managing these protected areas (Watson et al., 2014), including predator control.Furthermore, conservation reserves often face the "island dilemma," whereby if the surrounding area is unsuitable, the number of species the reserve can hold at equilibrium depends on its size and degree of isolation (Diamond, 1975).
In some cases, road edges, the area of public land between private property and roads, can act as dispersal corridors for animal species in agricultural landscapes (Dean et al., 2018;Ding & Eldridge, 2022) because these areas have higher levels of vegetation cover and plant abundance and diversity compared with the surrounding agricultural land (Ding & Eldridge, 2022).Wildlife can use these as corridors to move through disturbed landscapes (Bennett, 1991), which decreases the level of isolation between conservation reserves.However, in the highly fragmented native bush remaining in the wheat belt of Western Australia, little is known about the use of road edges by wildlife as a means of moving between conservation areas (Abensperg-Traun, 1991).
Tracking wildlife use and movement through environments requires direct observation of the species or of the traces they may leave behind (Kouakou et al., 2009) or require radio tracking (Kobryn et al., 2022).However, using these methods over large areas or in remote locations can be time and resource intensive (Campbell et al., 2011).Furthermore, some sampling procedures, such as live trapping, can harm animals (Putman, 1995).Therefore, noninvasive sampling techniques, such as camera trapping, audio recording, or DNA-based detection, can reduce the risks and costs of monitoring wildlife (Gogarten et al., 2020).One DNA-based method is invertebrate-derived DNA (iDNA) metabarcoding.This method has been used to document vertebrate diversity across the globe (Abrams et al., 2019;Calvignac-Spencer, Merkel et al., 2013;Lynggaard et al., 2019;Schnell et al., 2015).Invertebrates, such as leeches, mosquitoes, sand flies, and carrion flies, can be used as sources of vertebrate DNA (Massey et al., 2021) because they interact with vertebrate flesh, blood, or scat and thus carry these DNA laden substances on their bodies.Briefly, in iDNA metabarcod-ing the total DNA from invertebrate samples is extracted, and vertebrate DNA is amplified with polymerase chain reaction (PCR) with vertebrate-specific primers and nucleotide labeling before sequencing (Bohmann et al., 2021).The resulting sequences are matched against reference databases to identify the species detected (Taberlet et al., 2012).
There have been numerous recent studies that have validated and calibrated iDNA techniques for terrestrial mammal monitoring.For example, leeches have been used for occupancy modelling of vertebrate species (Ji et al., 2022) because they rarely move outside a very narrow range, which makes it possible to estimate occurrence ranges accurately.However, leeches are found in limited climatic and habitat zones (Abrams et al., 2019)-typically wet and tropical-meaning that using leech iDNA in drier environments may not be possible.Dipterans, such as carrion flies, mosquitoes, and sand flies, occur in drier habitats and are effective vertebrate iDNA reservoirs (Massey et al., 2021).Comparison studies show that carrion fly iDNA detects greater vertebrate diversity than iDNA from mosquitoes or sandflies (Massey et al., 2021;Saranholi et al., 2022).Carrion flies (Calliphoridae and Sarcophagidae, also known as blow and flesh flies) feed on carrion, open wounds of living animals, and fecal matter (Norris, 1965).Carrion flies are ubiquitous in many environments and easy to sample with inexpensive equipment (e.g., Calvignac-Spencer, Merkel, et al., 2013;Gogarten et al., 2020;Rodgers et al., 2017;Saranholi et al., 2022).Furthermore, carrion fly sampling provides a way to noninvasively sample vertebrate populations without risk of damage or theft of valuable equipment (such as cameras or audio recorders).Unlike mosquito or sandfly traps, there is no need for batteries or CO 2 (Schnell et al., 2015).Comparisons of carrion fly iDNA and camera trap methods show that iDNA detects more species than concurrent camera traps (Rodgers et al., 2017).Also, carrion fly iDNA disperses across minimal spatial ranges (up to 400 m) (Saranholi et al., 2022), making it ideal for measuring species occupancy and movement.
We used carrion fly iDNA to detect native and non-native mammals in the fragmented habitat of the wheat belt in southwestern Western Australia.We investigated whether the localization of the iDNA signal reflects the expected distribution of 4 native and 2 non-native mammal species in a highly fragmented landscape containing conservation reserves.We looked specifically at the composition and richness of mammals inside and outside conservation reserves.We predicted that the target native marsupial species-chuditch (Dasyurus geoffroii, Dasyuridae), numbat (Myrmecobius fasciatus, Myrmecobiidae), and woylie (Bettongia penicillata, Potoroidae)-would be restricted to conservation reserves and that there would be little evidence of dispersal across the spaces between the reserves.We predicted that echidna (Tachyglossus aculeatus, Tachyglossidae) would be most frequently detected in conservation reserves.However, because echidnas have wide home ranges in cropped and pastoral areas (Sprent & Nicol, 2012), we predicted that we would also detect them between the reserves.Finally, we expected invasive predator species-fox (Vulpes vulpes, Canidae) and feral cat (Felis catus, Felidae)-to be equally frequently detected across the landscape.

Study site
The wheat belt is southwest of Western Australia, covers 140,000 km 2 area, and is predominantly used for cereal growing.Wind speeds during sampling averaged 10 kmph in a southeasterly direction.More than 93% of the natural vegetation in this region has been replaced with exotic grasses and cereal crops.The remaining native vegetation is scattered throughout the region in thousands of remnants of varying sizes, shapes, and degrees of isolation (Saunders, 1989).
Three conservation reserves, Boyagin Nature Reserve (Boyagin NR), Dryandra Woodland National Park (Dryandra NP), and Lupton Conservation Park (Lupton CP) surrounded by crops and pastoral land, were chosen to represent islands of native vegetation within the farmland of the wheat belt (Figure 1).
Boyagin NR, in the north of our study area, is a 49-km 2 reserve located beyond the eastern edge of the Jarrah Forest (Schut et al., 2012).Boyagin NR has eastern and western halves separated by a 500-m-wide strip of farmland.Carrion fly collection was carried out in the western half of Boyagin NR, where foxes have been controlled since 1989.Woylies and numbats have been reintroduced in Boyagin NR, and populations have established in both halves of the reserve.Eighteen other mammal species occur in Boyagin NR, including our other native and non-native target species.
Dryandra NP, in the southern part of our study area (Figure 1), is the largest remnant of natural vegetation in the wheat belt, composed of 17 blocks, which equates to 280 km 2 of natural bushland.It is one of the most species-rich reserves in the region.Twenty-nine native mammal species occur in Dryandra NP, including all our target species.Dryandra NP plays a crucial role in conserving threatened species.Significant popu-lations of numbats and woylies occur in Dryandra NP (Friend et al., 1995;Pacioni et al., 2011).There is extensive predator control in the reserve and surrounding areas; fox baiting and feral cat management are ongoing and regular.However, there are records of both species in the conservation reserve (Marlow et al., 2015a).
Lupton CP is on the western edge of our study area (Figure 1).The park encompasses 87 km 2 but is contiguous with the rest of the surrounding jarrah (Eucalyptus marginata) forest.Several common mammal species inhabit the region, although limited records are available for the specific species that occur in Lupton CP (Department of Conservation and Land Management, 1987).Based on this limited information, we assumed that although numbat and woylies do not occur in Lupton CP or occur in meagre numbers, our other target species (native and non-native) do.In addition, feral cats and foxes are frequently detected in the jarrah forest, and predator control through baiting is a regular part of management in this area (Dundas et al., 2014).
In our study area, we selected 35 road edges between these conservation reserves, representing the farm matrix in which the conservation reserves are embedded.The road edges consisted of small patches of vegetation on either side.Most were <5-m wide with sparse vegetation, usually trees, with little to no natural understory and ground cover of exotic grasses.Road edges were chosen based on their accessibility.Mammal species records in this area were obtained through a search of NatureMap, a public database of species occurrence records provided by the Western Australian government (Appendix S1).

Sample collection
Traps were made from 2-L plastic bottles with the top and nozzle cut off and then inverted and inserted into the bottle.
Traps were sterilized for 30 min with a 20% bleach solution before deployment, and gloves were used to handle the sterilized traps.Approximately 20-30 mL of a commercial, organic fly bait (Magna Fly Bait, GEPRO, Australia) was put into each trap as an attractant and sampling fluid (Figure 1b).
Samples were set up and collected during January 2021 (Austral summer).Gloves were worn when traps were handled and were replaced between every trap.Five flytraps were placed approximately 500 m apart at each of the 3 conservation reserves in an X formation.These 5 traps were kept as separate samples.A single flytrap was set up at each of the 35 road edges throughout the region.The 35 road edge sampling sites were based on accessibility to public land and the availability of trees on which to hang flytraps.All flytraps were left out for 72 h.Sampling fluid (and flies) was decanted into falcon tubes.Traps were then thoroughly cleaned as described above, sampling fluid replaced, and the trap set up again.In each sampling location, the same trap was used.The 72-h fly sampling was carried out 3 times over 14 days.
Each trap and sampling period was considered a sample and extracted separately to increase the chance of detecting vertebrate species.This resulted in a total of 150 bulk carrion fly samples.The number of flies extracted in a sample ranged from 1 to 300, and the average was 24 files (SE 3).Samples were stored on ice during transport to the lab.Upon arrival at the lab, carrion flies in individual samples were gently washed with ultrapure water to remove traces of fly bait before being stored in 70% ethanol at −18 • C before DNA extraction.

Metabarcoding
Samples were processed using a nondestructive DNA extraction method (Nielsen et al., 2019).We used a modified DNEasy Blood and Tissue kit (Qiagen, Netherlands) for the final purification and concentration of DNA.The DNA extraction controls were included in every 23 samples.Detailed methods are in Appendix S2.
Two assays were used to amplify target species.A vertebratespecific assay was selected targeting the 12s rRNA region (Riaz et al., 2011) (vertebrate 12s), and a mammal-specific assay was selected targeting the 16s rRNA region (Taylor, 1996).Samples were amplified using fusion primers, gene-specific primers labeled on the forward and reverse with 6-8 bp molecular identification (MID) tags and Illumina sequencing adaptors (Bohmann et al., 2021).Samples were amplified in duplicate with the same tag combinations for each sample, including negative controls and a fly bait control.Samples were pooled in approximate equimolar concentrations.The library pool was sequenced on an Illumina MiSeq (Illumina, USA) with a single -end 300-cycle V2 kit per the manufacturer's directions (details in Appendix S2).

Bioinformatics and sequence processing
Sequence data for each primer set were processed separately.Sequences were demultiplexed using OBItools (Boyer et al., 2016) with no MID or primer sequence mismatches.Sequences were then length filtered for a minimum length of 50 bp.The DADA2 package (Callahan et al., 2015) in R 3.6.3(R Core Team, 2019) was used to further filter low-quality sequences.Sequences were filtered for quality with a maximum expected error of 2, and those identified as chimeras were removed.
Sequences were then de-replicated to produce amplicon sequence variants (ASVs).Singleton and doubleton ASVs were removed.The ASVs were matched to the NCBI GenBank reference database (www.ncbi.nlm.nih.gov/genbank/) with the Basic Local Alignment Search Tool (BLAST) for taxonomic assignment with a high-performance cluster computer (Pawsey Computing Centre, Perth, Western Australia).Taxonomic assignments were made to the lowest common ancestor (LCA) with the LCA script from eDNAFlow (Mousavi-Derazmahalleh et al., 2021) and a minimum query coverage of 100% and an identity threshold of 95%.Where the absolute value of the difference between percent identity of ASVs was <1, species taxonomy was not returned, and the ASV was assigned to the closest common ancestor.The specific ASVs found in the fly bait (cow [Bos taurus], pig [Sus scrofa], and sheep [Ovis aries]) and negative controls (cow and chicken [Gallus gallus]) were removed from the data set.We assumed ASVs contained a degree of haplotypic variation and thus removed only the specific ASVs associated with the controls.Because there are records of sheep and pigs in our study area (Appendix S1), ASVs associated with those species, but not with the fly bait, could belong to extant populations in this region.Any ASVs identified as human (Homo sapiens) were removed from the data set with the phyloseq package (McMurdie & Holmes, 2013).The ASVs from both metabarcoding assays were combined for further analyses.The ASVs were agglomerated at a species level and per trap to account for use of the same flytrap among the 3 sample collection times.The ASV table was transformed to presence or absence notation for community composition comparisons.

Statistical analyses
All statistical tests were run on R 3.6.3(R Core Team, 2019).Only ASVs identified as mammals were used for statistical analyses.The ASVs were categorized as native or non-native taxa.The ASV richness (observed number of ASVs) was calculated for these 2 groups at the sampling time with the phyloseq package (McMurdie & Holmes, 2013) in R 3.6.3(R Core Team, 2019).The ASV richness was transformed to the square root to meet the assumption of homogeneity of variance and normality.A 2-factor analysis of variance (ANOVA) was used to compare the square-root-transformed ASV richness of the groups of mammals (factor native vs. non-native, 2 levels) with the reserves (factor site, 4 levels [conservation reserves and road edges]).A Tukey's honest significant difference post hoc test was run using the agricolae (de Mendiburu & de Mendiburu, 2019) package to assess the significant differences among the groups.A 1-way permutational multiple analysis of variance (PERMANOVA) was conducted to compare the ASV presence-absence composition among the different reserves (factor: site) with the vegan package (Oksanen et al., 2019) with Jaccard similarity and 9999 permutations.Pairwise comparisons were performed using the PairwiseAdonis (Arbizu, 2020) package with a Benjamini-Hochberg correction for multiple comparisons.The R scripts and the accompanying data are available from https://doi.org/10.5281/zenodo.6767690.

Sequencing
In total, 3428 flies were captured during the sampling period.After removing the bait signal, 1,248,049 sequences were retained across both assays (mean [SE] of 13,287 [3686] sequences per sample).Nineteen ASVs were associated with the negative controls in both assays (18 ASVs in the 12s vertebrate assay and 1 ASV from the 16s mammal assay), and 22 ASVs were associated with the fly bait (10 ASVs from 12s vertebrate assay and 11 ASVs from the 16s mammal assay).These specific ASVs were removed from the data set.Across all samples, 45 vertebrate ASVs were detected (mean [SE] = 5 ASVs [0.55] per sample), of which 7 ASVs belonged to Aves and 36 belonged to Mammalia.Two ASVs could not be identified beyond the phylum Chordata and were removed from the data set before further analyses.Thirty-five ASVs were matched to a genuslevel identification, and 23 ASVs were matched at a species level.Taxa were agglomerated at a species level, meaning taxa with the same identification at lower taxonomic levels than species were merged.From these ASVs, we detected 8 of the 36 native mammal species (not including subspecies) and 1 of the 174 bird species recorded in the region (Table 1).
We detected 12 non-native taxa (including species in the fly bait) (Table 1).Because the wheat belt is predominantly agricultural, detections of the species found in the fly bait or controls (cow, pig, sheep) would not be unexpected.Therefore, although we removed the specific ASVs associated with the fly bait from our analyses, we kept other ASVs identified as these species for diversity and species composition assessments.We detected 16 ASVs from taxa classified as native to the region, including our 4 target species (Table 1).

Spatial distribution of target mammals
We detected all 4 native target species from carrion fly iDNA: chuditch, numbat, woylie, and echidna.Numbat detections were restricted to Dryandra NP (3 out of 5 traps) despite the species being previously recorded at Boyagin NR, albeit at lower population numbers.Woylie was detected at Dryandra NP (1 out of 5 traps) and Boyagin NR (1 out of 5 traps), and echidna was detected in Dryandra NP (1 out of 5 traps) and Lupton CP (2 out of 5 traps) but not at Boyagin NR despite being frequently observed there.Chuditch were only detected at Boyagin NR (1 out of 5 traps) (Figure 2c).We also detected woylies (1 out of 35 traps), numbats (1 out of 35 traps), and echidnas (2 out of 35 traps) from road edges adjacent to Dryandra.Woylies were detected up to 10 km away from the reserve (Figure 2a,b,d).Comparisons of relative sequence abundance showed a lower proportion of sequences from most of our target native species, except echidnas, outside conservation reserve areas than inside reserves (Appendix S3b-d).Foxes were detected in all conservation reserves (Boyagin NR, 3 out of 5 traps; Dryandra NP, 4 out of 5 traps; Lupton CP, 1 out of 5 traps), and feral cats were detected in Boyagin NR (1 out of 5 traps) and Dryandra NP (5 out of 5 traps) but not Lupton CP.These predators were also detected at many road edges (foxes, 18 out of 35 traps; cats, 10 out of 35 traps) (Figure 2e,f).A higher proportion of cat sequences were found in the Dryandra reserve than road edges; however, foxes were consistently found throughout the road edge and conservation reserve traps in high proportions (Appendix S3e,f).

Comparison between conservation reserves and road edges
The fully factorial ANOVA test on ASV richness showed there was no significant interaction between the factors native versus non-native and site (F 3, 98 = 1.77, p = 0.159).There was a significant effect of site (F 3, 98 = 5.91, p < 0.001) and a significant main effect of the factors native versus non-native (F 1, 98 = 78.75,p < 0.0001) (Figure 3b).The total mammal ASV richness was highest in Dryandra NP (Figures 3a & 4a), and there were no significant differences between the ASV richness at Lupton CP and Boyagin NR conservation reserves.The ASV richness was lowest at the road edges (Figure 3a).Boyagin NR and Dryandra NP had non-native ASV richness similar to the road edges, but Lupton CP had significantly lower nonnative ASV richness (Figure 3b).Dryandra NP and Lupton CP had higher native mammal ASV richness than Boyagin NR and the road edges, which were similar (Figure 3b).Native mammal ASV richness centered around the conservation reserves (Figure 4b), and this richness decreased as site distance from the reserve increased.Non-native mammal ASV richness was high throughout the region, including the road edges farthest away from conservation reserves (Figure 4c).

DISCUSSION
Using iDNA from carrion flies, we detected 14 of the 40 mammals previously recorded in our study area, including all our target species, native and non-native, from iDNA.Our predictions were supported in that the detected distributions of target native species were mainly restricted to conservation areas.In contrast, non-native feral cats and foxes were detected across the landscape.Furthermore, the composition of mammals in conservation reserves and road edges reflected our predictions of the different species richness and composition at these sites.These results support the value of conservation reserves in the region and reveal the differential response of native and nonnative mammals to the fragmented landscape.Moreover, our study lays the foundation for subsequent experiments and provides guidance related to experimental design, assays, and longer term monitoring with iDNA methods.
Three of our target native species-numbat, woylie, and echidna-were detected in conservation reserves and alongside road edges, but only from those roads close to the conservation reserves.For the numbats and woylies, the proportion of sequences was lower in samples collected from road edges compared with Dryandra NP, which contains a great abundance of these 2 species (Friend et al., 1995;Pacioni et al., 2011).The farthest detection from a reserve was woylie, on a road edge 10 km north of Dryandra NP.The detection of the target species reflects our prediction that the detection of native species would concentrate where they are mostly found.The lower proportion of sequences from target native animals outside Dryandra NP could indicate a reduced presence of these species.However, the ability to quantify species abundance from eDNA metabarcoding data in natural systems still needs to be fully understood, and caution must be exercised when basing any conclusions on the relative abundance of sequences (Fonseca, 2018;Lamb et al., 2019).Using species-specific assays may minimize the confounding variables of mixed species samples, allowing future iDNA studies to elucidate patterns of occurrence and habitat usage by mammal species.
The diversity of native mammals was centered on the conservation reserves, and in the central road edges of our sampling area there were no native mammals detected.This implies a degree of isolation.However, juvenile numbats and chuditch will begin to disperse to new habitats from December-January (Soderquist & Serena, 2000;Thorn et al., 2022), and it is probable that remnant vegetation along roadsides could act as corridors for this dispersal.The lack of detection of these species outside the conservation areas may not mean these species are completely absent from these areas and could instead reflect the limitations of our sampling method, such as the length of time of sampling, the sampling resolution, the lack of available carrion or feces as a food source, or that no animals were dispersing near where our samplers were located.Efforts have been made in and around the Dryandra reserve to maintain wildlife corridors-areas of retained bushland that allow for linkages between remnant areas (Hobbs & Saunders, 1991).
Although not every road edge, lined with remnant trees, can act as a wildlife corridor, the verges near conservation reserves may facilitate wildlife movement through these areas (Bennett, 1991).There was a woylie detection within 10 km of the edge of Dryandra NP, but it was within 2-3 km of nearby forested areas on the road connected to Dryandra NP, which would support the suggestion that roadside vegetation can serve as corridors for animal movement.However, it is still a considerable distance (>23 km) between the major populations at Dryandra NP and Boyagin NR.Therefore, future management strategies may need to assess the value of maintaining or enhancing roadside vegetation as wildlife corridors between conservation areas.
We expected that chuditch would be detected in Dryandra NP due to longstanding records of this species in this reserve (Orell, 2004), but we did not detect it in our samples.However, we detected chuditch in 1 sample at Boyagin NR.The chuditch is a solitary animal with a wide home range (Rayner et al., 2011).Therefore, the lack of chuditch detection may be because the iDNA sampling design needed to be optimized for detecting chuditch rather than the absence of this species from the area.We also expected that echidnas would be found throughout the region, inside and outside the conservation reserves, because echidnas can travel through pastoral areas (Sprent & Nicol,  2012).However, the detections of echidna also centered on conservation areas.A high proportion of sequences were detected from echidnas at a road edge site (Appendix 3), but this was a road edge close to Dryandra NP.As evidence suggests, minimal dispersion of the iDNA signal from carrion flies (Saranholi et al., 2022), using species-specific qPCR assays, or increasing the trapping intensity with more traps placed throughout the reserves or in the pastoral land itself could increase detections of these species.
Cats and foxes, as predicted, were detected in conservation areas and across the matrix of farmland between them.Predation by foxes and feral cats has been identified as one of the most significant causes of Australian mammal extinctions (Woinarski et al., 2015).Dryandra NP has carried out extensive fox eradication over the last 40 years (Marlow et al., 2015a); however, foxes were still detected in this reserve.Fox eradication in Dryandra NP is conducted through the regular deployment of 1080 poison baits (sodium fluoroacetate), a fast-acting and naturally occurring lethal toxin by which native animals in this region are unaffected (Marlow et al., 2015a;Thomson & Algar, 2000).Baiting occurred in the reserve 4 days before fly sampling began (Department of Biodiversity, Conservation, and Attractions, personal communication).If foxes were poisoned, the increase in carrion for flies to feed on might have influenced the detection of foxes from iDNA.However, feral cats are not as attracted to these baits, and declines in foxes can result in increased cat activity and predation rates (Marlow et al., 2015b), which can be detrimental to native mammal populations in these protected regions.Cats were not detected in Lupton CP, despite being contiguous with the rest of the jarrah forest region.This might be due to cat control in the area or reflect local, dynamic cat movements.
We detected fox and cat signals in the central road edges, where native animals were absent.Predator species have been shown to use road-edge habitats more frequently than macropods (Wysong et al., 2020), which implies that they could move through areas if they are well connected with roads.Therefore, managing foxes and feral cats should occur beyond reserve boundaries into the surrounding pastoral areas.Fox home ranges can average 704 ha, and inadequate baiting within this range can decrease the effectiveness of baiting programs because of decreased chance of uptake (Carter et al., 2011).There is also evidence that predator species using roads can cause avoidance of these areas by prey animals because of the increased predation risk (Latham et al., 2011).This may increase the isolation of the conservation reserves even if the habitat is maintained between them to facilitate movement between the areas.
Our results suggest a geographic limit for carrion flies to transport an iDNA signal.Although studies have shown that flies can disperse great distances (MacLeod & Donnelly, 1963), iDNA signals appear localized within an area very close to a vertebrate population (Saranholi et al., 2022), which might reflect carrion fly feeding patterns.Our results confirm a high degree of localization of the iDNA signal from carrion flies.In Dryandra NP, a fenced conservation sanctuary site, Barna Mia, contains several species of marsupial that are otherwise locally extinct, including the bilby (Macrotis lagotis).We detected bilby within 200 m of the Barna Mia fence.We also detected horse in both Boyagin NR and Dryandra NP.Because there is an established horse-riding trail through Dryandra NP, the detection of the horse could be explained.However, in Boyagin NR, the minimum distance from the fence line bordering pastoral land and the traps was 600-700 m.From this, we infer that within this study's temporal and spatial extent, flies were travelling no more than 500-700 m from the DNA source.This is the distance from the approximate center of the Barna Mia sanctuary area to the trap where this species was detected and the closest edge of the park fence line and where the horse signal was detected in Boyagin NR.The highly localized iDNA signal to the target population distributions suggests that carrion fly iDNA is viable for monitoring vertebrate diversity at a fine scale.Our finding that iDNA can be efficient and effective for species monitoring aligns with Calvignac-Spencer, Leendertz et al. (2013), Lee et al. (2016), Abrams et al. (2019), Gurtler (2020), and Cutajar & Rowley (2020), who recommend that iDNA could be incorporated into standardized temporal and spatial biodiversity studies.

Limitations, future directions, and conclusions
Understanding the strengths and limitations of carrion fly iDNA monitoring for species detection is essential for the future use of this method in conservation management.We demonstrated the utility of iDNA metabarcoding for monitoring the distributions of mammal populations.Although we could detect native biodiversity with iDNA, there were 26 mammal species our study did not detect.Our methods could be adjusted to detect more mammalian species or to target specific species.For example, using a passive baiting method increased the number of flies we collected, but the bait was a mammalian mix and overwhelmed the data set.Switching to a nonmammalian bait, adjusting trap designs to minimize fly contact with the baiting fluid, or using a blocking primer to block against bait species may be beneficial to increase species detections.Furthermore, a critical limitation was the paucity of native species in reference databases, which explains some of the lack of detections.For example, we could not detect the fat-tailed Dunnart (Sminthopsis crassicaudata), potentially not because the species was absent from the area but because the gene regions we targeted were not available for this species.Therefore, building the representation of species across multiple gene regions in reference databases will be essential.
We suggest that future researchers explore the effects of increasing sampling intensity spatially, temporally, or both when using carrion fly iDNA to detect mammals.Using other invertebrates from the study sites (e.g., mosquitoes and ticks) might broaden the taxonomic diversity detected (see Gurtler, 2020;Massey et al., 2021;Saranholi et al., 2022).We also were able to nondestructively extract vertebrate DNA from the carrion flies, meaning that future studies could look at how carrion fly species affect detections (e.g., examine associations with fly species and vertebrate species or how diversity and abundance of flies can affect vertebrate species detection).This is with the caveat that although metabarcoding studies can reduce the overall costs of conducting biodiversity surveys, molecular work can be expensive, and increasing sampling intensity or substrate can dramatically increase associated costs (Borrell et al., 2017).Previous work comparing carrion fly iDNA to camera traps shows that although iDNA could pick up more vertebrate biodiversity than concurrent camera traps, the iDNA method failed to detect very abundant species (Rodgers et al., 2017).Therefore, a combined approach in which iDNA is used to guide the design of more targeted surveying efforts may be more effective (Abrams et al., 2019).As sequencing costs continue to decrease and reference databases are populated (Tang et al., 2014), iDNA techniques could become a valuable tool in the wider eDNA toolbox for monitoring noninvasive mammal populations.
Open access publishing facilitated by Curtin University, as part of the Wiley -Curtin University agreement via the Council of Australian University Librarians.

FIGURE 1
FIGURE 1 (a) Flytrap locations alongside 35 road edges and in the 3 conservation reserves, Boyagin Nature Reserve (NR), Dryandra Woodland National Park (NP), and Lupton Conservation Park (CP), in Western Australia (dark gray, forested areas), (b) plastic bottle flytrap, and (c) typical habitat at Boyagin NR, Dryandra NP, Lupton CP, and road edges.

FIGURE 3
FIGURE 3 Mean (bars, SE) amplicon sequence variant (ASV) richness of (a) all 6 target mammals and (b) non-native and native target mammals at each site detected via invertebrate-derived DNA (differing letters indicate statistically different mammal ASV richness [i.e., number of ASVs] in a Tukey's honest significant difference [HSD] test).Carrion flies were sampled at sites along 35 road edges and in 5 sites in each of the 3 conservation reserves (x-axis) (NR, nature reserve; NP, national park; CP, conservation park in Western Australia).

FIGURE 4
FIGURE 4 (a) Overall mammal amplicon sequence variants (ASV) richness, (b) native mammal ASV richness, and (c) non-native mammal ASV richness values based on carrion-fly-derived DNA collected along 35 road edges and in 5 sites in each of the 3 conservation reserves (NR, nature reserve; NP, national park; CP, conservation park) in Western Australia (dark gray, forested areas; green borders conservation park boundaries).

TABLE 1
All native and non-native amplicon sequence variants (ASVs) identified to at least family level detected via carrion fly DNA.
Abbreviations: B, Boyagin Nature Reserve; D, Dryandra Woodland National Park; L, Lupton Conservation Park; R, road edge.b Taxa detected in the carrion fly bait.c Taxa investigated in this study.