Climate change is expected to lead to shifts in the phenology and distributions of species [1
]. Insects are expected to respond to changing temperatures particularly readily because of their short life cycles, which are heavily influenced by temperature [2
]. For example, significant poleward shifts have been reported in the ranges of non-migratory butterflies in Europe [3
], and a dramatic change in community structure of parasitic wasps over a period of 50–70 years has been documented in the Canadian subarctic [4
]. In the Arctic, changes in the composition of arthropod communities are projected to have a cascade effect on critical ecosystem parameters, such as decomposition, nutrient cycling, and primary productivity [5
]. Accurate and efficient arthropod biodiversity monitoring in the Arctic is needed to document changes in community structure and to detect the arrival of new species, including potential disease vectors. Although arthropod communities have been identified as a prospective model for studying general ecological responses to environmental variation, the lack of baseline data has been identified as an important shortfall [6
]. Specific responses of different arthropods to climate variation differ considerably, depending on taxon and locality [7
]; therefore, detailed knowledge of taxonomy is a necessary prerequisite for tracking the changes happening at the ecological community level. However, the deficient state of taxonomic knowledge of many abundant arthropod taxa and limited availability of expertise in morphological species identification restrict the scale and taxonomic scope of any monitoring schemes based on traditional approaches.
Biodiversity research in the Arctic has typically been conducted by field technicians and taxonomic experts flying in to conduct the work. However, there is a growing movement in biodiversity science toward community-based monitoring (CBM) [8
] which engages local residents in the research process. CBM incorporates traditional and local knowledge which provides valuable historical context of environmental changes and their impacts on communities over time [10
]. A collaborative approach is required to build collegial relationships between community members and scientific researchers, to promote learning, knowledge exchange, and joint use of research results to guide future planning and policy-making [10
In the Western Arctic, CBM has been instrumental in understanding the substantial changes in phenology, height, and relative species abundance of tundra vegetation associated with climate-induced increases in growing season length and active layer depth [8
]. Unfortunately, similar monitoring data on arthropods are not yet available. Arthropods have relatively high species diversity even in arctic environments, and the difficulty of species-level morphological identification of most taxa limits the potential of CBM. As a result, arthropod research in the Arctic has so far been conducted by external researchers; and the taxonomic, geographic, and temporal coverage of these studies has generally been limited. A DNA-based approach which bypasses morphological species identifications is not constrained by access to specialized taxonomic expertise and is thus better suited for monitoring arthropods on a large taxonomic scale [12
]. Furthermore, the involvement of local community members in data collection and interpretation, offers the potential for substantial increase in the spatial and temporal scales and increases societal relevance of such monitoring efforts.
DNA barcoding has become widely used in biodiversity assessment [14
], in studies of community ecology [17
], and food webs [19
]. Assessment of species diversity in poorly known and largely undescribed ecological communities is enabled through clustering DNA barcode sequences into molecular operational taxonomic units (MOTU), which can be used as a proxy for species. The Barcode Index Number (BIN) framework [21
] implemented in the Barcode of Life Data System (BOLD; http://boldsystems.org/
; ref. [22
]) is among the most widely used MOTU delimitation approaches. Comparisons between BINs and described species in well-studied arthropod faunas have shown that BINs reconstruct species boundaries with high accuracy [23
The Arctic BIOSCAN project (ARCBIO) is a partnership between the Centre for Biodiversity Genomics at the University of Guelph (CBG), Polar Knowledge Canada (POLAR), and counterpart organizations in the Kitikmeot Region of Nunavut. It aims to assemble a baseline DNA barcode reference library for the region, with a view to developing biodiversity monitoring protocols for the Canadian Arctic that draw from community leadership and employ high-throughput sequencing. Arthropods are among the key focus groups of ARCBIO, which builds on prior surveys, including arthropod community assessments [25
], seasonal changes and spatial patterns of spider (Araneae) assemblages [26
], biodiversity assessment of aquatic insect groups [16
], and distribution and temporal changes of dipteran assemblages [28
In this study, we assess the overall arthropod diversity in the vicinity of Ikaluktutiak (Cambridge Bay), Nunavut, using BINs as a proxy for species. We estimate beta-diversity (measured by BIN turnover) among two continuously monitored sites where a standardized sampling protocol was employed. Finally, we compare the relative performance of collecting methods in recovering taxonomically representative samples of arthropod diversity to determine which method or combination of methods would be prospective for CBM purposes, i.e., easy to deploy and low-maintenance, but still capturing a representative majority of arthropod diversity from the study site. We also present a baseline DNA barcode reference library for terrestrial arthropods (insects, arachnids, and springtails) for the Ikaluktutiak area of Victoria Island.
2. Materials and Methods
2.1. Study Sites
Field work was done from 7 July to 18 August 2018 in the Kitikmeot Region of Nunavut, in the vicinity of Ikaluktutiak (Cambridge Bay, 69.11° N 105.06° W), southern Victoria Island. The area is situated in the Middle Arctic Tundra ecoregion within Bioclimate Subzone D, as defined by the Circumpolar Arctic vegetation map (CAVM) [30
]. The dominant plant community type in this area is Nontussock-sedge, dwarf-shrub, moss tundra (CAVM G3) [30
]. Several finer-scale eco-sites are represented in the area, defined by winter snow cover and moisture regimes [31
]: exposed upland, xeric to mesic habitats are dominated by Dryas integrifolia
, Saxifraga oppositifolia
, and Carex rupestris
; snow-protected, mesic sites are characterized by Cassiope tetragona
, Dryas integrifolia
, and Salix reticulata
; protected lowland sites are dominated by Salix richardsonii
and Carex aquatilis
, and mosses. The area is underlain by cobbly bouldery limestone till with granite inclusions; in upland (xeric) habitats, stones occupy up to 90% of the surface area [31
Because of the goal to assemble a baseline DNA barcode reference library for arthropod diversity in the Ikaluktutiak area, efforts were made to cover different terrestrial eco-site types and freshwater bodies (tundra ponds, streams) opportunistically during general collecting activities. Moreover, two continuous monitoring sites have been established: one in proximity to the Canadian High Arctic Research Station (CHARS) Intensively Monitored Area (IMA site) on the northern shore of Grenier Lake [32
] (69.2186° N, 104.9258° W) and the second site near Water Lake (WL site, 69.1323° N, 105.0621° W)—a more disturbed habitat 1 km north of the hamlet of Ikaluktutiak. Following setup, the sites were checked every other day; however, deviations from this schedule had to be made for the less accessible IMA site because of weather and logistical constraints.
2.2. Collecting Methods
The following active collecting methods and stationary traps were used (Table 1
Malaise traps—three Townes-style Malaise traps (BugDorm, MegaView Science Co., Ltd., Taiwan) were installed at continuous monitoring sites: one near Water Lake and two near IMA. Samples were collected in 500-mL plastic bottles filled with ~350 mL of 95% ethanol, which were exchanged on a weekly basis. (NB: During this study, the polyester support strings used to anchor Malaise traps installed at the CHARS IMA site were damaged by arctic foxes, which, combined with strong wind gusts, caused the traps to collapse completely or partially between service dates. As a result, stock ropes were replaced with eight 1/16 inch galvanized “aircraft cables” (2.5 m in length). Cables were fitted with aluminum oval sleeves to form loops at the ends and a round swivel eye simplex snap was used to attach each cable to the trap. The distal end of each cable was anchored to the ground with a metal peg, reinforced as needed with heavy stones. These modifications prevented further collapsing of the traps for the remainder of the field season).
Pan traps—ten yellow shallow plastic bowls (9” diameter) were installed at each of the monitoring sites and half-filled with soapy water. Traps were checked every other day, and the two-day catch from each pan trap site was pooled into a single bulk sample. Collected specimens were strained through a Nitex nylon fabric with a mesh size of 50 µm and fixed in 95% ethanol.
Pitfall traps—a line of 10 translucent wide-mouth 500 mL plastic cups was installed at 3 m intervals at each monitoring site. The traps were half-filled with soapy water and capped with a 10 mm steel mesh to exclude vertebrate by-catch. The checking schedule and specimen collecting procedure was the same as for pan traps.
Sweep-netting—arthropods were collected from air and from vegetation with a standard insect sweep-net, then hand-sorted or aspirated, followed by placement in a kill jar or fixation with 95% ethanol. At each of the two monitoring sites, sweep-netting was performed weekly for five continuous minutes (six times at WL site and five times at IMA site).
Soil and litter sifting—three 30-m soil sifting transects were established parallel to the pitfall trap line at each of the two monitoring sites, with the middle transect adjacent to the pitfall traps, and one transect on either side of the pitfall line at a distance of 10 m. Ten ca. 1-L samples of litter and surface soil were taken with a trowel at 3-m intervals, sifted through a 5-mm mesh to remove larger debris, and pooled into a single bulk sample. The resulting soil and litter samples (ca. 3 L in volume) were placed in Berlese funnels for one week to extract specimens, which were collected in vials with 95% ethanol. The soil transects were sampled four times during the study period for each continuous monitoring site. In addition to the transects, samples of varying size were taken opportunistically by hand or with a trowel from soil surface, moss, debris, remains of dry vertebrate carcasses, or leaf litter in various habitats across the study area. Each sample was sifted on site through a 5 mm mesh, followed by hand sorting in the lab and fixation of all extracted organisms in 95% ethanol.
Aquatic sampling—arthropods were collected from the water column, submerged vegetation, and bottom sediments of tundra ponds and streams using a hand-held plankton net, D-net or small hand net, then strained from water and fixed with 95% ethanol.
Freehand collecting—individual arthropods were opportunistically picked by hand, with forceps, or hand-netted, followed by fixation with 95% ethanol.
Stationary traps were only used at continuous monitoring sites. Active methods were used opportunistically for general collecting throughout the study area, and in a regimented fashion at long-term monitoring sites, consistent with the Standardized Sampling protocols designed by the CBG’s field teams for monitoring arthropod biodiversity (deWaard et al. in prep.). A total of 450 bulk samples (lots) were obtained as a result of all arthropod collecting efforts.
2.3. Sample Processing and DNA Barcode Sequencing
Collection materials were accessioned and processed at the CBG, following the single-specimen barcoding workflows outlined in deWaard et al. [33
], with exceptions, as described below. Because the goal of the project was to create a DNA barcode reference library for future monitoring, bulk sample (lot) sorting was prioritized to maximize taxonomic diversity, while reducing the number of replicates per species through initial morphotaxonomic screening. Lots from continuously monitored sites were prioritized over lots from general collecting, except for Malaise trap materials, of which only lots collected in odd-numbered weeks (1, 3, 5, etc.,) were processed in order to reduce the costs of analysis. In total, 428 of the 450 lots were processed, of which 34 were sorted completely; the remainder had specimens remaining after sorting. Morphospecies sorting of each continuous monitoring sample began with picking all specimens representing taxa other than the highly abundant Diptera and Hymenoptera for analysis. A subset of each morphospecies of Diptera and Hymenoptera was then picked for analysis by senior technical staff with extensive experience in arthropod specimen sorting. General collecting samples were screened and morphospecies sorted by taxonomic experts in Araneae and Coleoptera, with a subset of remaining specimens from other groups chosen for analysis. A total of 24,198 arthropod specimens were selected for DNA barcode analysis. They belonged to three classes: Insecta (19,661), Collembola (904), and Arachnida (3633). Among arthropods isolated from aquatic samples, 264 specimens belonged to crustacean taxa. Although submitted for DNA barcoding analysis, they were not part of this study; therefore, they are excluded from the figures reported here.
DNA extraction, PCR amplification of the COI barcode region, and sequence analysis were done at the Canadian Centre for DNA Barcoding (CCDB, Guelph, Ontario, Canada; ccdb.ca)—the Genomics Unit of CBG. Single molecule, real-time sequencing was performed on the SEQUEL platform (Pacific Biosciences, Menlo Park, CA, USA) using the methodology described by Hebert et al. [34
2.4. BINs and Taxonomic Assignment
Taxonomic diversity among the successfully sequenced specimens was assessed by using the built-in BOLD algorithm for assigning DNA barcode sequences to Barcode Index Number clusters (BINs)—molecular operational taxonomic units used as a proxy for species [21
]. Using this algorithm, sequences over 300 base pairs in length can be assigned into an existing BIN, but the founding sequence of a new BIN must be at least 500 base pairs long [21
]. Further taxonomic assignments were made by checking the sequences against existing taxonomically identified reference data on BOLD, using built-in queries. When a BIN contained records with specimens identified to a concordant family, genus, or species, any unidentified specimens assigned into this BIN received this identification. Specimens representing a BIN new to BOLD were queried using the BOLD identification engine. The new BIN received a family-level identification if it showed <10% sequence divergence from previously identified sequences, and a genus-level identification if the sequence divergence was <5%. In addition, the query sequence had to be clustered within other BINs with the same taxonomic assignment in order to be assigned to that family or genus [33
]. Additional reviews were conducted using Neighbor-Joining trees coupled with images of the associated voucher specimens. When possible, representative specimens of some of the new BINs were also identified morphologically by taxonomic experts.
2.5. Archiving and Imaging
Specimens were archived in the natural history collection at CBG (BIOUG), although separated from the main archive, so that a synoptic collection could be devised for later deposition at the CHARS campus in Ikaluktutiak. Residual DNA extracts were deposited in the CBG DNA archive for quality control purposes. Specimens and DNA extracts are available for study upon request.
Records that gained BINs were queried within BOLD to determine which BINs lacked photographs. A single representative specimen was photographed from each new BIN. Efforts were made to select a specimen in good condition when possible. Specimens were photographed at high resolution and the images are accessible through both specimen and BIN pages under Creative Commons (BY-NC-CA) license.
2.6. Data Analyses
All specimen records and associated provenance data, as well as DNA barcode sequence data, were uploaded to BOLD (http://boldsystems.org
]. All analyses were restricted to the three major classes of Arthropoda which formed the vast majority of the processed material: Arachnida, Collembola, and Insecta. All specimens that failed to produce sequences of sufficient length and quality to get assigned to a BIN were excluded from analyses. Any sequences confirmed to be contaminated by non-target DNA were likewise excluded. The final dataset included 18,096 specimens with high-quality sequences, which gained BIN assignments by 31 October 2019.
The completeness of species coverage was estimated with accumulation curves based on the BIN data. With only a subset of each sample processed, multiple accumulation curves were created for different orders to account for the abundance bias and extrapolated to 100 sites with EstimateS 9.1.0 [35
]. Samples were grouped into sites (n
= 53) with distinct GPS coordinates to account for collective effort at one particular locality split among several samples (e.g., multiple jars/vials per collecting event) and for collecting at the same site on different dates. Chao1 estimates were not calculated for insect orders that were represented by less than 10 species (Ephemeroptera, Plecoptera, Trichoptera, Thysanoptera) and for orders of mites and collembolans where some order-level assignments were inconclusive or based solely on molecular data.
Comparative species diversity captured by different trap types and collecting methods and taxonomic overlap between them were estimated based on the presence or absence of BINs in the samples collected with the corresponding technique. The vocabulary used in field notes was standardized to aggregate records, based on the type of activity (i.e., passive or active trapping) and based on the targeted habitats (i.e., aquatic or terrestrial). The traps left unattended for a certain period of time, i.e., Malaise trap, pitfall trap and pan trap, were collectively considered passive terrestrial collecting methods. All aquatic collecting methods were considered active, because no traps targeting aquatic habitats were used. Active terrestrial methods included soil transects and sweep netting at the continuously monitored sites, as well as opportunistic sifting, sweep netting, and free hand collecting outside these sites.
2.7. Data Availability
All DNA barcode sequences and associated provenance data are accessible through a public dataset on BOLD (dx.doi.org/10.5883/DS-CHARS). The sequences are also available through NCBI GenBank, accession number range MN665381–MN683476. See Supplementary Material, Table S1
for detailed list of records.
Monitoring changes in the Arctic remains incomplete without data on arthropod communities. However, even baseline data for this speciose and ecologically important group are sparse and have limited taxonomic depth. A pilot effort to provide taxonomically comprehensive species-level biodiversity assessment of arthropods in the vicinity of Ikaluktutiak (Cambridge Bay), Nunavut, using DNA barcoding resulted in recording 1264 BINs (molecular proxy for species); Chao1 estimator was 1820 ± 73 BINs. Yellow pan traps alone captured 62% of the BINs, and their efficiency and ease of deployment makes them the preferred method for surveying and monitoring the terrestrial arctic arthropods. In order to sample ground-dwelling taxa, pan traps need to be complemented by other methods such as leaf litter and soil sampling.
Coupling community-based sample collection with high-throughput DNA barcoding has the potential to overcome or at least considerably alleviate some of the major obstacles to large-scale arthropod monitoring in the Arctic: travel cost of deploying research teams from southern institutions to conduct fieldwork, the difficulty of morphological identification of many arthropod species, and insufficient knowledge of taxonomy for many diverse taxa, such as soil mites. Data produced through such monitoring programs should be made available in real time to local stakeholders through online data portals, such as BOLD, providing required information about the state of biodiversity and facilitating data sharing and online collaboration with scientists, regulatory agencies, and between communities.