Monitoring the influx of new species through citizen science: the first introduced ant in Denmark

Biological data

During the Ant Hunt, families and schools across Denmark collected ants by conducting baiting experiments at a site of their choosing. Participants ranged across all ages, but the average participants were children aged 5–11 years accompanied by a grown-up aged 31–50. They set out six different resources (saltwater, sugar water, oil, dissolved protein powder, a cookie and water) on bait cards and waited for two hours before collecting all the ants that were foraging on the cards. Ants were then frozen and counted before they were placed in 96% ethanol and sent to the Natural History Museum of Denmark for identification. All experiments were registered in an online database with date and GPS-coordinates (see Supplemental Material S1 for detailed protocol). In total, families and schools completed 792 experiments, of which 566 contained ants.

The ants were identified using a variety of taxonomic keys (Collingwood, 1979; Seifert, 2007; Douwes et al., 2012; Lebas et al., 2016; Wagner et al., 2017). In total, participants had collected 16,985 specimens from 29 species (Table 1). Of these, specimens from two experiments could not be identified to species level due to missing body parts and specimens from two experiments were flagged as potentially new taxa for Denmark. These were T. immigrans and Technomyrmex albipes. The establishment of Technomyrmex albipes could not be confirmed. However, after the original discovery of T. immigrans during the Ant Hunt, trained scientists resurveyed the location several times throughout 2015–2019. The presences of T. immigrans was confirmed during every survey and the species was seen to expand to an area of approx. 40 m, with more than 20 nest entrances along the pavement.

Tetramorium immigrans (Figs. 1A and 1B) tends to have a larger overall body size, denser striation/sculpturation of the head, thorax and petiolar nodes, as well as a more pronounced microscopic scale pattern on the first gastral tergite than T. caespitum (Wagner et al., 2017). Because of the difficulty in distinguishing T. immigrans from other species in the T. caespitum complex, we visually inspected all specimens found of T. caespitum in the Ant Hunt and randomly selected 10 samples from a broad range of localities (Fig. 1C). These were then compared to the 67 specimens of Tetramorium workers from the Botanical garden in Copenhagen, which is part of the Natural History Museum of Denmark (Fig. 1D), which were examined visually for morphological characters distinguishing typical forms of T. immigrans and T. caespitum. Voucher specimens were stored at the Natural History Museum of Denmark (NHMD 0000188537).

One specimen was chosen to be inspected by X-ray micro computed tomography (MicroCT). The specimen was placed on a designed holder and placed inside a Zeiss Xradia 410 versa system. The system was operated with a high voltage of 40 kV and a power of 10 W. The data acquisition consisted of 3,201 images while rotating 360 degrees, each image had a pixel size of 4.14 µm. All data was reconstructed into a 4 mm by 4 mm by 4 mm 3D volume with a voxel size of 4.14 µm. The reconstructed image is shown in Fig. 1B and raw data is available through Figshare (Gundlach et al., 2020).

DNA analysis

For the DNA analysis of T. immigrans we selected two specimens from the confirmed find in the Botanical Gardens of Copenhagen, one from 2015 and one from 2018 and further selected 14 specimens of presumed T. caespitum, of which 10 had known coordinates and the remaining four were known to be from somewhere in Denmark.

Up until DNA extraction, all samples from the Ant Hunt were kept in 96% ethanol in a freezer, while samples from the Natural History Museum of Denmark had been kept as pinned specimens. We extracted DNA by cutting off a small piece of the middle leg of each specimen, to which we added 100 µl of 10% Chelex in Tris-HCI buffer. This was mixed and centrifuged for 10 min, after which the solution was heated to 99 °C for 15 min and centrifuged again. The supernatant was used as a template for PCR reactions. We used primers LCO1490 and HCO2198 (Folmer et al., 1994) to amplify mitochondrial COI gene and primer D2B and D3A-r (Saux, Fisher & Spicer, 2004) to amplify nuclear 28S rDNA gene. PCR reactions were carried out using RedTaq ReadyMix PCR Reaction Mix with 100 µg/mL Bovine serum albumin. The PCR reaction conditions consisted of an initial denaturing step of 94 °C for 5 min, followed by 35 cycles of 94 °C for 40 s, 48 °C (LCO1490/HCO2198) or 56 °C (D2B/D3A-r) for 40 s, and 72 °C for 60 s, and finally an extension step at 72 °C for 5 min. PCR products were purified using Invitek PCR clean-up MSB spin PCRapace kit and Sanger sequenced in both directions using the Mix2Seq from Eurofins.

Molecular identification was carried out by comparing samples to reference sequences from Wagner et al. (2017) (COI) and Schär et al. (2018) (28S). Raw sequences were edited and aligned using the software geneious v. 2019.0.4. A maximum likelihood tree with the best-fit model automatically selected by modelfinder and 1,000 rapid bootstrap replications was created for the alignment of COI sequences, using the program ‘IQ-TREE’ v. 1.6.1 (Nguyen et al., 2015).

Tetramorium immigrans has previously been found to be distinguished from similar species (including T. caespitum) by having a one base insertion (C) at site 438 of the 28S rDNA fragment amplified by the primers D2B and D3A-r (Saux, Fisher & Spicer, 2004; Schär et al., 2018). We aligned the sequence of the specimen from Copenhagen to the reference sequences mentioned above to see if the characteristic insertion of T. immigrans is present.

All genetic data referred to in this publication are available via the European Nucleotide Archive project PRJEB36036.

Climatic niches of T. immigrans and T. caespitum

We compared the environmental niche and predicted geographical suitability based on climate of T. immigrans and T. caespitum in Europe using occurrence data from AntMaps (Guénard et al., 2017) and the Ant Hunt along with 10 climatic variables from Worldclim 1.4 at 2.5 arc-min (~5 km) resolution (Hijmans et al., 2005).

Due to the recent distinction between T. caespitum and T. immigrans (Schlick-Steiner et al., 2006; Wagner et al., 2017), we only used data points from 2006 onwards from trusted AntMaps sources (Borowiec & Salata, 2018b; Espadaler, Pradera & Santana, 2018; Schär et al., 2018; Wagner et al., 2017, 2018). This resulted in 739 samples of T. caespitum and 187 samples of T. immigrans. Since T. immigrans was detected in only one location in Denmark, this one location does not contribute much to the overall niche analysis for the species.

We selected the climatic variables mean annual temperature, annual temperature seasonality, mean temperature of warmest quarter, mean temperature of coldest quarter, mean annual precipitation, annual precipitation seasonality, precipitation of wettest quarter, precipitation of driest quarter, precipitation of warmest quarter and precipitation of coldest quarter because these have been suggested to be the most relevant to T. immigrans (Schlick-Steiner et al., 2006; Steiner et al., 2008). We then tested for autocorrelation between the 10 climatic variables through a correlation matrix (Fig. S1) using the R packages ‘Hmisc’ (Harrell & Dupont, 2019) and ‘PerformanceAnalytics’ (Peterson & Carl, 2019). There was a strong correlation between mean annual temperature and mean temperature of warmest quarter (r = 0.90, p < 0.001) and mean temperature of coldest quarter (r = 0.94, p < 0.001). Annual precipitation was strongly correlated with precipitation of wettest quarter (r = 0.95) and precipitation of driest quarter (r = 0.86, p < 0.001). Precipitation of warmest quarter and precipitation of coldest quarter were also strongly correlated (r = 1, p < 0.001). We therefore discarded mean temperature of warmest quarter, mean temperature of coldest quarter, precipitation of wettest quarter, precipitation of driest quarter and precipitation of coldest quarter from the analysis and kept the remaining five variables.

Using the ecospat package (Broennimann, Cola & Guisan, 2018) in R Core Team (2018), we compared the environmental niches of T. caespitum and T. immigrans in a gridded environmental space, where each cell corresponds to a unique set of the five climatic variables. We first calculated the density of occurrences and the climatic variables along two climatic axes of a multivariate analysis and then measured the niche overlap along the gradients of this multivariate analysis as in Broennimann et al. (2012). Niche overlap was calculated using Schoener’s overlap metric ‘D’ (Schoener, 1968, 1970), which in this case compares the frequency of observations for each species within the chosen climatic categories (Schoener, 1968). However, we acknowledge that there are other ways to measure niche overlap (Seifert, 2017).

To test for niche equivalency and niche similarity, we compared the observed D metric with 1,000 simulated values of D. The niche equivalency test determines whether the overlap between the niches of the two species is higher than two random niches drawn from the same data pool. The niche similarity test determines whether the overlap between the two niches is higher than when one species’ niche is randomly drawn in the study area (Broennimann et al., 2012). If the observed value of D falls within the density of 95% of the simulated values, there is no detectable difference in the climatic niche of T. immigrans and T. caespitum. Finally, we projected the environmental niche of T. immigrans and T. caespitum onto Europe to visually compared the two species and pinpoint areas of suitable climate for T. immigrans to establish.

Habitat differences

We used the CORINE 100 × 100 m land cover raster dataset (Copernicus, 2012) and extracted land cover values for all data points for both species using the spatial analysis tool ‘extract values to points’ in ArcGIS (ESRI, 2010). The CORINE land cover dataset consists of 48 land cover types. For this analysis we excluded all data points that were labelled with no data or one of the water based land cover types (‘water bodies,’ ‘water courses,’ ‘sea and ocean’). The remaining 39 land cover types were reduced to six major classes (‘Artificial surface,’ ‘Agriculture,’ ‘Forest,’ ‘Scrub and/or herbaceous vegetation associations,’ ‘Open spaces with little or no vegetation’ and ‘Wetlands’) following the CORINE land cover nomenclature (Copernicus, 2015).

To test whether sampling of T. immigrans and T. caespitum were spatially biased, we compared the fraction of samples within each of the six land cover classes with the fraction of these land cover types in Europe using chi-square tests. We then did the same comparing the two species to each other to determine if T. immigrans and T. caespitum were being sampled in different habitats.

Monitoring for new species

In order to determine the suitability of citizen science for early detection and monitoring of introduced species, we compared the citizen scientist collected dataset of the Ant Hunt with a dataset collated from the Natural History Museum of Denmark, a personal collection by Sämi Schär and the Ph.D. course EuroAnts, which was collected in Denmark from 1990 to 2015, with 2015 being the most recent year with available data. We compared the datasets based on two measures, (1) distance to likely introduction sites and (2) sampling effort in different land cover types. We calculated the average distance from data points in each dataset to Denmark’s seven major cities (cities of 50.000 + inhabitants) and 31 major harbours (harbours with a yearly goods turnover of one million ton) in ArcGIS (ESRI, 2010) and compared the two datasets using a Mann Whitney U-test.

For the land cover analysis, we used the above-mentioned major categories, but also included the category Water as an approximation of samples collected close to the shore of water bodies (Copernicus, 2015). We summarised the number of samples collected in these seven major land cover classes and compared the observed values with expected values based on the availability of each land cover class using Chi-square tests. Based on availability of each land cover class we calculated the ratio of observed samples to expected samples to determine how much more or less a specific land cover type was sampled than what would be expected by chance.

Tetramorium immigrans

Tetramorium immigrans was first discovered in the Botanical Garden of the Natural History Museum of Denmark (55.69, 12.57) during a pilot run of the citizen science project in 2015 and was confirmed to be established and spreading in 2018 and 2019.

Although we did not collect traditional morphometric data, we regard the find of T. immigrans in Denmark to be verified. We base this determination on a maximum likelihood tree of the mitochondrial COI gene (Fig. 2A), the existence of a characteristic one base insertion (C) in the 28S rDNA fragment (Fig. 2B), and visual examination of diagnostic characters. This extends the northern limit of T. immigrans in Europe by three degrees latitude from Gmina Janowiec Wielkopolski, Poland (52.73, 17.50, Borowiec & Salata, 2018a) to Copenhagen, Denmark (55.69, 12.57), almost 460 km.

Climatic niches of T. immigrans and T. caespitum

Environmental niche overlap between T. immigrans and T. caespitum, measured as D, which compares the frequency of observations for each species within the chosen climatic categories, was 13% with only a slight difference of the niche centroid in environmental space (Fig. 3A). Based on the contribution of the five climatic variables along the two axes, T. caespitum is present in colder and wetter areas, compared to T. immigrans, which prefers warmer and drier conditions with stable temperatures (Fig. 3B). T. immigrans and T. caespitum differed in mean climatic values for eight of the original ten climatic variables (Table S1; Fig. S2), with no difference in precipitation seasonality and precipitation of coldest quarter. Despite these slight differences, the two species’ niches were more similar to each other than two randomly drawn niches within the same data pool (niche equivalency test, p = 1) and more similar than when a random niche of either species was drawn within the available climatic space (niche similarity test, p = 0.22).

Projection of the climatic niche into geographical space shows Eastern and Central Europe to be the most climatically suitable for both species along with the northern part of Southern Europe and the southern part of Northern Europe (Figs. 3C and 3D), although T. immigrans may have a slightly more southern distribution than T. caespitum.

Habitat differences

The current available data does not allow for an exact determination of the land use of T. caespitum and T. immigrans, but we can determine in which land cover types the two species are currently observed. Tetramorium caespitum was mostly observed in forest and agricultural land cover types (34.1% and 34.96% of observations, respectively, Table 2). T. immigrans was mostly found in land types with artificial surfaces (48.89%, Table 2). Both species were rarely found in wetland areas (T. immigrans: 1.11% and T. caespitum 0.14% of observations). Observations of both species differed significantly from what would be expected if land use reflected availability of the six land cover types in Europe (Chi square tests, χ² = 322.71, df = 5, p < 0.001 for T. immigrans and χ² = 243.95, df = 5, p < 0.001 for T. caespitum).

The two species also showed significant difference in the number of observations in each land cover type compared to each other (Chi-square test, χ² = 82.66, df = 5, p < 0.001). T. immigrans was found almost three times more often in land cover types with artificial surfaces than T. caespitum (ratio: 0.35; Table 2). On the other hand, T. caespitum was sampled over twice as often in forests than T. immigrans (ratio 2.56; Table 2).

Monitoring for new species

Overall, data collected by citizen scientists during the Ant Hunt was significantly closer to cities than data collected by scientists from 1990 to 2019 (mean 31.27 km ± 29.72 SD and 34.52 km ± 23.21 SD, respectively, W = 127,623, p < 0.001). For the Ant Hunt, 101 of 667 observations (15%) were within Denmark’s major cities compared to only 4 out of 448 (0.9%) scientist-collected samples. Data from the Ant Hunt was also significantly closer to major harbours than data from scientists (20.66 km ± 14.90 SD and 28.25 km ± 12.62 SD, respectively, W = 98,154, p < 0.001). Only six of the major harbours in Denmark were outside of cities with more than 5,000 residents (Fig. 4), suggesting a high correlation between industrial harbours and residential areas.

Both scientists and citizen scientists were significantly biased regarding which land cover classes they sampled within (Chi-square test, χ² = 653.75, df = 6, p < 0.001 and χ² = 3,094.4, df = 6, p < 0.001, respectively). However, although both datasets were biased towards areas with artificial surfaces, the effect was far more pronounced among citizen scientists, who sampled artificially surfaced areas eight times more than expected. Scientists only sampled artificial surface areas three times more than expected by random sampling. Both citizens and scientists avoided agricultural areas, but scientists sampled forests, scrub, coastal areas and wetlands 2–3 times more than expected by random sampling (Table 3).