Current knowledge of the Swedish insect fauna

To assess the current knowledge of the Swedish insect fauna, we used the content of DynTaxa (http://dyntaxa.se), the official checklist of the Swedish flora and fauna. Data were pulled from DynTaxa on February 8, 2017 (data provided in a public GitHub repo, see “Data availability”). We included all species recorded as reproducing ("Reproducerande"), accidental and reproducing ("Tillfälligt reproducerande"), uncertain ("Osäker förekomst"), accidental and not reproducing ("Tillfällig förekomst, ej reproducerande"), nationally extinct ("Nationellt utdöd"), no longer reproducing ("Ej längre reproducerande") or possibly nationally extinct ("Möjligen nationellt outdöd"). Due to lack of knowledge, very few insect species occurrence records in DynTaxa are currently classified into any of these detailed categories; the species are simply assumed to be “reproducing”. In well-known insect groups, a handful of species are known to be accidentals or belong to one of the other categories, but we chose to include them in the species counts to avoid systematic bias towards higher counts in poorly known groups.

Five species lacked occurrence status information in Dyntaxa: Dixella martini (Dixidae), Aphis violae (Aphididae), Diaspis boisduvalii (Diaspididae), Cosmopterix pulchrimella (Cosmopterigidae) and Haplothrips vuilleti (Phlaeothripidae). They were nevertheless included in our analysis as Swedish based on other data. The family placement of Sciarosoma nigriclava (Diptera; listed as the junior synonym S. borealis) is given as uncertain in Dyntaxa; it was included by us in the Diadocidiidae, following the classification in Fauna Europaea [13]. Five species in DynTaxa lack information about family classification, namely Dimorphopterus spinolae and Ischnodemus sabuleti in Blissidae, and three species of Cymus in Cymidae: C. claviculus, C. glandicolor, and C. melanocephalus. The family information was added for these records. For a few taxa, the number of Swedish species given in the DynTaxa 2017 checklist is obviously wrong and was corrected (details in S1 Table).

Historical knowledge of the Swedish insect fauna

Information was extracted from the three most comprehensive surveys published before the start of the Swedish Taxonomy Initiative. The first [14] lists all Swedish insect species known to Linnaeus in 1761. These species were mapped to currently valid species so that they could be placed in the classification used by us. The second [15] and third [12] do not provide species lists. Instead, they summarize the knowledge of the fauna at the time (1920–1922 and 2003, respectively) at higher taxonomic levels, mostly at the family level, and give the estimated number of known species of each taxon. We mapped these numbers onto the higher taxa in our classification as detailed in the “Taxonomic composition” section below and in S1 Table. In addition, the 2003 numbers were corrected for some groups where they are obviously wrong (see S1 Table for details).

For our analyses of data from the Swedish Malaise Trap Project, we needed explicit species lists from 2003 of the target taxa. These were reconstructed with the help of experts and the literature on each group. The resulting check lists from 2003 are provided in the public GitHub repo accompanying the paper (see “Data availability”).

Known European insect faunas

Data on the European insect fauna were obtained from the Fauna Europaea database (version 2.6, distribution 3) kindly provided by Yde de Jong. This version of the database was current as of February 2017. The database includes both taxonomic data and distribution data, detailing the occurrence of each species in each of the European countries and geographic regions recognized in Fauna Europaea. The database dump we used is provided in the GitHub repo.

The Fauna Europaea data were used to analyze the latitudinal gradient in the species richness of known European insect faunas. In this analysis, we excluded all areas outside of Europe, as well as European areas of countries that mostly lie outside of Europe (Russia and Turkey). We also excluded some exotic areas with faunas that are not typical of the country they belong to: Madeira, the Selvagens Islands, the Azores, Canary Islands, and Gibraltar. Country latitude data were taken from a file provided by Google [16]. Land area data were taken from the EuroStat NUTS survey in 2013 [17] and complemented with data from Wikipedia [18]. The raw data are provided in the GitHub repo.

Taxonomic composition

Taxonomic composition of the Swedish fauna was analyzed in terms of the families and orders recognized in the DynTaxa classification as of February 8, 2017. However, at the family level, we divided the three most species-rich insect families (Ichneumonidae, Braconidae and Staphylinidae) into subfamilies to make the family-level units more comparable in size. When data on the Swedish fauna were analyzed from older sources using a less detailed classification [14, 15], we split the listed species numbers based on the available literature from the time (see also S1 Table). The analysis of the European data, and of taxa expected to occur in Sweden, was based on the Fauna Europaea taxonomy [13].

Ecological composition

In the analysis of life-history traits, we focused on two traits that are conservative enough that they can be reasonably assumed, in most cases, to be homogeneous within the family-level groups we used: the main feeding niche and the main feeding (micro-)habitat. This is the niche and habitat of the immature stages (the main feeding stages), and may or may not be the same as the niche and habitat of more short-lived adult stages. Data were taken from standard works [15, 19–23] complemented with data from relevant taxonomic specialists and the primary literature.

Specifically, we defined the feeding niches as follows:

1. Parasite. This includes bloodsuckers, endoparasites (botflies) and exoparasites (lice).
2. Phytophage (plant feeder). This includes both chewers and sap suckers, as well as stem borers, leaf miners, root feeders and gall inducers.
3. Phytophage-parasitoid. This is restricted to all primary parasitoids of plant feeders.
4. Predator. This is restricted to free-living predators, it does not include parasitoids.
5. Predator-parasitoid. This includes both primary parasitoids of predators and all hyperparasitoids (parasitoids of parasitoids).
6. Saprophage (decomposer). This includes scavengers, decomposers, fungivores and microflora/bacterial grazers.
7. Saprophage-parasitoid. This is restricted to all primary parasitoids of decomposers.

The (micro-)habitats of the main feeding stages (larvae, nymphs) were defined as follows:

1. Fungi. Within fruiting bodies of macrofungi.
2. Homeothermic animals. On homeothermic animals (birds, mammals). The habitat of parasites (bloodsuckers, endoparasites and exoparasites).
3. Plants. In or on aerial parts of living plants, excluding wood.
4. Soil. In soil, including ground litter as well as living roots.
5. Temporary habitats. In or on ephemeral resources of rich nutrients, such as bird nests, carrion, dung and fermenting sap. Also includes indoors habitats.
6. Water. In or on water.
7. Wood. In wood.

Groups were coded for the dominant trait; notable exceptions and other comments are given in the tables. Because of different taxon or group circumscriptions, the traits differ in a few cases between the Swedish fauna (S1 Table), the groups used in analyzing inventory data (S4 Table), and the European fauna (S6 Table).

Expert estimates of the true fauna

Expert estimates of the total size of the Swedish fauna were based on the opinions of taxonomic specialists with intimate knowledge of the Swedish insect fauna. They were obtained for all family-level groups in the fauna at two occasions: in 2007, before the Malaise trap inventory material was processed, and in the spring of 2017 (current estimate). The 2007 estimates were based on less extensive surveys of experts, and we only give them here for the target groups used in the species richness analysis of the inventory data (S5 Table). The estimates from 2017 are detailed in S1 Table, together with the experts responsible for them.

For a few groups, we were not able to find local specialists willing to estimate the total size of the fauna. In those cases, MF, RH and FR estimated the total size of the Swedish fauna as follows. For small well-known groups (Diplura, Archaeognatha, Zygentoma, Dermaptera), the current number of known species was accepted as the estimate of the true fauna except in the case of Dermaptera, where we added one expanding species known from nearby countries (Labidura riparia). Protura are relatively well-known but rarely studied from a faunistic or systematic viewpoint, and are thus likely to contain undiscovered species. We estimated the true fauna to be 30% larger than the currently known one (an increase by a factor of 1.3).

For the two main groups of vertebrate parasites, Phthiraptera and Siphonaptera, we based our estimates on each family’s host preferences. Families of parasites exclusively or mainly on man and domestic mammals were considered well-known and we accepted the current number as final. For families of parasites on wild rodents, insectivores, bats and carnivores, plus siphonapteran parasites on birds, we increased current numbers by a factor of 1.4. For phthirapteran families including parasites of birds we used a factor of 1.6. This seemed reasonable considering the patterns of host range, the fauna of hosts in Sweden, the fauna of these families in better studied neighboring countries, and the relatively poor global taxonomic knowledge of phthirapteran bird parasites. In two species-poor families where these considerations suggested no addition, we nevertheless added expected species that are known from neighboring countries (one each in Enderleinellidae and Hoplopleuridae).

Malaise trap inventory

For critical groups, we checked the expert assessments against two independent species richness estimates based on data from the Swedish Malaise Trap Project. The project is the first systematic inventory of the whole diversity of the Swedish insect fauna, and it particularly targets the most poorly known groups in the orders Diptera and Hymenoptera [9, 10]. The material was collected using 73 Malaise traps, deployed at different sites throughout the country and representing a wide range of habitats. Most of the traps were run continuously from 2003–2006, but some traps were run for only part of this period, and a couple of traps were run in the period 2007–2009. Detailed data on the trap sites and the samples collected (1,919 samples in total) are provided in S2 Table (see also http://www.stationlinne.se/en/research/the-swedish-malaise-trap-project-smtp/). The trap sites were classified into four biogeographic regions and six habitat classes to simplify analyses of the data (S2 Table).

The inventory material (unsorted material and identified specimens) is deposited at the Swedish Museum of Natural History (http://www.nrm.se) through its collaboration with the Station Linné Field Station (http://stationlinne.se). The material is continuously sorted into more than 300 taxonomic fractions, mostly at the family or subfamily level (see http://www.stationlinne.se/en/research/the-swedish-malaise-trap-project-smtp/taxonomic-units-in-the-smtp/), which are sent out to experts for identification. The returned data are validated and cleaned, and identifications are matched to the national checklist (DynTaxa) as far as possible. Data on the sorted fractions is available in a separate dataset at the Global Biodiversity Information Facility portal (https://www.gbif.org), at the Swedish natural history collection portal (https://naturarv.se) and at the Swedish biodiversity data hub (https://bioatlas.se). For access to the material and for questions about the data, contact DK or JB.

Here, we analyze data from the first 165,000 specimens that have been identified, representing about 1% of the total catch [10]. The data analyzed here represent all the cleaned data files that were available in the spring of 2017 (S3 Table). In total, there are 127 files, most of which contain abundance data (117 files), but some of which only have incidence data (10 files). The incidence data include several groups (Braconidae: Cheloninae, Braconidae: Rogadinae, Cecidomyiidae: Porricondylinae (sensu lato) and Psylloidea) for which abundance data were not reported consistently for all samples and species. The files are provided in the Data Package in the Supporting Information.

We divided the data into 79 datasets, each corresponding to a set of samples of a particular taxon or assemblage of taxa (“analysis taxon”) processed by a different expert or group of experts (S4 Table). The data are available in the data files provided as Supporting Information. The analysis taxa mostly correspond to families or subfamilies of Diptera and Hymenoptera, and to orders or major portions of orders for other insect groups.

The analyzed information is estimated to represent from 2 to 67% of the entire catch of each analysis taxon (S4 Table). In total, the abundance data comprise 157,225 specimens (30,643 records) identified to 3,916 taxa. The incidence data represent another 1,516 observations of 279 taxa. The incidence data roughly corresponds to 7,800 specimens, if this material contained the same average number of specimens per species and sample as the material covered by the abundance data.

Specimens or records that were determined only to genus level (marked by “sp.” in the Species field in the data file provided in the Supporting Information) were removed prior to analyses of species richness and compositional trends. This affected 14,034 specimens (161 taxa) in the abundance data (most of these are females of Phoridae, which cannot at this point be identified to species) and 36 records (8 taxa) in the incidence data. In a few cases, determinations were to unspecified species that may or may not be distinct (affecting 40 records of abundance data marked by “sp. indet.” in the Author field); these records were included in the analyses under the assumption that they represent distinct species.

Performance of species richness estimators

To assess the performance of the species richness estimators, we used test data from two sources. First, we used well-studied groups, defined as those groups where the inventory has not discovered any new species (S4 Table). As the ‘true’ diversity for these groups, we used the number of species in the current expert estimate, which was typically very close to if not identical to the number of known species (S4 Table). As a second source of test data, we used the subset of species known in 2003 in the more poorly studied groups, that is, those groups where the inventory did discover new species (S4 Table). Specifically, the abundance or incidence data for the known species in those groups were used to estimate the total number of known species in the same groups. We detected no qualitative difference between these two data sources (well-known groups versus known species in poorly studied groups), except that the first set typically included a smaller number of specimens, and the sample represented a much smaller fraction of the total species pool. We only included groups with more than 100 specimens observed from more than 5 sites in the analyses. The accuracy of the estimators was measured using the squared error and bias of the log estimates, that is, and , respectively, where is the estimate and S the true value.

Statistical analyses and visualization

For statistical analyses and for generating the plots shown in the paper, we used standard functions in R, version 3.3.1 [35]. The R scripts are available in the Data Package of the Supporting Information.

Size of the fauna

The experts estimated that the Swedish insect fauna consists of approximately 33,000 species (S1 Table). This exceeds the 2003 estimate of the known fauna by 8,600 species and suggests that there are still around 5,500 species that remain to be discovered or described. Those species are concentrated, however, to a small number of taxa. Only ten of the 663 family-group taxa are expected to contain more than 100 of these missing species; only two of these taxa (Ichneumonidae: Cryptinae and Pteromalidae, both Hymenoptera) still await attention from taxonomists funded through the Swedish Taxonomy Initiative.

The identified specimens from the Malaise trap inventory (Fig 1) fall into 4,026 species, of which 689 have been putatively identified as new to science. Of the latter, only 87 have been described so far, forming part of the 301 new species to science recorded from Sweden since the start of the Swedish Taxonomy Initiative (S1 Table). A geographic and habitat breakdown of the inventory data for some key taxa of Diptera and Hymenoptera shows that many of the new discoveries, including species new to science, are made in mixed forests in the boreal and boreonemoral zones (Figs 1 and 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Results of the Malaise trap inventory by biogeographic region for ten poorly known groups of Hymenoptera and Diptera.

The species new to science, new to Sweden, and known previously in each biogeographic region are shown as a fraction of the total species pool of the group encountered in the inventory. The dots on the map mark trap sites. Sampling effort is given as the number of trap days represented by the processed samples for each taxon and region. Many of the new species discovered in the inventory are found in the boreal and boreonemoral zones. For details on the circumscription of each group, see S4 Table.

https://doi.org/10.1371/journal.pone.0228561.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Species found in the Malaise trap inventory by habitat.

Data are given for the same taxonomic groups as those in Fig 1. See S4 Table for detailed circumscription of each analysis group. Many of the new species are found in forest habitats.

https://doi.org/10.1371/journal.pone.0228561.g002

Of the inventory data we analyzed, the bulk (35 groups; 120,530 specimens) consists of abundance records for poorly known groups, where taxonomists have found new Swedish species in the material. Together, these groups account for 29% (2,525 of 8,655 species) of the fauna that experts assume is present in the country but that had not been recorded in 2003 (Table 1). Using the ratio of new to previously known species in the catch of these groups to estimate total species richness (MR estimator, analogous to mark-recapture approximation of population size) produces results that are generally in line with expert guesses for these groups (Tables 1 and S5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Species richness estimates based on abundance data for poorly known insect groups from the Malaise trap inventory.

“Known spp 2003” is the number of species known from Sweden prior to the inventory. For the inventory material, we give the proportion of the catch (Prop) and the number of specimens (N) processed, as well as the total number of species found (Spp), and the number of those species that were not recorded as Swedish prior to the inventory (New). Species richness estimates include expert guesses (Expert) from 2002, before the inventory (first number) and from 2017, after partial results were known (second number). Estimates also include extrapolations using mark-recapture (MR), Chao’s estimator based on the entire material (Chao1) or on occurrences of species across sites (Chao2); and the new combined non-parametric estimator introduced here (CNE). See S5 Table for more detailed results and additional species richness estimates. Data for Ichneumoninae do not include the tribe Phaeogenini.

https://doi.org/10.1371/journal.pone.0228561.t001

Before applying species richness estimators based on abundance and incidence data from the inventory, we tested the accuracy of select methods using data for well-known groups, and for the species pool known in 2003 of poorly known groups (S5 Table). In both cases, we thus know the total number of species in the pool, which allows us to compute the accuracy of the estimates. The tests show that estimates tend to become more accurate as the number of specimens (Fig 3A), sites (Fig 3B), or specimens per site (Fig 3C) increases. However, it is the coverage (proportion of the species pool that has been sampled) that largely determines accuracy (Fig 3D): it is only at relatively high coverage (50–60% or more of the total number of species in the pool) that the estimates become reasonably robust. When half the species have been sampled, one can expect a worst-case squared error of the log estimates of approximately 0.4, which means that the estimate may be off by a factor of 1.9 (close to half or double the real species number). The average estimate is considerably better, however.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Accuracy of different estimators in predicting the species richness of well-studied groups, and the species pool known in 2003 of more poorly studied groups.

The plots show the accuracy (measured as squared error of log estimates) of the species richness estimates as a function of the number of specimens (A), the number of sites (B), the number of specimens per site (C), and the fraction of the species pool sampled (D). The CNE outlier in (D) is Ichneumonidae: Adelognathinae, one of the groups with the smallest number of specimens per site.

https://doi.org/10.1371/journal.pone.0228561.g003

Chao1 and Chao2 consistently underestimate species diversity unless the coverage is very high, while CNE does better in terms of bias at the cost of increased variance (Figs 3–4). All examined estimators are biased downwards, that is, they underestimate the size of the fauna, especially when a small proportion of the species pool is sampled (Fig 4). The bias is less severe for the Preston and CNE estimators, but this comes at the expense of increased variance. The estimators that take heterogeneity across sites into account (Chao2 and CNE) do considerably better than the estimators that do not (Chao1 and Preston) when a small fraction of the species pool is sampled.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Bias in species richness estimators as a function of the fraction of the species pool sampled.

We show the bias in four species richness estimators: Chao1 (A), Chao2 (B), Preston (C) and CNE (D). Bias is measured on the log scale; the horizontal line represents absence of bias.

https://doi.org/10.1371/journal.pone.0228561.g004

The empirical correction proposed in [34] is based on the sampling effort, specifically the proportion of singletons and the number of specimens per species for estimators based on abundance data for the total sample. The corresponding measures of sampling effort for site-based data are the proportion of uniques, that is, the proportion of species that are encountered at a single site, and the number of site occurrences per species. These factors do correlate with the accuracy of estimators (S1 Fig), but not as strongly as the coverage of the species pool (Fig 3D). Even at high sampling intensities (the extreme right of the x axes in S1 Fig), the species richness estimates can be quite misleading. These are conditions when the corrected versions of the estimators are similar or identical to the original ones. There is actually a tendency for the non-parametric estimators to do worse under these settings, while there is no such trend for the parametric Preston estimator.

In terms of bias as a function of coverage of the species pool, the corrected Chao1 estimator based on abundance data for the entire sample does not do much better than the corresponding uncorrected versions (S2 Fig). However, the corrected estimators (Chao1P, Jack1P, Jack2P) based on site occurrence data show a slight but distinct improvement over the uncorrected versions (compare S2 FigB with Fig 4B).

The poorly-known groups in the inventory are represented by six times as many specimens as the well-known groups, and the inventory sample covers 50% or more of the estimated fauna (according to expert assessments) in many cases (S5 Table and Fig 5), suggesting that species richness estimates may be reliable or at least indicative of true species richness for these groups. The CNE estimates (Table 1; see also S5 Table) also tend to be in line with expert guesses and mark-recapture estimates, although there is considerable variation across groups. Conspicuous outliers are associated primarily with groups where the inventory sample does not cover the expected fauna well (S5 Table). Interestingly, the corrected versions of the estimators based on site incidence data come close to the CNE estimates, particularly Chao2P and Jack2P. The corrected Chao1p estimates, however, are not very different from the Chao1 estimates; both appear to significantly underestimate the species pool (S5 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Species accumulation curves for some groups studied in the Malaise trap inventory.

Accumulation curves are shown for the nine groups with the largest number of identified specimens (Tables 1 and S5). Accumulation curves are based on 10,000 random perturbations of the accumulation order of sites. The horizontal lines correspond to the expert guess from 2017 (Expert, dashed line), the estimate from the mark-recapture method (MR, dotted line) and the estimate from the combined nonparametric method (CNE, dashed and dotted line). Major disagreement among estimates tends to be associated with groups that are not covered well by the sample. In the Phoridae, the low coverage may be partly due to the small fraction (6%) of the material processed; in other cases, like the Symphyta (30% of the material processed), it appears instead that it is due to the inefficiency of Malaise traps in collecting these groups. DNA barcoding indicates that the accumulation curve for Phoridae is too steep because aberrant male-genitalia variants are sometimes misclassified by taxonomists as representing separate species [36]; this is likely to partly explain the poor performance of species richness estimators for this group.

https://doi.org/10.1371/journal.pone.0228561.g005

Composition of the fauna

Given the good agreement between the expert assessments and the estimates based on the inventory data for many of the critical groups, we assume in the following that the real fauna corresponds to the expert assessments. What can we then say about its composition? Taxonomically, the Swedish fauna is dominated by Hymenoptera and Diptera (Figs 6 and 7 (left)), both orders dwarfing the Coleoptera (beetles) and Lepidoptera (butterflies and moths). There are more than twice as many species of parasitic wasps alone as there are beetle species (Fig 6). With respect to ecological composition, our results indicate that about half of the fauna consists of phytophagous species and their parasitoids, while the decomposers and their parsitoids constitute a third (Fig 7). The remaining sixth largely consists of predators and their parasitoids. The species of plant feeders on average host nearly one parasitoid species each, while the decomposer and predator guilds are attacked by far fewer parasitoids. In terms of main feeding microhabitat, the fauna is dominated by species associated with plants (Fig 7).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Detailed view of the taxonomic composition of the true Swedish insect fauna, as estimated here.

Each colored fragment represents a different family (subfamily for Ichneumonidae, Braconidae and Staphylinidae); the width is proportional to the number of species. The families are grouped into monophyletic higher clades according to our current understanding of insect relationships. The inventory has shown that the fauna is much richer in Hymenoptera (sawflies, wasps, ants and bees) and Diptera (midges, gnats, mosquitoes and flies) than expected. These groups vastly outnumber Coleoptera (beetles), Lepidoptera (butterflies and moths) and other insect orders in terms of species diversity. Photographs (CC BY) John Hallmén.

https://doi.org/10.1371/journal.pone.0228561.g006

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Composition of the Swedish insect fauna, and changes in our understanding of the composition brought about by the inventory.

The taxonomic and ecological composition of the fauna, as estimated here (33,000 spp.), is given to the left. The composition of the species that were still unknown in 2003 (8,600 spp., many of which still await discovery), as predicted by their phylogenetic affinities (family or subfamily placement), is shown to the right.

https://doi.org/10.1371/journal.pone.0228561.g007

Historical trends in the knowledge of the fauna

Analysis of the historical records show that the knowledge of the fauna has always been worse than biologists realized at the time. The knowledge has also been strongly biased both taxonomically and ecologically. Less than 5% of the true fauna was known to Linnaeus [18], mainly conspicuous insects living on plants, such as butterflies, moths and herbivorous beetles (Fig 8). During the 20^th century [11, 19], as the fauna became better known, the proportion of Hymenoptera and Diptera increased dramatically, as well as the proportion of decomposers and parasitoids.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Changes in apparent composition of the Swedish insect fauna over time as it has become better known.

Over time, the apparent fraction of Hymenoptera and Diptera species has increased dramatically. The same is true for the community of decomposers and parasitoids (red bars). Changes in apparent microhabitat composition have been less pronounced; most insects live on or in plants. However, the proportion of species found in other microhabitats has increased over time.

https://doi.org/10.1371/journal.pone.0228561.g008

These trends are quite noticeable even when comparing the fauna known in 2003 with the real fauna estimated here (Fig 7 (right)). The new species belong to a large extent to the Hymenoptera and Diptera. Many of them are predicted to be decomposers or parasitoids, and more often associated with fungi or temporary habitats than other insect species. Thus, the new species substantially change our understanding of the composition of the fauna (see also Fig 6)

Latitudinal gradients in size and composition

If we assume a power law relating area to species diversity, then the Swedish Malaise trap inventory data show a decline in diversity from the nemoral to the boreal zone. Specifically, the logarithm of the number of species encountered in the inventory divided by the logarithm of the area of the corresponding biogeographic region decreases from 0.8 over 0.7 to 0.6 through the nemoral to boreal transition. There is also a significant increase in the proportion of Diptera species and in the proportion of decomposers with increasing latitude, but there is no such trend for Hymenoptera or parasitoids (Fig 9).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Compositional changes in the Swedish insect fauna over the nemoral to boreal transition, as indicated by the Malaise trap inventory data.

The plots show the changes in the proportion of species that belong to Diptera (A) and Hymenoptera (B), and the proportion of species that are decomposers (C) and parasitoids (D). The p values were computed by comparing the observed values with a reference distribution generated using 10,000 random permutations of the trap assignments to biogeographic region. The arctic zone was excluded from this comparison because there is an order of magnitude less inventory data for this zone, and data are lacking for key groups. Sample sizes: nemoral zone 18 traps, 1,557 species; boreonemoral zone 29 traps, 2,693 species; and boreal zone 22 traps, 1,691 species.

https://doi.org/10.1371/journal.pone.0228561.g009

The patterns in known European insect faunas are partly different. As expected, the species richness of national faunas is determined to a large extent by the land area of the country, as shown by a log–log regression analysis (Fig 10A). The residuals from this regression (Fig 10B) reveal that there are three outliers—Belarus (BY), San Marino (SM) and Iceland (IS)—that all have unexpectedly small insect faunas given their land areas. This may be because the fauna is little known (or poorly recorded in Fauna Europaea), or because of other factors. Controlling for the area effect reveals a significant but surprisingly weak latitudinal gradient in species richness (Fig 10C), and this effect disappears completely when all three outliers are removed (Fig 10D), or when just the most extreme outlier, Iceland, is omitted (plot not shown; r² = 0.03, p = 0.31).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Latitudinal gradient in overall species richness of national European insect faunas, as documented by the Fauna Europaea database.

(A) Relationship between country area and the species richness of its insect fauna (log-log plot). (B) Residuals from the previous plot. Note three outliers—Belarus (BY), San Marino (SM) and Iceland (IS)—with unexpectedly small insect faunas. (C) Latitudinal gradient in species richness after the area effect is removed, using all data. (D) Ditto, after removal of the three outliers.

https://doi.org/10.1371/journal.pone.0228561.g010

The Fauna Europaea data reveal a significant northern increase in the proportion of Diptera species (Fig 11A) and in the proportion of decomposers (Fig 11B), as indicated also by the Swedish Malaise trap inventory data (Fig 9). However, the European data also support a northern increase in the proportion of Hymenoptera species (Fig 11C and parasitoid species (Fig 11D). There is no suggestion of such trends in the Swedish data (Fig 9).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. Latitudinal trends in the composition of national European insect faunas, as documented by the Fauna Europaea database.

The plots show the latitudinal gradients in the proportion of species that belong to Diptera (A) and Hymenoptera (B), and the proportion of species that are decomposers (C) and parasitoids (D). Three outliers (see Fig 10B) were removed from these analyses.

https://doi.org/10.1371/journal.pone.0228561.g011