Print
The Rubens morph of Formica exsecta Nylander, 1846 and its separation from Formica fennica Seifert, 2000 (Hymenoptera, Formicidae)
expand article infoBernhard Seifert
‡ Senckenberg Museum of Natural History Görlitz, Görlitz, Germany
Open Access

Abstract

A study of numeric morphology-based alpha-taxonomy (NUMOBAT) considering the species Formica exsecta Nylander, 1846 and F. fennica Seifert, 2000 was performed in 166 nest samples with 485 worker individuals originating from 117 localities of the Palaearctic west of 59°E. The presence of intraspecific pilosity dimorphism is shown for F. exsecta. The setae-reduced phenotype, termed the Rubens morph, shows a frequency of about 25%, and the more abundant setae-rich phenotype, termed the Normal morph, one of 75%. The frequency of nests containing workers of both phenotypes is 15.5% in 58 samples from Denmark, Sweden, and Finland. Applying the DIMORPH test of Seifert (2016) on this territory, it is demonstrated that the association of Rubens and Normal phenotypes within the same nest cannot be interpreted as parabiosis of independent species (p=0.017) or as temporary (p=0.0004) and permanent (p=0.0001) socially parasitic association, whereas genetically mediated intraspecific dimorphism is most likely (p=0.659, all p data according to Fisher’s exact test). The Rubens morph of F. exsecta is phenotypically most similar to F. fennica but is safely separable by four different forms of exploratory data analyses using nest centroids (NC) as input data: NC-Ward, NC-part.hclust, NC-part.kmeans, and NC-NMDS-k-means. Data on zoogeography and the narrow climate niche indicate that F. fennica is unlikely to occur in Norway.

Key Words

cryptic species, dimorphism, numeric taxonomy, parabiosis

Introduction

The distributional range of Formica exsecta Nylander, 1846 covers the Palaearctic from Iberia to Kamchatka, includes the submeridional, temperate and boreal zones in the horizontal scale, and extends from the planar to the subalpine zone in the vertical scale (Seifert 2000, 2018, Seifert and Schultz 2007). Strong pilosity and color variation has led in the past to the description of various other taxa close to F. exsecta: F. exsecta rubens Forel, 1874, F. exsecta exsectopressilabris Forel, 1874, F. exsecta etrusca Emery, 1909, F. dalcqi Bondroit, 1918, F. exsecta sudetica Scholz, 1924, F. kontuniemii Betrem, 1954, F. nemoralis Dlussky, 1964, and F. mesasiatica Dlussky, 1965. Whereas the species status of F. mesasiatica was recognized by Seifert (2000), albeit with some caution, he synonymized the seven other taxa with F. exsecta. This assessment is in line with the current state of information after my examination of over 1100 samples of the subgenus Coptoformica from the whole Palaearctic, and with 800 of these samples being investigated by Numeric Morphology-Based Alpha-Taxonomy (NUMOBAT).

Seifert (2018) stated the presence of a pilosity dimorphism in F. exsecta. He distinguished a setae-reduced Rubens morph and a setae-rich Normal morph and reported that the Rubens morph is very similar to Formica fennica Seifert, 2000. Formica fennica has a narrow habitat spectrum and is known so far from only four sites in southern Finland and two sites in the Caucasus whereas F. exsecta shows the widest habitat spectrum, largest total abundance and biggest geographical range of any Coptoformica species. In a study from Finland, using microsatellite data, Hakala et al. (2018) found no genetic differences between two subjectively established morphological clusters. Within their material, they classified 38 samples as Formica exsecta and 33 samples as F.fennica”. The failing separation by microsatellite data in the samples of Hakala et al. is easily explained by the fact that >90% of their F.fennica” samples actually belonged to the Rubens morph of F. exsecta. Using the DIMORPH test of Seifert (2016), the first aim of this paper is to demonstrate that the Rubens and Normal morph of F. exsecta do not represent different species. Using NC-clustering methods (Seifert et al. 2013, Csősz and Fisher 2015), the second aim is to show that the Rubens morph is reliably separable and heterospecific from F. fennica. Prompted by the paper of Hakala et al. (2018), I perform here another attempt to convince morphology-based alpha-taxonomists to abandon idiosyncratic approaches and to adopt reproducible numeric character description and evaluation when aiming to investigate the real structure of biodiversity.

Material and methods

To have an improved separation between the Formica exsecta morphs and between the Rubens morph of F. exsecta from F. fennica, the materials of this study were restricted to the Palaearctic west of 59°E, that is, west of the Ural Mountains. NUMOBAT data were recorded in 166 nest samples and 485 worker individuals, the majority of which is deposited in the collection of Senckenberg Museum of Natural History Görlitz. This material listed below in the following sequence and format: site, date in the yyyy.mm.dd format, field sample number “field No” which is found on the mounted specimens [latitude and longitude in decimal format, meters above sea level]. The accuracy of coordinates is proportional to the number of decimal points and “xx” in the sampling date sequence mean missing data. In some samples without any direct or derived information of date, the assumed period is given and the collector is named if known.

Formica exsecta Nylander, 1846

A total of 152 nest samples with 431 workers were investigated.

Austria: Brunau, 2001.09.xx [47.23, 10.85, 700]; Fernpass, 1994.07.04, no. 20 [47.40, 10.90, 1420]; Gilfert, Lafaster-Alm, 1995.10.22 [47.26, 11.76, 1758]; Großglockner, Franz-Josephs-Höhe, 1968.xx.xx [47.075, 12.751, 2369]; Großglockner, Glocknerhaus, 1994.06.20 [47.070, 12.769, 2200]; Gschnitz, Sandertal, 1995.09.24 [47.00, 11.40, 1700]; Gschnitz, Schatthang, 1995.09.24 [47.00, 11.40, 1900]; Gschnitz, Blockfeld, 1995.09.24 [47.00, 11.40, 1900]; Kitzbühel, Bischof, 1994.08.22 [47.405, 12.512, 2000]; Kleiner Gamsstein, 1995.10.22 [47.733, 14.467, 1100]; National Park Kalkalpen, 2011.08.05 [47.750, 14.438, 990]; Pottenbrunn-3 km SSW, 1994.05.12, no. 043 [48.215, 15.718, 330]; Rofan, 1994.07.07, no. g23 [47.45, 11.78, 2000]; Seetaler Alp, Zirbitzkogel, 1971.09.04 [47.06, 14.55 2000]; Spitz-15 km W, Jauerling, 1990.07.05 [48.353, 15.332, 700]; Vikartal, 1995.10.08 [47.20, 11.45, 1500]. Bulgaria: Rhodopes: “Betova” [site not identifiable], 1985.07.03 [42?, 24.5?, 1300]; Rhodopes: Pamporovo, 1975.07.20 [41.61, 24.67, 1400]; Rila: Borovec, 1977.08.01 [42.26, 23.61, 1300]; Rila: Borovec, 1988.06.16 [42.26, 23.61, 1900]. Denmark: Jutland, Rye, 1986.08.xx [56.08, 9.73, 77]. Finland: Broända, 1996.07.07, no. 27, no. 130 [60.11, 24.27, 30]; Jyväskylä, Kypäramäki, 1998.06.01 [62.24, 25.70, 140]; Järajärvi-S, 2002.07.23 no. 039 [69.641, 29.046, 102]; Järajärvi-S, 2002.07.23 no. 040 [69.641, 29.045, 101]; Järajärvi-S, 2002.07.23 no. 041 [69.642, 29.045, 101]; Järajärvi-S, 2002.07.23 no. 042, no. 044 [69.642, 29.046, 100]; Kiannanniemi, 2002.07.19, no. 018 [65.165, 29.105, 207]; Kiiminki-26 km ENE, 1996.07.15, no. 146 [65.208, 26.325, 120]; Käylä-2 km NE, 1996.07.17, no. 147 [66.320, 29.170, 250]; Käylä-3 km E, no. 113, no. 159 [66.310, 29.203, 250]; Köyliö, 1999.06.25 [61.117, 22.300, 70]; Puhos-6.5 km WNW, 1996.07.13, no. 114 [62.107, 29.793, 97]; Renko-13 km WSW, 1996.07.11, no. 19 [60.836, 24.067, 136]; Sodankylä-26 km NNE, 1996.07.18, no. 55 [67.64, 26.74, 205]; Sodankylä-33 km N, 2002.07.21, no. 029, no. 031 [67.707, 26.749, 208]; Tammela-9 km ENE, 1996.07.11 [60.83, 23.93, 120]; Tronsböle-0.2 km NE, Oby, 2015.05.07, no. A, no. B [59.940, 23.198, 20]; Utajärvi-2 km NE, no. 120 [64.728, 24.426, 80]. France: Mt. Canigou, pre 1918, type F. dalcqi [42.51, 2.45, 2000]; Mt. Canigou, pre 1930 (leg. Weiss) [42.51, 2.45, 2200]; Refuge des Besines, 1998.09.17, no. 016 [42.604, 1.868, 2100]. Germany: Althüttendorf, 1985.05.07 [52.96, 13.80, 71]; Canthnitz-0.7 km E, 1997.08.23, no. 220 [53.376, 13.394, 103]; Dabelow, 1982.06.01 [53.248, 13.198, 64]; Dallgow-S, 1996.08.03 [52.52, 13.06, 41]; Eberswalde, 1987.06.12 [52.83, 13.79, 31]; Eberswalde-2.4 km E, 1997.08.21 [52.82, 13.85, 66]; Eberswalde-5 km N, 1995.07.10 [52.89, 13.88, 38]; Eberswalde-Buchholz, 1985.05.07 [52.88, 13.76, 70]; Eberswalde-Finow, 1987.04.12 [52.84, 13.73, 35]; Federsee: Wildes Ried, 1990.09.xx [48.045, 9.648, 583]; Federsee: Wildes Ried, 1991.05.xx [48.045, 9.648, 583]; Geesow, 1987.08.03 [53.239, 14.388, 25]; Meseberg, 1982.08.30 [52.95, 13.10, 50]; Müritzhof, 1988.07.27 [53.45, 12.74, 70]; Carwitz, NSG Hauptmannsberg, 2000.04.22, no. 014, no. 016, no. 017, no. 019, no. 024 [53.31, 13.455, 115]; Neulöwenberg 1982.06.xx [52.90, 13.19, 67]; Oberhersdorf-2 km ENE, 1991.09.04 [50.180, 6.542, 580]; Ödenwaldstetten, 1991.05.19, no. 19 [48.35, 9.39, 750]. Italy: Abetone, pre. 1920 [44.143, 10.666, 1400]; Abetone, Seletta, 1960.07.15 [44.132, 10.644, 1700]; Apennino Modenese: Le Pozze, 1941.07.26, [44, 11, 1000, guess]; Bosco di Corniglio, 1985.08.xx [44.44, 10.04, 880]; Corvara: Colfosco, 1993.08.22 [45.56, 11.86, 1900]; Monte Cimone, Lago Ninfa, 1959.xx.xx, no. 55–57 [44.21, 10.72, 1500]; Praccia, 1890.07.xx, type F. etrusca [44.061, 10.911, 750]; Pejo-6 km N, Stelvio NP, 2004.06.16, no. 18 [46.411, 10.688, 2400]; Toscanian Alps: Giulia (leg. Wolf) pre 1945 [44,10,1000, guess]. Netherlands: Bergen, 1994.09.13 [52.670, 4.677, 11]. Norway: Alta-9.5 km S, 2015.07.10, no. 1, no. 2 [69.911, 23.081, 90]; Gjoktbukmyra, 2016.07.11, no. 4 [69.147, 29.207, 64]; Hedmark: Dalholen, 1993.06.28 [62.191, 9.746, 820]; Klingenberg, 2015.08.20, no. 6 [61.017, 11.855, 493]; Klingenberg-E, 2012.07.13, no. 93 [61.001, 12.060, 460]; Klingenberg-Ulva, 2012.07.13, no. 92 [61.001, 12.059, 462]; Osen: Drageid: Seter-4.8 km E, 2012.04.29 [64.392, 10.589, 45]; Osen: Drageid, 2015.08.24, no. 1–3 [64.392, 10.589, 55]; Osen: Drageid, 2016.08.23, no. 3 [64.391, 10.590, 59]; Oyermoen, 2012.06.27, no. 71 [60.248, 12.442, 300]; Tysil: Torberget, 2015.08.21, no. 4, no. 5 [61.105, 12.019, 528]; Tysil: Torberget, 2016.07.23, no. 1, no. 2 [61.105, 12.019, 528]. Russia: Kormovische, 2001.xx.xx, no. U79 [56.83, 57.95, 250]; Svenigorod, 1985.08.xx, no. 623 [55.70, 36.72, 150]; Voronesh Zapovednik, 1962.08.29, no. 221, type F. nemoralis [51.809, 39.446, 130]. Slovenia: Loibl Pass-5 km S, 1994.07.11, no. g10, no. 12 [46.404, 14.277, 700]. Spain: Camprodon-10 km NW, 1994.07.13 [42.401, 2.304, 1600]; Sierra de Guadarrama, 2009.10.16 [40.823, -3.960, 1824]. Sweden: Aaland, 1998.xx.xx, no. S60 [60.23, 19.95, 20]; Abisko, 1951.07.xx [68.50, 18.66, 500]; Andrarum, 1992.06.09, no. g4, no. g17 [55.708, 13.966, 115]; Ange-WSW, 2002.08.03, no. 029–032 [62.422, 15.000, 269]; Arvidsjaur-Aljeplog, 2002.07.29, no. 013, no. 014 [65.926, 18.311, 464]; Attonträsk-4 km SW, 1996.07.29, no. 109, no. 125 [64.401, 18.004, 471]; Attoträsk-6 km NE, 1996.07.28, no. 49, no. 76, no. 137, no. 141 [64.461, 18.153, 430]; Degeberga- 6 km SSW, 1992.06.09, no. g89 [55.784, 14.045, 125]; Falkenberg, 2000.xx.xx [56.92, 12.49, 32]; Kalix, 2000.xx.xx, no. S96, no. S100, no. S101 [65.84, 23.10, 7]; Orsa-45 km N, 2002.08.04, no. 036, no. 037 [61.407, 14.819, 480], Orsa-45 km N, 2002.08.04, no. 039 [61.409, 14.821, 476]; Storuman, 2002.07.31, no. 018 [64.924, 17.034, 422]; Storuman, 2002.07.31, no. 019 [64.923, 17.033, 423]; Sweg, 2002.08.04, no. 034 [62.140, 13.984, 378]; Upland: Hallnäs, 1998.xx.xx, no. S75 [60.53, 17.87, 10]; Öland: Böda, 1992.06.14 [57.250, 17.06, 9]; Öland: Borgholm Slott, 2000.10.08, no. 008–010 [56.870, 16.640, 60]; Östernoret-3 km SW, 1996.07.29, no. 71, no. 126 [64.071, 17.290, 320]. Switzerland: Alp La Schera, 1998.07.26, no. 51 [47.648, 10.194, 2080]; Valle Fermaur: Apples, pre 1874, type F. rubens [46.550, 6.433, 622]; La Punt-3.7 km NW, no. 58, no. 106 [46.588, 9.900, 2150]; S-Scharl-0.5 km S, 1998.07.27, no. 184 [46.713, 10.336, 1900]; S-Scharl-0.8 km S, 1998.07.27, no. 243 [46.710, 10.334, 2100]; S-Scharl-0.9 km S, 1998.07.27, no. 182 [46.709, 10.333, 2215]; S-Scharl-2.6 km SSE, 1998.07.27, no. 146 [46.697, 10.350, 2140]; S-Scharl-2.6 km SSE, 1998.07.27, no. 247 [46.697, 10.347, 2175]; Stabelchod, 1998.07.28, no. 81 [46.661, 10.241, 1940]; Stabelchod, 1998.07.28, no. 174 [46.665, 10.243, 1990]; Stabelchod- 1 km W, 1998.07.28, no. 172 [46.662, 10.224, 1880]; Ticino: Piora, 1981.07.22 [46.544, 8.686, 1930]. Turkey: Gerede, 1976.02.27 [40.810, 32.192, 1600].

Formica fennica Seifert, 2000

A total of 14 nest samples with 54 workers were investigated.

Azerbaijan: Ilisu, 2006.06.03, no. 24 [41.457, 47.063, 1706]. Finland: Iisalmi Kotikylä, 1998.07 [63.450, 27.167, 100]; Iisalmi Kotikylä, 2009.07.24 [63.450, 27.167, 100]; Ilomantsi, Maukkula, pre 1996 (leg. Saaristo) [62.62, 30.84, 170]; Luhanpää-1.42 km NE, 1999.07.07, no. 273 [61.058, 25.050, 130]; Luhanpää-1.42 km NE, 2001.07.22, no. 1, no. 2 [61.058, 25.050, 130]; Puhos-6.5 km WNW, 1996.07.13, no. 86, no. 105, no. 119 (type F. fennica) [62.108, 29.800, 100]; Puhos-6.5 km WNW, 2002.07.19, no. 013–015 [62.108, 29.800, 100]. Georgia: Schenako, 1985.08.01 [42.732, 45.662, 1600].

Stereomicroscopic equipment and measurement procedures were as given in Seifert (2019).

NUMOBAT data were recorded in the primary characters CS, CL, CW, ClySet, ClyPub, EyeHL, F2L, F3L, nCox, nHTfl, nMet, nOce, OceD, SL, sqPDG, sqPDO, TERG, and T3f; the recording rules are explained in Seifert (2018: 399–407). It is repeated and emphasized here that the performance of species delimitation in Coptoformica is strongly dependent from identification of basal remains of broken-off setae by high-resolution microscopy. The standardized characters were used in the analyses as variable of absolute body size (CS), as shape variables (CL/CW, SL/CS, OceD/CS, F2L/CS, F3L/CS, F2L/F3L), as variables describing seta conditions (ClySet, ClyPub, EyeHL, nCox, nHTfl, nMet, nOce, OceD, TERG, T3f) and as variables describing pubescence conditions (ClyPub, sqPDG, sqPDO).

Hypotheses on phenotype dimorphism within the F. exsecta sample were formed by a Two-Step Cluster Analysis (TSCA). The TSCA was run with automatic determination of the number of clusters using the SPSS 15.0 software package. The first step of analysis is construction of a Cluster Features (CF) tree which provides a capsule summary of the data file. The second step is grouping the leaf nodes of the CF tree using an agglomerative clustering algorithm which produces a range of solutions which are then compared using Schwarz’s Bayesian Criterion to determine the “best” number of clusters. The hypothesis provided by the TSCA was checked and corrected by a single run of a linear discriminant analysis (LDA). LDA, Principal Component analysis (PCA), ANOVA and X² tests were run with the SPSS 15.0 software package. The DIMORPH test was performed according to Seifert (2016).

Nest samples of F. exsecta containing only worker individuals of the setae-reduced Rubens morph are under increased risk of being confused with F. fennica. In order to demonstrate a clear species separation, these samples were run together with those of F. fennica in different forms of exporatory data analyses using nest centroids as input data (NC clustering; Seifert et al. 2013). These were firstly hierarchical NC-Ward clustering, secondly and thirdly the hierarchical method NC-part.hclust and the iterative vector-quantization method NC-part.kmeans (the latter two methods are implemented in partitioning algorithms based on recursive thresholding, Csősz and Fisher 2015), and nonmetric multidimensional scaling combined with iterative vector-quantization NC-NMDS-k-means (Seifert et al. 2013). Checking samples with controversial classifications was done by an interaction of NC clustering and a controlling linear discriminant analysis (LDA) in which these samples were run as wild-cards following the rationale described in Seifert et al. (2013). The final classification (“final species hypothesis”) was established by the LDA in an iterative procedure and there remained no undecided cases even if their posterior probabilities were close to 0.5.

Standard air temperature (TAS) in °C of sampling sites 2 m above ground can be used as a rough approximation of the thermal niche component (Steiner et al. 2010, Seifert et al. 2014). Following Seifert and Pannier (2007), TAS was calculated as mean air temperature from 1 May to 31 August of the nearest three meteorological stations of the years 1961 to 1990 provided by Klimaabteilung der Zentralanstalt für Meteorologie und Geodynamik (1996). Values were corrected for an altitudinal temperature decrease of 0.661 °C per 100 m: TAS=-0.694*LAT+0.078*LON-0.00661*ALT+52.20.

Results and discussion

Existence of pilosity and pubescence dimorphism in Formica exsecta

The analysis of phenotype variance was restricted to the characters CS, CL/CW, SL/CS, ClySet, ClyPub, nOce, OceD/CS, EyeHL, sqPDO, sqPDG, TERG, nCox, nHTFl, and nMet because the full set of characters was not available for all Western Palaearctic samples. Yet, three of the four excluded characters (F2L/CS, F3L/CS, and F2L/F3L) did not contribute to morph separation (Table 1). Within 431 worker individuals investigated, the Two-Step Cluster Analysis (TSCA) distinguished only two clusters: the Normal morph and the Rubens morph. The final morph hypothesis was established by a single run of an LDA which confirmed 96.3% of the TSCA classifications. Based on this hypothesis, a stepwise character-reduced LDA was run which extracted a discriminant function of seven characters with

D(7) = 5.098*SL/CS + 1.609*ClySet-0.044*EyeHL-0.191*sqPDO-0.45*TERG + 0.100*nCox+0.064*nMet-7.9705.

Morphometrics of worker individuals of Formica fennica and the two pilosity morphs of F. exsecta. Data are given as arithmetic mean ± standard deviation [lower extreme, upper extreme]; n = number of individuals. The columns with data of an univariate ANOVA test (F values, significance levels p, and degrees of freedom df2) are placed between the columns of the compared entities. F values of most discriminative characters are given in heavy type.

F. exsecta Normal morph (n=288) ANOVA F, p df2 F. exsecta Rubens morph (n=143) ANOVA F, p df2 F. fennica (n=54)
CS [µm] 1390 ± 81 1.66, 0.198 1397± 83 65.03, 0.000 1287 ± 93
[1143,1586] 429 [1145,1608] 195 [1050,1468]
CL/CW 1.045 ± 0.023 15.50, 0.000 1.054 ± 0.020 18.52, 0.000 1.068 ± 0.021
[0.983,1.119] 429 [1.009,1.105] 195 [1.019,1.118]
SL/CS 1.028 ± 0.022 9.91, 0.002 1.020 ± 0.023 0.09, 0.762 1.022 ± 0.026
[0.966,1.085] 429 [0.965,1.071] 195 [0.962,1.084]
EyeHL 27.2 ± 6.9 0.89, 0.347 26.5 ± 8.0 5.17, 0.024 23.9 ± 2.3
[0.0,60.3] 429 [16.1,44.8] 195 [20.7,27.0]
TERG 1.03 ± 0.16 227.18, 0.000 1.92 ± 0.98 117.02, 0.000 3.48 ±0.64
[1.0,2.0] 429 [1.00,4.00] 195 [3.0,6.0]
nCOX 9.63 ± 2.99 399.87, 0.000 4.00 ± 2.26 97.07, 0.000 0.89 ±0.88
[2.5,18.5] 429 [0.5,13.0] 195 [0.0,3.0]
nHTFL 9.45 ± 1.89 69.28, 0.000 7.93± 1.65 6.69, 0.010 7.22 ±1.93
[5.1,15.5] 429 [3.8,13.1] 195 [4.2,13.9]
nMET 2.26 ± 2.15 100.89, 0.000 0.37 ± 0.75 13.09, 0.000 0.00 ± 0.00
[0.0,10.0] 429 [0.0,3.7] 195 [0,0.0]
sqPDO 5.38 ± 0.98 79.53, 0.000 6.43 ± 1.44 2.51, 0.115 6.77 ± 1.04
[3.21,8.91] 429 [3.74,13.40] 195 [4.52,9.92]
sqPDG 6.87 ± 1.20 12.64, 0.000 7.31 ± 1.36 2.35, 0.127 7.61 ± 0.56
[3.93,9.84] 429 [4.33,10.73] 195 [6.58,8.71]
ClySet 4.09 ± 0.56 1206.1, 0.000 2.17 ± 0.48 32.46, 0.000 1.72 ± 0.53
[2.0,5.0] 429 [1.0,3.0] 195 [1.0,3.0]
ClyPub 3.13 ± 1.04 0.70,0.404 3.01 ± 1.25 26.23, 0.000 2.07 ± 0.82
[0.8,6.5] 429 [0.5,6.5] 195 [0.5,4.3]
nOce 2.16 ± 0.56 95.59, 0.000 1.57 ± 0.65 195.10, 0.000 0.24 ± 0.42
[0.7,6.0] 429 [0.0,4.0] 195 [0.0,1.7]
OCED /CS 0.188 ± 0.010 0.00, 0.97 0.188 ± 0.011 20.27, 0.000 0.180 ± 0.010
[0.155,0.210] 429 [0.161,0.217] 195 [0.156,0.198]
T3f 0.97 ± 0.15 34.46, 0.000 0.77 ± 0.41 144.19, 0.000 0.06 ± 0.22
[0.00,1.00] 301 [0.00,1.00] 182 [0.00,1.00]
FL2/FL3 0.987 ± 0.040 3.03, 0.084 0.979 ± 0.040 20.60, 0.000 1.004 ± 0.033
[0.896,1.077] 142 [0.885,1.085] 144 [0.938,1.094]
FL2/CS 13.57 ±0.59 4.57, 0.034 13.37 ±0.53 0.30, 0.586 13.31 ±0.68
[12.43,14.77] 142 [12.08,14.63] 144 [11.98,15.21]

Individuals with D(7) <0 were classified as Rubens morph and those with larger values as Normal morph. This discriminant classified 94.9% of the 431 specimens with posterior probabilities > 0.90 (the recognition threshold used in the DIMORPH test, see below). Figure 1 presents a bivariate plot of D(7) and the first factor of a PCA considering these seven characters. Table 1 shows that the morphs are strongest separated by clypeal setae conditions (ClySet). The composition of the type samples is given in Table 2. Only Rubens morph workers were observed within the type series of F. rubens Forel, 1874 and F. etrusca Emery, 1909. Eight workers of the type serie of F. nemoralis Dlussky, 1964 belonged to Rubens but one worker to Normal whereas all three investigated type specimens of F. dalcqi Bondroit, 1918 belonged to Normal.

Figure 1. 

Discriminant value D(7) and 1st factor of a principal component analysis. Data of 288 workers of the Normal morph (white dots) and 143 workers of the Rubens morph (black dots) of Formica exsecta.

Morph composition in type series of taxa synonymized with Formica exsecta Nylander, 1846. D(7) is the discriminant value considering seven phenotypic characters, p(Norm) and p(Rube) are the posterior probability for the Normal and Rubens morph. Data indicating the Normal morph are shaded in grey.

Taxon; site; specimen no. D(7) p(Norm) p(Rube)
F. exsecta rubens Forel, 1874; Apples; no. 1 -1.692 0.0004 0.9996
F. exsecta rubens Forel, 1874; Apples; no. 2 -2.582 0.0000 1.0000
F. exsecta rubens Forel, 1874; Apples; no. 3 -0.665 0.0364 0.9636
F. exsecta rubens Forel, 1874; Apples; no. 4 -2.769 0.0000 1.0000
F. exsecta etrusca Emery, 1909; Praccia; no. 1 -2.396 0.0000 1.0000
F. exsecta etrusca Emery, 1909; Praccia; no. 2 -2.107 0.0001 0.9999
F. exsecta etrusca Emery, 1909; Praccia; no. 3 -1.862 0.0002 0.9998
F. exsecta etrusca Emery, 1909; Praccia; no. 4 -1.945 0.0001 0.9999
F. exsecta etrusca Emery, 1909; Praccia; no. 5 -0.3925 0.1114 0.8886
F. dalcqi Bondroit, 1918; Mount Canigou; no. 1 2.256 0.9999 0.0001
F. dalcqi Bondroit, 1918; Mount Canigou; no. 2 0.568 0.8954 0.1046
F. dalcqi Bondroit, 1918; Mount Canigou; no. 3 3.414 1.0000 0.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 1 -3.134 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 2 -2.601 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 3 -2.939 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 4 -2.874 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 5 -2.554 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 6 -1.684 0.0004 0.9996
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 7 1.082 0.9880 0.0120
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 8 -3.087 0.0000 1.0000
F. nemoralis Dlussky, 1964; Voronesh Zapovednik; no. 9 -2.023 0.0001 0.9999

Evidence for conspecificity of the Normal and Rubens morph

Within 152 nest samples from the Palaearctic west of 59°E, 104 samples contained only the Normal morph, 28 only the Rubens morph and 20 a mixture of both the Normal and Rubens morph. This figure of 13.2% of mixed nests already indicates heterospecificity to be most unlikely. Yet, checking the alternatives intraspecific dimorphism versus heterospecificity by the DIMORPH test is problematic considering the huge geographic area and heterogenous sampling philosophies of different collectors. Among the conditions allowing to run the DIMORPH test are panmictic behaviour and random sampling, i.e., no preference to collect certain phenotypes (Seifert 2016). Both conditions are not really fulfilled in the whole data set. As a matter of fact, F. exsecta samples with reduced setae conditions (suspicious during field sampling or during laboratory sorting to possibly represent F. bruni Kutter, 1967, F. fennica, or F. suecica Adlerz, 1902) were more frequently collected (and analyzed) than the much more frequent hairy samples. This distorts the estimates of allele frequencies in favour of the Rubens morph.

The only coherent geographic area with random, unbiased sampling and evaluation of F. exsecta samples included Finland, Sweden, and northern Denmark, largely based on the collecting activity of the Senckenberg Museum of Natural History Görlitz. Selecting this area makes also sense because the geographic reference is then well comparable with that in the study of Hakala et al. (2018).

The DIMORPH test was run for this area with the following basic data and parameters. Among 58 nest samples, 42 samples contained only the Normal morph, seven samples only the Rubens morph, and nine samples a mixture of both morphs. These data are based on a recognition threshold of p>0.90 in the LDA considering seven characters. Of the 157 individuals in the data set, 74.5% belonged to the Normal morph and 25.5% to the Rubens morph. Supposing a dominant-recessive inheritance and the Hardy-Weinberg model of population genetics, this would result in a frequency of the recessive ‘allele’ of 0.505 if Rubens is recessive and of 0.863 if Normal is recessive. Using these parameters, the DIMORPH test was run for both assumptions with 500 repeats each and the data were averaged. The clear results are presented in Table 3. Applying Fisher’s exact test, observed and predicticted morph distributions differed with p=0.017 for the scenario of parabiosis of independent species, with p=0.0004 for the scenario of temporary social parasitism and with p=0.0001 for the scenario of permanent social parasitism. In contrast, no significant difference was found between the observed distribution and the prediction for intraspecific dimorphism (Fisher’s test p=0.659, X² test p=0.547).

DIMORPH test comparing 58 observed within-nest phenotype compositions with four prediction models. Agreement of observation and predictions was tested by Fisher’s exact test (p) and the Chi-squared test (X², p).

within-nest phenotype composition observed prediction heterospecificity prediction intraspecific dimorphism
parabiosis temporary social parasitism permanent social parasitism
only Rubens 7 8.20 2.08 0.00 8.41
Rubens + Normal 9 0.58 0.09 1.16 12.20
only Normal 42 49.22 55.84 56.84 37.39
Fisher’s p 0.017 0.0004 0.0001 0.659
X², p 8.07, 0.018 13.4, 0.001 15.28, 0.0001 0.88, 0.547

The separation of the Rubens morph from Formica fennica

Nest samples of Formica exsecta containing only worker individuals of the Rubens morph are at increased risk of being confused with F. fennica. Only these critical samples, but this time originating from the whole of the Western Palaearctic and having T3f data available, were included in the following analyses. These were 21 nest samples of the Rubens morph with 82 workers and 14 samples of F. fennica with 54 workers.

In first analytic step all 18 NUMOBAT characters were used unselectively as input and for the exploratory data analyses NC-Ward, NC-part.hclust, and NC-NMDS-k-means. After running samples with classifications disagreeing between the different methods as wild cards, NC-Ward clustering and NC-part.hclust both showed a classification error of 0% with the latter exposing two samples as outliers (=5.7%), whereas NC-NMDS-k-means showed an error of 5.7%. NC-part.kmeans was not able to confirm the existence of two clusters for the complete unselected character set.

Accepting the former hypotheses, the character set was reduced by a stepwise LDA to the eight characters CS, CL/CW, ClySet, nOce, sqPDG, nCox, T3f, and F2L/F3L. Under this setting, all four NC-clustering algorithms achieved a classification error of 0% with NC-part.hclust exposing two samples (5.7%) as outliers (Fig. 2). On the individual level, the classification error of the LDA was 2.9% in 136 workers. The sample from Norway, Osen: Drageid: Seter-4.8 km E, 2012.04.29 [64.392N, 10.589E, 45 m] is determined under this setting with a sample mean of p = 0.9032 as Rubens morph if the three available specimens are run as wild-cards. A wild-card run considering all 18 characters results in a determination with p=0.9916. This should be emphasized as, based on a much weaker data basis available in the year 2012, I had erroneously determined this sample as F. fennica (published by Suvák 2013).

Figure 2. 

Three exploratory and a hypothesis-driven data analysis using nest centroids (NC) as input data. Evaluated were 21 nest samples of Formica exsecta containing only Rubens morph workers (grey bars right) and 14 nest samples of F. fennica (textured grey bars left). The classifications of NC-Ward and NC-part.kmeans coincide completey whereas NC-part.hclust is also in general agreement but exposes two outlier samples (black bar).

Data on zoogeography and climate niche do also not support the presence of F. fennica in Norway. Since F. fennica apparently does not spread north to the boreal zone and does not elevate to the subalpine zone (Seifert 2018), the high Skandinavian Fjäll should represent a strong distributional barrier for postglacial immigration from the east. The much narrower climate niche of F. fennica is explicitly indicated by the mean air temperature TAS of the sites from 1 May to 31 August. TAS is 11.98 ± 3.28 [4.75, 18.44] °C for 111 sites of F. exsecta and 13.78 ± 0.90 [12.72, 15.09] °C for the six known sites of F. fennica. Immigration to Norway, if at all, appears only possible via Skåne (Sweden) and along the Oslo Fjord.

Acknowlegements

Many thanks are given to Igor Antonov, Volker Assing, Peter Boer, Rolf Franke, Florian Glaser, Anya Goropashnaya, Alina Kupyanskaya, Wolfgang Münch, Frode Odegaard, Andreas Schulz, Roland Schultz, Jouni Sorvari, Uwe Sörensen, Dieter Stöckel, Martin Suvak, Jürgen Trettin, Kari Vepsäläinen, Bernd Wesenigk-Sturm, Seichi Yamane, and Lothar Zerche for donating samples. I also wish to thank two anonymous referees for useful comments on the manuscript, Robert Forsyth for careful copyediting and Dominique Zimmermann for straightforward handling of the submission and editing process.

References

  • Csősz S, Fisher BL (2015) Diagnostic survey of Malagasy Nesomyrmex species-groups and revision of hafahafa group species via morphology based cluster delimitation protocol. ZooKeys 526: 19–59. https://doi.org/10.3897/zookeys.526.6037
  • Hakala SM, Seppä P, Heikkilä M, Punttila P, Sorvari J, Helanterä H (2018) Genetic analysis reveals Finnish Formica fennica populations do not form a separate genetic entity from F. exsecta. PeerJ 6:e6013.https://doi.org/10.7717/peerj.6013
  • Schultz R, Seifert B (2007) The distribution of the subgenus Coptoformica Müller, 1923 (Hymenoptera: Formicidae) in the Palaearctic Region. Myrmecological News 10: 11–18.
  • Seifert B (2000) A taxonomic revision of the ant subgenus Coptoformica Mueller, 1923. Zoosystema 22(3): 517–568.
  • Seifert B (2014) A pragmatic species concept applicable to all eukaryotic organisms independent from their mode of reproduction or evolutionary history. Soil Organisms 86: 85–93.
  • Seifert B (2016) Analyzing large-scale and intranidal phenotype distributions in eusocial Hymenoptera – a taxonomic tool to distinguish intraspecific dimorphism from heterospecificity. Myrmecological News 23: 41–59.
  • Seifert B (2018) The Ants of Central and North Europe. Lutra Verlags- und Vertriebsgesellschaft, Tauer, 408 pp.
  • Seifert B, Ritz M, Csősz S (2013) Application of Exploratory Data Analyses opens a new perspective in morphology-based alpha-taxonomy of eusocial organisms. Myrmecological News 19: 1–15.
  • Seifert B, Csősz S, Schulz A (2014) NC-Clustering demonstrates heterospecificity of the cryptic ant species Temnothorax luteus (Forel, 1874) and T. racovitzai (Bondroit, 1918) (Hymenoptera: Formicidae). Contributions to Entomology (Beiträge zur Entomologie) 64: 47–57.
  • Steiner FM, Seifert B, Moder K, Schlick-Steiner BC (2010) A multisource solution for a complex problem in biodiversity research: description of the cryptic ant species Tetramorium alpestre sp.n. (Hymenoptera: Formicidae). Zoologischer Anzeiger 249: 223–254. https://doi.org/10.1016/j.jcz.2010.09.003
  • Suvák M (2013) First record of Formica fennica Seifert, 2000 (Hymenoptera, Formicidae) in Norway. Norwegian Journal of Entomology 60: 73–80.