1. A descriptive pollination study of Lysimachia nummularia in western Switzerland
The sampling of pollinators of L. nummularia was carried out during the months of June and July 2008 in ten different Swiss populations (see Tab. I and Fig. 1). In order to optimally describe associated pollinators, each population was visited at least twice during that period (a few sites were even visited three times), and on-site observations lasted between one and two hours. Observations were carried out by sunny weather at the hottest hours of the day (from 11h to 16h). If a pollination behavior was presumably observed, the insect was caught with a net and stocked in EtOH 70%. Determination of pollinators was done with the help of J.-P. Haenni for Diptera (Natural History Museum of Neuchâtel, Switzerland) and F. Amiet for Hymenoptera (Solothurn, Switzerland).
Table I: List of the ten Swiss L. nummularia populations studied, including the respective geographical and co-occurring Lysimachia species information.
Results and discussion
Our observations resulted in pollination records only in three sites (VIL, CHA and PRE2) as no pollinators were observed in the other populations. The insects observed can be classified into two main categories: hoverflies (Diptera, Syrphidae) and wild bees. As the sampled hoverflies belonged to generalist and common species (Melanostoma scalare Fabricius 1794, Platycheirus albimanus Fabricius 1781, Episyrphus balteatus De Geer 1776 and S. vitripennis Meigen 1822) they probably did not account for very effective pollinators, performing only short and superficial visits on the flowers. Wild bees pollinating L. nummularia in the present study belonged to two different genera: Lasioglossum Curtis 1833 (Hymenoptera, Halictidae) and Macropis Panzer 1809 (Hymenoptera, Melittidae).
Figure 1: Map of the L. nummularia populations studied, with the relative proportion and number for each type of pollinators
asioglossum was dominantly observed, with 12 individuals belonging to the six following species: L. morio Fabricius 1793 (the most frequent with six individuals captured), L. lineare Schenck 1869 (two individuals), L. calceatum Scopoli 1763, L. fulvicorne Kirby 1802, L. puncticolle Moravitz 1872 and finally L. rufitarse Zetterstedt 1838 (one single individual per species). Typical pollination behavior for Lasioglossum consisted in taking some pollen on back legs with regular movements towards the bottom of the flower. These are not the first observations of Lasioglossum species on L. nummularia (e.g. Teppner, 2005), and some species observed here (e.g. L. morio, L. calceatum and L. rufitarse) apparently pollinate a wide range of plants (e.g. from orchids such as Cypripedium calceolus, to the tomato Solanum lycopersicum; Teppner, 2005; Erneberg and Holm, 1999; Bittrich and Kaderheit, 1988). Consequently, L. nummularia should be unspecifically pollinated by species of Lasioglossum living near-at-hand. Finally we note that bees from this genus were also observed visiting other Lysimachia species, such as L. vulgaris (in numerous European populations, from Spain to Turkey (L. Bassin and Y. Triponez pers. obs.) and even L. nemorum (e.g. in population PRE; L. Bassin pers. obs.). The visits of these bees to L. nummularia are however not as frequent as in the relationship between Macropis and L. vulgaris.
Two individuals of the oil-collecting bee Macropis fulvipes Fabricius 1804 (previously reported as regular pollinator of L. nummularia; e.g Simpson and Neff, 1983; Vogel, 1986) were also caught, although in only one population (PRE2). Even if not strictly sympatric, we should note that this population stood less than 200 m. from a population of L. vulgaris (where M. fulvipes was also captured). In comparison, it is interesting to see that at the sympatric site VAU, M. fulvipes was caught only on L. vulgaris, neglecting nearby plants of L. nummularia. According to these observations M. fulvipes rarely visits L. nummularia and seems to give priority to L. vulgaris when available. Consequently, the assumed preference of M. fulvipes for L. nummularia (e.g. Michez, 2002) should be reconsidered, at least in Switzerland.
Our results add knowledge to the reproductive biology of L. nummularia. At first, the plant does not clearly depend on insects for its reproduction. Indeed, on most sites no pollinators were observed and, if pollinators were present, their visits seemed very sporadic. Asexual reproduction is therefore probably more frequent in L. nummularia than in other European Lysimachia species, although the seedset and quality of descendants is probably lower than with cross-pollinated sexual reproduction (Simpson & Neff, 1983; Bittrich & Kaderheit, 1988; Batygina, 2005; Hoffman, 2005). Moreover, according to our observations, pollinators could play a role in maintaining populations of L. nummularia fit, as populations without pollinators were often unhealthy and composed of a few individuals. Finally, our study confirms that, even if different native insects visit the flowers, small bees and more specifically specimens from the genus Lasioglossum should be the more effective vectors of L. nummularia pollen.
Batygina T. (2005) Sexual and asexual processes in reproductive systems, Acta Biol. Cracov. Bot. 47, 51-60.
Bittrich V., Kaderheit J. (1988) Cytogenetical and geographical aspects of sterility in Lysimachia nummularia, Nord. J. Bot. 4, 325-328.
Erneberg M, Holm B. (1999) Bee size and pollen transfer in Cypripedium calceolus (Orchidaceae), Nord. J. Bot. 19, 363-367.
Hoffmann F. (2005) Biodiversity and pollination. Flowering plants and flower-visiting insects in agricultural and semi-natural landscapes. University of Groningen.
Michez D. (2002) Monographie systématique, biogéographique et écologique des Melittidae (Hymenoptera, Apoidea) de l’Ancien Monde – Premières données et premières analyses. Faculté Universitaire des Sciences Agronomiques de Gembloux.
Simpson B.B., Neff J.L., et al. (1983) Floral biology and floral rewards of Lysimachia (Primulaceae), Am. Midl. Nat. 1102, 249-256.
Teppner H. (2005) Pollinators of tomato, Solanum lycopersicum (Solanaceae), in central Europe, Phyton-Ann. Rei Bot. A. 45, 217-235.
Vogel S. (1986) Ölblumen and Ölsammelnde Bienen: Zweite Folge Lysimachia und Macropis, in: Verl. F.S. (Ed.), Akademie der Wissenschaftenund der Literatur Mainz, Stuttgart, Germany.
2. Detailed methods for the present distribution modelling of Macropis
We extracted occurrences of the species in Europe from the GBIF database (www.gbif.org). We kept only the data with a minimal spatial accuracy of 5km. These data were pooled with the occurrences collected during fieldwork. Since the occurrences were highly aggregated in some parts of Europe, we randomly selected a subset of occurrences with a minimal distance of 50km. Most modeling techniques require not solely information about presences but also absences to determine the suitable conditions for a given species, thus we selected 10’000 pseudo-absences randomly. The modeling techniques discriminate the conditions for presences and absences, based on the background environment (Witz and Guisan, 2009). GBIF is known to be highly biased for western European countries (Yesson et al., 2007). Therefore, we did not select pseudo-absences in Eastern Europe countries but limited them to the Western and Balkanic countries. Selecting pseudo-absences in unsampled countries can significantly bias the response curves of the models (Phillips et al., 2009). The resulting presences and pseudo-absences were used in the further species distribution modeling.
We ran single-models with the seven following climatic variables from Worldclim (Hijmans et al., 2005) at a resolution of 2.5 arc-minute (around 5 km): maximum temperature of warmest month (bio5), minimum temperature of coldest month (bio6), mean temperature of warmest quarter (bio10), mean temperature of coldest quarter (bio11), precipitation seasonality (bio15), precipitation of the warmest quarter (bio18) and precipitation of the coldest quarter (bio19).
We modeled the distribution of the species using the BIOMOD package (Thuiller et al., 2009), as implemented in R CRAN (REFERENCE). We used eight different niche-based modeling techniques : (1) generalized linear model (GLM), (2) generalized additive model (GAM), (3) classification tree analysis (CTA), a classification method that selects the best trade-off between the number of leaves of the tree and the explained deviance, (4) artificial neural networks (ANN), a machine learning method, with the mean of three runs used to provide predictions and projections, as each simulation gives slightly different results, (5) mixture discriminant analysis (MDA), a classification method that uses MARS function for the regression part of the model, (6) multivariate adaptive regression splines (MARS), (7) generalized boosting model (GBM), a machine learning method which combines a boosting algorithm and a regression tree algorithm to construct an 'ensemble' of trees, and (8) Random Forest (RF), a machine learning method that is a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest.
In order to evaluate the predictive performance of the species distribution model, we used a random subset of 70% of the data to calibrate every model, and used the remaining 30% for the evaluation. Models were evaluated using a relative operating characteristic (ROC) curve and the Area Under the Curve (AUC) (Fielding and Bell, 1997). We replicated the data splitting ten times and calculated the average AUC of the repeated split-sample, which gives a more robust estimate of the predictive performance of each model.
Finally, each model was projected into current climate conditions with the Worldclim data of the CCSM circulation model for the last glacial maximum (LGM; -21’000 years) downscaled at a resolution of 2.5 arc-minute. In order to obtain the central tendency of these distributions, accounting for variations among modeling techniques, we applied a weighted average of the eight modeling techniques based on the predictive power (AUC). Ensemble forecasting approaches have been shown to significantly improve the accuracy of species distribution models (Marmion et al., 2009).
Fielding A.H., Bell J.F. (1997) A review of methods for the assessment of prediction errors in conservation presence/absence models, Environmental Conservation 24, 38-49.
Hijmans R.J., Cameron S.E., Parra J.L., Jones P.G., Jarvis A. (2005) Very high resolution interpolated climate surfaces for global land areas, International Journal of Climatology 25, 1965-1978.
Marmion M., Parviainen M., Luoto M., Heikkinen R.K., Thuiller W. (2009) Evaluation of consensus methods in predictive species distribution modeling, Diversity and Distributions 15, 59-69.
Phillips S.J., Dudik M., Elith J., Graham C., Lehmann A., Leathwick J., & Ferrier S. (2009) Sample selection bias and presence-only models of species distributions: implications for selection of background and pseudo-absences, Ecological Applications 19, 181-197.
Thuiller W., Lafourcade B., Engler R., Araújo M.B. (2009) BIOMOD - a platform for ensemble forecasting of species distributions, Ecography 32, 369-373.
Wisz M.S., Guisan A. (2009). Do pseudo-absence selection strategies influence species distribution models and their predictions? An information-theoretic approach based on simulated data, BMC Ecology 9:8.
Yesson C., Brewer P.W., Sutton T., Caithness N., Pahwa J.S., Burgess M., Gray W.A., White R.J., Jones A.C., Bisby F.A., Culham A. (2007) How Global Is the Global Biodiversity Information Facility?, PLoS ONE 2, e1124.
3. Niches’ overlap for M. europaea and M. fulvipes
In order to compare the niche of the two Macropis species, we extracted the climatic information for the occurrences from the bioclimatic variables of Worldclim (Hijmans et al., 2005). We selected several variables important for the ecology of the bees, in particular reflecting conditions of temperature and precipitation: bio5, bio6, bio10, bio11, bio18, bio19. Also, we calculated two additional layers from Worldclim data: degree-day (DDEG) and averaged moisture index (MIND) during the year. Recently, tools have been developed to compare the niche of species in different part of its range or to compare closely related species to investigate niche conservatism (Broennimann et al., submitted). Climatic niches of the two species, M. europaea and M. fulvipes were compared in a gridded climatic space using a principal component analysis (PCA). First we run the PCA on the climatic environment available represented by 30’000 pixels randomly selected throughout Europe. Each pixel in the environmental space corresponds to a unique set of environmental conditions present at one or more sites in the geographical space. The occurrences of the groups of populations were then projected in the climatic space available. Finally, we applied a kernel density function to determine the "smoothed" density of occurrences of each pixel in the environmental space for each bee species. We divided the density of occurrence by the density of the environment in each focal pixel to obtain a measure of the density of the species relative to the availability of climate. This ratio is finally rescaled between 0 and 1. This approach has been shown with virtual species to be very robust to compare climatic niche of species (Broennimann et al., submitted). The similarity between niches was afterwards measured using the Schoener’s D metric and its associated statistical tests (tests of niche conservatism), proposed originally by Warren et al. (2008). This metric, which ranges from 0 (no niche overlap) to 1 (complete overlap). In order to test if the niches of the groups are significantly different, occurrences are pooled and randomly split, maintaining the number of occurrences as in the original datasets. The niche overlap statistics (D, see Warren et al. 2008) are then recalculated. This process is repeated 100 times and a histogram of simulated values is constructed. The comparison of observed and simulated values of the test statistic allows significance testing for niche equivalency. If the observed value falls within the density of 95% of the simulated values, the null hypothesis of niche equivalency cannot be rejected. We also applied a niche similarity test in the way of Warren et al. (2008). Rejection of the null hypothesis indicates that the niche models of two species are more similar (or different) than would be expected by chance. Rejection of the null also indicates that the observed niche differentiation between species is a function of habitat selection and/or suitability rather than simply an artifact of the underlying environmental differences between the suite of habitats available to the two species (Peterson et al. 1999; Warren et al. 2008).
Results and discussion
Here we can see that the niche of M. fulvipes is slightly larger than that of M. europaea (represented by a larger surface of non-overlapping blue than of non-overlapping red of Fig. 1). According to the Worldclim variables, it can be deduced that M. fulvipes is found within habitats with colder and drier winters, with hotter summers and with more precipitations (typical conditions of the continental area). On the opposite, M. europaea depends on higher winter temperatures as well as on more winter precipitations, typical conditions of the oceanic climate.
Figure 1: Respective ecological niches of M. europaea (in red) and M. fulvipes (in blue) displayed on the two first coordinates of the environmental space delimited by the climatic variables of Worldclim. Niche overlap is represented in brown.
Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G. & Jarvis, A. (2005). Very high resolution interpolated climate surfaces for global land areas. Int. J. Climatol., 25, 1965-1978.
Peterson, A.T., Soberon, J. & Sanchez-Cordero, V. (1999). Conservatism of ecological niches in evolutionary time. Science, 285,1265-1267.
Warren, D.L., Glor, R.E. & Turelli, M. (2008). Environmental niche equivalency versus conservatism : quantitative approaches to niche evolution. Evolution, 62, 2868–2883.