Ecological data
To characterize the environment that stickleback occupy, we focused on a set of ecological variables which represent putative selective agents. First, we used the same five habitat classifications (warm, mined, pondweed, cladophorales, and rocky shore) previously described in (see supplement for details). We also collated data on ecological variables likely to reflect selective agents. These were: water temperature, water depth, stickleback CPUE, piscivorous bird density, and zooplankton abundances and community composition. These were chosen because temperature can affect metabolic processes, development, tolerance to parasite infections , as well as key life history traits , whilst depth can affect sensory processes invertebrate availability, and stickleback visibility to predators . Stickleback CPUE was used as a measure of intraspecific competition , piscivorous bird density as a measure of predation pressure , and invertebrate data as a measure of prey abundance and composition . It should be noted that our measures are only proxies for selection imposed by correlated ecological factors.
Temperature and water depth of each site were used as per . Average temperature at each site was measured between 30 June 2011 and 18 August 2011 with a temperature logger (iButton Maxim Integrated Products, San Jose, CA, USA), placed at mid-depth and recording at three-hour intervals. CPUE for each site was estimated using count data from the long-term monitoring study from June 2012. To measure piscivorous bird density, we used data collected during the waterfowl census conducted each year at Mývatn, during which all waterfowl observed from pre-determined vantage points with known survey areas are counted . We used count data collected between 15th May and 10th June 2012 on the following species known to predate on stickleback: horned grebe (Podiceps auritus ), red-breasted merganser (Mergus serrator ), great northern diver (Gavia immer ), red throated diver (Gavia stellata ) and goosander (Mergus merganser ). Note that the Arctic tern (Sterna paradisaea) is abundant at Mývatn, and predates on stickleback, but this species is not counted during the bird census. We calculated the density of piscivorous birds (summed across all taxa) in each surveyed segment of the lake (number/m2) (Figure S1).
We used invertebrate data from , which were collected by conducting surveys of the epibenthic and zooplanktonic community. Crustaceans (incl. Daphnia , copepods and epibenthic cladocerans) as well as rotifers are important food sources for stickleback in Mývatn (e.g. ). Although chironomid larvae are a main food source for Mývatn stickleback, data on midge larval abundance were not of sufficient spatial resolution to be used (see ). However, the benthic community is spatially correlated with the epibenthic and zooplankton community in the South basin , suggesting that measures of pelagic and epibenthic zooplankton may serve as a proxy measure for the benthic community in Mývatn. Briefly, three transects were conducted between June - July 2012, during which integrated vertical tows of the whole water column were made at each of 31 sites, spaced 500 – 600m apart (see Figure S2 for distribution, and for more details on sampling and sample processing). Each pooled sample of 15 L was filtered through 63-μm mesh and counted in entirety under a binocular microscope. We used data from the 2nd transect (25th July 2012) as the spatial resolution in this transect was the greatest. We used data collected from the closest site to each stickleback sampling site (distance to closest zooplankton site: min = 290 m, max = 1365 m). All stickleback sites were within 2.55 km of the nearest site used to collect invertebrate data, which was the distance at which zooplankton communities were found to be spatially autocorrelated. We used number per litre (n/L) of each taxon at the sites closest to stickleback sites.
To summarise variance between sites for use in downstream analyses, we ran a principal components analysis (PCA) using the native stats package in R version 4.1.2 . This summarised the invertebrate data in four general axes, described in detail in Table S1. PC1 (zPC1) described the overall abundance of crustaceans and rotifers, and explained 29% of the variation; PC2 (zPC2) described the negative covariance of rotifer sp. and Alona sp. with planktonic and epibenthic crustaceans, which explained 17% of the variation; PC3 (zPC3) described the negative covariance of the rotifer Keratella and the cladoceransAcroperus harpae and Chydorus sphaericus withDaphnia longispina , and explained 15% of the variation; and PC4 (zPC4) described the negative covariance of the cladoceransEurycercus lamellatus and Macrothrix hirsuticornis withDaphnia longispina , Cyclops abyssorum andAsplancha , explaining 11% of the variation in the data. Overall invertebrate abundance (described in zPC1), was highly negatively correlated with stickleback CPUE. We therefore used only CPUE and not zPC1 for downstream analyses.