Ecological speciation by temporal isolation in a population of the stonefly Leuctra hippopus (Plecoptera, Leuctridae)

Abstract Stream dwelling invertebrates are ideal candidates for the study of ecological speciation as they are often adapted to particular environmental conditions within a stream and inhabit only certain reaches of a drainage basin, separated by unsuitable habitat. We studied an atypical population of the stonefly Leuctra hippopus at a site in central Norway, the Isterfoss rapids, in relation to three nearby and two remote conspecific populations. Adults of this population emerge about a month earlier than those of nearby populations, live on large boulders emerging from the rapids, and are short‐lived. This population also has distinct morphological features and was studied earlier during the period 1975–1990. We reassessed morphological distinctness with new measurements and added several analyses of genetic distinctness based on mitochondrial and nuclear sequence markers, as well as AFLP fingerprinting and SNPs mined from RAD sequences. The Isterfoss population is shown to be most closely related to its geographical neighbors, yet clearly morphologically and genetically distinct and homogeneous. We conclude that this population is in the process of sympatric speciation, with temporal isolation being the most important direct barrier to gene flow. The shift in reproductive season results from the particular temperature and water level regime in the Isterfoss rapids. The distinct adult body shape and loss of flight are hypothesized to be an adaptation to the unusual habitat. Ecological diversification on small spatial and temporal scales is one of the likely causes of the high diversity of aquatic insects.


| INTRODUCTION
The adaptation of local populations to different environments can lead to spatial isolation and temporal isolation, both common factors in ecological speciation (Coyne & Orr, 2004: 179 ff.;Rundle & Nosil, 2005). Temporal isolation, which occurs when sympatric populations have nonoverlapping reproductive periods, has not been as thoroughly investigated as other isolating barriers (Coyne & Orr, 2004:204). Most examples concern plant populations that differ in flowering times (cited in Lowry, Modliszewski, Wright, Wu, & Willis, 2008) and allochronically spawning populations of fish and corals (e.g., Barson, Haugen, Vøllestad, & Primmer, 2009;Limborg, Waples, Seeb, & Seeb, 2014;Villanueva, 2016). In insects, periodic cicadas constitute a famous case (Marshall & Cooley, 2000), while Sota, Kagata, Ando, Utsumi, and Osono (2014) report incipient speciation in early and late winter populations of winter moths. Freshwater insects with synchronized adult emergence and a short adult life stage, particularly Ephemeroptera and Plecoptera, are expected to be good systems to study diversification by temporal isolation, but so far studies have been few and inconclusive (Dijkstra, Monaghan, & Pauls, 2014). Here, we present evidence of temporal isolation in a stream dwelling stonefly.
Stream dwelling invertebrates face habitat fragmentation on a variety of spatial scales. Their linear habitats generally cover less area and support smaller populations than most terrestrial or marine habitats (Whittaker & Likens, 1973), yet they harbor a relatively large proportion of all known animal species, particularly insects (Dijkstra et al., 2014;Dudgeon et al., 2006). Running waters (lotic habitat) are geologically more stable than still waters (lentic habitat), which may explain why lentic insects tend to be more capable of dispersion, while lotic insects demonstrate more genetic differentiation (Hof, Brändle, & Brandl, 2006;reviewed in Dijkstra et al., 2014). Genetic differentiation between drainage basins has been reported for many aquatic animals (e.g., Monaghan, Spaak, Robinson, & Ward, 2002;Phillipsen & Metcalf, 2009;Schmidt, Bart, Nyingi, & Gichuki, 2014) and is most pronounced in flightless species restricted to headwater habitats (Hughes, 2007). Within a drainage basin, drift leads to a predominantly downstream migration, while steep waterfalls constitute barriers to upstream dispersal for nonflying animals. The fauna of individual headwaters is therefore often more isolated and genetically less diverse than in higher-order streams (Alp, Keller, Westram, & Robinson, 2012;Wofford, Gresswell, & Banks, 2005), while the diversity among headwaters is high (Finn, Bonada, Múrria, & Hughes, 2011). Moving from headwaters to higher-order streams, habitat characteristics like aquatic and riparian vegetation, stream velocity and bottom sediment change, and organisms are typically adapted to particular parts of the drainage basin (McCabe, 2010).
In general, winged aquatic insects are capable of dispersal over a longer distance (c. 0.5-1 km) in the adult stage, but the majority of specimens do not fly far from their emergence site and generally follow the watercourse (e.g., Briers, Cariss, & Gee, 2002;Briers, Gee, Cariss, & Geoghegan, 2004;Griffith, Barrows, & Perry, 1998;Macneale, Peckarsky, & Likens, 2005;Petersen, Masters, Hildrew, & Ormerod, 2004). While incidental longer-range dispersal is important for the (re)colonization of new habitat patches, the limited gene flow resulting from short average dispersal allows for genetic adaptations to the local environment driven by differential selective pressure. In aquatic invertebrates, intraspecific genetic divergence is often found on small geographical scales (<50 km) (e.g., Alp et al., 2012;Yaegashi, Watanabe, Monaghan, & Omura, 2014). Watanabe, Kazama, Omura, and Monaghan (2014) found that divergence in three caddisflies and one mayfly at genetic loci under selection correlated most strongly with the ecological factor that was most limiting to the species' distribution (elevation, concentrations of algae, or ammonia nitrogen). Laboratory studies of stoneflies have identified a number of inheritable populationspecific life cycle characteristics, including the occurrence of egg diapause (Khoo, 1968), duration of nymphal growth period (Lillehammer, 1975b), and optimal temperatures for egg development and nymphal growth (Brittain, Lillehammer, & Saltveit, 1986;Lillehammer, 1987;Lillehammer, Brittain, Saltveit, & Nielsen, 1989).
While genetic divergence of populations in separate drainage basins may lead to allopatric speciation by genetic drift, divergence of populations adapting to contrasting environments within a drainage basin may lead to reproductive isolation, a process known as ecological speciation (Rundle & Nosil, 2005;Schluter, 2009). Assortative mating, immigrant inviability, and reduced fitness of hybrids enhance ecological speciation (Schluter, 2009). As an example of ecological selection, Yaegashi et al. (2014) found populations of the caddisfly Stenopsyche marmorata to be structured into upstream and downstream clusters, rather than according to catchments. Longitudinal structuring of populations within a drainage basin may be enhanced by lakes that constitute a barrier to gene flow in lotic species (Monaghan, Spaak, Robinson, & Ward, 2001). Such structuring may be an early stage of parapatric ecological speciation and is in line with the observation that congeneric aquatic insect species often inhabit overlapping reaches of a drainage basin (e.g., Kaćanski, 1971;Rupprecht, 1983;Gordon, Wallace, & Grubaugh, 1998;Statzner & Dolédec, 2011;Boumans & Murányi, 2014;reviewed in Dijkstra et al., 2014).
Our study addresses a case where adaptation to a small stretch of running water necessitated phenological adaptation, leading to temporal isolation and putative or incipient speciation. The widespread, spring-emerging West Palearctic stonefly, Leuctra hippopus Kempny 1899, has a morphologically and phenologically atypical population in the Isterfoss rapids in central Norway, previously described and studied by Lillehammer (1974Lillehammer ( , 1975bLillehammer ( , 1976Lillehammer ( , 1986Lillehammer ( , 1987 and Ørmen (1991). Our main hypothesis is that this population has become genetically isolated from conspecifics in nearby streams due to earlier emergence and short adult life, with the lack of gene flow enhancing genetic adaptation to local conditions. An alternative hypothesis is that the Isterfoss population has an historical origin that is different from other Scandinavian populations (Ørmen, 1991: 63-64).
In order to test the temporal isolation hypothesis, we reassessed the morphological distinctness of the L. hippopus population from Isterfoss with additional morphometric measurements. Subsequently, we analyzed mitochondrial and nuclear sequence data in order to assess the phylogenetic position of the Isterfoss population relative to Scandinavian and Western European conspecifics, testing the hypothesis of a distinct origin. Thirdly, we used genomewide sampled molecular markers (AFLP and SNPs) to establish the degree of genetic distinction and admixture of L. hippopus from Isterfoss and five additional sites at distances ranging from 8 to 1,280 km. With these analyses, we tested the hypothesis that Isterfoss is genetically homogeneous and distinct from its geographical neighbors. If the hypothesis of a distinct historical origin can be refuted, a low degree of admixture is considered indicative of (ongoing) local diversification and incipient speciation.

| Study species and locality
Leuctra hippopus (Plecoptera, Leuctridae; Figure 1) is a detritivore stonefly that is widely distributed in streams of the western Palaearctic (Graf et al., 2009). It has a 1-year life cycle, with a shortlived adult stage. The time of emergence in Fennoscandia varies from early March in coastal areas with oceanic climate, to June and July at high altitude and latitudes (Lillehammer, 1988: 148). Adults of typical populations live in the riparian zone and are capable of flight. A high degree of morphological variation in both males and females of L. hippopus has been documented for specimens collected from different localities in Norway, notably in the shape of abdominal tergites as well as wing length and wing venation (Lillehammer, 1974(Lillehammer, , 1976(Lillehammer, , 1985(Lillehammer, , 1986Ørmen, 1991: 36-41). While there is also within-site morphological variation, the species has a tendency to produce local forms.
The most characteristic and best documented local form occurs in the Isterfoss rapids.
While all closely studied Norwegian populations of L. hippopus show within-population variation in the shape of the abdominal sclerites, the population in Isterfoss is relatively homogeneous (Lillehammer, 1974(Lillehammer, , 1976. Likewise, Isterfoss is the only Norwegian population that has been reported to be homogeneously short-winged (Lillehammer, 1974(Lillehammer, , 1976. It is also relatively homogeneous in terms of variation in wing venation patterns, with a high proportion of individuals with an irregular pattern that occurs only at lower frequencies elsewhere (Lillehammer, 1974). The irregular wing venation is a heritable trait that was maintained in stoneflies reared in laboratory for two generations (Lillehammer, 1976(Lillehammer, , 1986. Unlike typical L. hippopus, the specimens from Isterfoss do not fly (Ørmen, 1991: 13; and own observations). Finally, they are more pigmented (Lillehammer, 1976) and more stockily built (Ørmen, 1991: 13) (Ørmen, 1991: 13). Also because the population is flightless, we assume that the eggs and nymphs de-  (Lillehammer, 1975b;Ørmen, 1991: 12). We obtained long-term, year-round measurements of temperature from nearby measurement stations from the Norwegian Water Resources and Energy Directorate (NVE). The temperature of the outflow of Lake Isteren at Isterfoss can be assumed to be similar to that measured at the outflow of Lake Femund, c. 5.3 km upstream from Isterfoss ( Figure   S1). (In this lake outflow, L. hippopus does not occur.) Temperature data were also available from the Atna River, c. 5 km upstream from our additional sample collecting site on the Atna River (Folldal, see below).
The temperature regimes of these two sites are typical for Norwegian lake outflows and mountain rivers, respectively (Asvall, 1994). The temperature is consistently higher in the lake outflow than in the river, the difference ranging from about 0.3°C in winter to 3.3°C in August and September ( Figure 2). Changes in water level at Isterfoss can be deduced from the daily discharge data collected downstream from Isterfoss at Lake Galtsjøen. Sudden water level rise resulting from snowmelt occurs in the first 3 weeks of May ( Figure 2).
In Isterfoss, adult L. hippopus are typically present during the second half of April and the first half of May (Lillehammer, 1975a(Lillehammer, , 1987, although emergence is earlier in some years (Ørmen, 1991: 11;Cleven, s.a. [1992]: 17). At the time of emergence at Isterfoss, the surrounding landscape is still partly snow-covered (own observation), and in some years, other streams are still covered with ice (Ørmen, 1991: 14). Emergence in Isterfoss is much earlier than in nearby streams (Lillehammer, 1987). For instance, in a stream in Tynset, 65 km away and 500 m a.s.l., adult L. hippopus are present F I G U R E 1 Leuctra hippopus Kempny, mating pair on one of the boulders in the Isterfoss rapids. Photograph taken by L. Boumans on 23 April 2011 in June (Lillehammer, 1975a). Specimens from the nearby stream Hølbekken (see below) also emerge about 1 month later than in Isterfoss (own observations). Laboratory experiments have shown that nymphal growth and timing of adult emergence of L. hippopus are determined by water temperature as well as population-specific genetic factors (Lillehammer, 1975b(Lillehammer, , 1987. The life span of adults from Isterfoss is only about a week, compared to 4 weeks or more for other Norwegian populations. As a consequence, sexual maturity is reached sooner after emergence, and females lay fewer eggs (Lillehammer, 1975b;Ørmen, 1991: 13-14, 22). The relatively high temperature of the lake outflow ( Figure 2) enables faster nymphal growth and an earlier emergence.

| Specimens
In order to establish the monophyly of the species and the phylogenetic position of the Isterfoss population in Scandinavia and Western Europe, we sequenced the mitochondrial marker cytochrome c oxidase subunit I (COI) from 55 L. hippopus specimens collected in northern, central, and southern Europe, either collected by ourselves or donated by colleagues. In addition, we included 12 specimens in total of the following closely related species as outgroup: L. andalusiaca Aubert 1962, L. elisabethae Ravizza 1985, L. hippopoides Kacanski & Zwick 1970, and L. pseudohippopus Rauser 1965 The nuclear markers 28S and ITS were sequenced for a subset of these. Specimens used in molecular research are preserved in 96% ethanol; all specimens are deposited in the DNA bank of the Natural History Museum, University of Oslo (NHMO). An overview of the specimens is presented in Table S1.
For the characterization of the Isterfoss population by means of morphometry, AFLP fingerprinting, and SNP analysis, we used adult specimens collected in Isterfoss on 23 April 2011 and 8 May 2012.
These were analyzed together with specimens from five additional collecting sites ( Figure 3, Table 1). We refer to the collecting sites in Table 1 as statistical "populations" in the context of AFLP and SNP sequence analysis, while they are not necessarily, or not all to the same degree, distinct populations in the biological sense.

| Morphometry
Morphometrical measurements were used to reassess the morphological distinctness of the Isterfoss population. Lillehammer (1974) measured the wing length relative to the whole body length. However, stoneflies are soft-bodied animals that are stretched to varying degrees as a result of egg production, desiccation, or postcollection treatment, which complicates the measurement of body length. For this reason, we reassessed wing length variation, this time dividing the wing length by the width of the head as a proxy for body size (Beer-Stiller & Zwick, 1995;Zwick, 1992). We measured specimens from Isterfoss and the additional sites (Table 1) Table S2 but not analyzed statistically due to the low number of samples. The ANOVA was run separately for female and male specimens.

| Sequence data
We extracted DNA from the head plus prothorax or, in the case of samples used for restriction site-associated DNA (RAD) sequencing, the entire body except the abdomen. The GeneMole extraction robot was used for all extractions used in AFLP or RAD tag protocols; for specimens used for sequence data, we also used the Qiagen DNeasy Blood and Tissue Kit. In both cases, we followed the manufacturers' protocols. Skeleton parts were not crushed but retrieved after DNA extraction and stored with the remainder of the specimen.
The mitochondrial marker COI of L. hippopus specimens and outgroup taxa was amplified with the protocol described in Boumans and Baumann (2012). Part of the COI sequences was produced at the sequencing facility of the Canadian Centre for DNA Barcoding in Guelph in the framework of the DNA barcoding projects NorBol-Freshwater Insects (NOEPT) and West Palaearctic Plecoptera (WPPLE), and retrieved from the Barcode of Life Data System (BOLD) (cf. Ratnasingham & Hebert, 2007 following the methods described in Boumans and Murányi (2014).
GenBank accession numbers of all sequences used are given in Table S1.
The package Geneious version 6.0.5 was used for sequence editing and alignment. Phylogenetic trees of the COI sequences were constructed with distance (neighbor joining, NJ) and maximumparsimony (MP) methods implemented in Paup*, and with Bayesian inference (BI) in MrBayes version 3.2 (Ronquist & Huelsenbeck, 2003).
Details are given in Appendix S1. The 28S and ITS data matrices were not suitable for phylogenetic inference due to, respectively, lack of sequence variation and variation occurring largely in indels. The most divergent ITS sequences are separated by two 3-to 5-bp-long indels whose homology and alignment is doubtful. However, to visualize the sequence divergence in ITS, we created a single most parsimonious tree based on an exhaustive tree search in which "gaps" were treated as a fifth character state.

| AFLP fingerprinting
We performed AFLP fingerprinting of 56 specimens, representing the collecting sites at Isterfoss, Femundsenden, Folldal, and Vardø in Norway and Rekem in Belgium (Table 1). (Samples from the Ringsaker site were not included in this analysis.) The AFLP laboratory protocol and data collecting procedure followed Alsos et al. (2007) and Skrede, Borgen, and Brochmann (2009) with minor modifications.
As scoring by means of visual inspection introduces a high degree of subjectivity (Bonin et al., 2004), we analyzed unedited peak heights in the size range from 50 to 500 bp output by GeneMapper with the R program AFLPScore version 1.4 (Whitlock, Hipperson, Mannarelli, Butlin, & Burke, 2008). After removing badly performing samples and markers, we retained 109 markers in 48 specimens representing the five above-mentioned populations. The average number of markers scored as present per specimen was 19.7 ± 4.5 (range 9-28). Error rates were calculated with the help of 13 replicated samples and ranged from 2.8% to 3.2% for three subsets produced with different selective primer pairs. Details of the laboratory and peak scoring protocols are given in Appendix S1.
The input files for these programs were created with the R program AFLPDat (Ehrich, 2006). Structure was used to infer the most likely clustering of sample specimens into population clusters. We allowed for individuals to have mixed ancestry (admixture model) and did not use the sampling sites as prior. The maximum number of inferred clusters was set to 7, allowing for the recognition of each sampling site as a cluster, in addition to some degree of site internal clustering. For each value of K in the range 1-7, 10 runs were performed each consisting of a burn-in phase of 500,000 repetitions followed by 1 million so as to identify the most divergent specimens and find the optimal clustering, respectively. The module "Spatial clustering of groups" was used to produce a Voronoi tessellation graphic (Corander, Sirén, & Arjas, 2008).

| Restriction site-associated DNA (RAD) markers
We prepared RAD sequencing libraries for 25 stoneflies from six different populations using the restriction enzyme SfbI After removing low-quality and incomplete reads, we retained approximately 11 million reads, corresponding to c. 438,000 reads per specimen. There were large differences between specimens, with the number of reads ranging from 41,000 to 1,348,000. After removal of lower-quality final sites, the reads were trimmed to a length of 86 nucleotides, including the first six invariable positions that belong to the restriction enzyme recognition site. Sufficient tags were obtained to allow for population genetics analysis, but measurements of heterozygosity may be biased. For this reason, we will not discuss population genetics measures based on heterozygosity, although we present these as supporting data (see Data accessibility).
We used the denovo_map.pl pipeline of programs in Stacks v1.29 (Catchen et al., 2011) Table 2). The read depth parameter is applied to alleles in the denovo and to individuals in the populations program (Stacks Manual p.24, 26). In order to avoid an artifactual surplus of real and erroneous homozygotes, we set the minimum read depth in populations to twice its value in denovo (i.e., to 6).
The amount of missing data is proportional to the genetic distance between taxa (Huang & Knowles, 2014). Preliminary analyses in Structure showed that the parameter setting that requires each locus to be present in at least two populations (-p 2) led to a surplus of missing loci in the two most distinct populations in Rekem and Vardø due to many loci being shared only between the more closely related southern Norwegian sites. This in turn led to a tendency to cluster Belgium and Vardø, which may be an artifact. Therefore, we performed two alternative analyses: Matrix 1 with the loci present in all six populations, and Matrix 2 without the samples from Vardø and the loci present in at least two of the other five populations (Table 2).
To create Structure input files for both matrices, we used whitelists of loci with 1-4 SNPs and no more than eight alleles (Matrix 1), or 1-3 SNPs and no more than six alleles (Matrix 2) with the export_ sql.pl program. In the program populations, each locus was required to occur at least once in a population. With the parameter "-write_sin-

| Morphometry
On average, females are larger than males, and they also have longer forewings relative to the width of the head (Figure 4, Table S2).
Leuctra hippopus of both sexes from Isterfoss have significantly wider heads than those from either "other Hedmark" and Vardø (Figure 4).
Likewise, the wings of both sexes are significantly shorter in Isterfoss than in the other Norwegian populations (Figure 4). Post hoc tests distinguish Isterfoss from either "other Hedmark" or Vardø with pvalues < .001 for head width and wing length in both sexes (Table S2).
Our measurements show no difference between the specimens from Vardø and those from the "other Hedmark" group. (not shown), but the ITS sequences, displaying much variation in the form of indels, support the monophyly of the Norwegian population.

| COI, 28S, and ITS sequences
In particular, the two ITS sequences from Isterfoss have the same ITS allele as the single sequence from Folldal ( Figure S6).

| AFLP
The analyses of the AFLP data in Structure show the highest posterior probability combined with a small standard deviation at K = 4, distinguishing (1) Vardø, (2) Isterfoss, (3) Folldal plus Femundsenden, and (4) Rekem (Figures 6a and S7). Evanno's measure ΔK reaches its highest value at K = 2, corresponding to the uppermost level of structuring in the dataset and separating Vardø from the West European populations. The AFLP data suggest that the distinction between Isterfoss and its geographical neighbors is larger than between the Belgian and the south Norwegian populations.
Clustering specimens or populations with the same data matrix in BAPS yielded three clusters, both with and without spatial information priors, if the maximum number of clusters K was set as 20: (1) Vardø, (2) Isterfoss, and (3) Folldal plus Femundsenden plus Rekem, corresponding to the Structure analysis with K = 3 (Figure 6a). With K set as 2, BAPS also distinguishes between Vardø and the western populations. BAPS did not detect admixture among the three clusters.
The Voronoi tessellation graph ( Figure S8) visualizes that the Isterfoss population differs from the neighboring populations.     Num Indv, average number of individuals per SNP; Q, frequency of the minor allele (1-P); π, estimate of nucleotide diversity based on pairwise number of nucleotide differences; Var, variance; Private, percentage of the total number of 825 private alleles found in each population.

| RAD-based SNPs
Femundsenden are subdivided, but no cluster corresponding to any of these collecting sites is formed ( Figure S9). For Matrix 2 (without Vardø), with each locus being represented in at least two of the five West European populations, the highest value of ΔK corresponds to K = 3 (Figures 6c and S10), identifying the Belgian samples, Isterfoss, and the remaining sites from Hedmark county as separate clusters.
When higher values of K are set, the Isterfoss cluster remains unchanged while the other clusters become subdivided. The histograms in Figures 6, S9 and S10 illustrate that the Isterfoss is more homogeneous than the other populations from southern Norway.
PCA and IBS clustering of the two smaller data matrices 1a and 2a (Table 2)  SNPs, the three specimens that have most missing data (75-91%) end up in unexpected positions ( Figure S16). This is an artifact that does not occur if, due to different data filtering, the dataset is slightly larger (data not shown). homogeneity of this coastal population is plausible. The collection site in Vardø is situated along a c. 10-km-long river on a peninsula on the northernmost edge of the continent (site FinLoc106 in Ekrem et al., 2012). Therefore, immigration is likely to contribute very little to local genetic variation in comparison with the collecting sites in southern Norway which are part of a large drainage basin.

| Ecological speciation
The Isterfoss population of L. hippopus illustrates the process of ecological speciation. It can be discussed whether the current Isterfoss population of L. hippopus should be considered as a distinct species. Ultimately, this is a matter of preference and practical considerations (see Notes on taxonomy and phylogeography, Appendix S2).
Scientifically, our interest is in demonstrating speciation by temporal isolation as an evolutionary process. Adaptation to the Isterfoss rapids is both morphological and phenological. The Isterfoss population is morphologically distinct from the other Norwegian populations in having wider heads and relatively short wings. The wider heads can be an expression of overall larger size, a more compact build ormost likely-a combination of both. The whole body is "fairly long" (Lillehammer, 1976: 171), and probably slightly longer than that of typical Norwegian populations (Lillehammer, 1987). The short wings measured in units of head width are likely due to a combination of wing reduction and a more stocky body shape. Both may have an adaptive value in that they enhance the adult stoneflies' ability to resist wind and cling to the boulders on which they aggregate and mate (Ørmen, 1991: 13). Typical Leuctra are slender long-winged insects and Isterfoss specimens blown away from their aggregation sites would be unlikely to reproduce. Typical L. hippopus adults hide in riparian vegetation and are much less exposed to wind.
Phenological adaptation to local environmental factors is reflected in the early emergence and short adult life span of the Isterfoss population. While the higher autumn and winter temperature of the Isterfoss rapids enables faster growth of the immature stages, the early emergence and short adult life span are likely to be adaptations to avoid spring snowmelt floods that threaten to splash or even cover the boulders during May. Typical populations of L. hippopus are not particularly exposed to flooding as they emerge after snow melt and peak discharge, live in the riparian zone, and are capable of flight. The hypothesis that the phenological characteristics of the Isterfoss population are driven by flood avoidance (Ørmen, 1991: 14) finds support in the longterm discharge data that show a sudden rise in water level in the sec- Many stonefly species use species-specific patterns of substrateborn vibrational signals for mate finding (Boumans & Johnsen, 2015;Stewart & Sandberg, 2005), and for this reason, Ørmen (1991) investigated whether variation in mating signals could be a barrier against gene flow between conspecific populations. He recorded male mating calls from Isterfoss as well as a number of other sites in Norway (Ørmen, 1991: 22-25), but no differences between populations were found (described in more detail by Cleven, 1992Cleven, s.a. [1992: 30-31).
No female signals of L. hippopus have ever been recorded; hence, it is not certain whether vibrational communication plays a role in mate finding or mate acceptance in this species. However, it almost certainly plays no role in the Isterfoss population as adults aggregate in high densities on nonresonant boulders. Besides, mating experiments revealed no ethological barrier to mating between males from Isterfoss and females from a typical population from Oslo (Ørmen, 1991: 25, 45-46).
It can be argued that Isterfoss is an example of sympatric speciation, depending on the scale in time and place. Typical adult L. hippopus disperse primarily along streams, mostly upstream, with few traveling farther than 150 m inland from the stream banks (Kuusela & Huusko, 1996;Müller, 1973;Petersen et al., 1999Petersen et al., , 2004 (Elliott, 1987). Ørmen's results are suggestive of intrinsic postzygotic isolation, but any conclusion must await a largerscale experiment with controls of same-site matings. Because the Isterfoss population is adapted to its environment in several ways, we expect that hybrids would suffer reduced viability either in Isterfoss or in a typical L. hippopus habitat (extrinsic postzygotic isolation).
The Isterfoss population must have diverged locally from other south Norwegian populations in the 10,000 years after the last glaciation. The four collecting sites in Hedmark (Ringsaker, Folldal, Femundsenden, and Isterfoss) have similar numbers of private alleles in the RAD-based SNPs. That only the specimens from Isterfoss form a distinct cluster appears to be due to their similarity, that is, more private and nonprivate alleles being shared among the specimens, rather than to their strong genetic divergence from the other Hedmark populations. Our interpretation is that genomewide allelic divergence is primarily a function of time and that the different south Norwegian populations are of similar age. The divergence of the Isterfoss population in morphology and life cycle characteristics does imply genetic changes, but these are likely to be few relative to the wide sampling of RAD loci.
While the Isterfoss rapids are now inhabited by an atypical form of L. hippopus, it has not in recent times been recolonized by the typical form, even though this stonefly is common in all suitable habitats in Scandinavia. The habitat characteristics of the rapids are so different from the species' typical habitat with riparian vegetation that it may not be a suitable habitat for the typical L. hippopus. So how did this species colonize the Isterfoss rapids in the first place? A possible explanation is gradual habitat isolation, similar to cospeciation of parasites or pollinators and their hosts (Coyne & Orr, 2004: 191). We propose About one-fifth of the European stonefly species are microendemics restricted to small isolated mountain ranges (Graf et al., 2009: 9). In particular, the families Nemouridae and Leuctridae include many short-range endemics (Aubert, 1959;Graf et al., 2009: 75-82;Vinçon & Ravizza, 2000). It is reasonable to assume that processes of ecological speciation similar to that in Isterfoss together with the generally restricted dispersal capacity of stoneflies explain this high incidence of endemism. Intraspecific genetic differentiation in different ranges of the same stream was found in some studies (Alp et al., 2012;Watanabe et al., 2014;Wofford et al., 2005), and there are several examples of congeneric species of aquatic insects inhabiting adjacent ranges. Differential longitudinal distribution and adaptation to the local environment is probably a common speciation mechanism in lotic invertebrates. Our study of the population in the Isterfoss rapids, which were glaciated until approximately 10,000 years ago, shows that genetic isolation and adaptation to very local circumstances can take place in sympatry over a relatively short time period.

ACKNOWLEDGMENTS
We thank Liv-Guro Kvernstuen and Gunnhild Marthinsen for their help with and advice on, respectively, AFLP and RAD sequencing techniques at the DNA laboratory at the Natural History Museum in  Figure S4). Ånund Kvambekk at the Norwegian Water Resources and Energy Directorate (NVE) provided the water temperature and discharge data series.

CONFLICT OF INTEREST
None declared.

DATA ACCESSIBILITY
Main characteristics of the samples included in this study are provided in Table S1, together with GenBank accession numbers for the associated DNA sequences; additional details are available from the BOLD online database (http://www.boldsystems.org/). Appendix S1 provides methodological details. Morphometric data are provided as Table S2.