may be among the key factors determining establishment success of new invaders. For the ‘S.Iberian’ lineage, we found a high level of shared haplotypes only in the marisma sites, and in populations assigned to the habitat type ‘open, close’, i.e. temporary ponds in open habitat situated within the Guadalquivir delta but located outside the marisma habitat. These are the habitats in which we predicted Triops to have the highest dispersal probabilities. This may suggest that in T. mauritanicus, new immigrants have a discernible effect on the local gene pool only where dispersal rates are exceptionally high.
Within the Guadalquivir delta, we found genetic diversity of Triops mauritanicus populations to differ among the three major habitat types: populations in the marismas showed the highest diversity, while populations in enclosed woodland or shrubs had the lowest. Populations in open habitat outside the marismas showed intermediate values of genetic diver- sity. This habitat type probably has been established in the Guadalquivir delta only recently, as a result of forest degradation and clearing following human settlement. Within the Doñana National Park, the natural forest vegetation was still present in 1636, but was rapidly replaced by small pioneer shrub species within a few decades, due to overgrazing, burning, charcoal making and lumbering (Granados Corona et al. 1988). Forest degradation is likely to have occurred earlier in adjacent areas, since the above- mentioned changes in vegetation resulted from a change in land use from an exclusively aristocratic hunting estate to a pasture for livestock (livestock from neighbouring towns was not admitted to the Doñana until 1628, see Granados Corona et al. 1988). The opening of formerly forested areas appears to have resulted in increased gene diversity of populations situated within these areas. It appears unlikely that the increase in gene diversity has been the result of increased population sizes due to higher primary productiv- ity following a reduction in shading (e.g. Mokany et al.
2008), because we did not find a positive correlation of habitat size to gene diversity despite the fact that our study ponds covered a considerable range of surface areas (approx.
60–124,480 m2, marisma sites not included). Instead, we
attribute the rise in gene diversity to increased dispersal probabilities. The fact that ponds in open habitat more than
75 km away from the marismas had similarly low diversity
to the forest/shrubland populations near the marismas supports this interpretation. Furthermore, the morphology of four of our study ponds in open habitat close to the marismas suggests that they are artificial [a comparison with aerial photographs taken in 1956 (available at http://www. juntadeandalucia.es) also suggests that at least two of these ponds did not exist at that time; M.K. pers. obs.] and thus were colonized only recently by two to three Triops haplotypes each, which supports our interpretation of high dispersal probabilities for ponds in open habitat.
The observed pattern is in line with our hypothesis that dispersal by waterbirds should result in an increased diversity in open habitats, but less likely in ponds surrounded by forests or shrub. Waterbirds may indeed have been the main dispersal vector leading to an increase in gene diversity in the open habitats, as we found a clear positive relation of waterbird abundance estimates to gene diversity (Fig. 4b). However, in these temporary ponds located within 2 km of the original (approximately year 1900) marisma borderline, wind dispersal and human activities such as ploughing could also have resulted in an increased diversity in the open habitat, but would be less likely in ponds surrounded by forest or shrubs. Thus, our data do not allow discriminating between these dispersal vectors, because all are likely to contribute to dispersal on a local scale. However, our data suggest that mammals are less effective dispersal vectors for Triops, because although they might represent a major dispersal vector within forest habitat (Vanschoenwinkel et al. 2008), their presence in forest and shrublands did not result in an increase in gene diversity (gene diversity was 0 in all but one of the forest/shrubland ponds studied). Consequently, waterbirds may represent the main vector for dispersal on a supra-local scale, as suggested for other large branchiopods (Hawes 2009).
The inferred lower dispersal probabilities in forested areas may have been a key factor for the evolution of Iberian lineages of T. mauritanicus, because regular strong fluctuations between forest and steppe vegetation have occurred in the Mediterranean region since approximately
2.5 million years before present (Thompson 2005) and the onset of these fluctuations roughly coincides with estimates of divergence times within these lineages (Korn et al.
2006). During phases with forest vegetation maxima,
dispersal may have been limited, promoting diversification into sublineages. During these periods, dispersal may have occurred mainly among a few temporary marshes and associations of large temporary ponds, which might have been the most attractive to waterbirds, whereas in areas with small scattered ponds Triops may have been much more isolated and prone to extinction, owing to small population sizes. The low genetic divergences within the Iberian T. mauritanicus lineages as compared to the high divergence observed for the northern African T. m. mauritanicus (Table 3; see also Korn et al. 2006) is indeed indicative of phases with increased extinction rates of Iberian populations and of resulting bottlenecks. In fact, for Triops in the south-west Iberian peninsula it might have been the forest maxima associated with glacial cycles, rather than the glaciation maxima or Heinrich events (episodes of massive iceberg discharge into the North Atlantic) with predominance of semi-desert vegetation (Fletcher and Sánchez Goñi 2008), that may have caused
these bottlenecks. The most unfavourable conditions may have been met in periods with closed forest under a temperate, moist, continental climate, as it appears to have formed in the study area during the late glacial Allerød interstadial, prior to the Younger Dryas (Fletcher et al.
2007). Similar conditions may also have occurred during preceding glacial cycles. To our knowledge, Triops has never been found in woodland ponds in regions with a temperate, moist climate, which suggests that such habitat is unsuitable. Contrary, the cold steppes found during glaciation maxima or Heinrich events may have been less unfavourable, as may be suggested by the north-eastward range extension of T. c. cancriformis under a cold continental climate in Russia, where the species has been reported from Syktyvkar (61°40′N) and Ukhta (63°34′N; Vekhoff 1993).
For the ‘S.Iberian’ lineage, the high number of private
haplotypes found in the Guadalquivir delta, in a small area in south central Portugal (roughly on a line connecting Mértola and Castro Verde), and in Extremadura suggests that these areas were already colonized prior to the last Ice Age. This is supported by the fact that the area in south central Portugal that hosts numerous private haplotypes in Triops is also characterized by a diverse fauna of other large branchiopod taxa, most importantly an endemic species of Tanymastigites (Cancela da Fonseca et al. 2008), a genus typical of semi-desert climates. Furthermore, the only known records of the ‘Gitanilla’ lineage are from Extre- madura. In both areas, a combination of local climatic conditions and soil type may have supported open-ground habitats even during forest maxima. This interpretation of our data suggests that the range of the ‘S.Iberian’ lineage before the last Ice Age might already have been similar to the presently occupied range, and thus might be the result of an earlier range expansion.
Three natural barriers (besides the Atlantic Ocean and the Mediterranean Sea) appear to have been limiting for this postulated earlier range expansion, but also for post-glacial recolonizations: first, a group of mountain ranges towards the eastern and northern range edges and a smaller one in southern Portugal, and second, areas already occupied by other main lineages of T. mauritanicus. A third factor may have been the main river systems, as we found no representatives of the ‘S.Iberian’ lineage south of the Guadalete River, nor between the Guadiana River and the Sierra Morena mountain range (the central mountain range in Fig. 3) within Extremadura. The possible effect of rivers as barriers to successful dispersal would imply that mammals were an important dispersal vector, since they might avoid crossing large rivers (e.g. Eriksson et al. 2004; Frantz et al. 2010) or might lose a high proportion of adherent mud with resting stages from temporary ponds (Thiéry 1987; Vanschoenwinkel et al. 2008) due to wash
off (or abrasion of macerated mud by river sediments or riverine vegetation). This appears to be in conflict with our interpretation that mammals were less effective dispersal vectors within the Guadalquivir delta (see above). Howev- er, mammals may have a relatively higher impact in areas with lower abundances of waterbirds. Furthermore, it should be noted that our interpretation of dispersal and recent gene flow is based on the present situation. Mammals might have had a higher potential for dispersal in prehistoric times, when Iberia had a diverse fauna of large herbivores, including wild horse (Equus caballus), steppe bison (Bison priscus), the bovine Leptobos steno- metopon, steppe rhinoceros (Stephanorhinus hemitoechus), woolly rhinoceros (Coelodonta antiquitatis), and straight- tusked elephant (Palaeoloxodon antiquus; Fortelius 2003; Kurtén 1968), and man-made barriers were still lacking. Especially during dry periods with dominance of steppe vegetation, Triops habitats may have been among the main water reservoirs, and thus may have attracted herds of large herbivores that often migrate over longer distances, poten- tially carrying Triops eggs with them via external transport in mud that is attached to the legs when approaching the ponds for drinking (Thiéry 1987; see also Vanschoenwinkel et al.
The inter-lineage differences observed in the amount of differentiation between populations (Table 5) and in the distribution ranges of the lineages (Fig. 3) may have resulted both from priority effects (the ‘S.Iberian’ lineage might have been the first to colonize the lower Guadalquivir valley, where high waterbird abundance may have resulted in an increase in dispersal probabilities for that lineage, see Table 4) and from individual dispersal abilities. Smaller resting propagules are generally more likely to be dispersed by birds than bigger propagules (Green and Figuerola 2005). In addition, for a certain body size of females, large sizes of resting eggs should translate into small clutch sizes and vice versa. This could clearly result in differing dispersal probabilities via birds that feed on adult females, especially at the beginning of the reproductive phase of the noto- stracans (or in ponds with short flooding phases), when clutch size is often limited to only a few propagules. We thus hypothesize that in Triops there is a trade-off between dispersal probability (favouring a reduction in the size of resting eggs) and probability of survival in local populations (promoting increased egg sizes, see above). Thus, the large sizes of resting eggs in the ‘Portugal’ and ‘Gitanilla’ lineages (Fig. 6) may have resulted in low dispersal abilities in these clades (Table 4).
Biogeography of Triops mauritanicus
Recently, the presence of a Triops mauritanicus population was reported for Ares del Maestre in northern Spain
(Zierold et al. 2007), based on COI, ATP6 and ATP8 sequence data. The authors classified that population under the invalid name T. cancriformis mauritanicus (the corresponding taxon had never been recorded from north- east Spain before), but did not indicate how they had determined that name. In their Table 1 Zierold et al. (2007) stated that the taxonomic identity was merely “inferred”, but did not elaborate, leaving molecular data as the likely but unconfirmed basis. [It had been demonstrated before that the former T. c. mauritanicus is paraphyletic (Korn et al. 2006), so that only a clear morphological classification would have allowed assigning the sample to either T. m. simplex or a sublineage of the former T. c. mauritanicus (characterized by considerably longer furcal spines; see Korn et al. 2006).] Thus, the actual taxonomic identity of the sample in question cannot be inferred from that study. However, a comparison of preliminary COI sequence data obtained from a representative subset of T. mauritanicus samples with the sequence by Zierold et al. (2007) retrieved from GenBank (accession number EF675908) clearly allows assignment of the Ares del Maestre sample to T. m. simplex (see Appendix: Fig. A1). Furthermore, in the preliminary ML phylogenetic reconstruction the northern Spanish sample branches off first within T. m. simplex, which suggests an origin of that taxon in the Iberian Peninsula. Apparently, T. c. cancriformis and T. m. simplex co-occur (on a regional scale) in northern Spain, which may explain why different authors disagree on the taxonomic identity of Triops populations in that region (see review in Korn et al. 2006). Clearly, a phylogeographic study including a sound morphological reinvestigation of northern Spanish populations is needed.
Based on our data and the interpretation of new data from northern Spanish populations (see above), we con- clude that out of the six known main lineages within Triops mauritanicus only a single one (T. m. mauritanicus) appears to be absent from the Iberian Peninsula, and four of those lineages appear to be Iberian endemics, among them the two lineages that are indicated by our phylogenetic analysis to have branched off first within T. mauritanicus (i.e.
‘Cádiz’ and ‘S.Iberian’ lineages; see Fig. 2). This strongly supports the hypothesis formulated by Korn et al. (2006) that a common ancestor of all T. mauritanicus lineages was located in the Iberian Peninsula. Korn et al. (2006) had hypothesized that T. m. simplex and T. m. mauritanicus may both have diverged during a range expansion into northern Africa. However, the presence of T. m. simplex both in northern Africa and in Spain, and the basal position of the northern Spanish sample within T. m. simplex (see above) rather suggest that this taxon has evolved in the Iberian Peninsula. It might have dispersed to northern Africa more recently, well after T. m. mauritanicus may have diverged from a common ancestor of T. m. simplex, T. m.
mauritanicus and the ’Portugal’ and ’Gitanilla’ lineages during a range expansion from Iberia to northern Africa (west of the Atlas Mountains). However, more molecular data from northern Spanish and northern African popula- tions of T. m. simplex are needed to corroborate this modified biogeographical scenario.
Dispersal success among populations in relation to levels of genetic divergence
Our data indicate a clear-cut difference in dispersal success in relation to levels of genetic divergence. Within the main phylogenetic lineages, gene flow as indicated by shared haplotypes appears to be high, at least on an evolutionary time scale. The fact that several of the ponds for which we observed indirect evidence of gene flow appear to be artificial and thus of recent origin (see above) suggests that gene flow may be frequent at least among populations in the Guadalquivir delta.
In contrast to this high level of exchange between populations of the same lineage, our results suggest that gene flow between the main phylogenetic lineages might be very low or even completely absent: despite the high number of 422 Iberian individuals investigated with molecular tools, we did not detect a single population with a co-occurrence of haplotypes from different main lineages. Indeed, the most divergent 12S haplotypes observed to co-occur were ‘S.Iberian’ haplotypes 3 and
24 (0.9% divergence as uncorrected p-distance, repre- senting 82% of the maximum intra-lineage divergence observed; data not shown), and ‘Cádiz’ haplotypes 2 and
5 (1.1% divergence, 85% of maximum intra-lineage divergence). Mean inter-lineage divergences are consid- erably higher than divergences among these co-occurring haplotypes, ranging from 2.9% to 5.1% (uncorrected p- distances; Table 3). The lack of shared haplotypes at this magnitude of differentiation suggests that the divergences among main lineages may have reached a level of differentiation where outbreeding depression (or the formation of pre- or postzygotic isolating mechanisms) may prevent ongoing gene flow. In contrast, especially for the ‘S.Iberian’ and ‘Cádiz’ lineages, low dispersal abilities per se could hardly have caused the lack of shared haplotypes among lineages, as both lineages show rather high accumulative dispersal distances (Table 4), indicating high potential for dispersal.
Furthermore, they are geographically located close to each other in a region with only weakly expressed geographic barriers (represented mainly by the Guadalete River), and with abundant suitable habitats (indicated by high abundances of populated ponds, see Fig. 3) as well as high abundances of migrating waterbirds (Rendón et al.
2008) as potential dispersal vectors. Several of the
populations of the ‘Cádiz’ lineage inhabit coastal temporary marshes associated with rivers and frequented by water- birds similar to those found in Doñana (e.g. Pérez-Hurtado et al. 1993). The probability of dispersal between these small temporary marshes and the Doñana therefore appears to be high. This supports our assumption that the lack of shared haplotypes among these regions likely is a result of low establishment success of new immigrants rather than very low dispersal rates. Although the use of mitochondrial markers would have failed to indicate successful inter- lineage dispersal by males, the apparent absence of such dispersal in females suggests that there is no free gene flow among the main lineages of Triops in SW Iberia.
A recently conducted molecular and morphological rein- vestigation of all subspecies formerly assigned to Triops cancriformis has clarified the gross phylogenetic relation- ships within this group (Korn et al. 2006). As a conse- quence, T. mauritanicus has regained full species status, and now includes the northern African populations of the former T. cancriformis simplex as a subspecies, T. mauri- tanicus simplex. Korn et al. (2006) assumed subspecific status for the lineages discovered in Portugal and Spain, but had only preliminary data for these geographic areas. Little was known about the substructure, position and status of the Iberian lineages, so that no formal subspecific names were assigned to them. The present study greatly improved our knowledge on the substructure within Iberian clades, identi- fying 36 further haplotypes in the 12S gene and five new haplotypes in the 16S gene (previously, only four 12S and three 16S haplotypes were known). More importantly, an additional clade was discovered in the southernmost region of Iberia. The resulting comprehensive set of sequence data indicates a clear differentiation of T. mauritanicus into six main lineages (Fig. 2; Table 3): T. m. mauritanicus, T. m. simplex, and four additional lineages of similar order that appear to be confined to the south-western part of the Iberian Peninsula.
The six lineages have diverged by an average of 1.2–2.8% in the 16S gene and 2.9–5.1% in the 12S gene (uncorrected p- distances; Table 3). Among well recognized and morpholog- ically differentiated species of Notostraca (Rogers 2001),
16S sequence divergences may be as low as 2.8% (between Lepidurus arcticus and L. apus apus; uncorrected p- distances, calculated for the sequences listed in Table 1),
3.0% (between L. lemmoni and L. arcticus) or 3.3% (between L. lemmoni and L. a. apus). For the 12S gene, a corresponding low value is 6.9% (between L. lemmoni and L. a. apus; we did not calculate 12S divergences for the other Lepidurus species due to insufficient overlap with our long 12S fragment). Thus, the divergence into six taxa
within Triops mauritanicus by 1.2–2.8% (16S; average uncorrected p-distances) and 2.9–5.1% (12S) reaches a magnitude reflecting separations at species level in other notostracan lineages (see also Korn and Hundsdoerfer
2006). Consequently, it appears plausible that the six main lineages within T. mauritanicus represent young, but separate species.
This assumption is supported by the clear morphological
differentiation observed within T. mauritanicus: three of the main lineages, namely T. m. simplex, the ‘Portuguese’ lineage and the ‘Gitanilla’ lineage, can be distinguished clearly by adult morphology alone (Table 9), and the latter two are further differentiated from the remaining lineages by their significantly larger resting eggs (Fig. 6). The remaining three lineages show less pronounced morpho- logical differentiation (Tables 7, 9). Nevertheless, the observed high level of morphological differentiation among several of the lineages within T. mauritanicus is remark- able, given the overall weak morphological differentiation among notostracan species. It is the stasis in gross morphology (e.g. Suno-Uchi et al. 1997), combined with typically high within-population variability in morphologi- cal key characters (Linder 1952; Longhurst 1955), that poses great difficulties for morphological classification of the group. This has resulted, for example, in the classification of Lepidurus packardi and L. couesii as two subspecies of L. apus by Longhurst (1955). Similarly, some southern African populations have been variously assigned to T. cancriformis (e.g. Barnard 1929; Hamer and Rayner 1995) or to T. granarius (e.g. Longhurst 1955), and the latter includes at least three separate, possibly cryptic species and may even be paraphyletic, with T. longicaudatus grouping within it (Korn and Hundsdoerfer 2006).
Lastly, our molecular data suggest that there is no free gene flow between the main phylogenetic lineages, as we found evidence for a high level of exchange between populations of the same lineage but failed to detect any indication of recent gene flow among the main lineages (see above). Taken together, our results demonstrate that a taxonomic revision of Triops mauritanicus appears justi- fied. We therefore reinstate T. m. simplex to full species status, as Triops simplex Ghigi 1921, and describe the Iberian lineages as new species below.
Our data confirm the general, previously recognized pattern of a lower dispersal probability in gonochoric lineages of Triops. However, we found additional evidence that dispersal in this taxon may be further complicated by a strong habitat-
dependence of dispersal probability, mediated by prevailing dispersal vectors. Similar effects are likely to occur also in other pond-dwelling invertebrates that are passively dispersed.
Morphological determination of samples
A well-known peculiarity of the Notostraca is their high variability in morphological key characters, even within pop ulations (e.g. L ongh urst 19 55 ). The r efore, a morphology-based determination appears reasonable only at the population level, whereas reliable determination of individual specimens requires the application of genetic markers. For the morphological determination of Iberian populations we recommend a minimum of 10 male specimens per population. Although our data suggest that the ranges of the south-west Iberian lineages do not overlap, and that occurrence of Triops simplex or T. m. mauritanicus in this region is unlikely, we cannot rule out the respective alternative possibilities. Thus, we recom- mend the use of both discriminant function models described above. We suggest to primarily classify the Iberian populations by use of the DFA model including only Iberian males, as this model shows higher discrim- inatory power. For a reliable determination of morpholog- ically indeterminable populations, we recommend the application of genetic markers. We further suggest con- firming all records that are clearly outside the known range of a species via molecular methods, including south- west Iberian populations that are assigned to T. simplex or T. m. mauritanicus by the DFA model including males of all lineages. An Excel file with the morphological dataset used for DFA in the present study is available for determination purposes as “Electronic Supplementary Material” from the online version of this paper. Regarding measurements of resting-egg sizes, we recommend using eggs that were extracted from natural sediments. If lab- grown eggs are considered, great care has to be taken that not only clutches from small individuals are included, otherwise resulting mean values might underestimate actual egg sizes [preliminary results suggest that the correlation of egg size with body size varies between populations (M.K. pers. obs., data not shown)].
Taxonomy of southern Iberian lineages of the Triops mauritanicus species group