However, 100% correct classification was achieved for populations of the ‘Portugal’ and ‘Gitanilla’ lineages also with this DFA model.
Size of resting eggs
This character divides the populations into four significant- ly different groups of taxonomic importance (ANOVA, p <
0.001; Tukey post-hoc test, p < 0.01). It separates all Triops mauritanicus lineages (population means of resting-egg diameter 412–553 μm) from T. c. cancriformis, which
shows consistently lower values (population means 367–
390 μm). Within T. mauritanicus, the ‘Gitanilla’ lineage exhibits the largest eggs, separating them from all the remaining lineages (Fig. 6). The populations of the
‘Portuguese’ lineage have the largest eggs of the remaining
lineages. Triops m. mauritanicus, T. m. simplex as well as the ‘S.Iberian’ and ‘Cádiz’ lineages assume intermediate positions between T. c. cancriformis and the ‘Gitanilla’ plus
‘Portuguese’ lineages. Within this group, T. m. simplex
tends to have the smallest eggs (observed population means
412 and 439 μm), while the ‘S.Iberian’ lineage tends to
Fig. 5 Unrooted NJ tree of squared Mahalanobis distances between group centroids obtained from discriminant function analysis of morphologi- cal data on adult males of all known Triops mauritanicus line- ages. Abbreviations: T.m.m. =
T. m. mauritanicus; T.m.s =
T. m. simplex
Table 8 Squared Mahalanobis distances between centroids (multi- variate means) of statistical groups (main phylogenetic lineages in Triops mauritanicus; lineage and taxon abbreviations as in Table 7) in the discriminant function analysis (DFA) of morphological characters; probabilities p < 0.001, except p = 0.004 for ‘S.I’ vs. ‘C’ in females (*)
Males of all lineages
Males of Iberian lineages
G P S.I C 17.8 22.6 6.5
G 37.4 18.0
Females of Iberian lineages
Fig. 6 Resting-egg size in populations of Triops cancriformis and main lineages of T. mauritanicus (C = ‘Cádiz’ lineage; G = ‘Gitanilla’ lineage; MM = T. m. mauritanicus; MS = T. m. simplex; P =
‘Portuguese’ lineage; S. I = ‘S.Iberian’ lineage; T.c.c. = Triops c. cancriformis). Eggs from populations 082, 084, 058, 108–111, 119,
cultures, remaining samples extracted from field-collected sediments; for details on populations see Table A1. Error bars indicate 95% confidence intervals
reach larger egg sizes, especially one of the four popula- tions from Extremadura, which showed a population mean diameter of 499 μm (population 057, see Table A1). Nauplius sizes were exemplarily measured for two of the populations (003, T. mauritanicus, ‘S.Iberian’ lineage;
121, T. c. cancriformis; data not shown). Total length of naupliae was significantly higher in the T. mauritanicus population (ANOVA, p <0.001, F = 99.8, df = 1) compared to the T. c. cancriformis population (and was approxi- mately 40–50% bigger than the diameter of resting eggs).
This demonstrates that bigger egg sizes are not merely the result of an increased thickness of the outer coating of resting eggs.
Morphological determination of populations
Among the lineages within Triops mauritanicus, we observed the lowest level of morphological differentiation
classification for the discrimi- nant function analysis (DFA)
in the ‘S.Iberian’ lineage (Table 7). Interestingly, its morphological differentiation from the northern African T. m. mauritanicus is lower than its differentiation from the
‘Cádiz’ lineage (Table 8), even though geographical separation is clear in the former case but next to none in the latter (Fig. 3). Consequently, the inclusion of samples of the nominotypical subspecies in discriminant function analysis clearly reduces classification success in the ‘S. Iberian’ lineage (Table 7) and even results in misclassifica- tions of some of the populations (Table 9) due to overlap in morphological characters. However, since our phylogeo- graphic data suggest a clear geographical separation between the ‘S.Iberian’ lineage and the nominotypical subspecies (with the Mediterranean Sea and the ‘Cádiz’ lineage situated in between), this overlap in morphological traits will cause fewer determination problems if geograph- ical origin of samples is taken into account. The probability that populations of T. m. mauritanicus, which appears to be endemic to western Morocco (Korn et al. 2006; Table A1), might occur within the known range of the ‘S.Iberian’ lineage appears to be low. Consequently, for the determi- nation of Iberian samples, a priori classification probabil- ities could be lowered for the nominotypical subspecies in order to improve classification success in Iberian popula- tions. Alternatively, Iberian populations could primarily be classified using the DFA model including only Iberian males, as this model has a higher discriminatory power. The DFA model including males of all lineages could then be additionally applied to identify Iberian populations that might be assigned to T. m. simplex or T. m. mauritanicus. The identity of such populations should be confirmed via molecular methods.
Evidence for asynchronous morphological evolution
Within Triops mauritanicus, morphological differentiation is highest for T. m. simplex, followed by the ‘Portugal’ and
‘Gitanilla’ lineages (Table 8; Fig. 5). A comparison with molecular phylogenetic reconstructions (Fig. 2) demon- strates that these lineages have branched off last within T. mauritanicus (Fig. 2c) or are among the lineages that have diverged last (Fig. 2a, b). Hence, the more recently diverged lineages are the ones with the highest level of morphological differentiation within T. mauritanicus. This suggests that this monophyletic group within T. maurita- nicus evolves faster morphologically than the ‘S.Iberian’ and ‘Cádiz’ lineages. The low level of morphological differentiation between T. m. mauritanicus and the ‘S. Iberian’ lineage might therefore be a secondary effect, assuming that the common ancestor of T. m. mauritanicus, the ‘Portugal’ and ‘Gitanilla’ lineages and T. m. simplex might have been morphologically more similar to the latter three.
Dispersal, gene flow and differentiation among populations
The marismas (natural temporary marshes) of Doñana are habitat for an outstanding number of waterbirds. With an average of more than 300,000 wintering waterbirds (Rendón et al. 2008) in the Doñana wetland complex, this is one of the most important wintering sites for migratory waterbirds in the Western Palaearctic (Rendón et al. 2008). There is increasing evidence that waterbirds may represent an important passive dispersal vector for a variety of invertebrates via internal transport (Frisch et al. 2007; Green and Figuerola 2005; Green et al. 2008). Among numerous other taxa, Triops have been successfully reared from material that was recovered from the lower digestive tract of domesticated ducks fed on a mixture of crustacean eggs extracted from dry natural playa sediments (Proctor
1964). Furthermore, Notostraca are known to form part of the natural diet of waterfowl (Krapu and Swanson 1977) and herons (Kazantzidis and Goutner 2005), and may even serve as the major food item during periods of high abundance (Lo 1991). As notostracan resting eggs are carried in brood pouches before they are deposited in sediments, it appears likely that at least part of the eggs from brood pouches (those which are almost ready for deposition) will survive gut passage (see, e.g., Sánchez et al. 2007 on the viability of anostracan eggs from ovisacs of ingested adults). Thus, predation by waterbirds should result in dispersal of a high number of Triops resting eggs, especially when temporary ponds are drying out and Triops are easy prey items.
Pasture use of the marismas should further increase dispersal probabilities, as external transport of Triops eggs on the feet of hoofed mammals has been demonstrated by Thiéry (1987). Furthermore, floating eggs could be trans- ported over wide distances during peak floods. In addition to the presence of numerous potential dispersal vectors, we also recorded Triops in all parts of the marismas, and eggs reached high densities in several of the sediment samples investigated (M.K. pers. obs.). We would thus expect to observe the highest frequency of dispersal within the marismas. Indeed, AMOVA showed that among all habitat types investigated, the amount of differentiation between populations was lowest in the marismas (FST= 0.10; Table 5). Accordingly, we found clear indirect evidence for recent gene flow among the marisma sites, indicated by commonly observed co-occurrences of geographically spread haplotypes (see above). Nevertheless, AMOVA indicated a significant level of diversification (FST= 0.10, p < 0.001) among the marisma populations studied.
This is supported by morphological data: variables used to characterize the size of the furcal spines (which represent an important part of the armature) showed significant differences among samples obtained from the eastern edge
and from central and western parts of the marismas (MANOVA, p < 0.05). This is unlikely to represent the result of local adaptation, as predation risk is expected to be of the same magnitude in these sites (waterbird abundance estimated at level 5, ‘highly abundant’, for all marisma sites). The differences in furcal spine morphology might rather be associated with relative abundances of a certain haplotype or group of haplotypes, suggesting that Triops from central to western and from eastern parts of the marismas originated from different source populations with differing morphology: the samples in which we observed the largest furcal spines (populations 002 and 005; Table A1) were dominated by haplotypes ‘S.Iberia 17, 18 and 24’, which together with haplotype 27 form a subclade within the ‘S.Iberian’ lineage (Fig. 2a). These haplotypes were not observed in the samples from the eastern edge of the marismas, which showed smaller furcal spines (pop- ulations 007 and 046). Korn et al. (2006) had demonstrated that the furcal spine size shows low within-lineage variability, which suggests that this character evolves slowly. In contrast, the size and number of dorsal carina spines shows high variability (Korn et al. 2006) and may quickly adapt to local predation regimes.
Additive effects of a large resting propagule bank and adaptation to local environmental conditions may reduce the impact of new immigrants and thus keep the popula- tions distinct despite high levels of dispersal, as proposed for zooplankton by De Meester et al. (2002). Although the capacity for local adaptation is expected to be strongest for zooplankton that reproduces via cyclical parthenogenesis, obligately sexually reproducing taxa are also predicted to have a high capacity for adaptation to local conditions (De Meester et al. 2002). While competition for resources and adaptation to local predation regimes are regarded as the main driving forces for local adaptation in zooplankton (De Meester et al. 2002), the special morphological features of Triops appear to allow for an additional mechanism that could enforce a fast selection for local adaptation, i.e. cannibalism among early instar larvae: in T. mauritanicus, juveniles of approximately 5 mm total length (which is usually reached within 1 week after hatching, M.K. pers. obs.) are already capable of feeding on metanaupliae (M.K. pers. obs.). These observations on cannibalism among juveniles were made in a small outdoor mesocosm with natural sediments (population 002), as well as on freshly field-collected specimens (population 086). Cannibalistic juveniles eliminate potential competitors and at the same time may increase their growth rates so that they may reproduce with a higher probability before predation pressure increases towards later stages of succession (Moorhead et al. 1998) or before the pond dries out. We thus hypothesize that in T. mauritanicus the timing of hatching and the size of resting eggs (and thus naupliae)