Fig 1.
Map of the study area and variation in shell morphotype frequencies.
The bottom panel (maps G-K) shows five geographical contact zones between M. edulis and M. trossulus, maps in the upper panel (A—F)—other studied areas. Pins depict sampling sites. Pie diagrams depict proportions of T-morphotypes (black sector) and E-morphotypes (white sector) in M. trossulus (diagrams with a red border) and M. edulis (those with a blue border) in combined samples from particular regions. If the data on salinity in sampling localities are available and considered in the analyses, it is indicated by the color of pins (light green–brackish, dark green–saline, white–salinity is unknown) and the proportions of the T-morphotypes in combined samples from brackish and saline localities are presented separately in diagrams placed on light and dark green background, respectively. Source data are given in S1 Table and S2 Table. Inkscape 0.92 [26] was used for producing the map.
Fig 2.
Frequency distributions of individual q-values in pooled samples from contact zones between M. edulis and M. trossulus.
Numbers of individuals are plotted on the ordinates, with q-values at 10% intervals as abscises. Red and blue bars indicate T- and E-morphotypes, correspondingly.
Fig 3.
Variation of PT, P(T|tros), P(T|edu), P(tros|T), P(edu|E) as functions of Ptros in the White Sea (WS), brackish Barents Sea (BL) and saline Barents Sea (BH).
Points–empirical estimates, their size is proportional to sample size (see S1 Table). Lines–regression model predictions, grey filling– 95% confidence intervals of regressions. (A) Proportions of T-morphotypes (PT) (Model 1). (B). Proportions of T-morphotypes among M. trossulus (P(T|tros), filled points) and M. edulis (P(T|edu), empty points) (Model 2). (C) Frequencies of M. trossulus among T-morphotypes (P(tros|T), filled points) and of M. edulis among E-morphotypes (P(edu|E), empty points) (Model 4). Vertical lines on B and C connect subsamples of M. trossulus and M. edulis from the same samples.
Fig 4.
Predictive power of the morphotype test in different contact zones.
(A) Dependence of proportion of M. trossulus (Ptros) on proportion of T-morphotypes (PT). Dotted lines are empirical regressions (Model 4). Solid gray lines–predictions of “Ptros by PT calculator” (Eq 3). Solid black lines represent Y = X dependence. (B) Probability to find a mussel with a T-morphotype among M. edulis (P(T|edu)) (empty points), and M. trossulus (P(T|tros)) (filled points) as a function of Ptros. Lines are empirical regressions (Model 5). (C) Probability of correct species identification by the morphotype test: M. trossulus by T-morphotype, P(tros|T) (filled points) and M. edulis by E-morphotype, P(edu|E) (empty points) as a function of Ptros. Dotted lines are empirical regressions (Model 6). Sold lines–predictions of “genotype by morphotype calculator” for M. trossulus (Eq 1, red line) and M. edulis (Eq 2, blue line). On each graph, dots represent the observed proportions in samples, and shaded areas around regression lines– 95% CI of regressions.
Table 1.
Proportions of M. trossulus among T-morphotypes (P(tros|T)) and proportions of M. edulis among E-morphotypes (P(edu|E)) in pooled samples (direct count) and in equally mixed samples (predictions by the regression Model 6) in different sample sets.
Table 2.
Formulas used for taxonomic and individual assignment using morphotype tests in different sample sets accordingly to the regression model coefficients represented in S3 Table.
Fig 5.
Correspondence between “Ptros by PT calculator” (Eq 3, left graph) and “genotype by morphotype calculator” predictions (Eqs 1 and 2, right graph) and regression Model 6 and Model 4, respectively.
Each point corresponds to a unique pair combination of samples from WSBL set. OX axis reflects dissimilarity of genetic structure in each pair (Delta) (for pure conspecific samples Delta takes a value of zero, for equally mixed samples– 0.5, for two pure heterospecific samples– 1). OY: goodness of correspondence between assessment of predictive values by equations and regression models.