Figure 1.
Schematics of the analysis carried out.
1- Dynamical classification of morphotypes (Mt) and quantification of the overall degree of synchronization: for each pair of morphotypes we calculated the average of Pearson correlations per period, the ensemble average , which resulted in 6 dynamics distance matrices, one for each period; 2a- Functional classification of the 36 morphotypes based on 4 functional traits (cell volume, longest linear dimension, silica use and motility) (Weithoff 2003), 2b- Distance matrices based on the traits: 500,000 combinations of weights summing to 1 were attributed to the functional traits. Gower's similarity index was used to calculate the functional distance between the pairs of morphotypes for each set of traits, which resulted in 500,000 functional distance matrices; 3- Comparison of the functional trait similarity and temporal dynamics distance matrices: Mantel test was used to compared the distances matrices culminating in a frequency distribution of the 500,000 different Mantel r's for each period.
Figure 2.
Frequency distribution of 500 000 Mantel r per period, based on 500 000 different distance matrices based on functional traits varying in their relative importance.
The dashed line indicates the median value of the frequency distribution and the corresponding values are shown in the box within the plot area. ** means p<0.01 and *** means p<0.001. Periods are (a) winter, (b) early spring, (c) late spring, (d) clear water phase, (e) summer and (f) autumn.
Figure 3.
Plots of functional trait distance vs. temporal dynamics distance at different periods of the year.
Each dot corresponds to a period ensemble average , per pair of morphotypes. Functional traits distances are measured for pairs of morphotypes using the Gower distance and temporal dynamic distance is based on a period correlation for (a) winter, (b) early spring, (c) late spring, (d) clear water phase, (e) summer and (f) autumn. The dashed horizontal line indicates the mean of ensemble averages
for each period and the solid thick grey trend line represents the running mean.
Figure 4.
Box plot of the standard deviations of pairwise Pearson correlation coefficients (“Morphotype pairs”) and of the null distribution generated by determining standard deviations from random species pairs and periods (“Random”).
Table 1.
Mean period ensemble average (F = 12.26, p<0.0001).
Table 2.
Mean and standard deviation of weights leading to the most positive Mantel r per period.