The depth diversity gradient (DDG) patterns of macrophytes
The mean depth pattern of submerged macrophytes’ alpha and gamma richness is hump-shaped, showing a peak between −1 and −2 m respectively (Fig. 2a,c). The mean alpha richness at the hump’s peak is 4.5 species (sd = 2.2), whereas the mean gamma richness peak is 11.4 species (sd = 5.1). In contrast, Beta richness shows a decreasing curve with its highest richness being 7.0 species (sd = 4.0) between surface and −1 m depth (Fig. 2b). However, all three richness components show high standard deviations in the depth classes. They vary across depth classes between 1.9 and 2.3 species for alpha richness, between 3.7 and 4.1 for beta richness and from 5.1 till 5.6 for gamma richness (see individual DDG curves for all lakes in Supporting information).
Comparing the richness components across depths revealed only significant differences between mid and greater depths, but not for shallower depth (Supporting information).
Plotting the DDG measures for all three richness components (Fig. 2d-f) we find a hump-shaped pattern for alpha richness and a bi-modal pattern for beta and gamma richness.
Looking at the DDG patterns of single field campaigns, for alpha richness, hump-shaped curves with a peak between -2 and -4 m are most frequent (52%) (Fig. 2g-i). For beta richness the majority are decreasing curves (40%) while for gamma richness hump-shaped curves with a peak between -1 and -2 m were slightly prevailing (39%)., All depth pattern types are found for all three measures.
The three depth-dependent richness measures are significantly correlated with one another ( < 0.05). The strongest correlation shows beta and gamma richness (cor = 0.95), followed by alpha and gamma richness (cor = 0.85) and alpha and beta richness (cor = 0.64). TheDmax and Rmax do not correlate within the respective richness components ( < 0.05). However, the Dmaxvalues across the three richness components correlates with each other. Similarly, Rmax also correlates across the three richness components ( < 0.05). For correlation coefficients see Supporting information. However, a Chi-square test shows, that pattern types and richness components are statistically significantly associated (p  = 0.0005).