Drivers of the depth diversity gradients
The DDG measures correlate with some of the abiotic variables
(Supporting information). R(α,max) correlates
highly significantly (p < 0.01) with area (cor =
0.53, p < 0.01), water level fluctuations (cor =
0.54, p < 0.01), conductivity (cor = 0.53,p < 0.01), NH4N (cor = -0.5,p < 0.01), SiO2 (cor = 0.62,p < 0.01) and spectral absorption coefficient(cor = 0.6, p < 0.01). R(β,max) correlates highly significantly (p < 0.01) witharea (cor = 0.55, p < 0.01).R(γ,max) correlates highly significantly
(p < 0.01) with area (cor = 0.57,p < 0.01) and water level fluctuations (cor =
0.56, p < 0.01). D(α,max) correlates highly significantly (p < 0.01) withO2 dissolved (cor = -0.54, p <
0.01), total phosporus content (cor = 0.6, p <
0.001), transparency (cor = -0.67, p < 0.001)
and tempsd (cor = 0.59, p < 0.01).D(β,max) and D(γ,max) do not
correlate highly significantly (p <0.01) with any of
the abiotic variables.
Abiotic and biotic variables are correlated with one another in a
complex fashion (Supporting information). Strongest positive
correlations (cor > 0.7 or < -0.7) within abiotic
data were found between Ntot andNO3N (cor = 0.92, p < 0.01),conductivity and SiO2 (cor = 0.83,p < 0.01), chloride and conductivity(cor = 0.73, p < 0.01). Strongest negative correlations
showed transparency and total phosphorus content(cor = -0.72, p < 0.01) and transparency andconductivity (cor = -0.71, p < 0.01).
Due to the high correlation coefficients between abiotic factors, we
performed a PCA (Supporting information). We use the first four axes
(81% of total variation – Fig. 3f-i) to address the DDG drivers. The
first axis, PC1, can be characterized as the“SiO2 & Conductivity axis” (both positive with
the axis), explaining 30.1% of the variance. The PC2, the second
axis, can be described as the “Temperature &
Ptot axis ” as both abiotic variables have the highest
(negative) impact (26.1% of variance). The third axis, PC3, can be
named the “Temperature sd – Chloride axis ” (13.3% of
variance) as it ranges from most negative variable temp_sd to
most positive variable chloride while the fourth axis, PC4 shows
the “O2diss – SAC axis ” (10.5% of variance)
spanned between O2diss (most negative) andSAC (most positive).
The GAMM showed that D(α, max) (R²=0.73)
significantly varies with all four PCA axes (Fig. 3a-d). TheD(α, max) decreases with PC2 (Temperature
& Ptot axis ) and PC3 (Tempsd – Chloride axis )
axes , slightly increases with PC4 axis (O2diss –
SAC axis ) and increases only for extreme positive values of PC1 axis
(SiO2 & Conductivity axis ). TheR(α,max) (R²=0.44) is only influenced by the PC1
axis (SiO2 & Conductivity axis ) with a positive
linear relationship (Fig. 3e). The GAMM analysis forD(β,max) ,R(β,max) ,D(γ,max) and R(γ,max) had
all R²<2.1% (see results in Supporting information).