Statistical analyses
The statistical analyses were performed in R version 4.2.2 (‘Innocent and Trusting’, R Core Team, 2022). The original data from 1996/98 (https://doi.org/10.1594/PANGAEA.864321) and 2012-14 (provided by I. Bartsch) were used for the time series analysis. Due to the logistical constraints of the intensive scuba diving campaign along the sublittoral gradient at Hansneset for all three time points of the time series, only a limited number of replicates was obtained per campaign (1996/1998: n = 2 (0m); n = 3 (5 – 15m); n = 4 (2.5m); 2012/2013: n = 3 (0 and 15m); n = 6 (2.5 – 10m); 2021: n = 3). The homogeneity of variances was tested using the Levene’s test before each ANOVA. As the absence of a species or group at a certain depth level results in zero values for that depth, only relevant depth levels were included in the statistical tests.
Time series analysis (1996/98-2012/13-2021): Separate two-factorial ANOVAs were performed to assess the effects of the fixed factors time, depth and their interaction for FW (log + 1 transformed data) and LAI of the biomass-dominant species or groups. When the ANOVA output revealed significant effects, a Tukey HSD post hoc test for uneven n was applied. In case homogeneity of variances could not be achieved by transformation but the results of the two-factorial ANOVA were highly significant, a non-parametric Kruskal – Wallis test followed by a pairwise Wilcoxon rank sum test was performed to test for differences between depth levels. This study focused on investigating changes over time and therefore the differences in species and group FW and LAI across depth levels in 2021 alone were not examined statistically.
Age and density comparison 2013 / 2021: To investigate differences between timepoints and kelp species for the mean age and density per m at 2.5m and 5m (2013 and 2021: n = 3), separate two-factorial ANOVAs were performed and significant effects were further investigated using a Tukey HSD post hoc test. For the analysis of mean density per m², data was log + 1 transformed to achieve homogeneity of variances. Juvenile specimens <1 year were excluded from the statistical analysis.
Adult kelp dry weight and biochemical investigations in 2021:Individual holdfast, blade and stipe DW, blade:stipe DW ratio as well as blade %carbon, %nitrogen and C:N ratio was compared between kelp species and across the relevant depth levels (2.5 – 5m) in 2021. Homogeneity of variances (Levene’s Test) and normal distribution (Shapiro-Wilk Test) could not be achieved through data transformation. The effect of the factors depth and species on the individual parameters were tested in separate non-parametric Kruskal-Wallis tests and by pairwise Wilcoxon rank sum tests with Bonferroni correction to reveal differences between species.
Carbon and nitrogen 2021: Mean extrapolated carbon and nitrogen stored in kelp blade DW per m² in summer 2021 was compared between depth levels and kelp species (2.5 and 5m; both n = 3) using separate two-factorial ANOVAs.
PAR and Turbidity time series: PAR and turbidity data were quality controlled according to an adapted ARGO standard for stationary sensors (Fischer et al., 2021; Waldmann et al., 2022). To analyze long-term changes and trends over the sampling period, the residuals of the observed weekly PAR and turbidity values to the expected PAR and turbidity values were calculated, using the means of the observed weeks across all years as expected values. The trend-analysis of PAR and turbidity over the sampling period was done by simple linear regression over time using the R base functions “lm” (R Core Team, 2019).