Statistical analyses
We performed all statistical tests in R version 3.6.1 (R Core Team 2019
p.). To fit distributions of response variables, we used the
fitdistplus-package
(Delignette-Muller
and Dutang 2015). The distribution for species richness was negative
binomial (on all levels). We tested if DLI, pH, total nitrogen, total
carbon, C:N-ratio, ammonium, nitrate, phosphorous, potassium, average
slope, aspect, and altitude affected species richness at the subplot
level using a generalized linear mixed model from the R-package lme4
(Bates
et al. 2015) with a negative binomial distribution from the R-package
MASS
(Venables
and Ripley 2002). Conditional R²-values were then calculated
(Nakagawa
and Schielzeth 2013). The same analyses were performed at the plot level
for those which were significant at the subplot level using generalized
linear models (GLM) where the response was negative binomial
distribution. Additionally, we determined if the structural complexity
measures DBHsd, TRI and SSCI affected any of the above-mentioned
environmental heterogeneity variables using GLMs, where light
heterogeneity was Gamma distributed while all other variables were
log-normally distributed. Next, we tested if species richness at an
aggregate level is affected by the environmental heterogeneity variables
(see above) using the GLM negative binomial distribution from the
R-package MASS (Venables and Ripley
2002).
For all heterogeneity variables we included the corresponding absolute
values to test whether possible effects are only induced by quantity.
Finally, we tested if the structural complexity variables had any effect
on species richness. This was tested using the GLM negative binomial
distribution from the R-package MASS (Venables and Ripley
2002).
To determine the model of best fit for each approach, model selection
was performed with R-package MuMIn (Bartoń
2019).
The package computes all possible models of the given variables. The
model of lowest Akaikes information criterion (AIC) is taken as valid.
We included variables which are ecologically important, and which do not
interfere with each other: forest community, DLI, soil-pH, C:N-ratio,
phosphorous, average slope, aspect and altitude.
Results