Local and landscape-level diversity effects on forest functioning

Abstract

Research of the past decades has shown that biodiversity is a fundamental driver of ecosystem functioning. However, most of this biodiversity-ecosystem functioning (BEF) research focused on experimental communities on small areas where environmental context was held constant. Whether the established BEF relationships also apply to natural or managed ecosystems that are embedded in variable landscape contexts remains unclear. In this study, we therefore investigated biodiversity effects on ecosystem functions in 36 forest stands that were located across a vast range of environmental conditions in managed landscapes of Central Europe (Switzerland). Specifically, we approximated forest productivity by leaf area index and forest phenology by growing-season length and tested effects of tree species richness and land-cover richness on these variables. We then examined the correlation and the confounding of these local and landscape-level diversity effects with environmental context variables related to forest stand structure (number of trees), landscape structure (land-cover edge density), climate (annual precipitation) and topography (mean altitude). We found that of all tested variables tree species richness was among the most important determinants of forest leaf area index and growing-season length. The positive effects of tree species richness on these two ecosystem variables were remarkably consistent across the different environmental conditions we investigated and we found little evidence of a context-dependent change in these biodiversity effects. Land-cover richness was not directly related to local forest functions but could nevertheless play a role via a positive effect on tree species richness.

Fig 1. Study design.

From a total of 1416 sites that are part of the Swiss biodiversity monitoring program (BDM Z9 plot network; biodiversitymonitoring.ch) and regularly spread across Switzerland (small black dots) we selected 36 sites that we classified as “coniferous”, “mixed” or “broadleaf” forests based on the Jaccard similarity of BDM-monitored plant communities (S1 Table; [34]). We selected two replicates of each forest type (FT) in each of the six biogeographic regions (BGR; [30]). We re-classified six forest types after executing the tree inventories, so that the final distribution of forest types is not fully orthogonal with biogeographic region (triangles = coniferous forests, n = 13; circles = mixed forests, n = 10; diamonds = broadleaf forests, n = 13).

Fig 2. Correlations among local and landscape-level predictors and local ecosystem functions.

Significance (P<0.05) is indicated in black, sign of correlation is indicated with red to blue coloring (blue = positive correlation, red = negative correlation). Bold black lines delimit groups of variables characterizing local and landscape-level diversity, landscape structure, climate and topography, forest stand structure and ecosystem functioning (local EF). log(SR): log-transformed tree species richness; LR: land-cover richness; ED: edge density; PD: patch density; F_frac: fraction of forested area; F_conn: connectivity of forest patches; alt: altitude; N-aspect: northerly aspect; temp: temperature; precip: precipitation; n_trees: number of trees, DBH_max: average DBH of three largest trees; BA: cumulative stem basal area in the forest stand. Number of study units n = 36 for leaf area index (LAI) and n = 22 for growing-season length (GSL).

Fig 3. Variance in local forest functions explained by diversity, environmental context and study design variables.

The coefficients of determination (R²) for models that include all combinations of 1–3 of all 8 predictors (listed on y-axis) are on average 16.2% for leaf area index (LAI) and 35.1% for growing-season length (GSL), respectively. The R² contributions of every predictor when averaged over all possible model-orderings are indicated for each model with light-grey dots. Results that were averaged across all models with 1–3 predictors are shown in black. Black dots represent average R² contributions over all possible model-orderings and black lines indicate the range of R² contributions when included first or last in the model, respectively. log(SR): log-transformed tree species richness; LR: land-cover richness; ED: edge density; alt: altitude; precip: precipitation; n_trees: number of trees, BGR: biogeographic region; FT: forest type. Number of study units n = 36 for LAI and n = 22 for GSL.