Scale dependence of the diversity effect.
Previous studies have showed that the relationship between diversity and productivity (biomass) changed with plot size (Chisholm et al.2013; Thompson et al. 2018; Luo et al. 2019). We also found that the explanatory power of diversity increased from the 400 to 1200 m2 grain size for biomass and productivity (Fig. 1). It is generally hypothesized that the biodiversity effect will first increase with larger plot size, and then show a decrease. At smaller plot sizes, environmental heterogeneity increases with increased grain size and thus biodiversity also accumulates, which lead to higher biodiversity effect. However, when diversity gradually saturates with further increase in grain size, the effect of diversity on ecosystem functions will weaken (Chisholm et al . 2013, Thompson et al. 2018). As for the grain size with the strongest diversity effect on forest biomass and productivity, (Chisholm et al. 2013) found that the optimal grain size was 0.04 ha, and that biodiversity effect may become null or even negative for plot size > 0.1 ha. However, later studies suggest that the optimal grain size is around 0.1 ha (Thompson et al. 2018). Our results suggest that the optimal grain size may be larger than 1200 m2 for forest biomass and productivity. A recent study also found that the diversity effect is the strongest at a grain size of 0.25 ha in a temperate forest (Luo et al. 2019). These results are clearly different from previous idea that biodiversity effect should be the strongest at the smallest grain size, where community processes (e.g. complimentary and sampling effect) play a dominant role (Chisholm et al. 2013).
Compared with forest biomass and productivity per se , the change of diversity effect on their stability with grain size has much less been reported with field data. However, Wang et al. (2014) proposed a theoretical model and predicted that ecosystem stability itselfshould increase from local to regional scales. As for diversity-stability relationship, their model predicts distinctive pattern between the local and regional scales: 1) at the local community scale, the diversity-stability relationship do not change with grain size; 2) while at the regional (metacommunity) scale, ecosystem stability should be higher with both increasing diversity and grain size. Interesting, our results obtained at the community scale also showed that the effect of diversity on biomass and productivity stability did not showed clear trend across grain sizes (Fig. 1), and thus provide support to their first prediction. Meanwhile, their second prediction is also consistent with a large-scale study, which found that the effect of biodiversity on productivity stability was stronger at a larger grid size of ~ 55 km (0.5 °) than the ~ 5 km grids (Mazzochini et al. 2019). Thus, biodiversity is crucial for ecosystem stability from local community to large scales.
In previous studies that examined the scale dependence of diversity effect, results based on subplots that differed greatly in sample size were generally compared directly (e.g. Chisholm et al. 2013, Thompson et al. 2018, Luo et al. 2019). Our analysis seems to be the first one that test the potential influence of sample size on the scale-dependence of biodiversity effects. Our test showed that while the results based on different subplot numbers (Fig. S3) showed some similarity with that based on random sampling of subplots (Fig. 1), there were also notable differences. For instances, for biomass and productivity, Fig. S3 can not provide a clear picture of increasing diversity effect with larger plot size, as revealed by Fig. 1. However, this difference seems not to be mainly caused by difference in sample size. Instead, it seems to be caused by the fact that: when the diversity effect is weak, diversity indices have higher probability to be excluded from the models, which result in the occasionally high explanatory power of diversity in Fig. S3A and S3B (this is also evident when comparing the effect of stand factors on ecosystem stability between Fig. 1 and Fig. S3). Thus, we suggest the random sampling method may provide a more robust way to examine the relative effect of diversity vs. other factors, and their scale dependence. It remains unclear whether the different optimal grain size found by previous studies (e.g. Chisholm et al. 2013, Thompson et al. 2018) from ours results were caused by this statistic issue. It is a pity that the maximum plot size in our dataset is not large enough. We suggest authors with larger plots to test our method, for a better understanding of the scale dependence of biodiversity effect.