Understanding patterns and potential drivers of forest diversity in northeastern China using machine‐learning algorithms

Abstract

Question: Forest ecosystems are the most important global repositories of terrestrial biodiversity. The mixed temperate forests in northeastern China constitute one of the most biodiverse temperate regions globally and provide nearly one-third of China's wood supply. We ask what are the spatial patterns and potential drivers of tree species diversity in mixed temperate forests.Location: Temperate, mixed forests of northeastern China.Methods: Using a large set of ground-source forest inventory data (FIN) and geospatial covariates derived from published raster layers, we compared different machinelearning and statistical models to study spatial patterns of tree species diversity and their underpinning drivers. Results:The spatial distribution of tree species diversity (species richness and evenness) varied greatly across northeastern China. Tree species diversity varied most with climatic (annual precipitation and annual mean temperature), topographic (elevation and slope), and anthropogenic factors. Anthropogenic factors affected tree species evenness (importance value = 13%) more than tree species richness (importance value = 9%). Based on these relationships, we mapped spatial patterns of tree diversity throughout the region at a 1 km × 1 km resolution. Conclusions:Our findings shed light on the processes behind community assembly and biodiversity patterns in mixed temperate forests in northeastern China, and provide a benchmark for future assessment of biodiversity. Our high-resolution tree species diversity maps can be useful to landowners and land management agencies in their decision-making processes about sustainable forest management, biodiversity conservation, and forest restoration -a priority task outlined by the recently implemented 2050 China National Forest Management Plan.

Environmental filtering drives biodiversity–spatial stability relationships in a large temperate forest region

“…The elevational range of the study region extends from 79 to 1255 m. The forests are fragmented and subject to chronic anthropogenic disturbance (Luo et al, 2021; Venter et al, 2016). The study region is characterized by a temperate continental climate, with distinct temperature and precipitation seasonality and drought stress during the driest quarter (Luo et al, 2021; Zhang et al, 2020). The regular changes in bioclimatic seasonality, drought stress during the driest quarter and chronic human disturbance have adverse effects on the ecosystems in the study region.…”

“…In 2017, a total of 456 circular sample plots, with a radius of 17.85 m (0.1 ha), were evenly distributed across the study area in a systematic sampling scheme (Figure 1). The elevational range of the study region extends from 79 to 1255 m. The forests are fragmented and subject to chronic anthropogenic disturbance (Luo et al, 2021; Venter et al, 2016). The study region is characterized by a temperate continental climate, with distinct temperature and precipitation seasonality and drought stress during the driest quarter (Luo et al, 2021; Zhang et al, 2020).…”

A comparative mapping of plant species diversity using ensemble learning algorithms combined with high accuracy surface modeling

“…Although various linear regression methods had been identified to be effective, it is still a challenge for the complicated non-linear relationship due to the spatial and temporal heterogeneity. XGBoost was an improved version of gradient boosting algorithm and has produced state-ofthe-art results in ecological applications (Li et al 2021;Luo et al 2021). Considering random subsets of features and sample data, RF showed a better performance than the other bagging methods in the generalization error.…”