Data from: Improving spatial predictions of taxonomic, functional and phylogenetic diversity

1. In this study, we compare two community modelling approaches to determine their ability to predict the taxonomic, functional and phylogenetic properties of plant assemblages along a broad elevation gradient and at a fine resolution. The first method is the standard stacking individual species distribution modelling (SSDM) approach, which applies a simple environmental filter to predict species assemblages. The second method couples the SSDM and macroecological modelling (MEM - SSDM-MEM) approaches to impose a limit on the number of species co-occurring at each site. Because the detection of diversity patterns can be influenced by different levels of phylogenetic or functional trees, we also examine whether performing our analyses from broad to more exact structures in the trees influences the performance of the two modelling approaches when calculating diversity indices.
2. We found that coupling the SSDM with the MEM improves the predictions for the diversity facets compared with those of the SSDM alone. The accuracy of the SSDM predictions for the diversity indices varied greatly along the elevation gradient, and when considering broad to more exact structure in the functional and phylogenetic trees, the SSDM-MEM predictions were more stable.
3. SSDM-MEM moderately but significantly improved the prediction of taxonomic diversity, which was mainly driven by the corrected number of predicted species. The performance of both modelling frameworks increased when predicting the functional and phylogenetic diversity indices. In particular, fair predictions of the taxonomic composition by SSDM-MEM led to increasingly accurate predictions of the functional and phylogenetic indices, suggesting that the compositional errors were associated with species that were functionally or phylogenetically close to the correct ones; this did not always hold for the SSDM predictions.
4. Synthesis. In this study, we tested the use of a recently published approach that couples species distribution and macroecological models to provide the first predictions of the distribution of multiple facets of plant diversity: taxonomic, functional and phylogenetic. Moderate but significant improvements were obtained; thus, our results open promising avenues for improving the predictions of different facets of biodiversityacross broad environmental gradients when functional and phylogenetic information is available.

1. 本研究对比了两种群落建模方法，以评估其在大尺度海拔梯度下、以精细分辨率预测植物群落分类学、功能学与系统发育学特征的能力。第一种方法为标准堆叠单物种分布模型（standard stacking individual species distribution modelling, SSDM），该方法通过简单的环境过滤实现物种群落的预测。第二种方法则将SSDM与宏观生态模型（macroecological modelling, MEM）相结合，形成SSDM-MEM框架，用于限制每个样地内的共存物种数量。由于多样性格局的检测结果可能受系统发育树或功能树不同层级的影响，本研究还探讨了在计算多样性指数时，从宽泛到精细的树结构开展分析，是否会对两种建模方法的性能产生影响。
2. 研究结果表明，相较于单独使用SSDM，将其与MEM相结合可显著提升各多样性维度的预测效果。单独SSDM对多样性指数的预测精度随海拔梯度发生大幅波动；而在考虑功能树与系统发育树从宽泛到精细的结构时，SSDM-MEM的预测结果更为稳定。
3. SSDM-MEM可适度但显著提升分类多样性的预测性能，这一效果主要源于预测物种数量的校正。两种建模框架在预测功能多样性与系统发育多样性指数时，性能均有所提升。具体而言，SSDM-MEM对物种分类组成的合理预测可带来功能与系统发育指数愈发精准的预测，这表明其组成误差多与功能或系统发育上接近真实物种的类群相关；而这一规律在单独SSDM的预测结果中并不总是成立。
4. 综合分析：本研究验证了一种新近发表的耦合物种分布模型与宏观生态模型的方法的应用效果，首次实现了植物多样性多维度（分类、功能与系统发育维度）分布格局的预测。本研究获得了适度但显著的预测性能提升；因此，在可获取功能与系统发育信息的前提下，本研究结果为提升跨大尺度环境梯度下生物多样性多维度的预测精度提供了极具前景的研究方向。