摘要：Traditionally, restoration ecologists and land managers have used the trial-and-error method to select candidate restoration species. This method, however, is time consuming (usually more than 3 years) and has a relatively low success rate. Recently, Wang et al. (2020) developed a trait-based species selection framework which can quickly (within 1 month) and successfully select many appropriate species for ecological restoration. They used 28 traits that are associated with tolerance to harsh environmental conditions to select candidate restoration species for a tropical coral island in Hainan Province, China. However, it is likely that some of the 28 traits used in this study may not be very important for species selection, providing the potential use of fewer overall traits. This is important since in many situations land managers will have limited data and resources on species traits. In this study, we used Wang et al. (2020)’s trait data to test which traits are necessary to achieve a similar success rate when screening species for restoration applications. We performed principal component analysis (PCA) to compute each trait's relative contribution. Then, we used the backward stepwise approach where the trait that had the least contribution among all remaining traits was removed one at a time, and the screening model was then run again using the smaller set of traits. Species which are proven very appropriate for ecological restoration in Wang et al. (2020) were the standard to quantify how many and which traits should be used to acquire similar screening results. We also classified all 28 traits into four types of functional traits to test if a small set of traits can mimic Wang et al. (2020)’s selection results. Our results indicate that it is hard to simultaneously reduce trait numbers and maintain the right screening results; especially for tree species. Likewise, vine species and herbaceous species still required most of the original traits used by Wang et al. (2020). Our results also indicate that multiple trait types are required, rather than one single group of functional traits. Our results reject the possibility of using fewer and more targeted traits for species screening. Although investigations in other ecosystems are needed to test the generality of the conclusion, our results suggest that multiple functional traits are still required to be measured to select appropriate plant species for the restoration of tropical coral islands.
