Influence And Evaluation Of Environmental Variables On Species Distribution Model

Cornus officinalis Sieb. et Zucc. is a perennial deciduous tree of cornaceae plants. Its air-dried sarcocarp is one of the common Chinese medicines that possesses effects of nourishing liver and kidney, astringing essence and solidifying collapse. With the increasing demands of C. officinalis, it has been widely cultivated. Consequently, it is necessary to research its suitable distribution and the response of environment change through scientific approach. It will provide scientific and rational basises and advices for cultivation and resource protection of C. officinalis.As the developing of species distribution models (SDMs), it usually used to simulate interactions between species and environments, which have been widely used in ecological fields. However, the "best" model doesn’t exist, we do not have a model that always can fit species distribution perfectly. To achieve better simulation effect, many model algorithms and improvements had emerged. Ensemble model is one of them that appeared recently. Meanwhile, to figure out how the different environment variables effects SDMs, they had been separated into three categories. Environment variables Ⅰ contains climate variables only, environment variables Ⅱ contains climatic, edaphic and topographical variables, and environment variables Ⅲ contains climatic, edaphic, topographical and remote sensing variables. Then combined with BIOMOD 2 to build models respectively. We selected SRE, CTA, RF, GLM, GAM, GBM, ANN, FDA and MARS to predict potential geographical distribution of C. officinalis, then built ensemble model with better SDMs based on comparison of model evaluation about 9 model we mentioned before. Analyzed of 10 model evaluations and potential geographical distribution predictions of C. officinalis, plotted response curve maps of dominate factors under 3 environment variables, clarify the scope of C. officinalis’ distribution. Besides, in order to define the influence of climate change on potential geographical distribution of C. officinalis, the climatic conditions under two different future era and the three different RCPs were used as environment variables, to build potential geographical distribution prediction model. To provide scientific basis for planted and protected of C. officinalis, we analyzed its variation range and tendency of distribution under future climate.Results show that RF and MARS have highly model evaluation of single SDMs with current environment variables, on the other hand, model evaluation of SRE is lower than most SDMs. All 3 models have good model predictions, and the more environment variables were used, the less predicted distribution was. Meanwhile, ensemble model has better model prediction ability than single SDMs. The best environment variables group among 3 categories was environment variables Ⅱ which contains climatic, edaphic and topographical variables. It also indicated that more variables types can improve model accuracy. The potential geographical distribution prediction of C. officinalis under different environment variables showed that its highly suitable district concentrate upon Gansu southeastern, Shaanxi southern, Henan western, the border of Shanxi southern and Henan, Hubei northwestern, Anhui southwestern and the border of Zhejiang northwestern and Anhui. Further counted and analyzed potential geographical distribution areas of C. officinalis under different environment variables. The results showed that suitable district areas of C. officinalis’ potential distribution under climatic variables are the largest. The other two model prediction areas had few difference. Hence, more environment variables should been used to describe different aspects of target species, it can increase model veracity of species distribution. Although there are three different environment variables, the dominate factors are the same. Bio6, Bio4 and Bio 17 were the most important variables, they all belonged to the climatic variables, and they had correlation with temperature, it meant that climatic variables had more effects on the potential geographical distribution of C. officinalis, and compared to precipitation, temperature had larger influence.With future climate conditions, model evaluation of RF was the best, followed by GBM. On the contrary, SRE and CTA had the bad results. In the 2050s, the highly suitable district of C. officinalis’potential geographical distribution will shift northward, and suitable district areas will be decreased. It showed obvious decreasing tendency of highly suitable district in China. Meanwhile, the northward tendency of highly suitable district became more clearly in Korean peninsula and Japan. In the 2070s, the northward trend of C. officinalis’ potential geographical distribution will decay, the highly suitable district further reduce, and suitable district have the same change. In the RCP2.6 scenario, the distribution region of C. officinalis begin to move southward. Contrarily, its tendency of shift northward become more obvious in the RCP8.5 scenario.