Planning priority conservation areas for biodiversity under climate change in topographically complex areas: A case study in Sichuan province, China

species information was collected from The Nature Conservancy and Institute of forestry research in Sichuan province (S1 Appendix). 2.3.2. Climate change data. Coupled with the Regional Climate Model and Beijing Cli- mate Center’s Climate System Model was employed to project monthly temperature and pre- cipitation for the period of 1981–2060 from the Fifth Report of the International Panel for Climate Change. Then, the model output was calibrated by using data from 204 meteorological stations in and around Sichuan Province (Fig 1). Subsequently, the Anusplin software (version 4.3.2) was used to the downscale temperature and precipitation data at a 1 km x 1 km spatial resolution. 2.3.3. Others. Identifying priority conservation areas involves datasets about species occurrence points, climate change data, digital elevation model (DEM), land use, and soil. Besides species information, and climate data, the land use, road, and settlement data were obtained from Landsat 7. The DEM of 90-m resolution was obtained from the United States Geological Survey (https://earthexplorer.usgs.gov/), which was used to generate the slope. Soil dataset was obtained from the Institute of Soil Science, Chinese Academy of Sciences. All spa- tial data were converted at a 90 m scale to plan priority conservation areas. 2.4. Models 2.4.1. Predicting species distributions. Maxent model has been widely used to analyze changes of species distributions under climate change, while it has been suggested to perform

best with few species records [25–28]. To characterize the species distributions of the study area, 21 environmental predictors were selected to assess species distributions, including 19 bioclimatic variables [29], the slope, and the soil types. The definitions of bioclimatic variables and calculation models were available online (http://www.worldclim.org/). To avoid over-fit- ting in the Maxent model, Spearman’s rank correlation coefficient was selected to test the cor- relation value between variables. When the threshold value of Spearman’s rank correlations were more than 0.75 according to other related studies [9], all highly correlated environmental predictors were removed. Additionally, we selected 80% data for model training and 20% data for model testing, keeping other values as default. To determine model variables, jackknife analyses were used to reduce the model reliability when omitted. The maxent model ran 1, 000 iterations of the pro- cesses or set the value of threshold to 0.00001. Meanwhile, the Receiver Operating Characteris- tic plot, the Area Under Curve (AUC) values, and the True Skill Statistics (TSS) were used to measure the accuracy of the Maxent model [30, 31]. The greater value of AUC represented the better performer [32]. Maxent was also used to project the distributions of biodiversity for the period of 2011–2040 and 2031–2060 based on future climate data. According to the previous studies [33, 34], there is a range of the value of the potential species distribution, which was

divided into 2 groups: unsuitable habitats (0–0.7); suitable habitats (0.7–1.0) for planning con- servation priority areas. 2.4.2. Evaluating the effect of topographic complexity. The species shift to similar cli- matic conditions for adaptation to climate change, and thus our study rests on two premises to evaluate refugia. First, this study assumed that temperature predominated habitat suitability under climate change. If species could disperse areas with similar temperature conditions, it means that biodiversity could find refugia to adapt. Second, the more topographic complex represents the higher spatial heterogeneity and more types of habitats available, leading to more opportunities for adaptation to climate change in a small area [35]. As shown in Fig 3a and 3b, the species could be greater opportunities to migrate refugia with similar climatic con- ditions in a small area of topographic complexity than those in a relatively flat area. PLOS ONE Planning priority conservation areas for biodiversity under climate change in topographically complex areas PLOS ONE | https://doi.org/10.1371/journal.pone.0243425 December 23, 2020 5 / 15