Fig. 1. The locations of (a) 7orders of migratory birds and (b) 23 different IUCN categories species.

Climate data and the scenarios

We only used static environmental variables to simulate the potential distribution of migratory birds, and focused on and analyzed the factors the static environmental factors that are known to affect birds’ movement and habitat changes (Arribas et al., 2019; Beumer et al., 2019). The data of bioclimate, temperature and precipitation were imported to the model from Worldclim 1.4 dataset(http://www.worldclim.org/ ) with 30 arc-seconds(about 1km) resolution ratio. Current climate scenario(average 1970-2000) is used to comprehend the latent distribution while two future scenarios(RCP2.6 and RCP8.5 during average 2040-2060) are used to estimate the future distribution shift. The RCP2.6 is an optimistic scenario and the RCP8.5 is an pessimistic scenario (Boisvert-Marsh et al., 2019; Rosen & Guenther, 2016). The land cover dataset and the NDVI dataset were downloaded from the resource and environment data cloud platform (http://www.resdc.cn/) with a resolution of 1 km. The reserves data were from the protected planet(https://www.protectedplanet.net/). The distance of each grid to protected area was calculated by Euclidian distance in ArcMap. Altitude and slope were compiled by using the Digital Elevation Model(DEM) from the Geospatial Data Cloud (http://www.gscloud.cn).
To make the result reasonable, the Pearson Correlation Coefficient (PCCs) was used to remove high correlated variables(Table S3) Moreover, 13 variables were retained in the model(Table1) after removing high correlated bioclimate variables(|r|>0.8).
Table1 List of variables used in MaxEnt modelling.