Species distribution model and inspection

All SDMs use the correlation between the species sample data and environmental variables to estimate the ecological niche of the species according to a specific method and show the correlation to the studied area (Panda & Behera, 2019; Thorson, 2019; Wang et al., 2019; Yu et al., 2019). We predicted the distribution of birds by using the MaxEnt model, which only uses existing events to study the principle of balance and simulates the absent data from the background(Manish & Pandit, 2019). It defines the correlation between species and the environment variables, and predict the distribution according the present sample data (Dudik et al., 2007; Dudik et al., 2004; Phillips et al., 2006; Phillips & Dudik, 2008). The distribution which has the highest entropy is selected to be the optimal distribution from all the eligible distribution.
We used the default settings and divided the data into two parts, 75% for modeling and 25% for evaluating. The jackknife test was done in MaxEnt to recognize the importance of variables, and the benefit of the test is to give the approximate confidence intervals for many parameters (Jacome et al., 2019; Shcheglovitova & Anderson, 2013).
The Area Under Curve (AUC), the proportion under the Receiver Operating Characteristic Curve (ROC), was used to evaluate the performance of model (Phillips et al., 2006). The AUC value is positively correlated with the model, so we picked the model with AUC value over 0.8 seeing at Table S2 (Elith et al., 2011; Fourcade et al., 2014).

Index of habitat change

2.4.1 The habitat loss
The decrease of habitat area is the main threat to the decline in biodiversity which is caused by climate change (Taubert et al., 2018). There is a continuous distribution suitability map coming from the model. To calculate the habitat loss, the “maximum training sensitivity plus specificity logistic threshold(MaxSSS)” was used to classify the continuous distribution suitability into the binary distribution map (Saupe et al., 2019; Vale et al., 2014). We also calculated the mean suitability of the map. The calculating of habitat loss is based on the comparison between the suitable area in current and future scenarios. It was described as habitat loss when the suitable area turns to the unsuitable area and the opposite is habitat gain..
2.4.2 The population centroid
The population centroid is regarded as the represented indicator reflecting the population movement process (Collins et al., 2017; Liu et al., 2019).We calculated the population centroid of the habitat for migratory birds in longitude (X) and latitude (Y) to explain the bias in future. After compared the population centroid of the species , we take the average offset distance in two scenarios
X=\(\frac{\sum{X_{i}P_{i}}}{P}\)Y=\(\frac{\sum{Y_{i}P_{i}}}{P}\)
where\(X_{i}\)and\(Y_{i}\ \)is the longitude and latitude of the site , and\(P_{i}\)is the sustainability at site, and P is the total sustainability .
  1. Results

    Spatial changes of habitat and the shift of population centroid

Based on the current and the two future scenarios, the models simulated the distribution of 7 orders of migratory birds and the endangered birds well by checking the AUC value(Table S1). The current and future potential geographical distribution of birds in China is shown in Fig.S1.
Table 2. Result of models assessing how climate influence the habitat change.