Statistical tests of different scales in the Nanling Mountains
In addition to the phylogeography of the 5 target species, we examined the occurrence records of all other bird species in the Nanling region to determine the long-term effects of the mountains on their distribution patterns. Bird occurrence in the Nanling region was determined from records downloaded from the CBRC and the eBird database. The data included species’ name, latitude and longitude and number of observations (Supporting information). Duration and distance of the observation were omitted in order to reconcile the formats of 2 datasets. Comparisons focused on the differences among three geographic subunits: north of the Nanling Mountains, south of the mountains, and the mountains themselves (Figure 1, Supporting information). The Nanling region as a whole, and consequently its birds, was delimited manually by including administrative counties that overlapped with the mountains (Figure 1), while the north-end was decides so as to form comparable size of area (North Nanling 177534 km2, Nanling itself 162695 km2 and south Nanling 175820 km2). The Nanling in the middle is rugged with low mountains, while in the north, there are plains with surrounding mountains. The south Nanling region comprises mountains, waters and coastlines. Before testing distribution patterns, we pruned wetland- and coastal birds based on the habitat data from AVONET (Tobias et al. 2022). The original datasets were based on observation events, which would lead to bias introduced by observers (Strimas-Mackey et al. 2020). We transformed the dataset by thinning the records omitting the time variance. For the 2 datasets from different databases, species were reduced to one record per 15 km radius. In the end, 27402 occurrence records for 446 species were used (Supporting information).
Two statistical tests from different taxonomic level were applied to examine differences in birds among the three geographic subunits. We first performed chi-square tests on the number of species in each family between each region pairs to locate significant differences in family composition. For this we used chisq.test in R core functions. Second, we performed one-tailed Wilcoxon’s rank tests in both directions number of each bird species records for each family to determine which family distributes differently among three geographic subunits. All tests were done using the wilcox.test in R core functions. To avoid abnormal estimates from very rare occurrence records, we omitted families with less than 10 occurrence records from the results.