Bird sampling
Bird species were sampled along the elevational gradient from 2,350 m and 4,950 m a.s.l. using the line transect method. Field sampling could not be done lower or higher because of the inaccessible topographies such as cliffs and glaciers. The entire 2,600 m elevational gradient was divided into 26 consecutive 100-m elevational bands. One transect (the length of each transect was 2.5 km) was placed in each elevational band (for a total of 26 transects, Figure 1). Bird censuses were carried out between 20 min after dawn and 10:00 hours and between 16:00 hours and 20 min before sunset (local time) by the same proficient observers. Birds within 50-m of the observers were recorded. The transects were sampled in a random order to reduce the temporal autocorrelation. Replicated bird censuses were made four times during the breeding season, from June to August 2018 and June to July 2019. For each transect, we made two replicated surveys in the morning and two at dusk, respectively. The taxonomic system of birds was followed Zheng (2011) (Zheng, 2011). We used breeding birds (including residents and summer migrants) for statistical analyses because seasonal and long-distance migratory birds could cause potential bias.
We interpolated the presence of species to 26 elevational bands according to their lowest and highest observed presences as the relatively small study area and strong movement ability of birds. This interpolation method is often used in elevational studies, and it can reduce the bias of underestimation of species richness due to insufficient sampling. We used species accumulation curves to evaluate whether species diversity was sampled sufficiently. If the species accumulation curve becomes an asymptote after a sharp rise and the rise is slow, the sampling is sufficient (Ugland, Gray, & Ellingsen, 2003). Estimator (Chao2) was used to compute estimated species richness. The Chao2 statistics of each 100-m elevational band were calculated in EstimateS 9.10 (http://purl.oclc.org/estimates) (R K Colwell, 2013). Pearson correlation analysis was used to assess the relationship between observed species richness and Chao2 to assess how well the species richness was sampled within each 100-m elevational band.