Seasonal variation in the dependence of avian order composition on vegetation characteristics: A case study from Heilongjiang Province, northernmost China

Based on the typical vegetation phenology in Heilongjiang Province, this study defined June to September as the growing season and January to March and December as the dormant season. These two phases represent markedly different vegetation conditions, allowing for an investigation of bird responses to seasonal vegetation dynamics. Bird observation data were obtained from the China Bird Report (CBR). To ensure data continuity and accessibility, observations from 2015 to 2023 were used. Each record includes geographic coordinates and bird species information. After excluding records with fewer than three species, a total of 746 bird observation records were retained for analysis.

To quantify vegetation characteristics surrounding bird observation points, we discretized the study area into hexagonal grids using the H3 index at resolution 7. To quantify vegetation vertical structure and land–water attributes, we used annual land-use maps for 2015–2023, which classify land cover into nine categories: cropland, forest, shrub, grassland, water, snow/ice, barren land, impervious surface, and wetland. For each hexagon, we calculated the proportional cover of “forest” as the Canopy Vegetation variable, the combined cover of “shrub” and “grassland” as the Understory Vegetation variable, and the combined cover of “wetland” and “water” as the Wetland Vegetation variable. These three variables therefore describe the spatial distribution of different vegetation layers and land–water settings, but they do not directly represent vegetation greenness or growth status.

To assess vegetation greenness, we used monthly NDVI data sourced from NASA’s MOD13A3 product. Consistent with the seasonal definition used for the bird data, for each year we averaged NDVI values from June to September to derive mean NDVI for the growing season, and averaged NDVI for January, February, March and December to derive mean NDVI for the dormant season in each hexagon unit. To further characterize vegetation growth status, we used the combined Leaf Area Index (LAI) product derived from the Moderate Resolution Imaging Spectroradiometer (MODIS). Monthly LAI composites were generated on the Google Earth Engine (GEE) platform and then averaged, in the same way as NDVI, to obtain growing-season and dormant-season mean LAI values for each hexagon unit. All vegetation variables were derived in ArcGIS v10.8.