Knowledge of vertical structural complexity (VSC) is important, because the resulting spatial partitioning is closely linked to resource utilization and environmental adaptation. However, the spatial pattern of VSC on large scales and its underlying mechanisms are poorly understood. Here, we systematically investigated 2,013 plant communities through grid sampling on the Tibetan Plateau (TP). VSC was quantified as the maximum plant height within a plot (Height-max), coefficient of variation of plant height (Height-var), and Shannon evenness of plant height (Height-even). Precipitation dominated the spatial variation in VSC in forests and shrublands, supporting the classic physiological tolerance hypothesis (PTH). In contrast, for alpine meadows, steppes, and desert grasslands in extreme environments, non-resource limiting factors (e.g., wide diurnal temperature ranges and strong winds) dominate VSC variation. Generally, with the shifting of climate from favorable to extreme, the effect of resource availability gradually decreases, but the effect of non-resource limiting factors gradually increases, and that the PTH only applicable in “favorable conditions”. With the help of machine learning models, maps of VSC at 1-km resolution were produced for the TP for the first time. Our new findings and maps of VSC provide new insights into macroecological studies, especially for adaptation mechanisms and model optimization.

Field sampling was conducted during the high-growth period from mid-July to late August in 2018, 2019, and 2020. According to the latitude and longitude, we divided the entire Tibetan Plateau (TP) into 1,000 grids of equal area (0.5° × 0.5°). For alpine grasslands (i.e., alpine meadows, alpine steppes, and alpine desert grasslands), we selected the dominant plant communities via visual observation in each grid and randomly set up three 1 m × 1 m plots for the field investigation. There were 1,527 plots for grassland communities, and the height of the tallest plant in each plot was measured. For each species within the plots, we recorded their coverage and then randomly selected three plants of each species to measure their height (all measurements for less than three individuals); in total, 19,320 plants were measured.

For forests (456 plots) and shrublands (30 plots), three 20 m× 20 m plots were randomly established in each grid, for a total of 486 plots. For forests, we recorded all vascular plants in these plots, including trees, shrubs, and herbaceous species. Diameter at breast height (DBH) was measured for trees with a DBH of ˃3 cm. A telescopic stick was used to measure the tree height, and 17,443 trees were measured. Only trees with a DBH > 3 cm were measured, as this is a common standard for recording tree layer-related data in forest community surveys . For shrublands, we recorded their basal diameter and height, and 339 plants were measured. Across all vegetation types, a total of 2,013 field plots were investigated, and the height of 37,102 plants was used for the analysis.