Abstract Climate change has driven extensive reorganisation of plant communities as species adjust their ranges in response to changing conditions. Existing studies focus on the role of temperature, documenting shifts in range margins, but neglect changes in abundance within the range. This results in unexplained patterns complicated by overlooked, interacting biotic and abiotic drivers, such as soil properties and herbivores, limiting our ability to predict forest response and vulnerability to climate change. Here, we develop the Abundance Trend Indicator (ATI), a machine learning approach that learns the conditions under which changes in species' abundance have occurred and, thus, the extent and location of the mismatch of a species' current range to its total realised habitat. The resulting abundance shift maps identify abundance shift directions, vulnerable species and their drivers. We validate the algorithm on New Zealand's forest inventory data (2821 sites, 77 woody species, collectively accounting for 75% of forest canopy nationally) and 37 predictor variables, which include climatic, topographic, edaphic and biological factors. ATI confirms globally observed trends of upward and poleward abundance shifts, but reveals drivers such as soil pH, grazing and climate stability for species shifting in opposite directions. Moreover, our findings suggest that vulnerability is primarily explained by a plant's ability to tolerate or avoid stress resulting from climatic fluctuations and limited migration capacity. Synthesis . Using existing forest inventory data, ATI models within‐range abundance shifts, rather than changes in range margins, resulting in earlier detection of environmental drivers and geographical directions of plant migration. This enables the early identification of vulnerable species.
Shabanov et al. (Fri,) studied this question.