ARTICLES https://doi.org/10.1038/s41558-020-0873-2 1 Department of Biology, University of Miami, Coral Gables, FL, USA. 2 Fairchild Tropical Botanic Garden, Coral Gables, FL, USA. 3 Universidad Nacional de Colombia Sede Medellín, Medellín, Colombia. 4 Forest Global Earth Observatory, Smithsonian Tropical Research Institute, Washington DC, DC, USA. ✉ e-mail: kjfeeley@gmail.com C limate change is altering the abundances and distributions of plant species worldwide 1–3 . Specifically, rising tempera- tures can lead to reduced fitness, and hence abundances, of species in the hotter portions of their ranges and increased fitness/ abundance of species in the colder portions of their ranges. In some cases, these differential changes in abundance may lead to range shifts, contractions or expansions as some species die back from the areas that become ‘too hot’ and/or invade areas that were previously ‘too cold’ 4–15 . Changes in other climate variables, such as precipita- tion and water availability, may likewise cause changes in species’ range limits as well as abundance distributions within the species’ ranges 14,16–18 . Collectively, these shifts in species distributions should cause widespread and directional changes in the composition of plant communities over time. In accordance with predictions, sev- eral studies have documented directional changes in local plant communities such that most have increasing relative abundances of heat-tolerant or heat-loving (thermophilic) species concurrent with rising temperatures (a process referred to as thermophiliza- tion) 19–25 . Other studies have documented shifts in composition due to changes in precipitation. For example, some lowland Amazonian forests are shifting in composition towards increased relative abun- dances of dry-affiliated species concurrent with an increasing fre- quency of droughts and decreasing water availability 26 . However, it remains unknown how these changes in local community compo- sition manifest at larger scales. In other words, is climate change causing broad-scale changes in the composition of regional plant communities? If so, what are the most important factors in deter- mining the direction and pace of change? To address these questions, we analysed changes in the floris- tic compositions of nearly 200 New World (North, Central and South America) ecoregions over the past 40 yr (1970–2011) using an expansive online database of plant collections and observation records (>20 million usable records after filtering for data quality and sample size criteria; see Methods and Supplementary Fig. 1). On the basis of collection locations, we estimated the optimal mean annual temperature (MAT opt ) and the optimal total annual precipitation (TAP opt ) of >17,000 ‘well-collected’ New World plant species 23 . We next characterized the plant community composition of each ecoregion in each year by calculating the community tem- perature index (CTI) and community precipitation index (CPI) 23 which reflect the relative abundances of thermophilic and meso- philic species (after correcting for collection biases; see Methods), respectively. CTI was strongly correlated with the ecoregions’ cur- rent mean annual temperature (MAT; Pearson’s R = 0.98; 95% con- fidence interval (CI) = 0.98–0.99; t = 72.14; d.f. = 189; P < 0.0001; Fig. 1a), CPI was strongly correlated with the ecoregions’ total annual precipitation (TAP; Pearson’s R = 0.87; 95% CI = 0.83–0.90; t = 24.59; d.f. = 189; P < 0.0001; Fig. 1b) and CTI and CPI were strongly positively correlated with each other (Pearson’s R = 0.72; 95% CI = 0.64–0.78 t = 14.04; d.f. = 189; P < 0.0001; Extended Data Fig. 1). These results reflect the availability of different climates (correlation between ecoregion MAT and TAP: Pearson’s R = 0.58; 95% CI = 0.48–0.67; t = 9.85; d.f. = 189; P < 0.0001) and show that both CTI and CPI can accurately characterize the climatic com- positions of ecoregion plant communities (Fig. 1 and Extended Data Fig. 1). Once we had estimated CTI and CPI for each ecoregion in each year, we then tested for changes in these metrics over time as indica- tors of shifts in plant species composition. For each of the ecoregions that had >10 yr of CTI and CPI estimates, we calculated the thermo- philization rate (TR) and the mesophilization rate (MR). A positive TR indicates increasing CTI due to increasing relative abundances of species with hotter MAT opt (that is, more-thermophilic species) and positive MR indicates increasing CPI due to increasing relative abundances of species with wetter TAP opt (that is, more-mesophilic species). As a separate measure of compositional changes over time, we compared each ecoregion’s initial versus final CTI/CPI as calcu- lated using the combined collections from the first and last 15 yr, respectively. We also identified the species that had been lost from the collection record of each ecoregion between these initial and Climate-driven changes in the composition of New World plant communities K. J. Feeley 1,2 ✉ , C. Bravo-Avila 1,2 , B. Fadrique 1 , T. M. Perez 1,2 and D. Zuleta 3,4 Climate change is altering the distributions of species, which in turn causes shifts in the composition of plant communities. Specifically, rising temperatures should cause increasing relative abundances of heat-loving or heat-tolerant species (that is, ‘thermophilization’) and changes in precipitation should cause altered abundances of water-demanding species. We analysed millions of records of thousands of species and found that the plant communities in most ecoregions in North, Central and South America have experienced thermophilization over the past four decades (1970–2011). Thermophilization was fastest in ecoregions with intermediate temperatures and was positively correlated with warming rates within many biomes. Changes in the relative abundances of water-demanding species were less consistent and were not correlated with changes in precipita- tion, meaning that the drought sensitivity of some ecoregions may be increasing despite decreasing rainfall and increasing probabilities of drought. Climate-driven changes in plant community composition will affect the function and stability of New World ecoregions. NATURE CLIMATE CHANGE | www.nature.com/natureclimatechange