Abstract
Aim Plant distributions and vegetation dynamics underpin key global phenomena, including biogeochemical cycling, ecosystem productivity and terrestrial biodiversity patterns. Aggregated and remotely collected ‘big data’ are required to forecast the effects of global change on plant communities. We synthesize advances in developing and exploiting big data in global change plant ecology, and identify challenges to their effective use in global change studies. Location Global. Methods We explored databases, catalogues and registries with respect to their accessibility, geographical and taxonomic extent, sample bias and other types of uncertainty, from the perspective both users and contributors. We identified four kinds of big data needed to predict the impacts of global change on plant populations and communities using spatially explicit models: remotely sensed and other environmental maps, species occurrence records, community composition (plots) and species traits, especially demographics. Results Digital environmental maps, including remotely sensed data, are the most mature class of big data discussed herein whereby protocols for archiving, discovering and analysing them have developed over three decades. Species locality records are being aggregated into databases that are easy to search and access, and while methods for addressing uncertainties are a major research focus, better spatial representation is still needed. Plot data from inventories have tremendous potential for monitoring and modelling the impacts of global change on plant communities but tend to be restricted to forests or concentrated in certain geographical areas. Ongoing efforts to aggregate plot and trait data from multiple sources are challenged by their heterogeneous coverage, attributes and protocols and a lack of data standards. Main conclusions Future goals include developing systematic frameworks for selecting geospatial data, improving tools for assessing the quality of species occurrence records and increased aggregation and discoverability of plot and trait data. Aggregated data collected by scientists, not sensors, provide more meaningful insights when data collectors are involved in analysis.