Aim: To map and interpret floristic and geoecological patterns across the Amazon basin by combining extensive field data with basin-wide Landsat imagery and climatic data Location: Amazonia Taxon: Ground truth data on ferns and lycophytes; remote sensing results reflect forest canopy properties Methods: We used field plot data to assess main ecological gradients across Amazonia and to relate floristic ordination axes to soil base cation concentration, CHELSA climatic variables and reflectance values from a basin-wide Landsat image composite with generalized linear models (GLM). Ordination axes were then predicted across all Amazonia using Landsat and CHELSA, and a regional subdivision was obtained using k-medoid classification. Results: The primary floristic gradient was strongly related to base cation concentration in the soil, and the secondary gradient to climatic variables. The Landsat image composite revealed a tapestry of broad-scale variation in canopy reflectance characteristics across Amazonia. Ordination axis scores predicted using Landsat and CHELSA variables produced spatial patterns consistent with existing knowledge on soils, geology and vegetation, but also suggested new floristic patterns. The clearest dichotomy was between central Amazonia and the peripheral areas, and the available data supported a classification into at least eight subregions. Main conclusions Landsat data are capable of predicting soil-related species compositional patterns of understory ferns and lycophytes across the Amazon basin with surprisingly high accuracy. Although the exact floristic relationships may differ among plant groups, the observed ecological gradients must be relevant for other plants as well, since surface reflectance recorded by satellites is mostly influenced by the tree canopy. This opens exciting prospects for species distribution modelling, conservation planning, and biogeographical and ecological studies on Amazonian biota. Our maps provide a preliminary geoecological subdivision of Amazonia that can now be tested and refined using field data of other plant groups and from hitherto unsampled areas.

Environmental and fern species data for Amazonian plotsSpecies and soil data were collected in the field by the researchers mentioned as data providers for each plot. Reflectance data come from a Landsat TM/ETM+ image composite based on images acquired in 2000-2009 and processed using R code written by Jasper Van doninck. CHELSA and SRTM data come from respective online portals. See README file for explanation of the data contained in each column.Tuomisto_etal_JBI-18-0631_data.csv