Amazonian biodiversity has yet to reveal many of its secrets, but is threatened by deforestation and a changing climate. Rapid and comprehensive mapping of biodiversity over this vast biome cannot be achieved based on field surveys alone, making remote sensing an indispensable tool. The Landsat satellite sensors with their global coverage at 30 m spatial resolution could play an important role in biodiversity modelling, but only register information from the upper layer of the forest canopy. It therefore remains unclear to what extent sub-canopy processes are represented by these this type of satellite data.
We created a composite image based on more than 16000 Landsat TM and ETM+ images acquired over a 10-year period over the entire Amazon biome. Using a field plot dataset of fern and lycophyte species and soil samples collected throughout Amazonia, we combined Landsat with climatic datasets to model species distribution, species composition, and base cation concentration in the soil.
Results/Conclusions
Combining a large number of individual Landsat scenes into a single composite image reduced noise, and allowed detecting subtle spatial patterns in surface canopy reflectance. A linear model between Landsat surface reflectance and soil cation concentration resulted in a high correlation between observed and predicted cation concentration values. The primary axis of a floristic ordination of fern and lycophyte field plots was strongly related to cation concentration and was best modelled using Landsat as explanatory variables. The secondary gradient was related to climatic variables. In species distribution models of 69 species, adding remotely sensed surface reflectance variables to climatic data layers improved the discriminatory ability of models between presence and absence sites.
Overall, Landsat data were able to predict soil-related distributional and compositional patterns of fern and lycophyte species across Amazonia. This is because both understory species composition and canopy species composition are to a large degree structured by Amazonian soil properties. This opens prospects for the use of freely available multispectral canopy reflectance data in biogeographical and ecological studies and in conservation planning.