Open AccessA Landsat composite covering all Amazonia for applications in ecology and conservation
Author(s) -
Van doninck Jasper,
Tuomisto Hanna
Publication year - 2018
Publication title -
remote sensing in ecology and conservation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.191
H-Index - 21
ISSN - 2056-3485
DOI - 10.1002/rse2.77
Subject(s) - edaphic , biome , amazon rainforest , compositing , remote sensing , vegetation (pathology) , amazonian , biodiversity , environmental science , multispectral image , geography , ecology , physical geography , ecosystem , soil science , soil water , biology , image (mathematics) , pathology , artificial intelligence , medicine , computer science
Studies at small spatial extents have shown that local floristic and edaphic patterns within the hyper‐diverse Amazonian forests can be identified at a high thematic resolution using Landsat imagery. This suggests that Landsat images have the potential to indicate ecologically relevant environmental and biotic variation in the forests also at the extent of the entire basin. However, the full potential of Landsat data for these purposes has not yet been exploited in ecological and biodiversity research or in conservation applications. This is largely because the artifactual noise that is introduced by atmospheric and directional effects into multi‐scene composite images can swamp the subtle spectral differences between different types of primary forest. Here, we present a new Landsat TM / ETM + image composite for the entire Amazon biome that largely overcomes these problems. It is based on more than 16 000 individual image acquisitions from the 10‐year period 2000–2009. The images were individually processed to directionally and topographically normalized surface reflectance and combined into 2.5 degree tiles using the medoid compositing criterion. Visual inspection showed that the resulting image composite is radiometrically clearly more consistent than other currently available Landsat composites. We tested the ecological relevance of the new Landsat composite by comparing its reflectance values with edaphic properties measured in more than 300 field sampling localities spread across 2000 km of Amazonia. We found a strong correlation between observed and predicted concentration of exchangeable base cations in the surface soil, which indicates that the compositing approach has succeeded in removing most of the artifactual noise. The Landsat composite image should be of great value for a multitude of applications in ecology, biodiversity research and conservation planning that require environmental data layers combining detailed spatial resolution, basin‐wide coverage and high radiometric accuracy.