V. Miranda, P. Pina, S. Heleno, S. Hong, H. Lee, G. Vieira

Abstract
Vegetation is a key-element of the terrestrial ecosystems of Antarctica [1]. Its extension, growth and species distribution, mainly lichens (e.g., Usnea antarctica, Usnea aurantiacoatra) and mosses (e.g., Andreaea, Sanionia), can be related to topoclimatic factors [2], being their monitoring particularly relevant as they can be used as bioindicators of climate change [3]. The only practical way of extensively mapping the vegetation is using satellite imagery, but its occurrence in relatively sparse and small patches makes this identification difficult and challenging [4], [5]. Only metric to sub-metric imagery (available for about a decade and a half) can provide direct adequate identifications of the vegetation [6]–[9], but on another hand only Landsat datasets (at 30m and lower) may allow its monitoring for a longer period (since the 1970s). Therefore, for large time and scale mapping objectives, it is crucial to first assess how Landsat imagery scale is adequate to identify the vegetation in Antarctica. This issue, still poorly addressed in Antarctic vegetation, is the main objective of this study. By understanding how vegetation classification can be accurately downscaled through the increasing size of the pixel of different sensors (UAV, WorldView, Sentinel and Landsat), one may infer changes in vegetation through a broader set of images and through a larger historic period of satellite coverage.

IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sensing Symposium

IEEE
2022 September

>> DOI