Unmasking Spatio-Temporal Variability of the Kerguelen Archipelago: Radiometric Landscapes As New Proxies

This study conducts the first regional mapping exercise of landscape dynamics of the Kerguelen Archipelago; a remote sub-Antarctic archipelago located in the southern Indian Ocean. It is based on an original adaptation of the radiometric landscape methodology applied to the long time series of MODIS NDVI data covering the period 2003-2022, allowing the temporal dimension to be explicitly integrated into the regional-scale landscape mapping. The aim is to provide a comprehensive framework for analyzing the structure, dynamics, and evolutionary trajectories of the landscapes of this sub-Antarctic archipelago, which remain largely unknown yet essential for understanding ecosystem functioning in isolated environments. The adopted approach is based on three complementary steps. First, synthetic variables were extracted from the NASA's MODIS NDVI product to summarize seasonal and interannual information. Next, spatial segmentation was used to group pixels into radiometrically homogeneous objects, constituting basic units of the landscape. Finally, unsupervised clustering was applied to classify these units and produce a coherent landscape typology. This original data-driven approach overcomes the limitations of supervised classifications based on predefined criteria by distinguishing landscapes that may appear similar at a given moment but have divergent NDVI value trajectories over two decades. The analysis led to the identification of five major landscape units, revealing different dynamics at the regional scale. The integration of outputs from the MAR regional climate model and the NASADEM digital elevation model made it possible to place these landscape units in their environmental and topographical context. The results show that landscape stratification follows a marked altitudinal gradient, which strongly influences the distribution and evolution of radiometric landscape units. Beyond producing unprecedented mapping, this approach constitutes the first regional, remote sensing data-driven geographical partitioning of a sub-Antarctic archipelago based on long-term vegetation trajectories. It provides a foundational environmental framework that can serve as a baseline for future field-based investigations, comparative analyses between/within the five identified landscape units, and multi-scale ecological studies focusing on specific environments. By capturing the specific values and evolutionary trajectories of vegetation, this methodology makes it possible to establishes an initial typology of the archipelago's landscapes, while providing insights to better understand the interactions between climate, relief, and vegetation. Our approach contributes to improving the knowledge and monitoring of landscapes in remote sub-Antarctic environments. As a territory managed by the Terres Australes et Antarctiques Fran & ccedil;aises (TAAF), the Kerguelen Archipelago constitutes a sentinel of environmental change in the Southern Hemisphere. The proposed methodology provides regional and national decision-makers with a robust tool for long-term ecosystem and landscape monitoring, supporting management and conservation strategies adapted to these particularly fragile environments that are highly vulnerable to climate change.