______________________________________________________________________________
Exploring the physicochemical limits of life and microbial adaptations to polyextreme ecosystems
Purificación López-García
Ecologie Systématique Evolution, CNRS & Université Paris-Saclay, 91190 Gif-sur-Yvette, France
In this talk, I will summarize recent studies of our team to determine the physicochemical limits of life along gradients of polyextreme conditions combining high concentrations of chaotropic salts, high temperature and low pH in the north Danakil Depression (Ethiopia). I will comment on the difficulties and confounding factors to unambiguously determine the occurrence or absence of microbial life, and show evidence suggesting that some of these highly chaotropic systems are devoid of life. I will then describe the structure of communities inhabiting the most polyextreme sites harboring life and discuss the adaptations of the archaea dominating these microbial communities as inferred from metagenomic data. Finally, I will comment about the discovery of novel archaeal families, their metabolism and their evolution, and show that the adaptation to hypersaline environments evolved convergently in archaea several times.
______________________________________________________________________________
Everything, Everywhere, all at once: the reasons for a systems approach towards sustainability
Evi Viza
University of the West of Scotland, UWS, School of Computing Engineering and Physical Sciences
The presentation gives an overview on certain factors that play a role towards achieving the three pillars of sustainability (people, planet, profit). These range from macro level such as geopolitics and business models to micro level such as everyday habits and awareness of the public. The case for a systems approach is made and that to achieve the sustainable development goals, a holistic approach is needed and a new mindset.
______________________________________________________________________________
Introduction to deep learning for Earth imagery
Mathieu Aubry
Imagine team, LIGM, École des Ponts ParisTech (ENPC)
In this presentation I will first introduce the basic concepts of supervised, unsupervised and deep learning on the example of satellite image time series classification [1]. I will then present how more complexe deep architectures can be used to perform more complex tasks such as semantic segmentation and change detection [2]. Finally, I will show some recent results for unsupervised object discovery in aerial LIDAR point clouds.
[1] Pixel-wise Agricultural Image Time Series Classification: Comparisons and a Deformable Prototype-based Approach, E. Vincent, J. Ponce, M. Aubry, ArXiv 2023
[2] Satellite Image Time Series Semantic Change Detection: Novel Architecture and Analysis of Domain Shift, E. Vincent, J. Ponce, M. Aubry, in submission
[3] Learnable Earth Parser: Discovering 3D Prototypes in Aerial Scans, R. Loiseau, E. Vincent, M. Aubry, L. Landrieu, CVPR 2024