Academia Sinica Logo

Abstract: Artificial intelligence is increasingly applied to analyze complex datasets, whether from Earth observation or astronomical instruments. In this talk, I will present recent work from the German Environment Agency using AI for environmental and climate challenges, including detecting radioactive elements via supervised dimensionality reduction, identifying wind turbines and solar panels from satellite imagery, and tracking illegal wildlife trade through automated online analysis: https://showcase.ai-env.de/en Many of the underlying methods are directly relevant to astronomy and physics. For example, dimensionality reduction techniques can improve element detection in spectra, while resource-efficient algorithms for object counting or data compression at the sensor can handle large-scale observational datasets. I will also highlight approaches for energy- and resource-efficient machine learning, demonstrating how to reduce computational cost without sacrificing accuracy. This talk illustrates how innovations in environmental AI can inform large-scale data challenges across scientific disciplines.