Descripción del proyecto
El aprendizaje automático llega a la modelización climática
Los modelos del sistema terrestre constituyen la base de la comprensión y de la proyección del cambio climático. A pesar del progreso en este ámbito, la capacidad de los modelos para simular las respuestas de los sistemas terrestres, tanto a nivel mundial como regional, se ve limitada por la representación de los procesos físicos y biológicos a pequeña escala. El proyecto USMILE, financiado con fondos europeos, utilizará el aprendizaje automático para mejorar la modelización y la comprensión del sistema terrestre. Los investigadores desarrollarán algoritmos de aprendizaje automático para mejorar los conjuntos de datos de observación de la Tierra que representan las covariaciones espaciotemporales, y parametrizaciones y submodelos basados en el aprendizaje automático para nubes y procesos de la superficie terrestre que han obstaculizado el avance de la modelización climática durante décadas. Además, detectarán y dilucidarán modos de variabilidad climática y de extremos multivariante y desvelará aspectos dinámicos del sistema terrestre con nuevas técnicas de aprendizaje profundo y descubrimiento causal.
Objetivo
Earth system models are fundamental to understand climate change. Although they have improved significantly, considerable biases and uncertainties in their projections remain. Process parameterisations limit the models’ ability to simulate both global and regional Earth system responses, which are key for assessing climate change and its impacts on ecosystems and society. In recent years, the volume of data from high-resolution models and observations has substantially increased to petabyte scales. Concomitantly, the field of machine learning (ML) has quickly developed, promising breakthroughs in detecting and analysing non-linear relationships and patterns in large multivariate datasets. Yet, traditionally, physical modelling and ML have been often treated as two different worlds with opposite scientific paradigms (theory-driven versus data-driven). Thus, despite its great potential, ML has not yet been widely adopted for addressing the urgent need of improved understanding and modelling of the Earth system. USMILE will combine multi-disciplinary expertise in ML and process-based atmosphere and land modelling to completely rethink model development and evaluation. ML will further allow us to define novel observational constraints on Earth system feedbacks and climate projections. We will (1) develop ML algorithms to enhance Earth observation datasets accounting for spatio-temporal covariations, (2) deploy ML-based parameterisations and sub-models for clouds and land-surface processes that have hindered progress in climate modelling for decades, and (3) detect and understand modes of climate variability, multivariate extremes and uncover dynamical aspects of the Earth system with novel deep learning and causal inference techniques. USMILE will drive a paradigm shift in the current modelling of the Earth system towards a new data-driven physics-aware science and to an unprecedented reduction of uncertainties in projections.
Ámbito científico
- natural sciencesbiological sciencesecologyecosystems
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
- natural sciencesearth and related environmental sciencesatmospheric sciencesclimatologyclimatic changes
- natural sciencescomputer and information sciencessoftwaresoftware applicationssimulation software
Programa(s)
Tema(s)
Régimen de financiación
ERC-SyG - Synergy grantInstitución de acogida
51147 Koln
Alemania