Fascinating. It's inspiring to see Europe's dedication to Earth observation. With the Living Planet Fellowship and missions like Sentinel-1D, how do you envision the increased data impacting climate models, and what role will advanced AI play in predicative analytics? Your insights are always so valuable.
You’re absolutely right - the increase in EO missions also means a huge increase in data volume. Fortunately, this “data overload” has been anticipated for years. That’s why there's such a strong push toward cloud-native formats and technologies that make it possible to:
- Store petabyte-scale archives efficiently
- Stream only the pixels we need
- Work interactively with huge multi-dimensional data cubes
The weather and climate institutions have really led the way here. They’ve been assimilating massive datasets for decades to produce operational forecasts. But it takes time and a clear strategy - robust infrastructure, serious compute, and the right tools - and many of those are still being rapidly developed. (A great example is ECMWF’s Polytope project - https://www.ecmwf.int/sites/default/files/medialibrary/2021-01/hawkes-polytope-poster.pdf)
Rather than running a full numerical simulation - which can be extremely computationally expensive - AI can accelerate parts of the workflow and deliver faster updates at higher resolution.
Another major development is the rise of foundation models for EO. And while self-supervised learning isn’t new, now we finally have the combination of high-quality data, compute power, and modern tooling needed to train these models. It’s taken time for all the pieces to come together - but agencies, institutions, and companies are definitely picking up the pace.
Europe’s Destination Earth (DestinE) initiative might be the best example of a strategy-driven approach that brings all these components together - multi-modal data, physics-based models, advanced AI, and massive compute - toward the shared goal of building reliable, actionable digital twins of our planet.
Fascinating. It's inspiring to see Europe's dedication to Earth observation. With the Living Planet Fellowship and missions like Sentinel-1D, how do you envision the increased data impacting climate models, and what role will advanced AI play in predicative analytics? Your insights are always so valuable.
Thanks for your comment!
You’re absolutely right - the increase in EO missions also means a huge increase in data volume. Fortunately, this “data overload” has been anticipated for years. That’s why there's such a strong push toward cloud-native formats and technologies that make it possible to:
- Store petabyte-scale archives efficiently
- Stream only the pixels we need
- Work interactively with huge multi-dimensional data cubes
The weather and climate institutions have really led the way here. They’ve been assimilating massive datasets for decades to produce operational forecasts. But it takes time and a clear strategy - robust infrastructure, serious compute, and the right tools - and many of those are still being rapidly developed. (A great example is ECMWF’s Polytope project - https://www.ecmwf.int/sites/default/files/medialibrary/2021-01/hawkes-polytope-poster.pdf)
On the AI front, again, weather and climate are paving the path. ECMWF’s AIFS (AI/Integrated Forecasting System) is an exciting step (https://www.ecmwf.int/en/about/media-centre/aifs-blog):
Rather than running a full numerical simulation - which can be extremely computationally expensive - AI can accelerate parts of the workflow and deliver faster updates at higher resolution.
Another major development is the rise of foundation models for EO. And while self-supervised learning isn’t new, now we finally have the combination of high-quality data, compute power, and modern tooling needed to train these models. It’s taken time for all the pieces to come together - but agencies, institutions, and companies are definitely picking up the pace.
Europe’s Destination Earth (DestinE) initiative might be the best example of a strategy-driven approach that brings all these components together - multi-modal data, physics-based models, advanced AI, and massive compute - toward the shared goal of building reliable, actionable digital twins of our planet.