We created an advanced AI model called Prithvi-EO-2.0 to help scientists better understand changes on Earth using satellite images. This model was trained on over 4 million image sequences collected over ten years from Landsat and Sentinel-2 satellites. It can recognize patterns in how the Earth’s surface changes over time—like seasons, natural disasters, and farming cycles.

The model uses a special architecture that learns both what an image shows and when and where it was taken. It comes in two sizes, with 300 million and 600 million parameters.

Prithvi-EO-2.0 has been tested on many real-world tasks such as mapping floods and wildfires, detecting landslides, identifying different crops, and estimating how much plants grow in ecosystems. It works with different kinds of satellite data and can handle both medium and high-resolution images.

Best of all, it’s open-source, meaning anyone can use and improve it through a platform called TerraTorch. Our talk will explain how we built and trained the model, how well it performs, and how it can be used to build more advanced Earth monitoring systems in the future.