SEVIR Dataset


There is no shortage of weather data available to the public: Weather radar, GEO & LEO satellite, surface observations, numerical weather prediction models, balloons, ocean buoys, aircraft measurements, lightning sensing, radio occultation, and others are being measured every second of every day to aid in our ability to analyze and forecast the weather. As a result, both the size and complexity of all of this data can make it difficult to work with. Many datasets, such as radar and satellite, are too large to process in bulk for those without access to massive compute and storage. Some datasets also require a level of domain expertise to decode, understand and process. These are some of the barriers for scientists and researchers who want to understand and use this data for their applications.

To help alleviate these issues, the SEVIR dataset was constructed to enable faster R&D in weather sensing, avoidance, short-term forecasting and other related applications. SEVIR combines and aligns multiple weather sensing modalities into a single, accessible dataset accessible for free on the cloud that can be used by meteorologists, data scientists and other researchers.


sevir data table


Take a look at our tutorial here.