02 / IT · Analysis
DeepMind releases precipitation research datasets
The company-published improvement concerns precipitation detection, while the datasets are available to researchers and developers.

Google DeepMind released a precipitation dataset with a company-published detection improvement of more than 30% for researchers and developers, AI Times reported.
According to AI Times, the release combines improved precipitation detection with access for outside users. That makes both the reported performance and the research use of the data central to the announcement.
AI Times also described the release of a dataset integrating satellite imagery. The scope of the detection improvement and the uses of the released datasets need to be considered separately.
The comparison concerns precipitation detection
According to AI Times, DeepMind compared its results with existing global precipitation observation data. The company-published improvement concerns the detection of light and moderate precipitation.
The figure reported by AI Times measures precipitation detection. It does not establish a corresponding improvement in the overall accuracy of weather or climate forecasts.
Observation data for research
AI Times described PARDIG as an AI-based global precipitation dataset. It also introduced SatWeave, a dataset that integrates imagery from geostationary satellites, which remain above the same area of Earth.
According to AI Times, the release aims to broaden the use of AI-based weather and climate forecasting technology. The company-published performance improvement concerns precipitation observations; using those observations in forecasting research is the next task.