UK Radar Precipitation Nowcasting Dataset (Met Office NIMROD 1 km, DeepMind DGMR)
- Provider
- Google DeepMind and the UK Met Office (Ravuri et al.)
- License
- CC-BY-4.0 (source)
- Stored at
- Google Cloud Storage, gs://dm-nowcasting/datasets (us)
- Access
- Free and public. No account with the data provider is needed.
- DOI
- 10.1038/s41586-021-03854-z
- Category
- Climate & Weather
- Source
- https://doi.org/10.1038/s41586-021-03854-z
About this dataset
Post-processed UK weather-radar rain-rate fields released with DGMR, the deep generative model of radar nowcasting in Ravuri et al. (Nature, 2021), for training and scoring precipitation forecasts at lead times of minutes to two hours. The source is the Met Office NIMROD radar composite over the United Kingdom, projected onto the 1 km Ordnance Survey National Grid (OSGB) and stored as sequences of frames at 5-minute spacing. The tree under datasets/nowcasting_open_source_osgb/nimrod_osgb_1000m_yearly_splits/radar/ splits by year into train, valid and test: training sequences are 256x256 random crops, evaluation sequences come both as full frames and as subsampled overlapping padded 512x512 tiles, each at a 20-minute stride. Measured 2026-11-15: 3,355 gzipped TFRecord shards, about 1.07 TB, with shard counts such as 712 full-frame and 499 tiled test sequences. Parse with tf.data using the feature spec in the deepmind-research nowcasting colab and pseudocode.zip. This row covers only datasets/; the bucket's tfhub_snapshots/ model weights are CC BY-NC 4.0 and are deliberately excluded. The datasets are CC BY 4.0 and contain public sector information licensed by the Met Office under the Open Government Licence v3.0; cite Ravuri et al. 2021.
How to access this dataset in Python
import gcsfs
fs = gcsfs.GCSFileSystem(token='anon')
fs.ls('dm-nowcasting/datasets/', detail=False)[:10]Lists the files with anonymous access. Runs on any machine, with no cloud account.
Data structure
Files
Example file: datasets/nowcasting_open_source_osgb/nimrod_osgb_1000m_yearly_splits/radar/20200718/test/full_frame_20min_stride/seq-24-00000-of-00712.tfrecord.gz
Path pattern: {prefix}/{seg1}/{seg2}/{seg3}/{YYYYMMDD}/{seg5}/{seg6}/{filename}
Analyze this dataset without downloading it
On Scigantic, UK Radar Precipitation Nowcasting Dataset (Met Office NIMROD 1 km, DeepMind DGMR) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Climate & Weather datasets
- BARRA v1: Bureau of Meteorology Atmospheric Regional Reanalysis for Australia (12 km + 1.5 km)
- Digital Typhoon: Long-Term Satellite Image Dataset of Tropical Cyclones
- NOAA GFS analysis, hourly
- Ocean Radar - Coffs Harbour site - Wave - Delayed mode
- NOAA Real-Time Mesoscale Analysis (RTMA) / Unrestricted Mesoscale Analysis (URMA)
- NOAA Fundamental Climate Data Records (FCDR)
- NOAA Atmospheric Climate Data Records
- BARPA: Bureau of Meteorology Atmospheric Regional Projections for Australia