This dataset demonstrates a prototype step towards addressing the challenges of weather data access by increasing the ease and speed with which you can get data. It does this by reformatting an existing weather dataset, NOAA's High Resolution Rapid Refresh or HRRR forecast, into a cloud optimized data zarr archive and hosting the zarr on Source Cooperative.

This demonstration dataset contains only

• One variable: Surface Air Temperature
• One month: June 2023
• The "analysis" data, the models best estimate of a value at each hour. It does not include a forecast.

The data covers the full HRRR model geographic domain covering the continental US and parts of northern Mexico and southern Canada.

The data to create this dataset was originally accessed from the The University of Utah's MesoWest archive of HRRR data in zarr format on AWS's Open Data Program, an impressively complete and unique archive. The zarr you have here is internally structured and compressed differently than the MesoWest zarrs to maintain access latencies to small subsets of the dataset, while enabling much higher throughput when accessing large swaths of the dataset.

Python example code using this zarr is available here.