Version 1.0

Initial version of assimilating microwave sounders and geostationary imagers to predict temperature and specific humidity.

• Based on the paper https://arxiv.org/abs/2407.11696
• Temperature and specific humidity taken from the Microwave Integrated Retrieval System (MIRS)
• Signal largely from the Advanced Technology Microwave Retrieval System (ATMS)
• Number of input satellite/sensors = 4 GEO + 3 LEO
• Output is the resolution of ATMS

Variables

• qv = Specific humidity (37 levels)
• temp = Temperature (37 levels)
• atms_brightness_temperature = Synthetic microwave radiances (22 channels)

Coordinates

• lat: Latitude, 0.16 degrees
• lon: Longitude, 0.16 degrees
• time: Forecast times hourly, starting at t=0
• level: Atmospheric pressure from 1000 to 1 hpa
• atms_channel: Channels corresponding to spectral wavelengths from ATMS

Location

AWS S3 - s3://zeusai-data/prod/earthnet/v1/forecast/{year}/{month}/{day}/earthnet.v1.forecast.6h.{year}{month}{day}{hour}00.zarr

Format

<xarray.Dataset> Size: 3GB
Dimensions:                      (level: 37, atms_channel: 22, lat: 1125,
                                  time: 7, lon: 2249)
Coordinates:
  * level                        (level) int64 296B 1000 975 950 925 ... 5 3 2 1
  * atms_channel                 (atms_channel) int64 176B 0 1 2 3 ... 19 20 21
  * lat                          (lat) float64 9kB -90.08 -89.92 ... 89.6 89.76
  * lon                          (lon) float64 18kB -180.0 -179.8 ... 179.7
  * time                         (time) datetime64[ns] 56B 2025-07-13T16:00:0...
Data variables:
    atms_brightness_temperature  (atms_channel, time, lat, lon) float16 779MB dask.array<chunksize=(3, 2, 282, 563), meta=np.ndarray>
    qv                           (level, time, lat, lon) float16 1GB dask.array<chunksize=(5, 2, 282, 563), meta=np.ndarray>
    temp                         (level, time, lat, lon) float16 1GB dask.array<chunksize=(5, 2, 282, 563), meta=np.ndarray>