Title: GREMLIN CONUS1 Dataset
Author: Kyle Hilburn (ORCID: 0000-0002-2078-9884)
Contact: Kyle Hilburn, kyle.hilburn@colostate.edu
Institution: Cooperative Institute for Research in the Atmosphere (CIRA) / Colorado State University (CSU)

Brief Summary: The CONUS1 dataset (conus1.nc) is a "toy" dataset that is small enough to be able to train a convolutional neural network on a laptop computer to do image translation from geostationary satellite images to ground-based radar images. It provides three input channels from GOES-16 ABI, one input channel from GOES-16 GLM, and one output channel from MRMS.

Associated article citations:

• The training dataset is described in:
  Hilburn, K., S. D. Miller, and M. Marchand, 2019: Using high-resolution observations from
  GOES-16 ABI to improve operational short-range forecasting. Joint Satellite Conference.
• The testing dataset is described in:
  Hilburn, K., M. Marchand, Y. Lee, C. Kummerow, and C. Alexander, 2018: Using GOES-16 to
  improve short-term forecasts. 98th American Meteorological Society Annual Meeting.
• The data preparation and GREMLIN model are described in:
  Hilburn, K. A., I. Ebert-Uphoff, and S. D. Miller, 2021: Development and interpretation
  of a neural-network-based synthetic radar reflectivity estimator using GOES-R satellite
  observations. J. Appl. Meteor. Climatol., 60, 3-21, doi: 10.1175/JAMC-D-20-0084.1.

Funding: GOES-R Program Award NA19OAR4320073

Acronyms:

• ABI = GOES Advanced Baseline Imager
• CIRA = Cooperative Institute for Research in the Atmosphere
• CLASS = NOAA Comprehensive Large Array-data Stewardship System
• C07 = GOES ABI 3.9 micrometer band
• C09 = GOES ABI 6.9 micrometer band
• C13 = GOES ABI 10.3 micrometer band
• CONUS = Contiguous United States
• GED = GLM Group Extent Density
• GLM = GOES Geostationary Lightning Mapper
• GOES = Geostationary Operational Environmental Satellite
• GREMLIN = GOES Radar Estimation via Machine Learning to Inform NWP
• HRRR = High Resolution Rapid Refresh
• MRMS = Multi-Radar/Multi-Sensor System
• NCEP = National Centers for Environmental Prediction
• NOAA = National Oceanic and Atmospheric Administration
• NWP = Numerical Weather Prediction
• REFC = Composite (vertical maximum) Radar Reflectivity Factor

Location where data were collected: Contiguous United States with bounding box 29.07 to 44.50 degrees North latitude and -109.02 to -79.09 degrees East longitude.

Spatial resolution of the data: Pixels are 3 km x 3 km on a Lambert conformal conic projection.

Time period during which data were collected: 2017-03-21T20:00:00Z to 2019-08-13T23:45:00Z.

Time sampling rate of the data: every 15 minutes

File format: NetCDF-4

Inputs (Xdata):

• Channel 0 = GOES-16 ABI C07 (3.9 micrometer) Brightness Temperature
• Channel 1 = GOES-16 ABI C09 (6.9 micrometer) Brightness Temperature
• Channel 2 = GOES-16 ABI C13 (10.3 micrometer) Brightness Temperature
• Channel 3 = GOES-16 GLM Lightning Group Extent Density

Output (Ydata):

• MRMS Composite Radar Reflectivity

Scalings: The data have already been linearly scaled into the range [0,1] using the values below. The images are fully covered by valid data (no fill values).
Chan Min Max Inverted Units
C07 200 300 True K
C09 200 250 True K
C13 200 300 True K
GLM 0.1 50 False groups/km^2/5-minutes
MRMS 0 60 False dBZ

Coordinates (lat,lon): The data are provided as 256 x 256 images on a 3 km grid (HRRR CONUS grid).

Methods: The methodology is described in detail by Hilburn et al. (2021). The ABI, GLM, and MRMS data sets were resampled to a common 3 km grid. A cloud height of 10 km was used for removing parallax displacements. Satellite and radar samples were matched in time with a maximum time difference of 2.5 minutes. GLM lightning groups were accumulated over 15-minute time periods.

Data Split: The training dataset consists of 9 cases and the testing dataset consists of 2 cases, where a case is a sequence of 24 images with 15-minute time step.
TRAINING DATASET:
Case Start Time Center Lon Center Lat Comment
1 2019-05-23T06:00Z -88 39 Illinois/Indiana, pre-existing squall line
2 2019-06-19T15:00Z -87 33 Alabama, pop-up convection
3 2019-06-20T06:00Z -88 35 Alabama, linear convection
4 2019-06-20T18:00Z -84 33 Georgia, linear convection
5 2019-07-09T00:00Z -99 40 Kansas/Nebraska, merging and propagation
6 2019-07-20T21:00Z -93 41 Iowa/Missouri, intensification of storm
7 2019-08-13T18:00Z -101 40 Nebraska/Kansas, isolated cell
8 2019-06-09T18:00Z -87 33 Alabama, disorganized pop-up
9 2019-07-05T21:00Z -104 40 Colorado, isolated cells with hail
TESTING DATASET:
Case Start Time Center Lon Center Lat Comment
1 2017-03-21T20:00Z -84 35 Georgia/South Carolina/Tennessee, convective initiation
2 2017-05-08T20:00Z -104 40 Colorado, billion-dollar hail disaster

Sharing/Access information

License information or restrictions placed on the data: The material is open access and distributed under the terms and conditions of the Creative Commons CC0 license that allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, with no conditions.

Data source: This dataset was created from the publicly available datasets:

• The L1b ABI data files are available from NOAA CLASS (https://www.avl.class.noaa.gov/saa/products/search?datatype\_family=GRABIPRD).
• The L2 GLM data files are available from NOAA CLASS (https://www.avl.class.noaa.gov/saa/products/search?datatype\_family=GRGLMPROD).
• The MRMS composite reflectivity data files are available from NCEP (https://mrms.ncep.noaa.gov/data/).