Pre-processed data and trained model weight in ADAF

Pre-proccesd data

The pre-proccesd data consists of input-target pairs. The inputs include surface weather observations within a 3-hour window, GOES-16 satellite imagery within a 3-hour window, HRRR forecast, and topography. The target is a combination of RTMA and surface weather observations. The table below summarizes the input and target datasets utilized in this study. All data were regularized to grids of size 512 $\times$ 1280 with a spatial resolution of 0.05 $\times$ 0.05 $^\circ$. 

Dataset
Source
Time window
Variables/Bands

Input
Surface weather observations
WeatherReal-Synoptic (Jin et al., 2024)
3 hours
Q, T2M, U10, V10

Input
Satellite imagery
GOES-16 (Tan et al., 2019)
3 hours
0.64, 3.9, 7.3, 11.2 $\mu m$

Input
Background
HRRR forecast (Dowell et al., 2022)
N/A
Q, T2M, U10, V10

Input
Topography
ERA5 (Hersbach et al., 2019)
N/A
Geopotential

Target
Analysis
RTMA (Pondeca et al., 2011)
N/A
Q, T2M, U10, V10

Target
Surface weather observations
WeatherReal-Synoptic (Jin et al., 2024)
N/A
Q, T2M, U10, V10

2022-10-01_06.nc is a sample of pre-proccesd data. The vairables in this file contain the input-target pairs mentioned above.

A sample file contains the following variables:

Variable
Decription
Dimension

z
Topography, normalized
[lat, lon]

rtma_t
T2M from RTMA, normalized
[lat, lon]

rtma_q
Q from RTMA, normalized
[lat, lon]

rtma_u10
U10 from RTMA, normalized
[lat, lon]

rtma_v10
V10 from RTMA, normalized
[lat, lon]

sta_t
T2M from station's observation, 0 means non-station, normalized
[obs_time_window, lat, lon]

sta_q
Q from station's observation, 0 means non-station, normalized
[obs_time_window, lat, lon]

sta_u10
U10 from station's observation, 0 means non-station, normalized
[obs_time_window, lat, lon]

sta_v10
V10 from station's observation, 0 means non-station, normalized
[obs_time_window, lat, lon]

CMI02
ABI Band 2: visible (red), normalized
[obs_time_window, lat, lon]

CMI07
ABI Band 7: shortwave infrared, normalized
[obs_time_window, lat, lon]

CMI10
ABI Band 10: low-level water vapor, normalized
[obs_time_window, lat, lon]

CMI14
ABI Bands 14: longwave infrared, normalized
[obs_time_window, lat, lon]

hrrr_t
T2M from HRRR 1-hour forecast
[lat, lon]

hrrr_q
Q from HRRR 1-hour forecast
[lat, lon]

hrrr_u_10
U10 from HRRR 1-hour forecast
[lat, lon]

hrrr_v_10
V10 from HRRR 1-hour forecast
[lat, lon]

Pre-comuted normalization statistics

stats.csv is pre-comuted normalization statistics.

Pre-trained model weights

best_ckpt.tar is the pre-trained model weights.

References

Jin, W. et al. WeatherReal: A Benchmark Based on In-Situ Observations for Evaluating Weather Models. (2024).

Dowell, D. et al. The High-Resolution Rapid Refresh (HRRR): An Hourly Updating Convection-Allowing Forecast Model. Part I: Motivation and System Description. Weather and Forecasting 37, (2022).

Tan, B., Dellomo, J., Wolfe, R. & Reth, A. GOES-16 and GOES-17 ABI INR assessment. in Earth Observing Systems XXIV vol. 11127 290–301 (SPIE, 2019).

Hersbach, H. et al. ERA5 monthly averaged data on single levels from 1979 to present. Copernicus Climate Change Service (C3S) Climate Data Store (CDS) 10, 252–266 (2019).

Pondeca, M. S. F. V. D. et al. The Real-Time Mesoscale Analysis at NOAA’s National Centers for Environmental Prediction: Current Status and Development. Weather and Forecasting 26, 593–612 (2011).