UChicago AI Weather Model Reforecast Archive

Data documentation

Common variables

These 18 variables are saved by every 0.25° model — GraphCast, Aurora 1.5, AIFS-single v2, AIFS-ENS v2, FGN, and Aurora 1.5 ENS — with the same units and temporal aggregation. The model sections below list only the variables each model adds on top of this set. NeuralGCM runs on a different grid with its own variable naming and is listed in full in its own section.

VariableUnitsTemporal aggregation
2 m temperature (2t)K6-hourly instantaneous
Mean sea level pressure (mslp)Pa6-hourly instantaneous
Specific humidity at 1000 hPa (q_1000)kg kg⁻¹6-hourly instantaneous
Specific humidity at 925 hPa (q_925)kg kg⁻¹6-hourly instantaneous
Specific humidity at 850 hPa (q_850)kg kg⁻¹6-hourly instantaneous
Temperature at 1000 hPa (t_1000)K6-hourly instantaneous
Temperature at 925 hPa (t_925)K6-hourly instantaneous
Temperature at 850 hPa (t_850)K6-hourly instantaneous
Total precipitation (tp)mDaily accumulation (UTC day)
U wind component at 50 hPa (u_50)m s⁻¹Daily mean (UTC day)
U wind component at 200 hPa (u_200)m s⁻¹Daily mean (UTC day)
U wind component at 850 hPa (u_850)m s⁻¹Daily mean (UTC day)
V wind component at 50 hPa (v_50)m s⁻¹Daily mean (UTC day)
V wind component at 200 hPa (v_200)m s⁻¹Daily mean (UTC day)
V wind component at 850 hPa (v_850)m s⁻¹Daily mean (UTC day)
Geopotential at 200 hPa (z_200)m² s⁻²Daily mean (UTC day)
Geopotential at 500 hPa (z_500)m² s⁻²Daily mean (UTC day)
Geopotential at 850 hPa (z_850)m² s⁻²Daily mean (UTC day)

Mean sea level pressure is named msl in AIFS-ENS v2 and mslp in the other models.

GraphCast

Class: deterministic Grid: 721 × 1440 (0.25°) Forecast length: 50 days

GraphCast saves the common variables only.

Aurora 1.5

Class: deterministic Grid: 720 × 1440 (0.25°) Forecast length: 50 days

VariableUnitsTemporal aggregation
2 m dewpoint temperature (2d)K6-hourly instantaneous
Surface pressure (sp)Pa6-hourly instantaneous
Soil temperature level 1 (stl1)K6-hourly instantaneous
Volumetric soil water layer 1 (swvl1)m³ m⁻³6-hourly instantaneous
Daily maximum 2 m temperature (2t_max)KDaily maximum (UTC day)
Daily minimum 2 m temperature (2t_min)KDaily minimum (UTC day)

AIFS-single v2

Class: deterministic Grid: 721 × 1440 (0.25°) Forecast length: 50 days

VariableUnitsTemporal aggregation
2 m dewpoint temperature (2d)K6-hourly instantaneous
Surface pressure (sp)Pa6-hourly instantaneous
Soil temperature level 1 (stl1)K6-hourly instantaneous
Soil temperature level 2 (stl2)K6-hourly instantaneous
Volumetric soil water layer 1 (swvl1)m³ m⁻³6-hourly instantaneous
Volumetric soil water layer 2 (swvl2)m³ m⁻³6-hourly instantaneous
Total column water (tcw)kg m⁻²Daily mean (UTC day)

AIFS-ENS v2

Class: ensemble Grid: 721 × 1440 (0.25°) Ensemble members: 51 Forecast length: 50 days

VariableUnitsTemporal aggregation
2 m dewpoint temperature (2d)K6-hourly instantaneous
Surface pressure (sp)Pa6-hourly instantaneous
Soil temperature level 1 (stl1)K6-hourly instantaneous
Soil temperature level 2 (stl2)K6-hourly instantaneous
Volumetric soil water layer 1 (swvl1)m³ m⁻³6-hourly instantaneous
Volumetric soil water layer 2 (swvl2)m³ m⁻³6-hourly instantaneous
Total column water (tcw)kg m⁻²Daily mean (UTC day)

FGN

Class: ensemble Grid: 721 × 1440 (0.25°) Ensemble members: 51 Forecast length: 50 days

VariableUnitsTemporal aggregation
Sea surface temperature (sst)K6-hourly instantaneous

Aurora 1.5 ENS

Class: ensemble Grid: 720 × 1440 (0.25°) Ensemble members: 51 Forecast length: 50 days

VariableUnitsTemporal aggregation
2 m dewpoint temperature (2d)K6-hourly instantaneous
Surface pressure (sp)Pa6-hourly instantaneous
Soil temperature level 1 (stl1)K6-hourly instantaneous
Volumetric soil water layer 1 (swvl1)m³ m⁻³6-hourly instantaneous
Daily maximum 2 m temperature (2t_max)KDaily maximum (UTC day)
Daily minimum 2 m temperature (2t_min)KDaily minimum (UTC day)

NeuralGCM

NeuralGCM is stored on a coarse grid with its own variable naming. The full save set is listed below.

Class: ensemble Grid: 64 × 128 (≈2.8°) Ensemble members: 51 Forecast length: 50 days

VariableUnitsTemporal aggregation
Temperature at 1000 hPa (temperature_1000)K6-hourly instantaneous
Temperature at 925 hPa (temperature_925)K6-hourly instantaneous
Specific humidity at 1000 hPa (specific_humidity_1000)kg kg⁻¹6-hourly instantaneous
Geopotential at 200 hPa (geopotential_200_avg)m² s⁻²Daily mean (UTC day)
Geopotential at 500 hPa (geopotential_500_avg)m² s⁻²Daily mean (UTC day)
Geopotential at 850 hPa (geopotential_850_avg)m² s⁻²Daily mean (UTC day)
Specific humidity at 925 hPa (specific_humidity_925_avg)kg kg⁻¹Daily mean (UTC day)
Specific humidity at 850 hPa (specific_humidity_850_avg)kg kg⁻¹Daily mean (UTC day)
Temperature at 200 hPa (temperature_200_avg)KDaily mean (UTC day)
Temperature at 500 hPa (temperature_500_avg)KDaily mean (UTC day)
Temperature at 850 hPa (temperature_850_avg)KDaily mean (UTC day)
U wind component at 50 hPa (u_component_of_wind_50_avg)m s⁻¹Daily mean (UTC day)
U wind component at 200 hPa (u_component_of_wind_200_avg)m s⁻¹Daily mean (UTC day)
U wind component at 500 hPa (u_component_of_wind_500_avg)m s⁻¹Daily mean (UTC day)
U wind component at 850 hPa (u_component_of_wind_850_avg)m s⁻¹Daily mean (UTC day)
V wind component at 50 hPa (v_component_of_wind_50_avg)m s⁻¹Daily mean (UTC day)
V wind component at 200 hPa (v_component_of_wind_200_avg)m s⁻¹Daily mean (UTC day)
V wind component at 500 hPa (v_component_of_wind_500_avg)m s⁻¹Daily mean (UTC day)
V wind component at 850 hPa (v_component_of_wind_850_avg)m s⁻¹Daily mean (UTC day)
Total precipitation, 24 h accumulation (total_precipitation_24hr)m24 h accumulation (UTC day)

Variable selection

Variables were selected for their utility in human-centered benchmarking of high-impact phenomena (heat waves, cold air outbreaks, heavy precipitation), for general geoscience research use, and for driver analysis of the large-scale atmospheric patterns that condition those phenomena — blocking and Rossby wave activity (z_500, z_200), jet position and storm tracks (u_200, v_200, u_850, v_850), stratospheric variability and its surface coupling (u_50, v_50), and lower-tropospheric moisture and thermal structure (q_1000q_850, t_1000t_850).

Temporal aggregation follows the same logic: fields used for extremes and threshold exceedance are kept 6-hourly, while circulation diagnostics are stored as daily means.

Archive structure & format

3.10 PiB
Daily Initializations — full 2000–2025 archive

Forecasts are stored as one zarr store per model per initialization: forecasts_<model>/init_YYYYMMDDT00.zarr. Dimensions are (time, [number,] prediction_timedelta, lat, lon) for 6-hourly fields and prediction_timedelta_daily for daily aggregates; number is the ensemble-member dimension. All data are float32.

Daily aggregates are given by: means of the four 6-hourly values in the UTC calendar day; precipitation is the daily accumulation; 2 m temperature max/min are daily extremes.