AIWeather126/CN-AEBench

Notice: This repository is currently public, and we will remove the “Access requests” option immediately upon paper acceptance. This setting is implemented solely because, unlike Science Data Bank, Hugging Face does not offer draft or release protection statuses.

If you are interested in our work, please contact AI_weather@126.com to obtain early access.

CN-AEBench is a comprehensive multi-source atmospheric & environmental dataset integrating ground meteorological observations, environmental monitoring, and ECMWF NWP IFS forecast data.

For more information, please visit the official repository: https://github.com/AIWeather126/CN-AEBench

Data for timeliness tasks are automatically uploaded to HuggingFace daily around 12:30 and 21:30. Please note that due to copyright restrictions on the raw data, there is a delay of approximately 2 days.

---

Detailed L3 Description

CN-AEBench L3 data is specifically designed for building end-to-end intelligent forecasting models and is currently at version 1.0.0.

Version History

Version	Release Type	Time Span	Resolution	Data	Method	Availability
0.2.0-alpha.1	Internal Test	2023090100-2025073123	1h	Partial atmospheric elements + 7 major environmental elements + Partial NWP variables	Simple fusion methods (e.g., IDW)	Private (Contact for access)
0.6.0-beta.1	Public Beta	2023090100-2025083123	1h	Full atmospheric elements + 7 major environmental elements + Partial NWP variables	Fusion methods (e.g., IDW, IDW+LightGBM)	Public
1.0.0-rc1	Public RC	2023090100-2025103123	1h	Full atmospheric elements + 7 major environmental elements + Full NWP variables	--	Disabled after 1.0.0 release
1.0.0 (Latest)	Public	2023090100-2025103123	1h	Full atmospheric elements + 7 major environmental elements + Full NWP variables	As per the fusion method in the paper	Public

* For checkpoints or subsets of the data, please contact us via email.

To ensure benchmark stability and comparability of research results, we release new versions only when significant improvements are made to accommodate new weather and environmental changes, with clear version numbering.

Variable Information

Static Descriptive Information Table

No.	Variable Name	Unit	Description
1	elevation	m	Station elevation
2	lon	degree	Station longitude
3	lat	degree	Station latitude
4	station_province	--	Province where station is located
5	station_city	--	City where station is located
6	station_id	--	Station identifier
7	type	--	Land use type at station location
8	ndvi	(-1 ~ 1)	ndvi value at station location

Multi-source Variable Description Table

No.	Variable Name	Unit	Description
1	ws_2min	m/s	2-minute average wind speed
2	ws_10min	m/s	10-minute average wind speed
3	wd_2min	degree	2-minute average wind direction
4	wd_10min	degree	10-minute average wind direction
5	wd_instant	degree	Instantaneous wind direction
6	ws_instant	m/s	Instantaneous wind speed
7	vis	m	Horizontal visibility
8	t	°C	Air temperature
9	dt	°C	Dew point temperature
10	precipitation	mm	Hourly precipitation
11	rh	%	Relative humidity
12	p	hPa	Atmospheric pressure
13	slp	hPa	Sea level pressure
14	vapor	hPa	Vapor pressure
15	phenomena	--	Weather phenomena
16	ec_vis	m	NWP horizontal visibility
17	ec_sh2	kg/kg	NWP 2m specific humidity
18	ec_t2m	°C	NWP 2m air temperature
19	ec_d2m	°C	NWP 2m dew point temperature
20	ec_sp	hPa	NWP surface pressure
21	ec_msl	hPa	NWP mean sea level pressure
22	ec_u10	m/s	NWP 10m u-component of wind
23	ec_v10	m/s	NWP 10m v-component of wind
24	ec_rh	%	NWP relative humidity (diagnostic variable)
25	ec_ws	m/s	NWP wind speed (diagnostic variable)
26	ec_wd	degree	NWP wind direction (diagnostic variable)
27	ec_cbh	m	NWP cloud base height
28	ec_sf	m of water equivalent	NWP snowfall
29	ec_blh	m	NWP boundary layer height
30	ec_fal	(0 ~ 1)	NWP albedo
31	ec_lcc	(0 ~ 1)	NWP low cloud cover
32	ec_mcc	(0 ~ 1)	NWP medium cloud cover
33	ec_hcc	(0 ~ 1)	NWP high cloud cover
34	ec_tp	m	NWP total precipitation
35	PM2.5	μg/m³	Hourly mean PM2.5 concentration
36	PM10	μg/m³	Hourly mean PM10 concentration
37	SO2	μg/m³	Hourly mean SO2 concentration
38	NO2	μg/m³	Hourly mean NO2 concentration
39	O3	μg/m³	Hourly mean O3 concentration
40	CO	mg/m³	Hourly mean CO concentration
41	AQI	--	Real-time AQI value

---

Detailed L1&L2 Description

1. CN-AEBench-L1 Description

CN-AEBench-L1 contains quality-controlled raw observational data, primarily designed for fundamental research applications including NWP data assimilation and gridding of observational data.

Usage Guidelines

CountryEnv - National Environmental Monitoring Station Data

• Historical Data:
  • Pre-2025.11.01: Batch processed and archived as CountryEnv-L1.parquet
  • Post-2025.11.01: Rolling updates
• Organization: Daily files with naming convention YYYY_MM_dd_HH.parquet
• File Structure:
  • Rows: Individual station records
  • Columns: Environmental parameters (AQI, CO, NO2, O3, PM10, PM2.5, SO2)

ProvinceEnv - Provincial Environmental Monitoring Station Data

• Format: Compressed Parquet files, compatible with pandas
• Naming Convention: ProvinceEnv-L1.parquet
• Processing: Direct pandas DataFrame operations supported

Atmo - Meteorological Observation Data

• Historical Data:
  • Pre-2025.11.01: Batch processed and archived as Atmo-L1.parquet
  • Post-2025.11.01: Rolling updates
• Organization: Daily files with naming convention YYYY_MM_dd_HH.parquet
• File Structure:
  • Rows: Individual station records
  • Columns: Meteorological variables

NWP - Numerical Weather Prediction Data

Raw forecast data are not included in this repository. Users can obtain L1-NWP products directly from our mail (AI_weather@126.com) or ecmwf.int.

2. CN-AEBench-L2 Description

CN-AEBench-L2 builds upon L1 with spatiotemporal alignment, missing data imputation, model data registration, and diagnostic variable computation. It is designed for domain-adaptive pre-training, statistical analysis, event characterization, and sequence interpolation tasks.

Usage Guidelines

CountryEnv - National Environmental Monitoring Station Data

• Historical Data:
  • Pre-2025.11.01: Batch processed and consolidated in CountryEnv-L2.parquet
  • Post-2025.11.01: Rolling updates
• Organization: Daily files with naming convention YYYY_MM_dd_HH.parquet
• File Structure:
  • Rows: Individual station records
  • Columns: Environmental parameters (AQI, CO, NO2, O3, PM10, PM2.5, SO2)

ProvinceEnv - Provincial Environmental Monitoring Station Data

• Format: Compressed Parquet files, compatible with pandas
• Naming Convention: ProvinceEnv-L2.parquet
• Processing: Direct pandas DataFrame operations supported

Atmo - Meteorological Observation Data

• Historical Data:
  • Pre-2025.11.01: Batch processed and consolidated in Atmo-L2.parquet
  • Post-2025.11.01: Rolling updates
• Organization: Daily files with naming convention YYYY_MM_dd_HH.parquet
• File Structure:
  • Rows: Individual station records
  • Columns: Meteorological variables

NWP - Numerical Weather Prediction Data

• Format: Compressed Parquet files, compatible with pandas
• Historical Data:
  • Pre-2025.11.01: Batch processed and consolidated in NWP-L2.parquet
  • Post-2025.11.01: Rolling updates
• Organization: Daily files with naming convention nwp_YYYYMMdd.parquet
• Processing: Direct pandas DataFrame operations supported