Question about pollutant units in the compact H5 dataset

[Vinci-ui]

Hi, I'm Vinci, thanks for the great benchmark and dataset! I have a question about the units of the 6 pollutants in the compact H5 files.

I understand from the README that metrics are computed in "original concentration units" after inverse normalization. However, when I checked the OpenAQ API parameter definitions, I noticed that each gaseous pollutant has two parameter entries — one in µg/m³ and one in ppm:

ID	name	units	displayName
3	o3	µg/m³	O₃ mass
4	co	µg/m³	CO mass
5	no2	µg/m³	NO₂ mass
6	so2	µg/m³	SO₂ mass
7	no2	ppm	NO₂
8	co	ppm	CO
9	so2	ppm	SO₂
10	o3	ppm	O₃

Since the preprocessing script (convert_csv_to_h5.py) matches by parameter name rather than ID, both unit variants for the same pollutant (e.g., no2 from ID=5 and ID=7) appear to be merged into one channel. I also noticed the script only reads the value column, not unit, from the raw OpenAQ CSVs.

When I analyzed the data by geographic region, I found that US stations have CO values around ~0.25 while European stations have values around ~317, consistent with ppm vs µg/m³. This suggests the gaseous pollutant channels in the H5 files contain mixed units depending on the reporting country.

Could you confirm:

1. Is this mixed-unit behavior expected / by design?
2. For the scaler.csv statistics (e.g., CO mean = 228.46), were these computed on the raw mixed-unit values?
3. Are there any plans to provide unit-normalized versions of the dataset?

Thanks for your time, and again for the excellent work on AirQualityBench!