PM2.5 calculation

I have been using the Sentinel 2A Aerosol Optical Depth band to follow particulate emissions but only for relative differences in the landscape. I am a retired research plant physiologist from the USDA and recognize I am far afield from my agricultural background. I would like to have quantitative values of PM 2.5 and/or PM10 calculated for each pixel. The 10m resolution of Sentinel 2 AOD is just what I need if it can be calibrated to ground-based PM2.5. Unfortunately there is a very weak correlation between AOD and ground-based PM2.5 using Sentinel 2a images. I have searched all the Sentinel toolboxes as well as tutorials without success. I realize this is processing after image collection. There are ample peer-reviewed articles modeling ground-level PM2.5 using various satellite platforms but well beyond this ‘citizen scientist’ skills.

By comparison, the MODIS satellite imagery has at least 2 websites that provide calculated PM2.5:
https://arset.gsfc.nasa.gov/airquality/python-scripts-aerosol-data-sets-merra-modis-and-omi

ftp://ftp.star.nesdis.noaa.gov/pub/smcd/hzhang/PM25SAT/

I have demonstrated that the MODIS data of PM2.5 ug/m3 is highly correlated to ground-based PM2.5 when there are continuous winds during the day for my region of the mid-Atlantic USA.
Unfortunately, the MODIS spatial resolution is in kilometers and not capable of identifying point-sources of PM emission–my particular interest.

I cannot find a source of such a sponsored model for the ESA and Sentinel 2 data. The MODIS sites make clear they are calculated values and the user takes all responsibility for interpretation, which is fine for my use.

Are there websites that I could access for the calculation of particulate concentration (PM2.5, PM10 or aerosol in general) using Sentinel 2 imagery?

Thank you very much for taking time to respond to my naïve question.

Sincerely,

D. Michael Glenn PhD
Retired Plant Physiologist USDA-ARS

I’m not sure if Sentinel-2 provides the appropriate data for doing this.
I think the Sentinel-5P - Missions (esa.int) could be worth a look. Here I have also no references of how it can be done.

Update: Now I see that the resolution doesn’t fit your needs too.

@davidglenn I have a couple of possible ideas for you:

  1. Would Sentinel-3 work a for you? It has 300m resolution across all of its bands, so you might be able to get a bit better resolution than MODIS which has a mix of bands between 250m and 1000m.
  2. Do you necessarily need the data to be fully correlated? If all you need is point sources, you can simply look at data the image and find the higher points. .
  3. Use multiple satellites. You could take an image from Sentinel-2, and down sample it to MODIS resolution, and then compare the measurements to a image taken by MODIS around the same time, and use this to calibrate the Sentinel-2 data. It won’t be perfect, but it might be enough.

Keep in mind I am just a hobbyist in the remote sensing world, so I may be way off course with these ideas.

Sentinel 5 P has an aerosol index. While it has a large spatial resolution it would be useful if I could download images as geotiff’s. I have never been able to save/download those images as geotiff to use in QGIS. I have tried resampling to a smaller image and SNAP will not export the image as geotiff. I’ve searched the forum’s for hints and haven’t found any that work. It appears, to me, that there are 3 distinct bands: longitude, latitude, and aerosol index. I can’t find how to merge them into a single image. Thanks for the suggestion, it pushed me back to trying to download it but with no success. Sentinel 2 downloads right in Copernicus but apparently Sentinel 5 does not.

I examined Sentinel 3 and could not find an aerosol optical density/thickness band, per se. I have learned that these AOD bands, as expected, are reflecting a very wide range of aerosol ingredients. In my experience working with measured PM2.5 from state air quality monitors in WV, MD and VA, the PM2.5 concentration is strongly and negatively correlated with wind speed–no surprise there. Yet AOD pixel values are not correlated with wind speed So there are some very complex and physics-based algorithms underlying the calculation of PM2.5 from MODIS data at these MODIS websites:

https://arset.gsfc.nasa.gov/airquality/python-scripts-aerosol-data-sets-merra-modis-and-omi

ftp://ftp.star.nesdis.noaa.gov/pub/smcd/hzhang/PM25SAT/

The aerosol index from Sentinel 5 utilizes UV bands. It would be useful if I could figure out how to download it even if it has km spatial resolution (see response to marpet). The other Sentinel platforms don’t measure the UV bands. As important and deleterious as PM2.5 is to human health, I’m surprised and frustrated that this valuable satellite technology, in particular Sentinel 2, hasn’t been moved from an academic interest to more of a technology transfer. I may have hit a stonewall at this point. But thanks for your suggestions.