Can this algorithms work for a specific city?

The granules size is 12,000 km by 12,000 km, I would love to try using what I learned in this competition for a specific city. Wouldn’t the fact that a granule size covers multiples cities even states make the predictions of pm 2.5 in my desired city inaccurate? Wouldn’t it generalize to multiple cities? When realistically I know (I can measure it) the amount of pm 2.5 in my city is extremely different than the neighbor cities.

Thanks in advance!

Hi @gziz! I’m not sure I understand your question. Some of the granules definitely cover large swaths, but provide data at high resolutions. For example, MODIS is provided at 1km pixel resolution and TROPOMI is provided at 5.5km x 3.5km resolution. A model should be able to generalize to other cities by making use of the nearest/most relevant points provided in the satellite data. Does that answer your question?