How to choose grid resolution for a daily AOD raster between fixed monitors

Pick the wrong grid cell size and you either waste storage on false precision or smear real haze gradients into noise. Neither helps when you're trying to show a regional office where a plume is building between the two stations that caught it first.
Start from the sensor, not your station spacing
Coarse-resolution multispectral satellite AOD products carry a native pixel size, typically 250 m to 1 km depending on the instrument and retrieval algorithm. That's the floor. Resampling a 500 m retrieval onto a 50 m grid doesn't manufacture new information about aerosol loading. It just interpolates between the same underlying values and makes the raster look sharper than the physics behind it.
So the first decision isn't "what resolution do I want." It's "what resolution did the retrieval actually produce." Check the product's documentation for nominal pixel size before you pick a GeoTIFF grid cell size. If you grid finer than the sensor's native resolution, say so in your metadata, because anyone pulling that raster into a GIS later will assume the pixel count means something it doesn't.
Match cell size to the gap you're filling, not the whole region
Fixed monitors in most networks sit anywhere from 5 km apart in dense urban cores to 30+ km apart in rural coverage areas. A proxy raster covers that in-between space with pixel-level estimates, sitting alongside the point measurements each monitor gives you at its fixed location. A grid cell of 500 m to 1 km usually does that job well for most inter-station gaps. You get enough cells across a 15 km span to see a gradient forming, without pretending the satellite pass resolved street-level variation it never measured.
If your station network is sparser, say 25 to 40 km between sites, you can afford a coarser grid, maybe 1 to 2 km, and still catch a haze event trending toward a monitor days before it crosses the threshold. The CSV time series per grid cell is where this matters most in practice: fewer, well-chosen cells give you a cleaner trend line per location, and a smaller file to pull into whatever time-series tool your team already uses. A finer grid multiplies cell count and CSV rows without adding real signal, and that's a maintenance cost someone on your team pays every day the refresh lands.
Coverage, cadence, and what between-monitor coverage buys you
A daily refresh cadence is the other half of this decision. Coarse-resolution platforms revisit a given swath on a roughly daily cycle, cloud cover permitting. That cadence sets a natural limit on how much value finer spatial resolution adds, because you're not resolving sub-daily movement regardless of pixel size. If you need sub-daily tracking of a specific plume, this product type isn't the right tool for that question. What it's built for is daily, between-station coverage: a grid you can check each morning against your two or three nearest fixed stations to see whether something is building in the gap before it shows up on either instrument.
A reasonable default for most regional networks: grid at or near the retrieval's native pixel size, keep cell boundaries aligned to a standard projection your GIS already uses, and resist the urge to downsample just because a finer grid looks more impressive in a report. The goal is a raster that tells you where to look next, not one that implies monitor-grade precision it can't deliver.
If you're weighing this for your own network, Air Quality Proxy delivers that daily AOD raster already gridded and ready to drop into the workflow, so you're deciding on cell size for your gaps rather than building the pipeline from scratch.
Worth a look if filling the space between your stations is the problem on your desk this quarter.