GeoTIFF vs NetCDF for gridded air quality data: which format fits your GIS

Two formats, two different jobs
If you've ever gotten a file from a modeling group or a satellite data provider and spent the first twenty minutes figuring out how to even open it, you already know the GeoTIFF vs NetCDF question isn't academic. It decides whether a gridded aerosol or PM layer shows up in your raster catalog this afternoon or sits in a "need to convert" folder until someone has time.
GeoTIFF is a flat, single-purpose format. One grid, one timestamp (or one band per timestamp if you stack them), coordinate reference system and extent baked right into the header. Drop it into ArcGIS Pro, QGIS, or a tile server and it renders. No driver install, no library version chase. That's the whole appeal: it behaves like every other raster layer already in your GIS, because structurally it is one.
NetCDF comes out of the atmospheric science world, and it shows. It's built to carry multiple dimensions in one file: latitude, longitude, time, sometimes vertical level, sometimes multiple variables like AOD and cloud fraction in the same container. CMAQ output, WRF-Chem runs, and a lot of satellite AOD archives (MODIS, VIIRS) ship as NetCDF because a modeler wants the whole month in one place, indexed by time, without juggling 30 separate files. That's a real advantage if you're running the model or doing the retrieval. It's a real headache if you just need Tuesday's grid on a map next to your fixed-station readings, because now you need xarray or rioxarray or a gdal_translate step to pull out one slice and give it a proper spatial reference your GIS will accept.
What this means when a state agency has to use the data, not just archive it
Here's the practical split: NetCDF is the right format for an archive, a modeling input, or a dataset that lives in its native scientific home for years. GeoTIFF is the right format for something that has to slot into an operational GIS workflow on a daily cadence, get published to an internal map service, and sit next to monitor station points without an analyst doing a conversion pass every morning.
The real friction shows up when a modeling group exports a gridded product in whatever format was fastest to generate on their end, and the file lands in the GIS shop's inbox needing a conversion pass before anyone can use it. A daily aerosol layer that arrives as NetCDF with CF-compliant metadata is technically correct and practically a second job.
There's also the time-series question. NetCDF handles time natively inside the file, which is elegant for a modeler querying a cube. But a lot of agency reporting doesn't need the cube. It needs "what was this grid cell doing over the last 90 days," which is a table, not a dimension. A GeoTIFF per day plus a separate CSV time series per grid cell covers both needs without forcing every user to learn multidimensional array slicing just to pull a trend line for a board memo.
Picking based on who opens the file
If the end user is a modeler writing Python against the data, NetCDF's dimensional structure earns its keep. If the end user is a GIS analyst filling the gap between two fixed monitors and needs the layer to overlay cleanly with exceedance zones, zoning boundaries, and station buffers in the software already on the desktop, GeoTIFF is the format that doesn't add a step. Delivery to a state agency almost always means the second case: someone needs the grid in the map, not in a notebook.
That's the reasoning behind how we package the daily aerosol optical depth layer: GeoTIFF for the raster, CSV for the per-cell time series, so it opens where your other agency layers already live instead of waiting on a conversion script. If you're trying to see what's forming in the space between your stations before the next reading comes in, that combination is worth a look.