Blog/

What a GeoTIFF is, and why gridded AOD data ships in one

Empty desert highway and power lines stretching toward the horizon under a wide open sky.

If you've spent any time pulling data into QGIS or ArcGIS from a satellite product, you've run into GeoTIFF. It's a regular TIFF image with georeferencing baked into the file header: a coordinate reference system, an affine transform that ties each pixel to a real-world coordinate, and usually a no-data value so you're not treating ocean or cloud-masked pixels as zero pollution. Open it in any GIS desktop or a Python script with rasterio and it just works, no separate world file, no guessing at projection.

For gridded aerosol optical depth (AOD) data, that matters more than it sounds like it should.

What a GeoTIFF actually carries for AOD

AOD is a column measurement: how much light gets scattered or absorbed by particles between the sensor and the surface. It's a proxy for particulate loading in that column. Turning it into a surface-level pollutant concentration takes extra work: humidity corrections, assumptions about aerosol type, boundary layer height. AOD comes off a satellite pass as a swath, a long curved strip tied to the orbit track, not a tidy rectangle you can drop straight into a project. To make that swath usable for your region, someone has to reproject it, resample it onto a regular grid, and clip it to your area of interest. A GeoTIFF is the container that holds the result of that work: one band (or one file per day) of AOD values, one pixel per grid cell, each pixel carrying a float value and a known location.

That's the part that saves you time. You're not reverse-engineering a swath file's scan geometry or writing your own gridding routine before you can even look at a haze event. You load the raster, and the pixel at your station's coordinates is already the pixel you want.

Why rasters beat point lists for filling network gaps

A fixed-station network gives you excellent data at a handful of points and nothing in between. If your two closest monitors are 40 km apart and a dust plume or wildfire smoke layer is tracking through that gap, your network won't see it until it's already overhead at one station or the other. A gridded raster covers the whole space, so you can watch a haze event build at a grid cell with no monitor on it, days or hours before it reaches a station that does report.

This is also why a CSV of point values alone doesn't replace the raster. A per-cell time series is useful for trend work, for comparing one grid cell's AOD against a nearby monitor's PM2.5 reading over a season, but it only tells you about the cells you've already decided to track. The raster is what lets you scan the whole region and notice a cell you weren't watching start to spike. Most workflows end up using both: the GeoTIFF for the spatial picture, the per-cell CSV for the time-series analysis once you've identified which cells matter.

Where this fits next to your existing data

Coarse-resolution AOD, at 250 meters and coarser, carries real limitations around cloud cover, surface brightness, and the conversion from column AOD to surface-level concentration. It works as a between-monitor proxy, flagging where to look before a fixed station catches a plume, rather than standing in as a regulatory PM2.5 reading. Analysts who already know this distinction get the most out of it: they expect the raster to cover the space a monitor can't reach, and they leave the regulatory reporting to the station itself.

If you're already pulling multispectral satellite products for other work, a daily AOD raster delivered as GeoTIFF with a matching per-cell CSV slots into that pipeline without much friction. Air Quality Proxy builds exactly that: a daily gridded raster over your region plus the time series underneath it, so the gap between your stations stops being a blind spot.

Worth a look the next time a haze event forms somewhere your network can't see it.

Get started

← Back to the blog