I've thought about this, it would be a lot of work but definitely doable.

You have cameras set up in different places at fixed locations, looking upward. Each plane crossing the field of view is analyzed for the opacity and size of the trail over time. Sightings are correlated with flight databases to get the plane type which yhou can then use the apparent optical size to get its altitude reasonably accurately. Then you take tons of data.

Once you have tons of data in different locations you look at the distributions of size, opacity, length of time (perhaps use a combined metric) with different altitudes and atmospheric conditions. If the chemtrail hypothesis is correct, you'd expect to see a separate population of data points from the normal contrails (ie longer coverage times for the same altitude / humidity conditions)