Detecting cold pools with a hyper-local weather network

These plots are very difficult for me to read. I'm not sure why the color maps span 90F+ when discussing nighttime radiative cooling on the order of 12F. I'm also not sure how a regression line with a negative slope constitutes a "weak" temperature inversion -- it's not an inversion at all.

It would be interesting to see the station data integrated with an urban land cover map. Surface air temperature is sensitive to the locations of buildings, roads, soil type/water/ground cover, etc., not just elevation.