Start a community thread about "what are good weather-based features" – for example, should we combine airmass, seeing, cloud cover, and sky brightness into a single SNR number and use this? Separating features where they lead to different kinds of behavior is desireable (i.e. if skybrightness makes you change filter, while cloudiness makes you move the telescope .. and you have a reason to think you'd act differently based on these different pieces of information .. SNR, in different filters, for example, could still drive you to the same conclusion as having these pieces of information separately). https://community.lsst.org/t/what-weather-based-features-should-the-scheduler-consider/684