I thought I would toss this out as an idea in case anyone was interested in burning several hours on a might work / might not.
So, using Openclaw (if you are not familiar with it, Google / watch the MeTube). I have it using my security cameras to image recognize snow and how much snow we have gotten. It has a camera that is looking at a specific area that the snow blower clears but using image recognition it can measure if the area has snow and approximately how much / close enough for what I need anyway. I have this camera setup sensitive enough that it will trigger when snowing so that will tell Openclaw to start keeping an eye on things. Once it sees we have at least one inch of snow it uses Bluestar App Emulator and runs the job I have for cleaning the driveway via the Yarbo app. Much more accurate than the Yarbo scheduled weather feature.
Have fun if you decide to try out.
My setup in case you want to know:
Clawbot on Ubuntu using Windows as a node, but you could just run on Windows (I have other reasons for this configuration).
Opus 4.6 for the default AI
Google Flash for Image Recognition
Bluestar for App Emulator
Blue Iris for Camera - I am sure anything would work so long as Openclaw can access
Have fun if you decide to try out.