There are two aspects to the weather model: clouds and seeing.
For the clouds, you can take a look at what I put together for Rubin/LSST simulations here. These can then be turned into a reduction in t_eff (or a fraction of exposures rendered useless) using statistics from this tech note. In practice, for this exercise, I think this is far more elaborate than is needed, and also for the wrong location (Cerro Tololo, Chile, rather than Stone Edge Observatory).
You can, however, get weather data for locations in the U. S. from the NOAA here, with documentation here. I was able to download a CSV of the data for the past couple of years from the Santa Rosa airport weather station (which seems to be the one closest to SEO) this morning.
For our purposes, I think it's good enough to approximate all time as either "useless" (SCT, BKN, OVC anywhere in the "HourlySkyConditions" column, maybe) or "perfect" (anything else). The simplest thing to do, I think, would be to download as much weather data as we can from NOAA (the site claims to go back to 2000, but when I asked for all of that at once the request failed. Doing just a few recent years worked) and either replay it as is, or take statistics by month and use that to randomly generate as much data as we want (good enough for this project).
There are two aspects to the weather model: clouds and seeing. For the clouds, you can take a look at what I put together for Rubin/LSST simulations here. These can then be turned into a reduction in t_eff (or a fraction of exposures rendered useless) using statistics from this tech note. In practice, for this exercise, I think this is far more elaborate than is needed, and also for the wrong location (Cerro Tololo, Chile, rather than Stone Edge Observatory).
You can, however, get weather data for locations in the U. S. from the NOAA here, with documentation here. I was able to download a CSV of the data for the past couple of years from the Santa Rosa airport weather station (which seems to be the one closest to SEO) this morning.
For our purposes, I think it's good enough to approximate all time as either "useless" (SCT, BKN, OVC anywhere in the "HourlySkyConditions" column, maybe) or "perfect" (anything else). The simplest thing to do, I think, would be to download as much weather data as we can from NOAA (the site claims to go back to 2000, but when I asked for all of that at once the request failed. Doing just a few recent years worked) and either replay it as is, or take statistics by month and use that to randomly generate as much data as we want (good enough for this project).
Originally posted by @ehneilsen in https://github.com/deepskies/TelescopePositioningSimulation/discussions/35#discussioncomment-6617758