Right now
Next three days
daylight hoursWhere the uncertainty comes from
Dive log
Import history
one line per dive
Paste past dives, one per line — date, time, site, visibility. Order does not matter and
the parser is lenient: 2025-06-14, 09:00, Buffels, 6
and 2025/07/02 14h bakoven 3.5 both work.
Each dive gets its own ten days of archived swell and wind pulled from ERA5, so the fit is reproducible.
Raw values
3-hourly · ensemble median · SASTSite profile
These are my guesses about your sites, and they are the weakest part of the model. Move them until the output matches dives you remember, or log dives and let the fit move the ones it can actually see. Anything you change here becomes the prior the fit starts from.