Okay I think I have an idea how to do that.
First I needed to understand how the %watering actually works. The Zimmerman is simply changing in the end in a sophisticated way the value and does this just before the start of a water event. Regardless the switch: use weather adjustment means that a program takes the %watering into account. So if the %watering is changed manually (but not by a weather adjustment method) the watering times are still adjusted if the value is not 100%.
Hence it is quite simple. Run a cron job that monitors evapotranspiration and changes the %watering via an api call.
e.g. http://ipadress:8080/co?pw=mypassword&o23=77 (sets the value to 77%).
So I think what should work beautifully is the following:
Set the sprinkling system to a regular schedule that waters you garden with say (just an example)
10mm of water every 5 days which refills it to 100%
Everyday, check the evapotranspration up to the last sprinkling event and if it is above 10 mm then water back to the 10 mm level.
If during the 5 days the sum of evaporation was below 10mm we again fill up to 10mm. (only what is really needed)
If levels are getting low but forecast has a high chance of rain in the next 24 hours we wait.
Best would be here to have a rain sensor to check if it has rained.
Finally if it rains we “reset” the schedule and start to count the 5 days again (amount of rain needs to be above 10 mm, if this is the reference).
So we simply need a script in say python that grabs all the bom data and changes the water levels/times according to those rules.
A minor difficulty is that there seems to be a lag from the BOM data (the evapotranspiration for the day before is not available before 3 pm afternoon the following day). So there needs to be some buffer in the fill ups to cover that half day.
One more (maybe stupid) question
When is the best time to water (early morning or later in the evening)
I thought in the morning just after sunset, but not sure where I read that.
Cheers