We only use nationally observed data to derive ET. We used to use forecasted data for ET and allowed PWS observed data for ET but the forecasted waterings were always widely different then observed waterings since we would use the “observed” ET from the PWS to decide if we were watering or not for that day. The biggest discrepancy was wind observed from PWS stations which was usually much lower than wind from properly setup national stations. This would have us watering about half as much as was recommended since the recorded ET from PWS stations was usually much less. There is a thread about this exact topic somewhere in the community history 
Went back and forth with this and the approach we’ve adopted seems to be the most predictable.
