@pybonacci That makes sense. Thanks.

(I had sussed the accumulation from the grib message attributes, which are effectively the info in the idx, but I hadn’t worked out the pattern. )

I am left intrigued by the logic of this particular strategy of information production. Given these are hourly forecast fields, it would make sense for “the last one hour” in each, or the “last n hours”, but a variable number out to each six hour cycle makes little sense to me But there will be a reason ...

Is there anyone else out there using NOAA #GFS data as described here https://www.nco.ncep.noaa.gov/pmb/products/gfs/gfs.t00z.pgrb2.0p25.f003.shtml who can explain what is going on with the accumulated precip fields 596 and 597?

It appears that for the 596 message, at the 3h forecast (f003), it’s a 0,3 hour accumulation, at 6h (f006), it’s a 0,6 hour accumulation, and at 9h forecast (f009) it’s back to a 3h accumulation 6,9.

Is this behaviour documented anywhere, and why?

Exciting new opportunity for #PhD students who want to work within our new “Advancing the Frontiers of Earth System Prediction” (AFESP) programme.

Potential projects will range across fundamental new research in the physical, mathematical and computational sciences required to address next-generation challenges in Earth system prediction and operational weather and climate forecasting. PhD students will be trained and embedded within the overall AFESP programme.

https://www.reading.ac.uk/graduate-school/funding/phd-studentships/earth-system-prediction-doctoral-training-programme