[CF-metadata] NMAT

From: Jonathan Gregory <j.m.gregory>
Date: Sun, 17 Dec 2006 09:27:58 +0000

Dear Julian

> if this was to be used for data discovery and
> automatic processing we would have to introduce an element of
> standardisation in how we put the name of the climatology into the
> source attribute.

This question can certainly be revisited when it is clearer what to do. It
may be that standardisation for particular projects is adequate.

> One of the suggestions from my colleagues here is that the climatology
> and its uncertainties should always be appended to any anomaly field and
> its uncertainties and that the absence of a climatology field in an
> anomalies file should be treated as an error.

That sounds like a good practical approach for those situations where it can
be done without making the file too large.

> We have also realised that air_temperature_difference_from_climatology
> is not necessarily the same as air_temperature_anomaly.
>
> air_temperature_difference_from_climatology is well suited to a system
> such as MOHMAT where we calculate an actual and then remove a
> climatology. However, CRUTEM3 calculates gridded anomalies from station
> anomalies which are the station actuals with a station climatology
> removed. In the latter case a "climatology" is unknown and so CRUTEM3
> would be more accurately described as air_temperature_anomaly.

OK. Could you suggest an alternative? "anomaly" alone is not clear enough,
I think. Your anomaly is a difference in some kind of way from a time-mean,
although I see it is not a simple as subtracting a mean.

> We have only just begun to move our historic datasets away from PP. ;;-)

Good for you!

To repeat (so we don't lose the point): Apart from deciding what the stdname
modifier is called, we also need to introduce "(night)" for time-means in
cell_methods.

Best wishes

Jonathan
Received on Sun Dec 17 2006 - 02:27:58 GMT