[CF-metadata] statistic indices

From: Jonathan Gregory <j.m.gregory>
Date: Sun, 28 Oct 2007 10:41:20 +0000

Dear Heinke

> the percentile case is not so easy.
I agree - it is hard!

> > You mean the threshold is a percentile of a climatology, and then you use the
> > threshold to evaluate the probability of exceedance in a different climate.
> >
> I think 'evaluate the probability of exceedance' is a further step. The
> data are
> 'percent' of time per time period or number and not the 'propabiltiy'.
> ... we can't be sure that the number of samples is
> sufficient to create a probability.

> > I think it's quite difficult to describe this systematically and economically.
> >
> The percentile is not only a value for example 0 degC. It is not
> enough to say the percentile is a threshold with value '90 th' and period
> 1961-1990.' It is a field. The percentile depends on lat and lon.
> What can we do ?

In the general case, the air_temperature_threshold is simply a given field, and
the quantity you want is defined as the fraction of occasions which exceeded
this threshold field, which is a function of lat and lon. In that case the
air_temperature_threshold would not be given as a coordinate value; some other
method would be needed to point to the air_temperature_threshold variable. The
description of the statistic itself would be simpler, while the
air_temperature_threshold field would describe itself. That would be more
convenient, because the field would have its own coordinate variables and so on
and these could be used to specify the time-period, the percentile, etc.

Would that be a possible approach? (Is it clear what I mean?)

I'm still interested in what you think of this:

> > Let me ask a hypothetical question: how would
> > you identify these various indices etc. in netCDF without trying to use CF
> > standard names? Let us consider only those quantities which you actually have
> > a present need for, as I am sure we can imagine indefinitely complicated cases
> > without trying hard!

Best wishes

Jonathan
Received on Sun Oct 28 2007 - 04:41:20 GMT