[CF-metadata] statistic indices

From: Jonathan Gregory <j.m.gregory>
Date: Thu, 6 Sep 2007 17:04:21 +0100

Dear Heinke

Thanks for your email. I am not commenting here on every name, but on
illustrative cases

(1)
> double threshold
> threshold:standard_name = "air_temperature_threshold" ;
> threshold:units = "degC" ;
>
> frost_days_index_per_time_period ->
> number_of_days_with_air_temperature_below_threshold
> threshold = 0.
> cell method = "time: minimum"

> Is this correct ?
Yes.
The use of cell_methods in this way is an extension of current use. In fact
I think I would refine it to be "time: minimum within days" (as suggested
later in my last email). That makes it clearer what the minimum applies to.

> Why do we need special threshold CF names ?
> Can we use air_temperature as threshold standard name ?
It has a slightly different meaning, don't you think? "air_temperature" is
the standard name for a quantity we measure or model, but "air_temperature_
threshold" is a defined parameter.

> air_temperature_threshold is in Kelvin not in Celsius. This should not
> be a problem but it is a little bit confusing. What do you think ?
It could be in either degC or K, since it has a units attribute of its own.

(2)
> warm_spell_days_index_wrt_90th_percentile_of_reference_period ->
> number_of_days_with_mean_air_temperature_above_percentile_and_spell_length_above_threshold
>
> double percentile
> threshold:standard_name = "percentile_function_of_air_temperature" ;
> threshold:units = "percent" ;
>
> double threshold
> threshold:standard_name = "spell_length_threshold" ;
> threshold:units = "day" ;
> data :
> percentile = 90. ;
> threshold = 6 ;
>
> variable attribute e.g. 1961-1990 (climate reference period)
>
> I would prefer the name percentile_function_of_air_temperature instead
> of cumulative_probability_of_air_temperature.
> What do you think ?

I'm not sure that's correct. I would understand "percentile function" to
mean a function which gave a percentile, so a percentile function of air
temperature would be a particular air temperature value. The cumulative
probability is the independent variable i.e.

90th percentile of air temperature = percentile_function (air_temperature
values, cumulative_probability = 90%)

> It is not the maximum. It is the mean temperature.
We can omit "mean" or "maximum" from the standard name if we put it in
the cell_methods as "time: mean|maximum within days".

> The name is very long.
Yes - but it is informative!

(3)
> growing_season_length_index -> growing_season_duration
>
> Counted are the number of days per calendar year between the first
> occurrence of at least 6 consecutive days where the daily mean
> temperature is above 5 degree Celsius and the first occurrence of at
> least 6 consecutive days after 1st of July where the daily mean
> temperature is below 5 degree Celsius.

This is very detailed, and I guess it is very specific and other people will
use different definitions. I wonder whether it is really useful to have a
standard name for a quantity with such a complicated definition involving
many choices.

(4)
> What do you think about number_of_frost_days_with_no_snow_cover ?
Since we are not using "frost days" elsewhere, I would suggest
number_of_days_with_air_temperature_below_threshold_and_zero_surface_snow_amount

I think it is still worth considering the alternative approach for these
quantities with one or two thresholds of simple standard names such as
number_of_days and fraction_of_days, and rely on the coordinate variables
alone to indicate the conditions that apply.

Best wishes

Jonathan
Received on Thu Sep 06 2007 - 10:04:21 BST