⇐ ⇒

[CF-metadata] interpreting time and vertical axis

From: Nan Galbraith <ngalbraith>
Date: Fri, 12 Nov 2010 09:41:04 -0500

Hello Yilmaz -

I was really hoping someone else would answer this, because
there are others on this list who are more familiar with COARDS,
so I'm replying based on what we do with CF for current velocity
and other ocean/met data.

Your data file - which I assume you're reading, not writing - should
have some information about how the data was collected; the answers
to your questions have more to do with that than with the CF spec.

Specifically, if this data is from a profiler (ADCP or other) then the
assumption is that bin averaging has taken place. If this is a time
series of data from current meters, then the Z axis reflects point
measurements. (Note: this is one of the problems with the point
obs convention, lumping together 2d time series and profiles)

Regarding the time variable, I don't think CF provides enough info
to adequately describe the timing of instrument data, at least for
sensors with any sample scheme other than spot measurements.
But that's for another thread.

More in-line below ...

On 11/9/10 9:59 AM, Yilmaz Arslanoglu wrote:
> I had a number of questions regarding interpretation of time and
> vertical axis in a given NetCDF file
> (vertical especially for ocean current data)
>
> 1) What should be the default interpretation of the timestamps on a
> given time axis
> when there is no explicit CF convention based information in a given
> file (in COARDS format we can say)
>
> Should they be taken the centers of time intervals, or the starting
> point of the intervals ?
The default, I believe, if there is no cell method provided, is that
this is an instantaneous sample; and, for that reason, it seems most
logical to assume that time represents the center of the sample.
> 2) Should the corresponding values be assumed to be the average of
> values on that interval or measurements
> at that specific timestamp?
The latter, I think. I don't see this specifically addressed in the CF
docs, though.
> 3) For vertical layer if I have the following depths: [0.0, 10.0,
> 50.0], should this be interpreted as three layers with
> boundaries [0.0, 10.0] [10.0, 50.0] [50.0, +inf] ?
Instrument dependent, and not part of the standard, as far as I know.
> 4) And finally, is it usual to apply interpolation in these axes
> (time, vertical) or just locating the closest time record
> and vertical layer is the general approach since they are in general
> the averaged values on that time interval or on that vertical layer?
This really depends on the data generator. In my lab, data is carefully
matched to the time, but depth is almost impossible to specify with any
accuracy, because we work with moored surface buoys. So again, this
is not something that's part of the CF standard, or COARDS, as far as
I know.

Regards -
Nan Galbraith

-- 
*******************************************************
* Nan Galbraith                        (508) 289-2444 *
* Upper Ocean Processes Group            Mail Stop 29 *
* Woods Hole Oceanographic Institution                *
* Woods Hole, MA 02543                                *
*******************************************************
Received on Fri Nov 12 2010 - 07:41:04 GMT

This archive was generated by hypermail 2.3.0 : Tue Sep 13 2022 - 23:02:41 BST

⇐ ⇒