[CF-metadata] Recording "day of year on which something happens"

John Helly hellyj at ucsd.edu
Thu Mar 16 13:36:06 MDT 2017


Sorry to jump in here but isn't this just the Julian day?

J.


On 3/16/17 8:24 AM, Nan Galbraith wrote:
> I agree that there's a lot of interest, and I have 2 questions.
>
> To make the data most useful, shouldn't the time coordinate variable be
> Jan 1, and shouldn't the 'days since' (data) variable represent the 
> yearday
> within that year?
>
> My specific concerns with Jim's approach:
>
> first_freeze_date:units = "days since 1900-01-01 00:00:00"   - This 
> doesn't seem
> to me to provide the most easily used data point, wouldn't the 
> year-day be more
> convenient, for seeing how this value varies over the years?
>
> And with Antoio's:
>
> first_freeze_date:coordinates="threshold time"; - I don't see how 
> threshold,
> which is a temperature, can be a coordinate of this variable. Also, 
> I'd like to know
> why setting   time:units="days since 2000-6-1"; is preferable to using 
> 2000-1-1;
> doesn't this invite errors in using the time in applications like 
> matlab and python?
>
> Actually, the metadata doesn't tell me how to interpret the values in 
> first_freeze_date -
> the short name implies that they're dates, the units implies they're 
> elapsed days, but
> without a reference date to enable decoding.
>
> Cheers - Nan
>
>
> On 3/16/17 8:45 AM, Jim Biard wrote:
>>
>> Hi.
>>
>> There is clearly interest here! I agree that day_in_year is rather 
>> generic, and there should probably be a more precise term. I'm not so 
>> sure about the cell_methods that were suggested below. In my 
>> particular case the values are derived from a daily Tmin product. 
>> Each value is the date of the first Tmin < 0 C within the time 
>> bounds. If it was a spell length, such as growing season length, then 
>> I can see the need for a more climatological cell_method.
>>
>> We can keep this up and work up some standard_name definitions to 
>> propose. I'm sure the results will be better if we collaborate 
>> compared to what I'd do on my own.
>>
>> Grace and peace,
>>
>> Jim
>>
>>
>> On 3/16/17 7:23 AM, Antonio S. Cofi�o wrote:
>>> Dear all,
>>> There is no standard_name for the concept but there are 2 different 
>>> ones which delimit the approach that it could be used as templates 
>>> for the new one:
>>> *time_when_flood_water_falls_below_threshold 
>>> *(time_when_flood_water_rises_above_threshold and 
>>> time_of_maximum_flood_depth are also good examples )
>>> http://cfconventions.org/Data/cf-standard-names/41/build/cf-standard-name-table.html#time_when_flood_water_falls_below_threshold_tr 
>>>
>>>> The quantity with standard name 
>>>> *time_when_flood_water_falls_below_threshold*: is the time elapsed 
>>>> between the breaking of a levee (origin of flood water simulation) 
>>>> and the instant when the depth falls below a given threshold for 
>>>> the last time, having already risen to its maximum depth, at a 
>>>> given point in space. If a threshold is supplied, it should be 
>>>> specified by associating a coordinate variable or scalar coordinate 
>>>> variable with the data variable and giving the coordinate variable 
>>>> a standard name of flood_water_thickness. The values of the 
>>>> coordinate variable are the threshold values for the corresponding 
>>>> subarrays of the data variable. If no threshold is specified, its 
>>>> value is taken to be zero. Flood water is water that covers land 
>>>> which is normally not covered by water.
>>> the problem is the event definition, which is quite different to the 
>>> one it's been considered here which is more like a climatological 
>>> statistics. The good thing is the CF already has some good 
>>> definitions for those climatological statistics, like Example 7.11 
>>> on CF1.6 document:
>>> http://cfconventions.org/cf-conventions/v1.6.0/cf-conventions.html#extreme-statistics-and-spell-lengths-ex 
>>>
>>>
>>> And more convenient definition of this climatological statistics 
>>> could be:
>>> http://cfconventions.org/Data/cf-standard-names/41/build/cf-standard-name-table.html#spell_length_of_days_with_air_temperature_above_threshold_tr 
>>>
>>>> Air temperature is the bulk temperature of the air, not the surface 
>>>> (skin) temperature. A spell is the number of consecutive days on 
>>>> which the condition X_below|above_threshold is satisified. A 
>>>> variable whose standard name has the form 
>>>> spell_length_of_days_with_X_below|above_threshold *must have a 
>>>> coordinate variable or scalar coordinate variable with the a 
>>>> standard name of X to supply the threshold*(s).*It must have a 
>>>> climatological time variable, and a cell_method entry* for within 
>>>> days which describes the processing of quantity X before the 
>>>> threshold is applied. A spell_length_of_days is an intensive 
>>>> quantity in time, and the cell_methods entry for over days can be 
>>>> any of the methods listed in Appendix E appropriate for intensive 
>>>> quantities e.g. "maximum", "minimum" or "mean".
>>>
>>> And this definition gives a more appropriate way to encode the date 
>>> of freezing days using a auxiliary coordinate to specify the 
>>> threshold and use a cell_methods attribute along with the 
>>> climatology_bounds attribute on time coordinate to specify an 
>>> statistics over a period.
>>>
>>> The standard_name should be more like the definition for 
>>> spell_length_of_days, but removing using 'time' as general instead 
>>> of days. This what I would suggest with respect to the encoding:
>>>
>>> variables:
>>>   float first_freeze_date(lat,lon);
>>> first_freeze_date:standard_name="time_when_air_temperature_below_threshold"; 
>>>
>>>     first_freeze_date:coordinates="threshold time";
>>>     first_freeze_date:cell_methods="time: minimum within days time: 
>>> minimum over days";
>>>     first_freeze_date:units="days";
>>>   float last_freeze_date(lat,lon);
>>> last_freeze_date:standard_name="time_when_air_temperature_below_threshold"; 
>>>
>>> last_freeze_date:coordinates="threshold time";
>>> last_freeze_date:cell_methods="time: minimum within days time: 
>>> maximum over days";
>>> last_freeze_date:units="days";
>>>   float threshold;
>>>     threshold:standard_name="air_temperature";
>>>     threshold:units="degC";
>>>   double time;
>>>     time:climatology="climatology_bounds";
>>>     time:units="days since 2000-6-1";
>>>   double climatology_bounds(time,nv);
>>> data: // time coordinates translated to date/time string type format
>>>   time="2008-01-16T00:00";
>>>   climatology_bounds="2007-08-01T00:00", "2008-05-31T00:00";
>>>   threshold=0.;
>>>
>>> The time: minimum over days, on first_freeze_date cell_methods 
>>> attribute represents the shortest time minimum daily temperature 
>>> (time: minimum within days) is below threshold.
>>> Equivalent for the last_freeze_date, but in this cas represents the 
>>> longest time (time: maximum over days).
>>>
>>> Regards
>>>
>>> Antonio
>>>
>>>
>>>
>>>
>>> -- 
>>> Antonio S. Cofi�o
>>> Associate Professor and Researcher
>>> Grupo de Meteorolog�a de Santander
>>> Dep. of Applied Mathematics and Computer Sciences
>>> Universidad de Cantabria (Spain)
>>>
>>> Academic Visitor
>>> National Centre for Atmospheric Science
>>> Department of Meteorology
>>> School of Mathematical, Physical and Computational Sciences
>>> University of Reading (UK)
>>>
>>> http://antonio.cofino.es
>>> On 15/03/17 18:16, Jim Biard wrote:
>>>>
>>>> Dan,
>>>>
>>>> How about that? I'm working on similar products. We haven't even 
>>>> considered standard names for them.
>>>>
>>>> I went ahead and used 'days since YYYY-MM-DD 00:00:00' for my first 
>>>> and last frost dates, since they are valid dates. My files are 
>>>> structured as (example for first frost date):
>>>>
>>>>     dimensions:
>>>>             time = UNLIMITED ; // (56 currently)
>>>>             lon = 960 ;
>>>>             lat = 490 ;
>>>>             bnds = 2 ;
>>>>     variables:
>>>>             double time(time) ;
>>>>                     time:standard_name = "time" ;
>>>>                     time:long_name = "time" ;
>>>>                     time:axis = "T" ;
>>>>                     time:units = "days since 1900-01-01 00:00:00" ;
>>>>                     time:calendar = "gregorian" ;
>>>>                     time:bounds = "time_bounds" ;
>>>>             double time_bounds(time, bnds) ;
>>>>             double lon(lon) ;
>>>>                     lon:standard_name = "longitude" ;
>>>>                     lon:long_name = "longitude" ;
>>>>                     lon:units = "degrees_east" ;
>>>>                     lon:modulo = 360. ;
>>>>                     lon:axis = "X" ;
>>>>                     lon:bounds = "lon_bounds" ;
>>>>             double lon_bounds(lon, bnds) ;
>>>>             double lat(lat) ;
>>>>                     lat:standard_name = "latitude" ;
>>>>                     lat:long_name = "latitude" ;
>>>>                     lat:units = "degrees_north" ;
>>>>                     lat:axis = "Y" ;
>>>>                     lat:bounds = "lat_bounds" ;
>>>>             double lat_bounds(lat, bnds) ;
>>>>             float first_freeze_date(time, lat, lon) ;
>>>>                     first_freeze_date:_FillValue = 1.e+20f ;
>>>>                     first_freeze_date:missing_value = 1.e+20f ;
>>>>                     first_freeze_date:comment = "Date of the first
>>>>     day with a minimum temperature at or below 0 degrees C over the
>>>>     9 month period starting Aug 1 of each year." ;
>>>>                     first_freeze_date:flag_meanings =
>>>>     "No_Freeze_Following" ;
>>>>                     first_freeze_date:long_name = "First freeze 
>>>> date" ;
>>>>                     first_freeze_date:valid_min = 0. ;
>>>>                     first_freeze_date:flag_values = -2. ;
>>>>                     first_freeze_date:units = "days since 1900-01-01
>>>>     00:00:00" ;
>>>>                     first_freeze_date:calendar = "standard" ;
>>>>
>>>> with the time bounds reflecting 1 Aug to 1 May for each year.
>>>>
>>>> On 3/15/17 1:50 PM, Hollis, Dan wrote:
>>>>>
>>>>> Hi Jon,
>>>>>
>>>>> I�d be interested to know how to tackle this problem too. I�ve 
>>>>> recently been generating some datasets of �date of first frost� 
>>>>> and �date of last frost� and have no idea how to describe them in 
>>>>> a CF-compliant way.
>>>>>
>>>>> Jim�s suggestion of �day_of_year� is better than just �days�, 
>>>>> however this doesn�t capture what the �something� is that has 
>>>>> happened, nor that is the first/last/Nth occurrence of that event. 
>>>>> What sort of events are you looking at?
>>>>>
>>>>> In my application I�m just looking at UK data, hence my �year� 
>>>>> runs from 1^st July to 30^th June (to span the N Hemisphere 
>>>>> winter). It�s easy enough to use the bounds to indicate this, but 
>>>>> I�m then not sure what values to store in the data array. Number 
>>>>> of days since 1^st July maybe? Or ordinal date (1^st Jan = 1, 
>>>>> 31^st Dec = 365)?
>>>>>
>>>>> Dan
>>>>>
>>>>> PS I have a whole bunch of other metrics that I�m looking at e.g. 
>>>>> length of the longest spell, number of spells greater then N days 
>>>>> etc. These seem even more complicated to describe using CF. 
>>>>> Something for another post I think...
>>>>>
>>>>> *From:*CF-metadata [mailto:cf-metadata-bounces at cgd.ucar.edu] *On 
>>>>> Behalf Of *Jim Biard
>>>>> *Sent:* 15 March 2017 16:28
>>>>> *To:* cf-metadata at cgd.ucar.edu
>>>>> *Subject:* Re: [CF-metadata] Recording "day of year on which 
>>>>> something happens"
>>>>>
>>>>> Jon,
>>>>>
>>>>> I agree that a cell_methods attribute doesn't seem to be 
>>>>> necessary. A new standard_name like 'day_in_year' or 'day_of_year' 
>>>>> would likely make things clearer.
>>>>>
>>>>> Jim
>>>>>
>>>>> On 3/15/17 11:22 AM, Jon Blower wrote:
>>>>>
>>>>>     Thanks Jim, that�s very helpful. Is cell_methods necessary in
>>>>>     this case (for the time axis bounds) � probably not since this
>>>>>     isn�t a statistical quantity like an average, but a value
>>>>>     that�s �representative� of the year.
>>>>>
>>>>>     I seem to remember from a while back that there was a proposal
>>>>>     to allow time axes to use �calendar years since X� (as opposed
>>>>>     to �years since X�, which uses a fixed-length UDUNITS year),
>>>>>     which might handle this use case. I have been out of the loop
>>>>>     for a while, but I can�t find mention of that in the CF spec,
>>>>>     so maybe that didn�t go through.
>>>>>
>>>>>     I might consider requesting a new standard name � �days� is
>>>>>     good, but I wonder if a more specific one would be helpful.
>>>>>
>>>>>     Best wishes,
>>>>>     Jon
>>>>>
>>>>>     *From: *CF-metadata <cf-metadata-bounces at cgd.ucar.edu>
>>>>>     <mailto:cf-metadata-bounces at cgd.ucar.edu> on behalf of Jim
>>>>>     Biard <jbiard at cicsnc.org> <mailto:jbiard at cicsnc.org>
>>>>>     *Date: *Tuesday, 14 March 2017 15:12
>>>>>     *To: *"cf-metadata at cgd.ucar.edu"
>>>>>     <mailto:cf-metadata at cgd.ucar.edu> <cf-metadata at cgd.ucar.edu>
>>>>>     <mailto:cf-metadata at cgd.ucar.edu>
>>>>>     *Subject: *Re: [CF-metadata] Recording "day of year on which
>>>>>     something happens"
>>>>>
>>>>>     Jon,
>>>>>
>>>>>     1) I'd use 'days'. It is a valid standard name apart from the
>>>>>     'days since date' formalism. It's not perfect, but it's legal.
>>>>>     You could, alternatively, request a new standard name.
>>>>>
>>>>>     2) Use a time_bounds variable. I would tend to set the time to
>>>>>     be July 1 at midnight for each year, and set the bounds for
>>>>>     each year to Jan 1 of that year and Jan 1 of the next year.
>>>>>
>>>>>     Grace and peace,
>>>>>
>>>>>     Jim
>>>>>
>>>>>     On 3/14/17 10:43 AM, Jon Blower wrote:
>>>>>
>>>>>         Hi all,
>>>>>
>>>>>
>>>>>         We need to structure a NetCDF file that will hold a 
>>>>> variable that represents the day of the year on which an event 
>>>>> happened (integers from 0 to 366). This value is recorded every 
>>>>> year for a number of years. I have a couple of questions about how 
>>>>> best to do this:
>>>>>
>>>>>
>>>>>         1. What is the best standard name to use for the day of 
>>>>> the year? I didn�t find anything in the standard name table, 
>>>>> although I might have missed it.
>>>>>
>>>>>
>>>>>         2. What would be the best way to define the time axis? 
>>>>> Each point along the axis would represent a whole year, rather 
>>>>> than an instant in time. I could simply pick an arbitrary instant 
>>>>> (e.g. midnight on 1st Jan) to represent the year, but is there a 
>>>>> better way?
>>>>>
>>>>>
>>>>>         Thanks in advance for any help!
>>>>>
>>>>>
>>>>>         Jon
>>>>>
>>>>>
>>>>>
>
>

-- 
John Helly, University of California, San Diego / San Diego Supercomputer Center / Scripps Institution of Oceanography / 760 840 8660 mobile / http://www.sdsc.edu/~hellyj
ORCID ID: orcid.org/0000-0002-3779-0603

-------------- next part --------------
A non-text attachment was scrubbed...
Name: hellyj.vcf
Type: text/x-vcard
Size: 158 bytes
Desc: not available
URL: <http://mailman.cgd.ucar.edu/pipermail/cf-metadata/attachments/20170316/e46a82a0/attachment-0001.vcf>


More information about the CF-metadata mailing list