[Go-essp-tech] Incorrect use of missing value in some MRI ocean data -- correction
V. Balaji
V.Balaji at noaa.gov
Fri Dec 9 07:41:46 MST 2011
Hi Karl, I'm not convinced that the "0 or missing" approach will work
in many cases: most NC tools are built to recognize and discard missing
values but I don't think they can be expected to recognize 0 as an
"unreasonable" value.
Also, I'm not sure it's guaranteed that even native-grid data will have
sftof of 0 or 1. If land and ocean/ice are on independent grids, they
will discretize the land-sea mask differently: then either sftof or
sftlf will have to have fractional values to ensure global area
conservation. (GFDL chooses to keep sftof 0/1 and clip the land cells:
but this is an arbitrary choice).
Karl Taylor writes:
> Hello Jamie,
>
> For the surface layer of the ocean and for variables at the actual interface
> between the ocean and the atmosphere, I think it is clear that sftof, if
> properly reported, can be used in all cases to determine which grid cells are
> ocean (or sea ice), and which are not. For models reported on their native
> grids, the values of sftof will I think either be 0. or 100. (%); for models
> regridded to latxlon grids, the values will vary between 0. and 100. This
> will need to be communicated to users, and I'm open to suggestions about
> that.
>
> For the ocean layers below the surface, there are several options, but
> perhaps no general solution. For many fields (e.g., theto, so, masscello,
> etc.), a value of 0. clearly will indicate that the value is missing (because
> a reasonably realistic ocean would never have a value of 0.), or perhaps it
> will be set to "missing", so a user could use any of these to infer the mask.
> I'm not sure how volcello is being reported when a grid cell is on land or
> below the bottom of the ocean. I would think it would also be either 0.0 or
> "missing", so it could also be used, but I'm not sure whether for some ocean
> grids, cells can appear and disappear (at the bottom of the ocean). If so,
> then this field can't be used in general because it isn't time-dependent. I
> think we need to find out from the modeling groups whether volcello would
> work in all cases.
>
> regards
> Karl
>
> On 12/6/11 1:23 AM, Kettleborough, Jamie wrote:
>> Hello Martin, Karl,
>> aren't there other things coming out of this:
>> 1. who is responsible for spotting cases where land points are represented
>> by 0.0, and cases where they are represented by missing value. I think the
>> implication is at the moment its the users. (I guess it could have been
>> CMOR - though thats too late now - it could be something in 'QC' [though I
>> may have used not quite the right term in saying 'QC']).
>> 2. how are users to know the best ways of dealing with the cases? ie to
>> use sftof? (I guess the danger in giving advice in general on these sort
>> of things is the advise turns out to be wrong, or not quite applicable in a
>> particular context).
>> Jamie
>>
>> ------------------------------------------------------------------------
>> *From:* go-essp-tech-bounces at ucar.edu
>> [mailto:go-essp-tech-bounces at ucar.edu] *On Behalf Of *Karl Taylor
>> *Sent:* 05 December 2011 18:59
>> *To:* martin.juckes at stfc.ac.uk
>> *Cc:* cmip5-users at jiscmail.ac.uk; go-essp-tech at ucar.edu;
>> graham.parton at stfc.ac.uk
>> *Subject:* Re: [Go-essp-tech] Incorrect use of missing value in
>> some MRI ocean data -- correction
>>
>> Hi Martin,
>>
>> Right again.
>>
>> thanks,
>> Karl
>>
>> On 12/5/11 10:39 AM, martin.juckes at stfc.ac.uk wrote:
>>> Hello Karl,
>>>
>>> Seiji (from MRI) suggested people use sea-area fraction"sftof": I
>>> think "areacello" (standard name "cell_area") refers to the actual grid
>>> cell, which may be be partially or completely land. "volcello" (standard
>>> name "ocean_volume") doesn't have a definition in the current standard
>>> name table -- but from the name I would expect to refer to the volume of
>>> ocean, so that should be OK for the height dependent fields,
>>>
>>> cheers,
>>> Martin
>>>
>>> ________________________________
>>> From: Karl Taylor [taylor13 at llnl.gov]
>>> Sent: 05 December 2011 18:21
>>> To: Juckes, Martin (STFC,RAL,RALSP)
>>> Cc:go-essp-tech at ucar.edu; Parton, Graham
>>> (STFC,RAL,RALSP);cmip5-users at jiscmail.ac.uk
>>> Subject: Re: [Go-essp-tech] Incorrect use of missing value in some MRI
>>> ocean data -- correction
>>>
>>> Hi Martin,
>>>
>>> You're quite right. It states:
>>>
>>> "2. Unless otherwise specified, the ocean and sea-ice output
>>> (including Oclim, Oyr, Omon, and OImon) represents a mean over only the
>>> sea portion of each grid cell (i.e., it is interpreted as "where ocean
>>> over ocean"), and a value of 0.0 should be reported where the sea fraction
>>> is 0."
>>>
>>> I'm can't recall whether this specification was the result of some
>>> requirement, or if it was simply a mistake. Thinking about it now, I
>>> think the recommendation should have been to set it to the "missing"
>>> value, not 0.0, since over land the values are undefined.
>>>
>>> Given that some groups will follow the specified rule, and others will
>>> put in the missing value for land points, users would be advised to obtain
>>> the areacello variable (or volcello) which should be set to 0.0 where
>>> there is no ocean.
>>>
>>> Any other ideas?
>>>
>>> Best regards,
>>> Karl
>>>
>>> On 12/5/11 9:45
>>> AM,martin.juckes at stfc.ac.uk<mailto:martin.juckes at stfc.ac.uk> wrote:
>>>
>>> Hello Karl,
>>>
>>> In the following:
>>> http://cmip-pcmdi.llnl.gov/cmip5/docs/standard_output.xls, sheet
>>> "general", column A, row 15 and 17.
>>>
>>> It says "the value 0.0 should be reported where the sea fraction is
>>> 0",
>>>
>>> regards,
>>> Martin
>>>
>>> ________________________________
>>> From: Karl Taylor [taylor13 at llnl.gov<mailto:taylor13 at llnl.gov>]
>>> Sent: 05 December 2011 16:42
>>> To: Juckes, Martin (STFC,RAL,RALSP)
>>> Cc:go-essp-tech at ucar.edu<mailto:go-essp-tech at ucar.edu>; Parton, Graham
>>> (STFC,RAL,RALSP);cmip5-users at jiscmail.ac.uk<mailto:cmip5-users at jiscmail.ac.uk>
>>> Subject: Re: [Go-essp-tech] Incorrect use of missing value in some MRI
>>> ocean data -- correction
>>>
>>> Dear Martin and all,
>>>
>>> For 3-d ocean fields like thetao, I think the missing value (not 0.0)
>>> should invariably be assigned to all land grid cells and for grid cells
>>> below the bottom of the ocean. I can't find where in the document pointed
>>> to below it says otherwise. I would like to correct the document, so
>>> please tell me where it needs revision.
>>>
>>> thanks,
>>> Karl
>>>
>>> On 12/4/11 8:07
>>> PM,martin.juckes at stfc.ac.uk<mailto:martin.juckes at stfc.ac.uk><mailto:martin.juckes at stfc.ac.uk><mailto:martin.juckes at stfc.ac.uk>
>>> wrote:
>>>
>>> I owe MRI an apology -- the use of zero instead of a missing value is
>>> in fact recommended in
>>> (http://cmip-pcmdi.llnl.gov/cmip5/docs/standard_output.xls),
>>>
>>> regards,
>>> Martin
>>> ________________________________________
>>> From: Juckes, Martin (STFC,RAL,RALSP)
>>> Sent: 04 December 2011 08:45
>>> To:go-essp-tech at ucar.edu<mailto:go-essp-tech at ucar.edu><mailto:go-essp-tech at ucar.edu><mailto:go-essp-tech at ucar.edu>;
>>> Parton, Graham
>>> (STFC,RAL,SSTD);cmip5-users at jiscmail.ac.uk<mailto:cmip5-users at jiscmail.ac.uk><mailto:cmip5-users at jiscmail.ac.uk><mailto:cmip5-users at jiscmail.ac.uk>
>>> Cc: Pascoe, Stephen (STFC,RAL,RALSP)
>>> Subject: Incorrect use of missing value in some MRI ocean data.
>>>
>>> Hello,
>>>
>>> It has come to light that some MRI ocean data has land grid points set
>>> to zero rather than to the declared missing value. This is likely to
>>> confuse some applications reading and plotting the data -- I've notified
>>> Seiji Yukimoto at MRI,
>>>
>>> sincerely,
>>> Martin
>>>
>>
>
--
V. Balaji Office: +1-609-452-6516
Head, Modeling Systems Group, GFDL Home: +1-212-253-6662
Princeton University Email: v.balaji at noaa.gov
More information about the GO-ESSP-TECH
mailing list