[Thread Prev][Thread Next][Index]

Re: [las_users] How to configure time axis with ISO 8601 datetime strings?



Thanks for the responses, Steve and Roland. I suspected the strings were a problem, especially since I later found similar errors in other clients (ERDDAP and DChart). The strings are being produced in my NcML aggregation in TDS when I use the dateFormatMark feature to extract a time axis from the dataset's filenames (because the data don't contain a time variable). I've inquired with Unidata, but I don't believe I have a choice of units on the time axis that gets created. But maybe, like you say, there is some way of converting it with additional NcML? We're also investigating the possibility of re-writing these data to include a time variable, for which we would use the appropriate units (i.e. not strings) and side-step all this messiness.
Thanks!,
John Maurer

----- Original Message -----
From: Steve Hankin <steven.c.hankin@xxxxxxxx>
Date: Tuesday, December 21, 2010 1:22 pm
Subject: Re: [las_users] How to configure time axis with ISO 8601 datetime strings?
To: "John Maurer, IV" <jmaurer@xxxxxxxxxx>
Cc: las_users <oar.pmel.las_users@xxxxxxxx>, Ansley Manke <Ansley.B.Manke@xxxxxxxx>

> Hi John,
>
> The data set
>     http://oos.soest.hawaii.edu/thredds/idd/satellite.html?dataset=dhw
> does not have a valid CF time axis.  Use of strings as dates is not valid.
>
> I'm guessing that the Unidata Java netCDF libraries evidently support this encoding of time, even though CF does not.  As a result the Java-based addXML utility succeeds in configuring the dataset in LAS.  But the Ferret application, when called by LAS,  cannot read the times.  (Behind the scenes explanation:  when Ferret finds a problem in a coordinate variable -- coordinates out of sequence, embedded missing values, string values instead of numeric values, etc. -- it synthesizes a generic array of indices.   The value of "T=-6.3143E+10" in the error message is the translation of "2000-12-02" into a time stamp value using a default algorithm -- not very helpful as an error clue under the circumstances.)
>
> Before discussing work-arounds, a question:  where the file comes from and whether it is too late to suggest valid CF time axes for these files?  If too late, then the next best idea is to use ncml to replace the time axis of strings with time axis of typical CF time step values encoded using units  such as "days since <t0-date>".
>
>     - Steve
>
> ======================================================
>
>
> On 12/21/2010 2:13 PM, John Maurer, IV wrote:

Hi All,
> I have a dataset whose time axis includes ISO 8601 datetime strings (e.g. "2010-12-21T00:00:00Z") that is not displaying in LAS. I get an error:
>  **ERROR: illegal limits: CRW_HOTSPOT does not exist at T=-6.3143E+10 >           Axis extremes are T=0.5:1042.5 > 
This error shows axis extremes (0.5:1042.5) that use only the time variable's indices. That plus the strange T=-6.3143E+10 seems to suggest that LAS is not understanding the format of the time axis strings. How can I tell it that the axis is using ISO 8601 datetime strings? When configuring this dataset in LAS, I used this but I tried various permutations of the units setting:
>
>   <time-t-dhw type="t" units="days" display_lo="2000-12-02" display_hi="2010-12-13" default="last">
>     <arange start="2000-12-02" size="1042" step="3.5" />
>   </time-t-dhw>
>
> And here is the link to the dataset in our THREDDS Data Server (TDS):
>
> http://oos.soest.hawaii.edu/thredds/idd/satellite.html?dataset=dhw

>
> An additional wrinkle to getting this dataset into LAS is that the time axis is bi-weekly. Will LAS limit the list of available days in the pull-down to those that are actually contained in the dataset?, or will there be a lot of days listed for which there are no data and the user will get an error?
> Thanks,
> John Maurer
> Hawaii Ocean Observing System (HiOOS)
> University of Hawaii
>


[Thread Prev][Thread Next][Index]


Contact Us
Dept of Commerce / NOAA / OAR / PMEL / TMAP

Privacy Policy | Disclaimer | Accessibility Statement