Dear All,

The topic of swath observations is of particular interest here at NOAA's
NCDC.  Ken Knapp, a scientist within our Remote Sensing Applications
Division, has worked with CF for some time now and has expressed
interest in a "channel" dimension, similar to John Caron's "wavelength"
dimensions described below.  He also mentioned the desire to address
support for calibration tables.  More details are forthcoming as I gain
additional feedback and understanding.

Regards,
Ken Roberts
> Message: 2
> Date: Fri, 20 Nov 2009 05:24:01 -0700
> From: John Caron <[email protected]>
> Subject: Re: [CF-metadata] Swath observational data
> To: "[email protected]" <[email protected]>
> Message-ID: <[email protected]>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> Raskin, Rob (388M) wrote:
>   
>> While the Point observational conventions document is undergoing final 
>> review, I want to initiate a discussion on a complementary topic - Swath 
>> observational conventions. This model addresses satellite observational 
>> measurements and potentially airborne measurements.
>>
>> The Swath conceptual model is essentially a grid in spacecraft coordinates. 
>> One dimension of this grid ("along_track") follows the path of the 
>> satellite. Normally there are one or two additional dimensions: 
>> "cross_track" and/or "vertical". The "cross_track" dimension is 
>> perpendicular to the satellite path, as the instrument typically makes "side 
>> views" of the surface rather than just at the nadir. The "vertical" 
>> dimension is present when a vertical profiler instrument is used. 
>> CF:FeatureType will need to account for each possible combination of these 
>> 2-D and 3-D swaths.
>>
>> Typically, time is explicitly stored and associated only with the 
>> along-track dimension. Spatial resolution generally will differ in the 
>> along_track and cross_track directions. 
>>
>> Orbits are not mapped to files in any consistent way: a file might 
>> correspond to a complete orbit, a half-orbit, or some other value. However, 
>> it is common to explicitly consider yet another dimension: "satellite_node", 
>> with values "ascending" (crosses equator going northward) and "descending" 
>> (crosses equator going southward).  
>>
>> Common satellites are in sun-synchronous polar orbits such that the 
>> ascending node remains at a near constant Local Solar Time (LST), while the 
>> descending node remains at a near constant LST shifted by 12 hours. For 
>> example, the ascending node may be at 6am LST and the descending node at 6pm 
>> LST. Often gridded data products are produced from these swaths, with 
>> separate grids corresponding to the AM and PM cases. A new CF time 
>> representation for "LST" is required to indicate that the global data are 
>> all at a time such as 6am LST.
>>
>> Unrelated to the swath geometry, some measurements use spectral band as an 
>> independent variable, as they sample at multiple "channels". This capability 
>> requires a new standard name for "spectral_band" or "spectral_channel" with 
>> values that may be numeric, a numeric range, or string.
>>
>> Swath data include many new dependent variables that correspond to 
>> engineering parameters of the retrieval rather than geophysical parameters 
>> (point spread function is a common example). If these names are standardized 
>> at all, they should be indicated as being of the engineering type.
>>
>> In the case of an airborne (rather than satellite) measurement, there is 
>> more commonality with the "trajectory" representation from the Point 
>> observation model. Hence, the focus here is on spacecraft measurements.
>>
>> Finally, on an unrelated note, I have semantically mapped the entire CF 
>> Standard Name list to an ontological representation. But that is the subject 
>> of a separate communication.
>>
>> -Rob
>>
>>     
>
> Hi Rob, thanks for starting this up.
>
> We have done some preliminary thinking about the "swath feature type" in the 
> CDM data model, though we dont have any implementations.
>
> A prototype coordinate system would look something like:  
>
> dimension:
>   scan = 1234;
>   xscan = 987;
>   wavelength = 123;
>
> variables:
>   double lon(scan, xscan);
>   double lat(scan, xscan);
>   double alt(scan, xscan);
>   double time(scan);
>   double wavelength(wavelength);
>
>   byte data( scan, xscan);
>     data:coordinates = "lon lat time alt";
>  
>   byte spectral( scan, xscan, wavelength);
>     spectral:coordinates = "lon lat time alt";
>
>
> I think this should handle zigzags or grids, although perhaps adding a "scan 
> strategy" attribute would be good.
>
> The geometry of each point is an interesting wrinkle, and may need some new 
> conventions. would a rotated ellipse work (3 params) or do we need a more 
> general polygon? Does it have to be specified per point, or can is be common 
> to all points? I would imagine that quick visualizers might ignore the 
> details of this (essentially assuming a tesselating grid), but more 
> sophisticated and specialized tools would need this.
>   
-- 
Ken P. Roberts
Programmer Analyst, STG, Inc., Government Contractor
Remote Sensing & Applications Division
National Climatic Data Center
151 Patton Ave.
Asheville, NC 28801-5001
Phone: (828) 271-4083
Fax: (828) 271-4328
[email protected]

_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata

Reply via email to