VIIRS SDR navigation not working over (north) pole

Errors and unexpected results
Post Reply
User avatar
hproe
Posts: 504
Joined: Sat Nov 27, 2010 3:46 pm

VIIRS SDR navigation not working over (north) pole

Post by hproe »

Hi -

VIIRS SDR granules from CLASS that include the north pole do not display. I just get the layer information with the range of the colour bar [missing,missing]. It is obvious that the image navigation gets confused at the north (and south?) pole. Metop AVHRR swaths over the pole display correctly - there is just a small hole where the swath touches the pole.

I note a difference in how CLASS data for VIIRS/AVHRR swaths cover the polar region. While with VIIRS the pole is inside the granules, with AVHRR the swath just touches the pole.

I have not found any trace of the is problem being listed.

cheers, HP
User avatar
bobc
Posts: 988
Joined: Mon Nov 15, 2010 5:57 pm

Re: VIIRS SDR navigation not working over (north) pole

Post by bobc »

Hi HP -

I picked up 3 consecutive pre-aggregated granules from CLASS where the middle granule crosses over the north pole:
  • GMODO-SVM14_npp_d20170219_t2049591_e2055395_b27545_c20170221181452800042_noaa_ops.h5
  • GMODO-SVM14_npp_d20170219_t2055408_e2101211_b27545_c20170221181442709984_noaa_ops.h5
  • GMODO-SVM14_npp_d20170219_t2101224_e2107028_b27545_c20170221181454040842_noaa_ops.h5

I tried several ways of displaying this, including:
  • With only the middle granule (which crosses over the north pole), display at low resolution (no geographic subsetting). There's a large gap in the data right at the north pole, but outside of that gap there's valid data and the enhancement range reflects this.
  • With only the middle granule, display at high resolution (subset of granule that crosses over the north pole). There's a small hole at the north pole and a projection seam at 0 degrees longitude (from the North Pole projection), but everything else is valid data.
  • I aggregated the 3 granules above together and displayed the full domain at low resolution and also subsetted a region over the north pole, and the results matched the two points above when working with the single granule.
  • I ordered the same granules above with the geolocation detached and this behaved the same way.
  • The granules above are ascending, but I downloaded a descending set of granules and things worked the same way.
  • With the descending granules, I ordered them with the terrain-corrected geolocation and things still worked.

A few questions for you:
  • What band(s) are you working with? For example, are you using SVI01 - SVI05, SVM01 - SVM16, or DNB?
  • Are you doing any subsetting in the Region tab before displaying the data?
  • Have you seen this problem with other granules crossing over the north pole?
  • Can you please post your data to our ftp on the chance that I can't find any granules that replicate the problem?

Here's an image of what I'm seeing when I subset a region over the north pole, display the data, and set the texture quality in the Layer Controls to High.
SVM14 over north pole
SVM14 over north pole

Thanks -
Bob Carp
McIDAS Help Desk
User avatar
hproe
Posts: 504
Joined: Sat Nov 27, 2010 3:46 pm

Re: VIIRS SDR navigation not working over (north) pole

Post by hproe »

Hi Bob -

Thanks for giving this much consideration. I have been working with bands M15 and DNB. I have uploaded them to your ftp server as file VIIRSnorthpole.7z (geolocation detached).

By and large you see what I see.Testing it again I have found some discrepancy btw M15 and DNB (always using method swathToGrid!).
M15
Thumbnail with large black spot at the pole and a black line along 0° longitude (see image in uploaded zip file).
Display without subsetting -> no image, colour bar range [missing,missing], no native projection available
Display with subsetting around polar black spot -> no image, colour bar range [missing,missing]
Display with subsetting well outside pole but across 0° longitude -> good image
DNB
Thumbnail showing NO black marks
Otherwise behaviour identical to M15 above.

Note the differing thumbnails and that whenever there is no display, also the native projection is unavailable (which seems quite logic). I have not looked (yet) at other instances of granules straddling the north pole.

HP
User avatar
bobc
Posts: 988
Joined: Mon Nov 15, 2010 5:57 pm

Re: VIIRS SDR navigation not working over (north) pole

Post by bobc »

Hi HP -

Thanks for the data and the thorough description of what you are doing. Here's a summary of the problems you noted:

  1. With the SVM15 data, the preview image in the Region tab of the Field Selector shows a gap at the prime meridian as well as a large gap/hole over the north pole where there should only be a small hole.
    * I think this is all related to Inquiry 1804, which has a summary of "The preview image in the Region tab of the Field Selector shows a gap over the prime meridian when an NPP or MODIS granule crosses it.". I added your data to our testing repository for this inquiry and added mention of the large gap in the preview image over the north pole. I'm thinking that the north pole gap can be addressed at the same time as the prime meridian seam, but that would have to be addressed at the time a programmer digs into the code a bit.
  2. Display the granules using swathToGrid() without subsetting and the enhancement range goes from missing to missing.
    * I think this is related to Inquiry 1446, which has a summary of "The swathToGrid formula from the VIIRS plugin only works if you subset a region.".
    * I replicated that with just the one file you gave I ended up with a missing to missing enhancement range and no data displayed.
    * I ordered the data/geolocation for the granule following yours (SVM15_npp_d20160825_t0126084_e0131470_b25008_c20160825073149268381_noaa_ops.h5) that also includes the north pole hole and things behaved the way that Inquiry 1446 mentions. Instead of ending up with a low resolution solid display of data, you end up with just a bunch of squares of data:
    swathToGrid two consecutive files from CLASS with no subsetting
    swathToGrid two consecutive files from CLASS with no subsetting
  3. When using swathToGrid(), subset a region around the north pole hole/gap and you end up with a range of missing to missing and no data displays.
    * I replicated this problem and wrote it up as Inquiry 2494, which has a summary of "When subsetting SVM granule over Region tab's gap over the north (and south?) pole with swathToGrid does not display the data correctly."
    * Using the same granule that comes right after yours (mentioned in number 2 above) the behavior is slightly different. I still see the same north pole hole in the Region tab, but when I subset over the hole the behavior is slightly different. Only a very small sliver of data displays and the enhancement range only matches data values over this small displayed area.
    * I added both of these cases to Inquiry 2494.

Please let me know if this doesn't address all of your concerns.

Thanks -
Bob
User avatar
hproe
Posts: 504
Joined: Sat Nov 27, 2010 3:46 pm

Re: VIIRS SDR navigation not working over (north) pole

Post by hproe »

Hi Bob -

Your summary is fine as concerns SVM data. Have you also looked into SVDNB data, of which I uploaded also a sample granule? DNB behaves slightly different. The thumbnail is complete and shows no black marks. I also have looked at next DNB granule (t0126...) like you did for SVM15. A subset spanning the whole granule displays fine. There is no north pole included as you say in your point 2.
Have you seen my remark in another post about a possible link to Inquiry 1333 (date line)?

HP
User avatar
bobc
Posts: 988
Joined: Mon Nov 15, 2010 5:57 pm

Re: VIIRS SDR navigation not working over (north) pole

Post by bobc »

Hi HP -

Yes, I did replicate your findings with the single DNB granule you provided. I haven't ordered the following granule as of yet.

I'm wondering if the hole over the north pole in the SVM15 preview is an artifact of the way that McIDAS-V is handling the bow-tie deletion with the data over that region. There is no bow-tie deletion with the DNB data (per page 12 of CIRA's Beginner's Guide to VIIRS Imagery Data), and I'm thinking that this is why the preview image looks good with that data. I added some specifics to Inquiry 2494 to cover the differences with the DNB data compared to other SVM bands.

I cross referenced Inquiry 2494 with Inquiry 1333, which you mentioned in your other forum post. It is very possible that the 'missing' to 'missing' enhancement problem could be related to subsetting across the dateline.

Thanks again -
Bob
Post Reply