Hi HP -
Thanks for reporting this and for also posting your data. Using your CLASS granule, I am replicating your issue. One thing that I noticed about the granule is that if I display the full thing without any subsetting, the very upper left corner of the granule crosses over the dateline. There are some known problems with subsetting across the dateline, which is
inquiry 1333. In most cases that I've seen, when granules cross the dateline, I can still draw a bounding box on the eastern side of the dateline, but that's not the case here. I'm wondering if this is because such a small portion of the granule crosses the dateline.
You mentioned that you didn't see this problem with the granule loaded from PEATE. I downloaded the data from PEATE and I was able to replicate the problem. Here's the granules I downloaded:
SVM15_npp_d20140929_t1235172_e1236414_b15140_c20140929185732037690_noaa_ops.h5
SVM15_npp_d20140929_t1236426_e1238068_b15140_c20140929185732037690_noaa_ops.h5
SVM15_npp_d20140929_t1238080_e1239322_b15140_c20140929185732037690_noaa_ops.h5
SVM15_npp_d20140929_t1239334_e1240558_b15140_c20140929185732037690_noaa_ops.h5
... along with the GMTCO geolocation files for each SVM15 data granule.
These four granules from PEATE (when aggregated together) match up with the data you sent from CLASS and I can't subset here, either. The problem is with the *_t1235172 granule, which is the one that barely crosses the dateline. Interestingly, I ordered the granule from one timestep earlier (which crosses the dateline some more):
SVM15_npp_d20140929_t1233518_e1235160_b15140_c20140929185714562424_noaa_ops.h5
... and subsetting works as I would expect it to when the data crosses the dateline (it's much easier to subset a region).
I'll add more details of this to Inquiry 1333, and I'll meet with our NPP programmer to make him aware of this (new) dateline-related subsetting issue.
Thanks again for making us aware of this issue -
Bob Carp