karlmsmith
commented
6 years ago Comment by @AnsleyManke on 14 Jan 2016 23:28 UTC This was a precision problem in the time computations. For instance observations that should have been 2 minutes after the start of a month were mistakenly counted in the previous month. The unweighted (number-of-obs) and weighted (number-of-cruises) are computed using two different functions in Ferret, and the per-cruise calculation had a lack of precision in the time binning.
This is a duplicate of Ferret ticket http://dunkel.pmel.noaa.gov/trac/ferret/ticket/2312 and was fixed in the Ferret external functions tracks2grid_mask_ave_xyt and tracks2grid_std_xyt
Reported by @AnsleyManke on 14 Jan 2016 23:17 UTC Matthias Tuma reported in an August 31 email that the grid cells with missing and non-missing data differed in the V2 gridded dataset.
''On the note of the upcoming v3, I also wanted to briefly let you know that during my work with the v2 gridded monthly dataset, I saw that the missingness patterns for the cruise-weighted data are slightly different from those of the unweighted data for the majority of the covered months. I am mentioning it because this caused some of my consistency checks to fail which I had originally meant to check my own work. I did not follow up on this in-depth (I'll be out of office until Sep. 14), but wanted to mention it to you in case it points to an inconsistency that would influence a tiny fraction of v3 data points as well. It may well have an explanation that I am missing because of lacking familiarity with the dataset, but I still wanted to mention it.''
Migrated-From: http://dunkel.pmel.noaa.gov/trac/las/ticket/1720