Are the apparent gaps data rejected by QC? I downloaded the data as visualized via ERDDAP and almost all (not just a few) wind10m values come over as NaN. All wind10m_qc_agg values = 2 (not evaluated). All wind10m_qc_tests = 2147483647.
The east and north observed winds seem fine, so I’ll work with those. But it is nice to have 10m for the purposes of comparing to model forecast winds, which is what we’re doing.
Also, in my experience, when downloading to Matlab ERDDAP always converts time to MATLAB datenum, but that has not happened with these data. I’ll try to figure what’s up using some other download options.
I see the same gaps in the CSV files created by ERDDAP. It looks like the time interval in the data file is 1 minute, but values for 10m winds are only reported every 1 hour (at the :30 mark).
Also, in case it helps, this command in Matlab gives times consistent with the CSV file: datetime(ooi_cp04ossm_sbd11_06_metbka000.time, 'ConvertFrom', 'posixtime')
Thanks Tom. I see the issue. An ERDDAP collection has only one time coordinate, so the full resolution data are at 1 minute intervals whereas the 10m-wind (presumably, being a derived quantity) is only reported at 1 hour intervals. Trimming to only the hourly times I see all valid wind10m as displayed in dataexplorer. But the gaps are still there, so my question about QC eliminating those remains.
I was wrong about the time coordinate in MATLAB. The conversion to datenum was something I coded internally to my erdddap_read.m function.
So, issues solved but I remain curious why there are so many gaps in u10 winds in the derived data. They seem to occur at the end of every second day (see here)
I’m not sure of the reason, but there are more data in the recovered dataset compared with the telemetered dataset. Although there still some gaps. Perhaps someone who knows more about the data telemetry can answer.
I suspect the issue here is with the hourly flux datastreams that you have plotted there. When I plot the 1-minute resolution data that these are based on, I don’t see any difference between the wind vectors for the two data delivery methods. Since the hourly flux data pulls in multiple parameters for the calculations, including from the completely separate instrument the seawater point velocity instrument (VELPT), there is likely some missing data within one of those sensors which results in the differences seen here.
My suggestion would be to use the recovered host data stream, or whichever has the most complete record. They both pull from the same source (the mooring computer’s storage) so they should be identical values for each time point. I’ll see if I can pinpoint exactly where the differences arise.
