Data from the Laser Optical Plankton Counter instrument on the MBARI Dorado AUV
are routinely processed into NetCDF files. These data need to be loaded into
STOQS database and used for assessing the size distributions of particles.
An example such data set is here:
http://dods.mbari.org/opendap/data/ssdsdata/ssds/generated/netcdf/files/ssds.sho
re.mbari.org/auvctd/missionlogs/2014/2014295/2014.295.00/lopc.nc.html
The sepCountList and mepCountList variables are spatial-temporal trajectories
of a spectrum of counts in size classes ranging from 0.1 to 15 mm.
In order to produce counts from the Multi Element Plankton an integration over
time must be done. The default integration period is 10 seconds. Ideally, we
would like to have the LOPC data line up with the 2 second data that we
currently load from the other Dorado instruments. There is also the issue of
the fourth dimension of "bin" and how that can fit into the STOQS data model.
There are 2 big issues to resolve:
1. Temporal matching of the LOPC to other data. Note that 10 seconds may not be
the optimal integration time. Whatever scheme designed must accommodate
variable integration time periods
2. Storing the whole size class spectrum. There are 994 bins. Should each be a
different parameter name? Is a schema change needed?
Parallel development of Machine Learning techniques may help us evaluate the
best approaches for these issues.
Original issue reported on code.google.com by MBARIm...@gmail.com on 11 Feb 2015 at 12:09
Original issue reported on code.google.com by
MBARIm...@gmail.com
on 11 Feb 2015 at 12:09