Before any eyes glaze over at the mere hint of a data quality assurance and quality control (QA/QC) discussion, consider the unsung heroes of the marine world . . . buoy technicians. I imagine they would rather spend time welding moorings than digging through time series charts looking for sensor degradation. At the UNC-W Coastal Ocean Research and Monitoring Program (CORMP), we have Quality Assurance / Quality Control of Real Time Oceanographic Data (QARTOD) running under the hood as data arrive to automate that QA/QC process and to alert technicians when something begins to look suspicious.
Chris LaClair and Brett Bolton (not pictured) of CORMP preparing for the next buoy turnaround.
The QARTOD community has drafted manuals that specify data quality tests for a wealth of parameters. Required tests generally include checks for: timing, syntax, location, sensor range, climatological (seasonal) range, spike, rate, and flatlining. Did the data come in on time and from the right location? Is it within the sensor manufacturer's min/max? Is it within the min/max that makes sense for this time of year? Does the current reading make sense when compared to its previous one? How about ones over the past few hours?
All of these QARTOD checks are done inside the CORMP database and are presented to technicians in three ways. (1) If a sensor stops reporting altogether or its location breaches its watch circle, an all-hands-on-deck email is sent to a distribution list. (2) A nightly email is sent that includes an at-a-glance data quality summary table which gives technicians an opportunity to spot potential problem sensors.
An alert email indicating no-data-received for 6 hours (on left). A nightly email containing a data quality summary table (on right).
The table from the nightly email contains links to a (3) QARTOD dashboard website where technicians may take a deeper dive to see individual quality flags alongside data readings to determine if a sensor may be showing signs of failure.
A screenshot of the QARTOD dashboard indicating failed and suspect QARTOD flags.
While such detailed QA/QC may be valuable to buoy slingers and technicians, the general public is only presented with a single CORMP "rollup" data quality flag: are the data good, bad, or suspect. The QARTOD community does not provide direction on how to roll their flags into one, but CORMP and Second Creek have created a set of rules to automate that decision. The ability for technicians to override the automatically generated CORMP rollup flag is important, however.
Hurricane Matthew (Oct 2016), for example, "delivered large amounts of rain along the U.S. southeast Atlantic coastline. Extensive freshwater run-off affected coastal waters along the South Carolina coast and caused water temperature, salinity, and conductivity data from the CAP2 mooring to be flagged [as suspect] every hour for almost two weeks."
Hurricane Matthew as seen on the Weather Channel.
With the help of South Carolina Department of Natural Resources (SC DNR), it was determined "that the CTD data from the CAP2 mooring were accurate. Large volumes of fresh water had reduced salinity, causing the climatology test to be suspect. Once it was confirmed that the data from the mooring were correct, project personnel marked the data as 'pass' in the roll-up flag. The 'suspect' flag for climatology, however, remains in place because the data were below the climatological range for that location." From Status and Near-Term Plans for the U.S. IOOS Quality Assurance / Quality Control of Real-time Oceanographic Data (QARTOD) Project.
A time series graph of CAP2 salinity during Hurricane Matthew.
QARTOD tests of CAP2 salinity during Hurricane Matthew indicating a suspect climatological test.
I consider it a success if I can increase a client's efficiency, but I am under no illusion that automated solutions are any substitute for the human touch. By providing quick and automated quality checks for data as they arrive, technicians aren't required to manually check-in as often. Instead they can concentrate on dreaming up the next generation of telemetry and power saving technologies for the next buoy rotation. When, of course, they aren't glued to buoy pages or the Weather Channel during a weather event to make sure everything is running smoothly.
Comments