Overview
   Data Flow
   Enhancement
   Our philosophy is to flag (not remove) suspect data at the parametric level
   The RVSMDC system uses both automated and visual data inspection
   Automated flagging
      - Pre-process for realistic ranges, time sequence, etc. (see handbook)
      - New statistical spike/step flagging tool
   Visual inspection completed using VIDAT (VIsual Data Assessment Tool), a software tool developed in-house (see handbook)
      - Visualize multiple data streams
      - Map positions/climatologies
      - Check automated flagging
      - Analyst adds additional flags
   Upon completion of QC, the RVSMDC provides feedback to vessel operators
  Data flow begins with original data and documentation being combined into a single file (netCDF format)
   Output from each QC process (flags) combined into data quality report
   Report and value-added data (with flags) released to public
   New automated procedures have been developed to flag systematic errors
      - Based on analyst experience using VIDAT
      - Greatly increases QC efficiency (less analyst hours per vessel)
   Example: Stack exhaust impacts
      - With certain ship-relative winds, exhaust influences temperature and humidity measurements
   Visual QC identified increases in air temperature when ship-relative winds are near 180 degrees (from stern)
   Using earlier QC system, the analyst manually flagged suspect temperatures
   Today, our QC system takes advantage of automated identification of suspect regions
   The new automated detection algorithm identifies and flags spikes, steps, and statistically suspect values
      - The method examines the difference in near-neighbor values
      - Flags are assigned based on a threshold derived from observations
   Graphical representation of the flagged values identifies flow conditions with severe problems
      - Flags (red = spike, blue = suspect, green = step) are plotted as a function of ship-relative wind
      - The percent flagged in each wind bin is shown on the outermost ring (pink represents > 7.5% flagged)
   The analyst then determines a range of data to autoflag (For the Meteor, 145 - 210 degrees was chosen)
   Flagging all values in the range determined using the polar plot is preferred, since one cannot be certain that realistic looking temperature values are accurate when the wind is relative to the stern for the Meteor
   Analysis has shown the final result to be similar to analyst-added flags, but with substantial time savings
Analyst manual visual flags | ||
Flags applied by statistical auto-flagger | ||
Flags assigned to suspect ship-relative wind directions |