Marine Meteorology Quality Control at the Florida State University

1
Marine Meteorology Quality Control at the Florida
State University

• Shawn R. Smith

Research Vessel Surface Meteorology Data
Center Center for Ocean-Atmospheric Prediction
Studies Florida State University www.coaps.fsu.edu
/RVSMDC
2
Who We Are

• Data center specializing in the quality control
  (QC) of marine observations collected by
  automated instrumentation on research vessels
  (R/Vs)
• We employ quality control procedures developed
  in-house to create value added data products
• We freely distribute all products to science
  community and apply them to current scientific
  problems

3
History of RVSMDC

• David M. Legler and James J. OBrien formed the
  Data Assembly Center (DAC) for WOCE in 1993
• Final WOCE archive contains meteorology data from
  over 439 hydrographic cruises (82 of completed
  cruises)
• Expanded early on to include all surface
  meteorology data from TOGA/COARE
• Late1990s, added data from select international,
  UNOLS, and NOAA R/Vs
• With expansion beyond WOCE, renamed archive R/V
  Surface Meteorology Data Center (RVSMDC)
• In 2004 we initiated the Shipboard Automated
  Meteorological and Oceanographic System (SAMOS)
  initiative

4
Experience

• Our archive contains many high-time resolution
  (lt15 min.) meteorology data sets

• 1990-1998 archive includes over 100 cruises with
  sampling lt 15 min. intervals
• 30 of data are poleward of 40S and 50N.

5
Current Archive

• Over 11, 000 ship days have been quality
  controlled
• Variables evaluated vessel navigation, air and
  sea temperatures, pressure, moisture parameters,
  ship-relative and true winds, radiation, and
  precipitation.
• Current focus 1-minute observations from U.S. RVs
  participating in SAMOS

6
Importance of metadata

• Accurate metadata are essential for scientific
  application of marine observations
• RVSMDC files contain detailed metadata that
  include instrument height and sensor type, units,
  time averaging period, ship ID, cruise ID (when
  available), and the facility that provided the
  data
• Communication with data providers essential to
  collection of accurate metadata

7
Quality-Control overview

• The goal of our quality control (QC) is to
  provide well-documented, reliable, and consistent
  research vessel data to the scientific community.
  
• Flags are applied values at the parametric level.
• This means that each individual observation (for
  which QC is applied) will have a single quality
  control flag.
• Alternative is to treat all values collected at a
  single time as one record and flag whole record
• Our philosophy is to flag suspect values, not
  remove them from the data files.
• Accepting or excluding flagged values may vary
  depending on the user's scientific goals our
  system retains all values for that purpose.

8
Quality-Control overview

• Two primary types of QC Real-time and Scientific
• Real-time
• Primarily used to meet needs of operational
  forecasting and modeling
• Includes simplified, aggregate flag structure
• 0 - good
• 1 - suspect
• 2 - erroneous
• 3 - not evaluated
• 4 - parameter missing
• Not ideal for climate research or future
  scientific applications
• Example From QARTOD http//nautilus.baruch.sc.edu
  /twiki/bin/view

9
Quality-Control overview

• Scientific
• Type of QC used at the RVSMDC and for SAMOS
• Each flag corresponds to a specific quality test
• For example, in our system we verify the
  relationship that the TTwTd. If this test fails
  a D flag is applied to the appropriate T, Tw, or
  Td values
• This method provides user with greater detail as
  to why a value was flagged.
• The method also allows for flags that do not
  indicate problems, but interesting features
  (e.g., frontal passages, pressure minima, etc.).

10
Quality-Control overview

• Our system uses both automated and visual data
  inspection
• Automated flagging
• Pre-process for realistic ranges, time sequence,
  etc.
• New statistical spike/step flagging tool (SASSI)
• VIDAT (VIsual Data Assessment Tool (software
  developed in-house)
• Visualize multiple data streams
• Map positions/climatologies
• Check automated flagging
• Analyst adds additional flags
• Provide feedback to vessel operators
• Two way communication with data providers is
  essential to understand problems and have them
  corrected

11
Quality-Control data flow

• Original data/documentation combined into single
  file (netCDF)
• Output from each QC process (flags) combined into
  data quality report
• Report and value-added data (with flags) released
  to public

SASSI
Pre-Screen
12
Quality-Control visual inspection

• Identifies systematic errors (e.g., severe flow
  distortion, sensor heating, and acceleration
  errors)
• Finds problems and features that are unique to
  new system deployments

13
Quality-Control enhancement

• New automated procedures developed to flag
  systematic errors
• Based on experience from VIDAT
• Greatly increases QC efficiency (less analyst
  hours per vessel)

• Example Stack exhaust impacts
• With certain ship-relative winds, exhaust
  influences temperature and humidity

14
Quality-Control spike/step

• Increases in air temperature visually identified
  when ship-relative winds near 180 deg. (from
  stern)
• Early QC Analyst manually flagged suspect
  temperatures
• QC today Takes advantage of automated
  identification of suspect regions

15
Quality-Control spike/step

• Spikes, steps, suspect values identified
  (flagged)
• Examines difference in near-neighbor values
• Flags based on threshold derived from
  observations
• Graphical Representation
• Identifies flow conditions w/ severe problems
• Flags plotted as function of ship-relative wind
• flagged in each wind bin on outer ring
• Analyst determines range of data to autoflag

16
Quality-Control spike/step

• Analyst manual visual flags (top)
• Flags applied by statistical auto-flagger
  (center)
• Flags assigned to suspect ship-relative wind
  directions (bottom)
• Final result similar to analyst flags, but w/
  substantial time savings

17
Quality Control true winds

• Another common problem with automated marine
  instruments is incorrect estimation of the true
  (earth-relative) winds
• Quality true winds (green) show no signal of ship
  motion (black)
• 180 deg. Error in reported ship-relative winds
  (blue) results in incorrect true winds (red)

18
Quality Control true winds

• Common causes for true wind errors
• Incorrect anemometer installation
• Failure to document wind direction convention
  (meteorological, oceanographic, merchant marine)
• Incorrect code to compute true wind (must remove
  motion induced by ship from ship-relative wind
  data)
• Confusing definitions for navigation parameters
  (course, heading, and speed)

19
Quality Control true winds

• Several vessels and agencies take advantage of
  our true winds analysis
• R/V Polarstern (Germany) and R/V Wecoma (U.S.)
  modified data collection and reporting systems
  based upon our recommendations
• Hankuk University of Foreign Studies (S. Korea)
  using our recommendations to improve winds on
  ferries
• Some yacht clubs and small marine companies are
  using our advice to improve instrumentation on
  recreational vessels

20
Final Thoughts

• Although we still conduct delayed-mode QC (3-6
  month lag from collection to distribution), we
  now focus on near real-time QC through SAMOS
  Initiative.
• Quality procedures undergo constant revision and
  updates. Future enhancements may include
• Automated ship heating algorithms
• Improved radiation QC (only range checks and
  visual inspection now)
• Procedures developed by the RVSMDC have been
  successfully applied to surface marine
  observations on multiple time scales (for both
  ships and buoys)