Scripps High Resolution XBT Network: QA/QC Procedures

1. Scripps High Resolution XBT Network: QA/QC Procedures Lisa Lehmann, Dean Roemmich, and Glenn Pezzoli Scripps Institution of Oceanography GTSPP Workshop, Oostende Belgium May 2010

2. Outline • Data quality begins at sea. • Delayed-mode quality control. • Future directions. The SIO High Resolution XBT Program deploys 6000 XBTs per year, as part of the global HRX partnership.

3. Data quality begins at sea • XBTs are deployed from a stern-mounted autolauncher, resulting in many fewer wire problems than bridge-launching. The autolauncher can be re-positioned at sea according to conditions. • Probes in the autolauncher stay close to SST, minimizing temperature bias seen in heated or cooled probes. • Routine calibration using high precision resistors (test cannister) identifies problems occurring anywhere from the MK21 to the autolauncher (electronics, cables, and connectors). • Experienced ship-riders oversee data collection at sea, so problems are discovered rapidly. • Immediate automated checking compares each profile with the previous one, alerting the ship-rider to failures and unusual features for quick re-drops.

4. Data quality begins at sea: re-drops. Eastern subtropical North Pacific: Drop 44 is quite different from 43 and earlier profiles. A re-drop (45) about 2 km along-track shows that 44 is ok, with subsequent drops “filling in” the transition. Tropical North Pacific: Drop 182 is quite different from 181 and earlier profiles. A re-drop (183) about 2 km along-track confirms that 182 is bad.

5. Delayed-mode Quality Control • Read ship-rider report • Check for and remove false splashes • Edit data. Compare using: • Neighboring profiles • Climatology • Buddy profiles • Regional oceanographic features • Renavigate position based on previous and post GPS location

6. Delayed-mode QC: Read Ship-rider report Important to understand cruise and weather conditions. Eastern subtropical North Pacific: Drop 001 (in gray) shows record rain runoff. If ship-rider did not document that fact, I would have considered the surface suspect. The following profiles also show deeper mixed layer than in previous transects, likely due to huge swells in area during this time. PX37 South, January 2010

7. Remove false splashes False splashes are caused by ground faults in the wiring. These are generally caused by moisture in the equipment. Our ship-riders are trained to detect and fix problems as they occur, thus ensuring a cleaner dataset. We run any profiles with false splash detects through an operator driven filter which shows the raw ascii data and allows us to choose the actual splash. Then we can overlay the re-created profile with it’s neighbors and climatology to determine if the data is good. Profile 44 (red) shows false splash detect Profile 44 (red) after removing false splash

8. Delayed-mode QC: Neighbors Important to realize best source of QC information for each profile in high resolution sampling is the neighboring profiles. This sequence shows the temperature inversion at the base of the mixed layer just north of Fiji. PX09 Oct 2009

9. Delayed-mode QC: Neighbors This cluster of neighbors in the Kuroshio illustrates how closely spaced profiles (~15 km apart) reveal the highly structured character of the temperature field in a strong current. PX44 January 2009

10. Delayed-mode QC: Climatology We use a climatology developed from the HRX transects for many checks. Here it is used to determine where profile 247 (Top, green) went bad. Profile 247 (Bottom, black) is compared to climatology. It appears to have gone bad at 600 m, where it veers abruptly warm to climatology. • PX37 June 2008 • Profiles from various • PX37 transects within 1 degree • latitude and longitude to create • climatology.

11. Delayed-mode QC: Buddies We can view “buddies” of a profile individually to see if a particular feature has been see in the past (or future). Profile 170 (Top, red). Is that an inversion or wirestretch at 580m? Profile 170 (bottom, black) appears to veer warm to Climatology.. However… PX10 January 2004 220 profiles used to create climatology

12. Delayed-mode QC: Buddies PX10 January 2004-Profile 170 (black, all 3 figures) is shown with buddies from different cruises. We can find many profiles that match it very well over time. We’ll call this a PIA (Inversion Probable) Shown with PX10 February 1994 Shown with PX10 April 2003 Shown with PX10 August 1995

13. Delayed-mode QC: Edit data: Regional features If a feature (top: profile 027 in black, bottom: profile 036 in black) is not present in neighboring profiles and does not have a close geographical buddy in past cruises, then consider regional oceanographic features. PX37 Feb 2008 PX37 Nov 2008

14. Delayed-mode QC: Edit data: Regional features This (top, right) and other previous transects show confirmed eddies in the region of the previous two unconfirmed profiles. Therefore we consider the two profiles on the previous slide as good data (class 1). PX37 Jul 1996 At right, a very similar feature, occurring nearby in a CTD transect at 37oN, 128oW, was described as a California Undercurrent eddy by Cornuelle et al (2000, JGR, 105, 1227-1243.)

15. Future directions: XBT and Argo • High Resolution XBT transects and Argo are a valuable combination scientifically for estimating the time-varying heat transport and storage in large ocean regions (e.g. Douglass et al., 2010, DSR II). • Need to identify and remove systematic errors (fall-rate, wire-related problems) for consistency of the datasets. • Argo will be useful in quality control of HRX data by providing: • Global climatologies of temperature and its variability (e.g. Roemmich and Gilson, 2009, Progress in Oceanography). • Across-track gradients to enable comparison along non-collinear HRX ship tracks. All ship tracks have some variability. PX37/10/44: 72 transects 1991 - 2009 Argo 6-year mean temperature at 600 dbar