Editing RT QC flag in delayed mode ? Virginie Thierry DMQC 4 Toulouse, 28 septembre 2009.

Slides:



Advertisements
Similar presentations
DRAFT GUIDANCE NOTE FOR MEMBER STATES ON FRAUD RISK ASSESSMENT
Advertisements

Configuration management
Configuration management
ADMT10 - Toulouse, 30 Sept 2009 ADMT9 Action 24 “Develop a common method for determining the positions and observation times at DACs”
Groom-gliders data-management workshop Brest, December 2012 Groom gliders data management n In 2012 : vertical profiles from 26 platforms.
Brian A. Harris-Kojetin, Ph.D. Statistical and Science Policy
Copyright 2010, The World Bank Group. All Rights Reserved. Statistical Project Monitoring Section B 1.
Retail Market Subcommittee Update to TAC February 5, 2004.
Pierre Jaccard1 MyOcean Quality Control for Ferryboxes MyOcean Tutorial, NERSC, Bergen.
Argo a year ago. 500 floats. Regional coverage only.
Argo Real-time Quality Control Process NOAA/AOML: Y.-H. DANESHZADEH, R. MOLINARI, R. SABINA, C. SCHMID CIMAS/UM: E. FORTEZA, X. XIA, H. YANG.
© Crown copyright Met Office The EN dataset Simon Good and Claire Bartholomew.
Argo QC with an emphasis on the North Atlantic Justin Buck British Oceanographic Data Centre Joseph Proudman Building 6 Brownlow Street Liverpool L3 5DA,
UNFCCC How to address DOE Liability in context of PoA: Perspective of a Project Developer 8 th May 2011 Gareth Phillips Chairman, Project Developer Forum.
1 presented by: Tim Haithcoat University of Missouri Columbia QA/QC and Risk Management.
QARTOD III Scripps Institution of Oceanography La Jolla, CA Quality Control, Quality Assurance, and Quality Flags Mark Bushnell, NOAA/NOS/CO-OPS November.
Course Technology Chapter 3: Project Integration Management.
Configuration Management
ISO 9000 Certification ISO 9001 and ISO
10th Argo data management 2009 Toulouse User’s manual comment Here is a list of comments received on User’s manual version 2.2 since August 28 th. Jean-Philippe.
Quality Control Standards for SeaDataNet Review status at 1 st Annual Meeting (March 2007) Review developments over last year Current status Future work.
IQuOD An International Quality Control Effort Tim Boyer EDM workshop September 10, 2014.
Planning an Internal Audit JM García Merced. Brainstorm.
TASK TEAM ON WIGOS QUALITY MANAGEMENT (TT-WQM). Topics:  Activities and discussions of the task team  Proposal for WIGOS Quality Monitoring System ST-QM.
Chapter 1 Database Systems. Good decisions require good information derived from raw facts Data is managed most efficiently when stored in a database.
EFRAG’s preliminary position on the IASB Supplementary Document Financial Instruments: Impairment Draft comment letter 28 February 2011.
Copyright Course Technology 1999
 To explain the importance of software configuration management (CM)  To describe key CM activities namely CM planning, change management, version management.
NO FRAUD LEFT BEHIND The Effect of New Risk Assessment Auditing Standards on Schools Runyon Kersteen Ouellette.
1 Reportnet for Noise: Feedback from member countries Colin Nugent Eionet National Reference Centres for Noise meeting Copenhagen October 2009.
© Crown copyright Met Office The EN QC system for temperature and salinity profiles Simon Good.
Argo workshop in Ghana, December North Atlantic Regional Centre S Pouliquen, V Thierry, C Coatanoan, F Gaillard, PM Poulain & al.
Database Security Outline.. Introduction Security requirement Reliability and Integrity Sensitive data Inference Multilevel databases Multilevel security.
Argo workshop in Ghana, December Argo data status & data access.
Quality Control for the World Ocean Database GSOP Quality Control Workshop June 12, 2013.
Argo & New Telecom. Systems Technical Workshop Introduction and Summary. ADMT#11 October 2010, Hamburg/Germany.
CBS-SSB STATISTICS NETHERLANDS – STATISTICS NORWAY Work Session on Statistical Data Editing Oslo, Norway, September 2012 Jeroen Pannekoek and Li-Chun.
Statistical Expertise for Sound Decision Making Quality Assurance for Census Data Processing Jean-Michel Durr 28/1/20111Fourth meeting of the TCG - Lubjana.
Syllabus Talbot Middle School 8thGrade Science Syllabus
RESEARCH METHODS Lecture 29. DATA ANALYSIS Data Analysis Data processing and analysis is part of research design – decisions already made. During analysis.
Developments at the US GDAC ADMT #10 – Toulouse, France 30 September - 2 October 2009 Mark Ignaszewski FNMOC ADMT September – 2 October 2009.
The APEX Truncated Negative Pressure Drifts (TNPDs)
10 th Argo data management 2009 Toulouse What is new at GDACs ?
10 Chapter 101 Using Menus and Validating Input Programming Logic and Design, Second Edition, Comprehensive 10.
CABLING SYSTEM WARRANTY REGISTRATION. PURPOSE OF CABLING REGISTRATION.
Summary of issues and conclusions from DMQC-4 1.Pressure Issues, especially Surface pressure offset, Druck microleak, APEXs that truncate negative pressures.
Surface pressure PROVOR/ARVOR floats Coriolis DAC V. Thierry DMQC 4 – Toulouse – September 2009.
Cognitive & Organizational Challenges of Big Data in Cyber Defence. YALAVARTHI ANUSHA 1.
CTD Data Processing Current BIO Procedure. Current Processing Software Matlab Migrating to R & Python Code Version Control SVN Migrating to GitHub.
MANAGEMENT INFORMATION SYSTEM
SeaDataNet Technical Task Group meeting JRA1 Standards Development Task 1.2 Common Data Management Protocol (for dissemination to all NODCs and JRA3) Data.
TAIYO KOBAYASHI and Shinya Minato
EIA approval process, Management plan and Monitoring
Let Auditing Be Your Superpower
On the use of the CRC for decoding Argo data
Status on anomalies detected with altimetry
Enhanced File Consistency Checking
Argo Delayed-Mode Salinity Data
Different Types of Testing
Outline RTQC goals to achieve Description of current proposal Tests
Comparisons of Argo profiles and nearby high resolution CTD stations
ServiceNow Assessments
LDV Real Driving Emissions: - Drafting of physical PEMS protocol –
GODAE Quality Control Pilot Project
International Argo Data Management Team
Present Status of APEX Float Technology
Overview of Database Security
Workshop on Erroneously-Filed Elements and Parts
Chapter 3: Project Integration Management
Sample Assessment & Governance Results
Presentation transcript:

Editing RT QC flag in delayed mode ? Virginie Thierry DMQC 4 Toulouse, 28 septembre 2009

28 Septembre 2009DMQC4 - Toulouse Description of the problem RT QC flag attributed by automatic procedures and in some cases by an operator after visual inspection In “D” mode, DM operator control those flags and change them before correcting data for sensors drift or inertial thermal mass. When QC flags are re-examined and edited in DMQC, should these long and fastidious edits be made to the RAW QC fields or to the ADJUSTED QC fields ? Only data with a QC flag of 1, 2, 3 are used by the OW method (to detects of a conductivity sensor drift).

28 Septembre 2009DMQC4 - Toulouse Example : – cycle 44 Near surface salinity data bad: should be flagged to 4 Generic problem for the CTS3 PROVOR float: the last conductivity measurement between 0 and 5db is not good. Green = flag 1 TEMPSIGMA0PSAL

28 Septembre 2009DMQC4 - Toulouse Example : – Cycle 7 Salinity offset for part of the profile. The data are not correctable : data should be flagged to 4 to not being considered by the OW method. Green = flag 1Orange = flag 3 TEMPPSALSIGMA0

28 Septembre 2009DMQC4 - Toulouse Example : – Cycle 113 Few bad salinity measurements with a RT QC flag set to 1, 3 or 4 : they all should be flagged to 4 Red = flag 4Orange = flag 3 TEMPPSALSIGMA0 Green = flag 1

28 Septembre 2009DMQC4 - Toulouse Other examples Density inversion incorrectly flagged The position QC are sometimes erroneously set to 4 Hook at the base of the profiles Flag 4 assigned to good data Flag 3 assigned to suspicious data by an operator. After verification those data are good. Etc….

28 Septembre 2009DMQC4 - Toulouse AST8 – March 2007 Most DACs followed Argo data policy of preserving the R/T QC flags on the raw fields and editing the QC flags in the adjusted fields. This practice may be inadequate in the long-term, as future efforts (say by an ARC or scientist) to revisit the drift adjustments will have to disentangle QC changes associated with poor R/T QC screening and QC flags associated with the quality of drift adjustments/thermal lag adjustments. We should consider changing the raw QC flags in DMQC so that this man-power intensive component (despiking, flagging deep hooks, correcting R/T QC errors) is made distinct from the other DMQC adjustments - thermal-lag and drift assessment. This practice could be implemented immediately.

28 Septembre 2009DMQC4 - Toulouse Brian and Susan’ recommendation When QC flags are re-examined and edited in DMQC, these edits should be made to the RAW QC fields and not the ADJUSTED QC fields. RAW QC flags 1 & 2 should then be propagated to ADJUSTED, 3 and 4 should be set to 4 in ADJUSTED and filled with missing values as per the DMQC manual.

28 Septembre 2009DMQC4 - Toulouse DMQC3 – September 2008 A discussion was carried out on whether delayed-mode operators should adjust real-time qc flags. […] Opinion from workshop participants on this AST action item was mixed, with several operators supportive of this resolution, several operators strongly against it, and the majority of the operators feeling neutral or uncertain. The reasons that some operators are against this resolution include the fact that editing real-time qc flags by subjective judgement can lead to inconsistency and can potentially make the real-time qc flags worse. Also results from the real-time qc tests will be lost. Overall the workshop participants felt that more open discussions on the ramification of this action item were needed between the AST, the real- time DACs, and the delayed-mode operators. Annie Wong will compose a letter to the AST summarising the concerns of delayed-mode operators on editing real-time qc flags, so that this topic can be discussed further at the next AST meeting. Action 9.2.6: Annie Wong to compose a letter to the AST summarising the concerns of delayed-mode operators on editing real-time qc flags, so that this topic can be discussed further at the next AST meeting.

28 Septembre 2009DMQC4 - Toulouse ADMT9 – November 2008 Before starting DMQC, DM operators look at the real time flags and correct them when necessary before running the OW method. There is discussion within DM-operator whether or not these RT flags should be provided back to DACs and overwrite the automatic flags assigned in RT. This will be discussed on argo-dm-dm mailing list. The RT Dac operators recommend to transfer these corrected RT flags from DM operators to them to clean up this RT datasets.

28 Septembre 2009DMQC4 - Toulouse AST10 – March 2009 Per the DMQC-3 workshop’s request, the AST provided guidance on the changing of real time flags when delayed mode correction is done. Action item 11: AST recommends changing of RT flags with DM correction

28 Septembre 2009DMQC4 - Toulouse Editing or not editing RT flag ? That is the question It’s time to take a decision ! Need an agreement on what to do and clear rules to ensure consistency among DM operators Adding new parameters is not the solution Although some voices are against editing RT flags (e.g. Greg Johnson), discussions on argo-dm-dm and past recommendations suggest that most of the people agree with the necessity of editing RT QC flags.

28 Septembre 2009DMQC4 - Toulouse New definition of PARAM and PARAM_QC ? Definition of PARAM (=RT data) and PARAM_QC (= RT QC flag) are misleading New definition: –PARAM contains the raw values transmitted from the floats or data converted from the raw transmitted values (DOXY for instance) –PARAM_QC contains qc flags that pertain to the values in PARAM Values in PARAM_QC are set initially in 'R' and 'A' modes by the automatic real-time tests. They are later modified in 'D' mode at levels where the qc flags are set incorrectly by the real-time procedures, and where erroneous data are not detected by the real-time procedures. PARAM_QC should reflect the best knowledge we have at a given time on the quality of PARAM

28 Septembre 2009DMQC4 - Toulouse Rules for bad/uncorrectable data and QC flag of 4 Edits of the PARAM_QC flags should be made to affect a QC flag of 4 to bad and non correctable data that were not detected by the real time tests Edits of the PARAM_QC flags should be made to affect a QC flag of 1 or 2 to good data that were identified as bad or probably bad data by the real time tests

28 Septembre 2009DMQC4 - Toulouse Case of the bad but correctable data And when a drift is detected and corrected in the adjusted fields ? What shall we do for the raw data and the associated QC ? Definition of flag 3 in R mode according to the User’s manual –Meaning: Bad data that are potentially correctable –RT comments: Test 15 (greylist) or Test 16 (gross sensor drift) or Test 17 (visual inspection) failed and all other real-time QC tests passed. These data are not to be used without scientific correction. A flag ‘3’ may be assigned by an operator during additional visual QC for bad data that may be corrected in delayed mode.

28 Septembre 2009DMQC4 - Toulouse Rules for setting PARAM_QC TO 3 My suggestion: Set PARAM_QC to 3 when the data are corrected in delayed mode –Drift or offset: the entire profile is flagged to 3; flags greater than 3 are left unchanged  spikes, hooks, density inversion identified during the first step are still flagged with a QC = 4 –Thermal mass: only the corrected level are concerned ? Ensure consistency with other procedures (grey list, gross drift test, visual inspection by RT operator) and QC definition Inform users that the raw data are bad but correctable and corrected if the data are in “D” mode.

28 Septembre 2009DMQC4 - Toulouse Need to review Brian and Susan’ recommendations When QC flags are re-examined and edited in DMQC, these edits should be made to the RAW QC fields and not the ADJUSTED QC fields. RAW QC flags 1 & 2 should then be propagated to ADJUSTED New: RAW QC flags 3 should be set to 1 or 2 in ADJUSTED and associated ADJUSTED fields should be corrected for sensor drift or thermal mass. RAW QC flags 4 should be set to 4 in ADJUSTED and filled with missing values as per the DMQC manual.