March 6, 2013iQuam v21 In situ Quality Monitor (iQuam) version 2 Near-real time online Quality Control, Monitoring, and Data Serving tool for SST Cal/Val Sasha Ignatov and Feng Xu NOAA/NESDIS SST from Polar Orbiters: Use of NWP Outputs 5-7 March 2013, OSI SAF Workshop, Lannion, France

iQuam v22 Use of in situ SST CAL(L2): Derive regression coefficients/Adjust bias in RTM VAL (L2/L3/L4): Monitor Accuracy/Precision of satellite SST L2/L3/L4 SST products LEO (AVHRR, MODIS, VIIRS, ABI) & GEO (GOES, MSG, MTSAT, Electro) data are matched with in situ data SQUAM analyzes VAL statistics for L2 (ACSPO, IDPS, NAVO, OSI SAF; Geo); L3 (Pathfinder); and 10+ Level 4 products Near-Real Time NCEP GTS Data Global Telecommunication System data available in NRT Data quality is non-uniform & suboptimal, QC not available Ad hoc/Simplistic/Non-uniform QC often used in satellite Cal/Val Use of in situ SSTs at NESDIS March 6, 2013

iQuam v23 Establish near-real time online in situ Quality Monitor (iQuam) which performs the following 3 functions  Quality Control: Perform accurate/flexible QC, maximally consistent with wider Meteorological and Oceanographic communities  Monitor statistical summaries of in situ minus reference SST (stratified by ships, drifters, tropical & coastal moored; ARGO; and also for individual platforms)  Serve Qced Data online for wider SST community -Currently, only Drifters and Tropical Moorings are employed in NESDIS operational Cal/Val -Ships and coastal moorings are also included in iQuam, and ARGO floats are being added, to explore their potential use Objectives of in situ SST Quality Monitor March 6, 2013

iQuam v24 Quality Control – Consistent with UK MO CategoryCheckType of error handledPhysical basis PreprocessingDuplicate Removal Duplicates arise from multiple transmission or data set merging Identical space/time/ID PlausibilityPlausibility checks Unreasonable field valuesRange of single fields & Relationships among them Internal consistency TrackingPoints falling out of trackTravel speed exceeds limit Spike checkDiscontinuities in SST time series along track SST gradient exceeds limit External consistency Reference Check Measurements deviating far away from reference Bayesian approach (*) (Ref. SST: Daily OI SST v2) Mutual consistency Cross- platform Check Mutual verification with nearby measurements (“buddies check”) Bayesian approach (*) based on space/time correlation of SST field (Correlation model: 2-scale SOAR, Martin et al., 2002) (*) Lorenc and Hammon, 1988; Ingleby and Haddleston, 2007

CMS Blacklist of drifters are updated every 10 days iQuam QC: add flag indicating status on the CMS blacklist UK MO Black list available from Jonah Jones-Smith or Wemma Fiedler(?) Create iQuam’s “blacklist”: Performance History (PH) Add in stiu platform to ‘blacklist’ if the fraction of ‘bad’ data exceeds certain threshold (e.g. 30%) in the preceding time window (e.g. 30days) – Currently under development March 6, 2013iQuam v25 Version 2 Upgrade: Incorporate CMS and UK MO Blacklists

Comparative Analysis: to show the effectiveness of iQuam QC and CMS buoy blacklist and their difference. Drifter data are separated into three subsets: –IQ x BL: passed both iQuam QC and CMS buoy blacklist check –IQ – BL: passed iQuam QC but failed CMS blacklist check (“iQuam leakage” or “CMS false alarm”) –BL – IQ: failed iQuam QC but passed CMS blacklist check (“CMS leakage” or iQuam “false alarm”) One year data of 2011 used for this analysis Statistics of (in situ SST vs. Reynolds) are calculated March 6, 2013iQuam v26 CMS Blacklist vs. iQuam QC

No. of Obs. March 6, 2013iQuam v27 CMS Blacklist Analysis Fraction of buoys captured by CMS BL but not captured by iQuam is ~0.6%. Fraction of buoys captured by iQuam QC but not captured by the CMS black list is 4.9% of population.

March 6, 2013iQuam v28 Most BL-IQ are ‘bad’ data, indicating blacklist has leakage Bias SD IQ-BL are more like ‘noisy’ data, however, the population is negligible. CMS Blacklist Analysis

–CMS buoy blacklist appears to provide 0.6% addition to the iQuam QC –iQuam WC appears to screen out ~5% extra pixels, compared with CMS black list –More analyses are needed to understand the increase noise in data complements –iQuam QC was designed to be more conservative, to remove all potentially bad data – may screen out some diurnal warming events during the daytime –Collaboration with CMS underway to understand and resolve the differences March 6, 2013iQuam v29 Preliminary Observations from CMS Blacklist Analysis

3,654 profiles today; Good global coverage 10-day cycle (~3 SST/month) March 6, 2013iQuam v210 Version 2 Upgrade: Add ARGO How to extract SST from Argo Profiles?

–[Martin et al., 2012]: the shallowest valid data point between 3-5m depth –iQuam: The shallowest valid data point between 3-8m March 6, Add ARGO: Depth considerations iQuam v2

Use inherited ARGO QC –time/lat/lon pass ARGO QC –temperature/pressure pass ARGO QC Additional ARGO QC –Vertical spike check: temperature vs. pressure gradient –Valid range: 3dbar < pressure < 8dbar Extracted ARGO SST go through the same iQuam QC as other in situ platforms March 6, Add ARGO: QC considerations iQuam v2

March 6, 2013iQuam v213 ARGO has the most uniform coverage

March 6, 2013iQuam v214 Monthly Statistical Summaries Outliers detected by each QC check Moments of ΔT S =T in situ - T Reynolds Histograms of ΔT S

March 6, 2013 iQuam v2 15 Time Series of Monthly Statistics (1991-pr) No of Platforms No of Observations Bias in ΔT S SD of ΔT S ships drifters moorings ARGO ARGO has more platforms but smallest number of observations ARGO compare w/Reynolds less favorably than drifter or tropical moorings – Reynolds assimilated drifters but not ARGO floats?

March 6, 2013iQuam v216 Monitoring individual platforms List of platforms & individual statistics Error Rate History Time Series of ΔT S Monthly Trajectory

March 6, 2013iQuam v217 Data for Download Last monthly file updated in NRT every 6hrs. Initial QC performed on the fly. Final QC requires ~7 days. QC’ed data in HDF format available for download (1991-pr)

March 6, 2013iQuam v218 Use of iQuam data in SST Quality Monitor VAL against iQuam now available in SQUAM

March 6, 2013iQuam v219 Summary  Online in situ Quality Monitor (iQuam) -QC’es in situ data. QC consistent with UK MO (Lorenc and Hammon, 1988; Ingleby and Haddleston, 2007) -Reports statistical summaries stratified by data types (ships, drifters, tropical & coastal moored) and for individual platforms -Serves QC’ed in situ SSTs to users via Http and aftp (Latency ~6 hrs; best QC achieved with ~7days latency)  iQuam v2 under testing, to be released soon -Include ARGO floats -Incorporate CMS blacklist

March 6, 2013iQuam v220 Future Work  Initial Documentation available ate -  Enhancements -complete adding ARGO floats -Extend time series back to 1980 (currently, 1991) -Add track obs (work with Helen Beggs) -Get rid of flash player -Explore diurnal correction in reference check -Explore adding three-way error analyses (O’Carroll et al)  Use -Generate MDBs w/AVHRR, MODIS, VIIRS ACSPO’ IDPS; OSI SAF products -Explore ships & coastal moorings in satellite CAL/VAL

Backup Slides March 6, 2013iQuam v221

March 6, 2013iQuam v222 Inherited ARGO QC The Argo data system has three levels of quality control. –The first level is the real-time system that performs a set of agreed automatic checks on all float measurements. Real-time data with assigned quality flags are available to users within the hrs timeframe. –The second level of quality control is the delayed-mode system. –The third level of quality control is regional scientific analyses of all float data with other available data. The procedures for regional analyses are still to be determined. QC are done on Time, Location(lat/lon), Measurements (Temp/Pres/PSAL)

–Argo Real-Time QC iQuam v223 Inherited ARGO QC OrderTest NameDescription 1Deepest Pressure Test Check if pressure exceeds the deepest possible pressure of that float 2Platform IdentificationWMO allocated number 3Impossible Date Test 4Impossible Location Test−90 to 90 ; −180 to 180 5Position on Land Test 6Impossible Speed TestDrifting speed <3m/s 7Global Range Test a gross filter on observed values for pressure, temperature and salinity :  Pressure cannot be less than −5 dbar  Temperature in range −2.5 to 40.0°C  Salinity in range 2 to 41.0 PSU 8Regional Range Test specific ranges for observations from the Mediterranean and Red Seas further restrict what are considered sensible values 9Pressure Increasing Test requires that the profile has pressures that are monotonically increasing 10Spike Test Difference between sequential measurements, where one measurement is quite different than adjacent ones, is a spike in both size and gradient March 6, 2013

–Argo Real-Time QC (cont’l) iQuam v224 Inherited ARGO QC OrderTest NameDescription 11 Top and Bottom Spike Test obselete 12Gradient Test Check if difference between vertically adjacent measurements is too steep 13Digit Rollover TestCheck digit rollover and correct it 14Stuck Value Test Check all measurements of temperature or salinity in a profile being identical 15Density Inversion compares potential density between measurements in a profile 16Grey List The decision to insert a float parameter in the grey list comes from the PI or the delayed-mode operator. 17 Gross salinity or temperature sensor drift detect a sudden and significant sensor drift: average temperature of last 100dBar vs. previous profile, difference < 1deg 18Frozen profile detect a float that reproduces the same profile (with very small deviations) over and over again. 19Visual QC Subjective visual inspection of float values by an operator March 6, 2013

iQuam v225 Inherited ARGO QC For Pres/Temp/Psal Subjective assessment of profiles of temp vs. pres, psal vs pres or temp vs. psal, with references to nearby floats and historical data. (…like manual mutual-consistency check…) The purpose is to identify: –(a) erroneous data points that cannot be detected by the real-time qc tests –(b) vertical profiles that have the wrong shape. –Argo Delayed-Mode QC