SQUAM System

1
(c)
The SST Quality Monitor (SQUAM) Alexander
Ignatov1, Prasanjit Dash2 1NOAA/NESDIS/STAR
2Colorado State University-CIRA
7. Level-4 SST comparison
3. Benefit NRT SST diagnostics
5. AVHRR SST monitoring
Requirement

• Requirement of an online near real-time (NRT)
  tool to monitor satellite SST products for
  stability and cross-platform consistency led to
  the development of the SST Quality Monitor
  (SQUAM).
• SQUAM analyzes statistics of deviations in
  satellite SST (TS) with respect to several global
  reference SST fields (TR), e.g., Reynolds SST,
  RTG, OSTIA and ODYSSEA.
• Currently, SQUAM provides NRT diagnostics of SST
  from
• two NESDIS AVHRR systems the heritage Main Unit
  Task (MUT) and the Advanced Clear-Sky Processor
  for Oceans (ACSPO), for five platforms
  NOAA-16,-17,-18,-19 MetOp-A
• EUMETSAT OSI SAF MetOp-2 FRAC (analyses
  underway)
• Intercomparison of daily Level-4 (L4) SST
  products Reynolds SST, RTG, OSTIA and ODYSSEA

• Discussions with NCEP and presentations at GHRSST
  motivated to add L4 SST comparisons. Analyses are
  made with Hovmöller plots, maps, histograms, and
  time-series (examples given below)

(1) Trend statistical parameters in time
(check satellite SST stability and cross-platform
consistency)
Reference In situ SST
Reference Reynolds SST
Median Bias
(b)
Robust Standard Dev.
Cross-platform consistency (using
double-differencing technique)
1. Science
(c)
(a)
Fig. 1 NOAA-19 MUT night SST minus OSTIA. More
analyses at http//www.star.nesdis.noaa.gov/sod/s
st/squam/
Cross-consistency

• Customarily, satellite SSTs are validated against
  in situ SSTs. However, in situ data are sparse,
  geographically biased, of non-uniform quality,
  and have limited availability in NRT.
• SQUAM employs global analyses SST fields instead,
  as a reference. These fields cover full retrieval
  domain with a more uniform quality, and are
  available in quasi-NRT.
• Monitoring of SST is done in global difference
  space, i.e., satellite SST (TS) minus reference
  SST (TR), ?TSTS-TR.
• Probability density functions of ?TS are
  near-Gaussian (even though TS and TR are highly
  skewed). Fig. 1 shows example ?TS map for NOAA-19
  MUT minus OSTIA SST.
• Outliers are handled using robust statistics
  (Dash et al., 2009).
• Fig. 2 shows ?TS histograms before and after
  removing outliers.
• Daily L4 SSTs are also compared using one of them
  as reference.

Fig. 6 L4 SST comparisons. a) daily Reynolds
RTG (low), b) daily Reynolds OSTIA,
Dec-31-2009, c) Mean of L4 SSTs - daily
Reynolds (AVHRR AMSR-E). More L4 analyses are
available at http//www.star.nesdis.noaa.gov/sod/
sst/squam/L4
4. Outliers and histograms
Fig. 3 Time-series statistics of ?TS (MUT
AVHRR). Double-differences (bottom panels) check
cross-platform consistency.

• Global distribution of ?TS is near-Gaussian.
  First 4 moments (Mean, Standard deviation,
  Skewness, Kurtosis) are used for monitoring (cf.,
  Fig. 3).
• The outliers are removed based on Median
  4RSD criterion.

(2) Plot mean SST difference vs. observational
parameters (check satellite SST for
self-consistency)
Summary

• SQUAM is fully operational with two NESDIS AVHRR
  SST products, the heritage MUT and the newer
  ACSPO, from five platforms (NOAA-16, -17, -18,
  -19, and MetOp-A).
• Six daily L4 SSTs routinely intercompared in
  SQUAM.
• SQUAM also monitors OSI SAF FRAC SST (not
  shown).
• SEVIRI SST products were tested for GOES-R ABI
  preparedness.

MUT NOAA-18 AVHRR Night ?TS histogram
Fig. 4 MUT night ?TS vs. view zenith angle (VZA)
for two periods. A misallocation in VZA before
January 2006 was detected (left) and corrected
(right).
AVHRR SST in situ
2. Data operational and test
6. SEVIRI preliminary analyses
Partnerships / Links
Before outlier removal
After outlier removal

• Satellite SSTs NOAA-16, 17, 18, 19, MetOp-A,
  MSG
• MUT SST from 2004 to present (e.g., Ignatov et
  al., 2004)
• ACSPO SST from September 2008
• OSI SAF FRAC SST from July 2008
  (www.osi-saf.org)
• 2 months of SEVIRI SST (June 2008, January 2009)
  testing
• Reference SSTs
• Global analysis fields Reynolds (weekly and
  daily), RTG (low and high resolution), OSTIA,
  ODYSSEA
• Climate Pathfinder (ACSPO), Bauer-Robinson 1985
  (MUT)
• In situ bulk (http//www.star.nesdis.noaa.gov/sod/
  sst/calval/)

SEVIRI SSTs were also tested as a proxy for
GOES-R ABI. For SEVIRI, in addition to regression
SST, a physical (Merchant et al., 2009) and a
hybrid SST are also produced.

• OSDPD (ACSPO), NCEP (L4), GHRSST (L4), OSI SAF
  (FRAC)

Next steps / Transition path
Reference In situ match-up
Reference Daily Reynolds SST
AVHRR SST OSTIA

• Apply diurnal model and extend for ACSPO FRAC,
  NAVOCEANO
• Transition path not foreseen web-based system
  quasi-operational

Mean Bias (K)
References
Fig. 2 Probability function of MUT NOAA-18 AVHRR
night ?TS. Statistical parameters annotated are
trended in time.
Std. Dev. (K)

• Autret Piollé, 2007 ODYSSEA User manual,
  CERSAT IFREMER
• Dash et al., 2009 submitted to J.Tech.
• Gemmill et al., 2007 RTG-HR. NOAA/NWS/NCEP/MMAB
  260, 39pp
• Ignatov et al., 2004 13th AMS Conf., Norfolk,
  VA, 20-24 Sep., 2004
• Kilpatrick et al., 2001 JGR, 106, 9179-9198

• Merchant et al., 2009 RSE, 113, 445-457
• Reynolds et al., 2002 JClim, 15, 1609-1625
• Reynolds et al., 2007 JClim, 20, 5473-5496
• Stark et al., 2007 OSTIA.Oceans 07IEEE, 18-22
  June 2007, Scotland
• Thiébaux et al., 2003 RTG-R. BAMS, 84, 645-656

Fig. 5 Validation of SEVIRI SST vs. in situ and
Reynolds SST.