On the use of Satellite Altimeter data in Argo quality control

1
On the use of Satellite Altimeter data in Argo
quality control

• Stéphanie Guinehut
• CLS, Space Oceanography Division

Argo Delayed-Mode Quality Control Workshop
(DMQC-3) 10-12/09/2008 Seattle - 1 -
2
Objective

• Background idea to check the data before the
  DMQC flag the data more quickly as soon as a
  malfunction is detected to get a cleaner data
  set in real-time global consistency check
• As, Sea Level Anomalies (SLA) from altimeter
  measurements and Dynamic Height Anomalies (DHA)
  calculated from in-situ T and S profiles are
  complementary but also strongly correlated
• ? Satellite altimeter measurements are used to
  check the quality of the Argo profiling floats
  time series
• Altimeter measurements represent the mesoscale
  and the interannual variability
  more efficient than the use of climatological
  fields
• The main idea is to compare co-located SLA and
  DHA to detect systematic or punctual errors in
  the Argo data sets time series
• Analysis are performed for each float time series
• S. Guinehut, C. Coatanoan, A.-L. Dhomps, P.-Y.
  Le Traon and G. Larnicol, 2008 On the use of
  satellite altimeter data in Argo quality control,
  accepted in JAOT.

3
Data and Method

• The main idea is to compare co-located
• Altimeter Sea Level Anomalies (SLA)
• and Dynamic Height Anomalies (DHA) from Argo T/S
  profiles
• for each Argo float time series
• Method
• DHA DH Mean-DH / SLA
• 2 times series co-located in time and space
• SLA AVISO combined maps
• DHA Argo Coriolis-GDAC data base (acquired
  around the 12th of February 2008)
• DH calculated from T/S profiles
  using a reference level at 900-m depth
• only data with POSITION_QC 0,
  1, 5
• JULD_QC 0, 1, 5
• PRES/TEMP/PSAL_QC 1 (DATA_MODER)
• PRES_ADJ/TEMP_ADJ/PSAL_ADJ_QC 1
  (DATA_MODEA/D)
• Mean-DH Levitus annual climatology,
  contemporaneous Argo climatology (2-steps
  approach)

4
Data and Method

• Very good consistencies between the two time
  series
• Impact of the delayed-mode and real-time
  adjustment

5
Method

• Method
• DHA DH Mean-DH / SLA
• Differences between DHA and SLA can arises from
• Differences in the physical content of the two
  data sets
• Problems in SLA (assumed to be perfect for the
  study)
• Problems in the Mean-DH / Inconsistencies between
  Mean-DH and DH
• Problems in DH (i.e. the Argo data set)
• In order to minimize the problems in the Mean-DH,
  when have used a 2-steps approach
• 1st Mean-DH Levitus annual climatology,
  comparisons, questionable Argo floats separated
• 2st calculation of an Argo Mean-DH consistent
  with the Argo period
• comparisons on the all data sets
• In order to take into account the differences in
  the physical content of the two data sets, mean
  representative statistics of these differences
  have first been computed

6
The Argo Mean dynamic height
Correction to the Levitus Mean Dynamic Height

General statistics (243 834 observations) Lev
itus Argo Correlation 0.81 0.83 Mean-diff -1.18
0.03 Rms-diff 35.4 30.6 (
Rms-SLA) questionable Argo floats separated
7
Impact of the Argo Mean dynamic height
Levitus mean dynamic height ? offset of 5.7 cm
due to the mean ? bad float

8
Impact of the Argo Mean dynamic height
Argo mean dynamic height ? offset reduced ?
good float

!!! The Mean-DH has a very important impact for
bias identification !!!
9
Mean representative statistics

• Computed using the same data set questionable
  floats separated
• Correlation coefficient (DHA/SLA)
• Rms of the differences (SLA-DHA)
  
  as
  of SLA variance

0.0 0.2 0.4 0.6 0.8
1.0

1. 30 50 70 90 110 130 150

10
Global results

• One point represents a time series at its mean
  position ( 4100 floats)
• Correlation coefficient (DHA/SLA)
• Rms of the differences (SLA-DHA)
  
  as
  of SLA variance

? Questionable floats can already be extracted by
comparing to the neighbours
0.0 0.2 0.4 0.6 0.8
1.0

1. 30 50 70 90 110 130 150

11
Global results

• Comparisons with the mean representative
  statistics

12
Global results

• Comparisons with the mean representative
  statistics
  ? extraction of 111
  anomalous floats

1. 30 50 70 90
   110 130 150

13
Global results

• ftp//ftp.ifremer.fr/ifremer/argo/etc/argo-ast9-it
  em13-AltimeterComparison
• List of floats to be checked
• DAC WMO INST-TYPE TYPE OF ANOM
• -------------------------------------------------
  -----------------------------
• kma 2900434 846 spikes
• meds 4900116 846 offset
• meds 51886 831 offset
• meds 51887 831 offset
• incois 2900783 846 offset
• coriolis 1900651 846 spike
• coriolis 5900198 842 ?
• coriolis 6900399 841 offset
• coriolis 69039 842 drift
• bodc 1900141 842 spike
• bodc 1900454 842 spikes
• .
• The AIC monthly report for May

14
Global results very good consistency

• The majority of floats !

Float 1900586 r 0.96 rms-diff 12.53
mean-diff -2.27 cm samples 90
15
Global results very good consistency
Float 3900133 r 0.91 rms-diff 20.44
mean-diff -0.73 cm samples 147
16
Global results very good consistency
Float 2900138 r 0.94 rms-diff 6.53
mean-diff 1.20 cm samples 112
17
Global results representative anomalies

• Spike

PSAL_ADJUSTED 0.0
Float 3900412 r 0.83 rms-diff 45.96
mean-diff -0.70 cm samples 95

• Delayed-mode values to be qualified
• GUI might help

18
Global results representative anomalies

• Problem in the Adjusted time series
• Delayed-mode value S-offset 0.015
• Real-time adjusted value S-offset 0.092

Float 3900225 r 0.45 rms-diff 142.75
mean-diff 0.20 cm samples 147

• Adjusted values in real-time to be qualified
• Float now corrected

19
Global results representative anomalies

• Systematic bias of 13 cm

Float 51886 r 0.57 rms-diff 334.00
mean-diff -13.28 cm samples 106
20
Global results representative anomalies

• Progressive drift of the salinity/pressure
  sensors

Float 1900249 r 0.00 rms-diff 1538.0
mean-diff -8.98 cm samples 152
21
Results

• Malfunction of the salinity sensor
• Salinity only partially rejected
• Comparison with previous cycle not applied??

22
Global results representative anomalies

• Float in grey list but PSAL_QC1

Float 2900783 samples 9
Float in grey list 2900783,PSAL,20070709,,4,Sensor
Problem,IN
23
Update of the results

• To come soon in AIC monthly report for
  August/September / on Coriolis ftp
• Floats have been corrected (Argo SIO program)
• New floats have been detected

Not detected by the real-time QC (comparison with
previous cycle not applied??)
24
Feedbacks from DM operators

• Only feedbacks from Argo SIO program
• ? some floats have been corrected
• ? 2 floats have been detected by the Altimetry QC
  method but data are considered good ? ex. of the
  limitation of the method

Offset of -8.9 cm for 11 measurements  High 
differences because of the Mean-DH ? frontal
zone ?? Next cycles not analyzed because
shallower than 900-m ? method to be adapted to
other depth
25
Conclusions and Perspectives

• Errors mainly detected in the real-time data set
  big big errors
• Only a few isolated example for adjusted values ?
  qualification needed
• Only a few isolated example for delayed-mode
  files ? qualification needed
• The method is efficient for big big errors
  (spikes, offset, drift) but only says that one
  of the field has a problem P?, S?,
• The method is complementary to the real-time and
  delayed-mode QC
• What the method is not able to do
• Extract small errors in high variability regions
• And very small bias (2-3 cm) in lower
  variability regions
• Future plans
• What kind of signals (in term of T/S/P) the
  method is able to detect ??
• ? Might depend on the area
• Method to be adapted to the mean max depth of
  each float

26
Feedbacks on the method

• Did you got the chance to look at the results ?
• Do you think that the analysis are useful ?
  confusing ?
• Will you be interested on having regular updates
  ?
• Other comments