Development of New SSES for AATSR L2P

1
Development of New SSES for AATSR L2P

• Gary Corlett
• AATSR Validation Scientist

2
Content

• Reminder of current SSES scheme
• SSTs from AATSR
• What affects the quality of AATSR SSTs?
• An alternate scheme using view differences
• Examples
• Analysis of GHRSST-PP MDB
• Future work
• Summary Conclusions

3
Revised (Realistic) Content

• Reminder of current SSES scheme
• SSTs from AATSR
• What affects the quality of AATSR SSTs?
• An alternate scheme using view differences
• Examples
• Analysis of GHRSST-PP MDB
• Future work
• Summary Conclusions

4
What Affects AATSR Data Quality?
5
What Affects the Quality of AATSR SSTs?

• For December 2005 coefficients
• D2-D3 bias (the latitude dependent correction)
  owing to limited number of channels
• Small residual bias ( 0.15 K at T lt 280 K) most
  likely due to temperature dependent emissivity
• These issues are being addressed in (A)RC project
• Suspected 12 µm calibration error
• In process of analysing flight-spare FPA
• Clouds and aerosols (claerosols)
• AATSR cloud screening is over-zealous
• AATSR is designed for climate SSTs
• Much better during day owing to 1.6 µm channel
• Observable edge effects at night
• Dual-view corrects for tropospheric aerosol
• But not quite!
• Being addressed in (A)RC project
• Relative alignment of nadir and forward views
• New results suggest systematic offset
• Currently being investigated by AATSR QWG

6
Suspected 12 µm Error Implication for SSTs

• Calibration offset seen between AATSR and ATSR-2
  12 µm clear-sky SEA BTs
• ATS_MET_2P and AT2_MET_2P products analysed
• Validation results suggest AATSR is incorrect
• Results in increase in latitude dependent D2-D3
  bias
• Effect on current D3 SSTs is 0.05 K
• Effect on current D2 SSTs is - 0.04 K

7
Distribution of M-AERI Matchups
LEGEND All ve D-N -ve D-N

• Bi-modal distribution of differences
• 1 distribution due to -ve dual-nadir SST
  differences
• 1 distribution due to ve dual-nadir SST
  differences

• Matchups
• Day Overpass 60 min 1x1 Block

8
The Affect of Saharan Dust on AATSR SSTs (30)
Nov 2003
July 2003
9
Caribbean SST differences for clean months
November to MarchData is derived from 30 product
Day D2-N2
Threshold -0.03 K
Night D3-N3
Threshold 0.26 K
10
Caribbean SST differences for clean months
November to MarchData is derived from 1-km
product
Day D2-N2
Threshold 0.25 K
Night D3-N3
Threshold 0.26 K
11
Pacific Ocean clean months
12
BT
13
Schematic of BT Effects
14
Relative alignment of forward and nadir views (1)
Forward View 3.7 micron
Nadir View 3.7 micron
15
Relative alignment of forward and nadir views (2)
Forward minus Nadir View 3.7 micron
Forward minus Nadir View 3.7 micron with shift
16
Basis of New SSES Scheme
17
Requirements for AATSR SSES

• Need, per pixel
• Bias
• Retrieval bias, or
• Bias to a common standard GHRSST match-up
  database
• Standard deviation
• What exactly is a single pixel standard
  deviation?
• Quality flag
• Scale of 1-5 agreed in Melbourne
• 1 Cloudy
• 2 Bad
• 3 Acceptable (defined by provider)
• 4 Good
• 5 Excellent

18
Updated QF Scheme Theoretical Thresholds
19
Updated QF Scheme Experimental Thresholds

• T2U 0.04 0.09 C
• T2L - 1.53 0.09 C
• T3U 0.51 0.07 C
• T3L -0.51 0.07 C

20
Updated PCF Scheme - Stratification
21
Potential SSES Scheme Summary

• Effect of clouds and aerosols can be accounted
  for using D-N SST differences
• Should get well-defined normal distribution
• Can set thresholds to remove data not in normal
  distribution
• Suspected alignment error between forward and
  nadir views will contribute to D-N differences
• Significant positive edge structures still seen
  after view-difference shift
• D-N test is designed for current AATSR data
• Will need revising if and when relative alignment
  is corrected
• 12 µm calibration error will contribute to D-N
  differences
• D-N test is designed for current AATSR data
• Will need revising if and when 12 µm calibration
  error is corrected

22
Examples
23
Example 1

• AATSR Orbit ATS_NR__2PNPDE20060516_012117_0000499
  12047_00403_22000_0504.N1
• Area of interest 102 ºE to 118 ºE by 12 ºS to 4
  ºS
• Day time descending node
• The day time example demonstrates the
  over-zealous cloud-screening, as the updated PCF
  mask is mainly indicating good data, while the
  larger-scale bad data is most likely affected by
  aerosol rather than cloud. The data are for 16th
  May 2006 over Indonesia, and it is noted that the
  volcano Mount Merapi, in central Java, was fairly
  active during May 2006.

24
Comparison of N1 and L2P SSTs
25
L2P Quality Control
26
N1 D-N New QF
27
Change in QF
28
Summary of Main Changes Due to New Scheme

• Large number of pixels classified as QF 2 and QF
  3 data are now classed as QF 5
• Little change for QF 4 and QF 5 (except night
  time)
• A lot of pixels identified as bad (QF 2) by new
  scheme was previously classed as QF 3

29
Analysis of GHRSST-PP MDB
30
Definitions of SST
SSTskin SST10m (K) 0 0.5
1.0 1.5 2.0 2.5
3.0
10mm
10mm
1 mm
1 mm
Infrared sensors (dT 10s)
1 m
1 m

Depth
Depth
Contact thermometers Ships/Buoys (dT minutes)
10 m
10 m

• Night time situation, light wind
• AATSR 3 channel

(b) Day time situation, strong solar
radiation and light winds AATSR 2-channel
31
Location of GHRSST-PP Match-ups (1)
32
Location of GHRSST-PP Match-ups (2)
Others
Drifting Buoys
33
Example Day time Good data low wind speed
34
Example Night time high data low wind speed
35
Statistics from GHRSST-PP MDB Analysis
A Data below lower threshold B Data within
upper and lower thresholds C Data above upper
threshold
36
New Confidence Values and Statistics
Biases have been converted to skin biases by
adding 0.17 K
37
Summary Conclusions (1)

• Current PCF scheme used in L2P applied to AATSR
  data is appropriate for nadir-only IR with poor
  cloud screening
• Does not account for over-zealous operational
  AATSR cloud screening
• Does not account for impact of claerosols on
  forward view
• Have developed improved quality flag for AATSR
• Use D-N SST difference to identify retrievals
  affected by claerosols
• Threshold mode 3 sigma
• Also stratification according to wind speed
• Future scheme could take into account spatial and
  temporal variation in SST retrievals

38
Summary Conclusions (2)

• Issues
• Data classed as others
• No day time wind speed dependence
• May imply no significant diurnal warming has
  started or diurnal warming cancels out cool-skin
  effect
• Night-time sub-skin correction
• No current agreement on how to correct
  (consistently)
• Would only recommend QF 5 data if using AATSR as
  reference
• Requires changes to L2P product
• Addition of nadir-only SST for each dual-view SST
• Addition of flag to identify 3-channel retrievals

39
Acknowledgements

• Defra, the UK Department of Environment, Food and
  Rural Affairs, who funded AATSR to support their
  programme of climate prediction and research,
  which in turn provides inputs to their
  policy-making processes
• Funding agencies in Australia who made
  significant contributions
• The European Space Agency
• This work was funded by ESA (ESA-NOCS Contract
  AO/1-4362/03/I-LG) and Defra (Space ConneXions
  Contract 2004-03-001/CPEG10)

40
Current SSES Scheme
41
Current SSES scheme for AATSR L2P
42
(No Transcript)
43
(No Transcript)
44
(No Transcript)
45
(No Transcript)
46
(No Transcript)
47
AATSR SSTs
48
What is AATSR?

• Imaging infrared and visible radiometer on
  ENVISAT
• Similar channels to AVHRR MODIS.
• Dual view (nadir and 55 to nadir)
• Along-track scanning, two views of same scene at
  different angles, for better atmospheric
  correction
• On-board calibration
• 2 on-board black bodies for IR calibration
• VISCAL unit for visible channel calibration
• 500 km swath
• 1 km IFOV at nadir

• Stirling Cycle Coolers, cooling low noise
  detectors to 80K, for optimum signal-to-noise
  ratios

49
AATSR SST Retrievals

• Retrievals by radiative transfer modelling
• Usually of the form
• where n is the number of channels
• For AATSR we have 2 channels during day and 3
  during night
• 3.7 µm not used during day owing to solar
  contamination
• We have two views, so we have four SST retrievals
  in total
• We also have two product formats
• 1-km gridded (swath)
• Spatially averaged (gridded)

Four Possible Retrievals Nadir 2-channel
N2 Nadir 3-channel N3 Dual 2-channel D2 Dual
3-channel D3
50
Retrieval Coefficients 1-km
51
Ongoing Work
52
Spatial Pattern of AR Differences to Buoys
Data from Anne OCarroll (Met Office)
53
Further Refinement of SSES

• The (A)ATSR L2P project is aiming to make a few
  further improvement to the SSES
• The two main ones are
• To use high quality ICOADS in situ data
• Introduce spatial and temporal variation to bias
  and standard deviations

54
L2P Reformatting (2)
45 degree
20 degree
10 degree
55
L2P Reformatting (3)
45 degree
20 degree
10 degree
56
(No Transcript)
57
(No Transcript)
58
(No Transcript)
59
(No Transcript)
60
(No Transcript)
61
Example 2

• AATSR Orbit ATS_NR__2PNPDE20060516_012117_0000499
  12047_00403_22000_0504.N1
• Area of interest 70 ºW to 50 ºW by 25 ºN to 35
  ºN
• Night time ascending node
• The night time example in Section 5.1.2 clearly
  demonstrates the edge of pixel cloud
  contamination that affects AATSR data, and is
  clearly not picked up by the current PCF scheme.
  In addition, as for the day time data, there is
  no reason to suspect the rest of the SST data,
  which has been cloud screened twice and retrieved
  using two different SST retrievals. Therefore
  the data which passes the D-N test is set as good
  data and the PCF mask for these pixels is set to
  5.

62
(No Transcript)
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
(No Transcript)
68
(No Transcript)
69
(No Transcript)
70
(No Transcript)
71
AATSR and ATSR-2 Simulated Bias Centre of Swath
From Owen Embury (University of Edinburgh)
72
Compare AATSR and ATSR-2 BTs Case 1
Nadir View
Forward View