Pathfinder –> MODIS -> VIIRS Evolution of a CDR

Robert Evans, Peter Minnett, Guillermo PodestaKay Kilpatrick (retired), Sue Walsh,

Vicki Halliwell, Liz Williams

OCRT – May 4-6, 2009

Standard SST Retrieval Equation(Pathfinder, MODIS and VIIRS)

• SST = a + b*T4 + c*(T4-T5)*Tsurface + d*(sec(q)-1)*(T4-T5)– where q is the zenith angle of the instrument– T4 and T5 are the brightness temperatures from AVHRR

channels 4 and 5, or channels 31 and 32 for MODIS– Two set of monthly coefficients are determined for

• T4 - T5 <= 0.7 (temperate to polar)• T4 - T5 > 0.7 (equatorial to temperate)

• Quality Levels– Pathfinder defines eight quality level (0-7) where 7 indicates

the highest confidence of an accurate retrieval, (0-3 for MODIS)

• Current Challenges – Mid-high latitude seasonal anomalies, – Dust aerosols and high water vapor

V5 Path – HadSST2(night)

5.1 for N7

Operational AVHRR – HadSST2

(night)

V5 Path – HadSST2

(day)

Operational AVHRR – HadSST2

(day)

+N7 +N9 +N11 +N9 +N14 +N16 +N18

Pathfinder–HADSST2 residuals for 1982 through 2007, now includes NOAA-7

Next Generation SST Improvements• Minimize High Latitude Seasonal Oscillation• Approach:

– Estimate Coefficients using 20o zonal bandscentered at the Equator, ± 2.5o transition between

bands- Coefficients estimated monthly and repeat for each year for

a given sensor (volcano periods to be handled separately)- Reference SST – Reynolds ¼o, daily Uses AMSR, Pathfinder and In situ observations

retains SST in high gradient regions

New algorithms are first developed and tested for MODIS (improved sensor characterization and more stable operating environment)

Latband algorithms will be transitioned to AVHRR as part of our SDS program

Application of LATBAND to MODIS AQUA 11 μm SST

Median of SST residuals (VALIDATION set) by quarter (2002-2008). Each line corresponds to a latitude band. Upper panel corresponds to latitude-specific SST residuals; lower panel is current

SST (CSST).

MODIS algorithm produces a ‘skin temperature’ product

Note high latitudeseasonal oscillations

Latband implementation removes seasonal residual oscillations

Application of LATBAND to MODIS AQUA 11 μm SST

Standard Deviation of SST residuals (VALIDATION set) by quarter (2002-2008). Each line corresponds to a latitude band. Upper panel corresponds to latitude-specific SST; lower panel is

current MODIS SST retrieval-buoy (CSST).

Latband S.D. order 0.4

Current algorithm S.D. order 0.5

Next Generation MODIS + VIIRS SST algorithm

• Reference SST field is the V2 daily, 0.25o Reynolds OI analysis that incorporates AMSR, Pathfinder and In Situ observations.

• Use of this higher spatial and temporal resolution field enables retention of Pathfinder and MODIS retrievals in high gradient regions.

• For MODIS and VIIRS, a 3band algorithm permits more accurate retrievals in regions of high water vapor and dust aerosols.

• The 3band algorithm significantly minimizes the presence of cold fringes around clouds.

• For MODIS (VIIRS TBD) the use of the 3-4 μm bands extends the available retrievals in the presence of high water vapor and dust aerosols relative to the coverage provided by the 11-12 μm bands.

MODIS AQUA 2 and 3 band (8.6, 11, 12μ) Night - Latband SST- Quality 1

2 Band02 Jul 06

3 Band,02 Jul 06

Difference Field

Reynolds V2-MODIS

Yellow, RedMODIS Cold

Use of Q0data

removes cold

retrievals

MODIS AQUA SST4 2 & 3 band (Night) – (3.75, 3.95, 4.05 μm)Quality 1 (Reynolds V2 – MODIS)

Latband2 band

Latband3 band

Difference Field

Reynolds V2-MODIS

Yellow,RedMODISCold

Next generation MODIS SST algorithm results

02 JUL 06Nighttime4 um SSTUsing Latband +3band algo

4um SST +cloudyareasfilled withReynoldsV2

Histograms of MODIS AQUA SST residuals Latband algorithm has reduced cold residuals

Top panel shows distribution of latitude SST residuals, bottom panel shows old SST (CSST). The vertical lines correspond to, from left to right, percentiles 0.05, 0.25, 0.50, 0.75, and 0.95. From MODIS satellite-in situ match-up database.

Residuals calculated as Satellite - Buoy

Histograms comparing Latband retrievals for 2 and 3-band nighttime SST, MODIS

AQUA night 02 Jul 06Reynolds V2 – (2 or 3) band Q0 residuals

Dashed curve is 2-band algorithm (‘cold tail’)Solid curve is 3-band algorithm (more symmetric)2-band algorithm has increased number of colder retrievals

3 band algorithm reduces

cold retrievals

AQUA MODIS Summary statistics for SST and SST4Referenced to Buoys

Current Algorithm StdDev SST ~.45-0.50 SST4 ~ 0.38 – 0.40

Data SetMin Q1 Median Mean Q3 MaxStdDev NLatitude SST

Training -2.509 -0.399 -0.164 -0.178 0.056 2.114 0.387 75011Validation -2.285 -0.403 -0.170 -0.181 0.055 2.348 0.389 40391

Dust SSTTraining -2.454 -0.432 -0.186 -0.182 0.041 1.646 0.362 75011Validation -2.688 -0.403 -0.181 -0.180 0.044 1.742 0.363 40391

Latitude SST4Training -2.218 -0.356 -0.177 -0.208 -0.017 2.237 0.346 81906Validation -2.123 -0.358 -0.177 -0.208 -0.018 2.176 0.346 44104

Dust SST4Training -2.577 -0.330 -0.198 -0.211 -0.072 0.903 0.240 81906Validation -2.516 -0.329 -0.198 -0.211 -0.071 0.927 0.242 44104

V6 MODIS & Pathfinder Status

• SEADAS code base is being used to support MODIS and is being modified to support AVHRR Pathfinder, will facilitate distribution to interested users

• MODIS Latband code recently delivered to GSFC for incorporation into SEADAS

• For AVHRR– L2GEN Pathfinder (SEADAS) has been delivered to Ken Casey,

will need to be integrated into current GSFC L2GEN version. Tested on N16, 17, 18

– Add GHRSST format output files including hypercube– Latband for NOAA sensors will be integrated when NOAA

SDS funding becomes available

Conclusions

• New algorithm approaches (Latband) has resulted in a significant reduction of uncertainty in IR satellite SST retrievals

• S.D. reduced from 0.5 to <0.4 for 11μm band retrievals, 0.4 to < 0.35 for 4μm band retrievals• Future implementation of 3 band algorithm for MODIS and

VIIRS suggests that cold fringes around clouds and aerosols can be detected and correction in these conditions is significantly improved, S.D. ~0.36 @11um, ~0.24@4um.

• Latband and 3 band algorithm approaches will be implemented for VIIRS.

• AVHRR, MODIS and VIIRS satellite observations will be available through community accessible SEADAS programs.

• Pathfinder processing is being transferred to NODC to ensure continued availability of the multi-decade Pathfinder SST time series.

END