1. Page 1© Crown copyright 2004 Three-way error analysis between AATSR, AMSR-E and in situ sea surface temperature observations anne.ocarroll@metoffice.gov.uk May 2007 Anne O’Carroll, John Eyre, Roger Saunders

2. Page 2© Crown copyright 2004 Outline Aims and introduction Description of data used in analysis Methodology Results of comparison Theory of error analysis Results from error analysis Conclusions

3. Page 3© Crown copyright 2004 Aims of work  Using co-locations of three independent SST observation types to estimate the standard deviation of error on each observation type.  SST observations: AATSR; in situ; AMSR-E  Assume errors are not correlated. Attempt made to validate this assumption.

4. Page 4© Crown copyright 2004 Advanced Along Track Scanning Radiometer (AATSR) on Envisat  AATSR brightness temperature data provided on 1/6 th degree resolution in near-real time  Converted to a ‘bulk SST’ using Fairall (1996). Dual-view retrievals Uses channels 3.7,11 and 12 µm -> Dual-view, 3- channel retrievals used in this analysis

5. Page 5© Crown copyright 2004 In situ sea surface temperatures  Moored and drifting buoys used, downloaded in near-real time from the GTS Distribution of AATSR/buoy matchups for 2003 Pink = moored Blue = drifting

6. Page 6© Crown copyright 2004 Advanced Microwave Scanning Radiometer (AMSR-E) on Aqua  ¼ degree spatial resolution  Sub-skin sea surface temperature  Retrievals possible under non- precipitating cloud AMSR-E data are produced by Remote Sensing Systems and sponsored by the NASA Earth Science REASoN DISCOVER Project and the AMSR-E Science Team. Data are available at www.remss.com.

7. Page 7© Crown copyright 2004 Methodology  AATSR SSTs have been routinely collocated to buoy SST observations on a weekly basis in near-real time since September 2002.  Used AATSR/buoy matchup database for 2003.  Collocated each AATSR/buoy matchup to AMSR-E SST observation.  Calculated daily differences & then overall yearly mean differences & standard deviations between:  AATSR bulk D3n SST – AMSRE SST  Buoy – AMSRE SST  AATSR bulk D3n SST – buoy SST

8. Page 8© Crown copyright 2004 Experiments  Eight different experiments performed where certain observation and/or matchup criteria is varied to investigate whether the assumption that the errors are uncorrelated is valid. ExptRegionAATSR/buoy matchup cutoff period (hrs) Buoy type 1Global3Moored & drifting 2Global3Moored 3Global3Drifting 4Global1Moored & drifting 50º to 90ºN; 0º to 180ºW3Moored & drifting 690ºS to 0º; 0º to180ºE3Moored & drifting 7As Expt 1, but AMSR-E SSTs interpolated to AATSR location 8 As Expt 2, but AMSR-E SSTs interpolated to buoy location

9. Page 9© Crown copyright 2004 AATSR (bulk D3n) SST – AMSRE SST July 2003  Differences increase in regions ~45ºN and ~45ºS where AATSR cooler than AMSR-E and AATSR warmer than AMSR-E respectively (regions of high windspeed)

10. Page 10© Crown copyright 2004 AATSR/buoy/AMSR-E 3 point statistics  Bias corrected AATSR bulk D3, AMSR-E SSTs & buoy SSTs are co-located and the global mean differences calculated for 2003 (experiment 1):  AATSR – buoy SST = 0.00K, sd 0.28K (experiments 2-8 have standard devs ranging 0.27 to 0.30K)  AATSR – AMSR-E SST = 0.03K, sd 0.45K (experiments 2-8 have standard devs ranging 0.42 to 0.53K – highest for expt 2 moored buoys only)  buoy – AMSR-E SST = 0.03K, sd 0.48K (experiments 2-8 have standard devs ranging 0.48 to 0.57K – highest for expt 2 moored buoys only)

11. Page 11© Crown copyright 2004 Theory of error analysis  The error in observation X i, of type i, can be expressed as X i = X T + b i + E i where X T is the true value of variable X b i is the bias (mean error) in the observation E i is the random error in the observation  We can say, assuming the errors in the 3 observation types are uncorrelated, that: sd²(a,b) = (error in a)² + (error in b)² sd²(a,c) = (error in a)² + (error in c)² sd²(b,c) = (error in b)² + (error in c)²  Therefore: (error in a)²= ½(sd(a,b)²) + ½(sd(a,c)²) –½(sd(b,c)²)

12. Page 12© Crown copyright 2004 Errors calculated from 3-point analysis  Calculated error for each observation type (experiment 1): AATSR bulk D3 SST= 0.16K Buoy SST = 0.23K AMSR-E SST= 0.42K  Similar trends are seen for the other experiments, ranging:  0.12K <= error in AATSR SST <= 0.16K  0.22K <= error in buoy SST <= 0.27K  0.42K <= error in AMSR-E SST <= 0.51K

13. Page 13© Crown copyright 2004 Conclusions  Standard deviation for AATSR bulk (D3) SST observations very small at 0.16K, followed by 0.23K for buoys and 0.42K for AMSR-E SSTs.  Varying the co-location citeria produces similar values of error throughout the experiments for each observation type. Can conclude that the assumption that the errors are not correlated is valid.  Globally, differences between AATSR and AMSR-E are less than 0.5K. However, at around 45ºN the AATSR SSTs are cooler than the AMSR-E SSTs by up to 2K; whilst at around 45ºS the AATSR SSTs are warmer than AMSRE SSTs by up to 2K.

14. Page 14© Crown copyright 2004 Paper submitted to Journal of Atmospheric and Oceanic Technology