High resolution SST products for 2001

1. High resolution SST products for 2001 • Satellite SST products and coverage • In situ observations, coverage • Quality control procedures • Satellite error statistics • Optimal interpolation • Covariance models • Product delivered to Bjarne & ODON • Elliptical Covariances • Conclusions

2. Introduction Objective • Construct an interpolated high-resolution SST product for year 2001. • Assess the data availability • Satellite products: • O&SI SAF (NOAA 14+16), resolution 2 km • BSH SST (NOAA 12) • Baltic Sea, resolution 1.2 km • North Sea, resolution 1.5 km Only nighttime observations. Day-night difference up to 1.5 deg in the mean and annual variation

3. Satellite coverage Averaged in 20 km and 3 days Nov 1- June 1 June 1-Nov 1

4. Adjust BSH SST to SAF SST • Compare coinciding SAF and BSH satellite observations • Annual variation in the differences, largest for the Baltic SST product • Adjust BSH to SAF on daily basis to obtain a consistent satellite product

5. In situ SST observations Averaged in 20 km and 3 days • From: GTS, Bouys, Ferrybox, Research cruises, etc. • ~200000 SST observations for 2001 Green: Bouys Blue: Surface Red: Profiles

6. Quality control • Construct a pseudo climatology • Cloud erosion filter • Variability criteria and standard deviation filter: No hard limits, all QC based upon statistics. • Removed 5-10 % of both satellite and in situ data. • QC: Improved the comparisons 0.2 degree std dev., equally important on in situ and satellite observations.

7. Satellite error statistics 1) Errors of the satellite products Bias used to skin-bulk SST correction Accuracy of corrected SST

8. Satellite error statistics 2) Other conclusions from report, which will be available within 2 weeks: • Differences vary substantially between different types of in situ observations • Annual variation in the in situ – satellite differences • Larger errors at cold and warm SST • Daytime observations have smaller bias and larger std dev (0.66 – 0.79 deg) Mean difference Standard deviations

9. Fill the gaps: Optimal Interpolation • Uses steady state statistics • x, y and t covariances calculated from satellite observations • Spatially varying covariance models are fitted to data • Multiplatform, individual noise on observations • Same covariance model for satellite and in situ observations • OI is performed on anomalies (annual and semiannual variations subtracted)

10. Covariance models Model determined from mean correlations In x and y r(x) = exp(ax0.5) In time r(t) = exp(at0.85)

11. Spatial variation of parameters • Regional differences in the x,y and t correlation model parameters

12. OI product to ODON/Bjarne • 10 by 10 km, grid every 12 hours in 2001 • Used 8 or 4 observations • Search radius 3 days • Areas outside domain filled with ECSSTs Predictions compared against independent observations:

13. Elliptical covariances nature not x or y oriented

14. Conclusions • SAF and BSH satellite products have been validated with in situ observations • Satellite Coverage is about 67% for North Sea and 61% for the Baltic Sea (for 20 km and 3 days averaging) • In situ coverage is 10% for the North Sea, 5 % for the Baltic Sea. • A skin—bulk correction has been applied to the satellite SSTs and the accuracy is 0.6oC. • Quality control procedures are important both for in situ and satellite observations. • Optimal interpolation in time and space has been used to fill the gaps where no in situ or satellite observation exist • 12 hourly SST fields delivered to the ODON project • The accuracy of the OI product is max 0.82oC • Elliptical covariances may be important in some areas in the region