Publications

Verification and Inter-Comparison

Title: BLUElink> Regional High-Resolution SST Analysis System - Verification and Inter-Compression

Authors: Helen Beggs BMRC, Bureau of Meteorology (BoM)
Address:GPO Box 1289, Melbourne, Vic 3001, Australia
Email: h.beggs@bom.gov.au »

ABSTRACT

As part of the BLUElink> Ocean Forecasting Australia project », the Australian Bureau of Meteorology (BoM) has modified its existing operational optimal interpolation SST analysis system (Smith et al., 1999 ») to produce 1/12° resolution, daily SST analyses over the Australian region (20°N - 70°S, 60°E - 170°W).  The new BLUElink> high-resolution analysis system combines SST data from infrared (AVHRR and AATSR) and microwave (AMSR-E) sensors on polar-orbiting satellites with in situ measurements to produce daily "foundation" SST (SSTfnd) estimates, largely free of nocturnal cooling and diurnal warming effects.   The method used in blending these data streams is described in the Regional SST Analysis PDF ».  By ~0230 UT each day, the analyses of the previous day's observations can be downloaded as GDS v1.7 L4 files from GODAE ».

The advantage for an SST analysis system in correcting input data streams to a common foundation SST is illustrated in Table 1, which indicates that filtering satellite SST observations for suspected diurnal warming events results in significant reduction in RMS error when compared with buoy observations.

Table 1.   Comparisons of SST measurements from each satellite data stream used in the BLUElink> analysis system with collocated drifting and moored buoy SST measurements over the region 20°N - 70°S, 60°E - 170°W for the period 1-30 November 2006. Data are considered "matched" if measured within same 24-hour period and centres of observations are within half the satellite sensor resolution.  The first set of columns give statistics (mean, standard deviation and total number of match-ups) for measurements at the SST data product native depth (skin, sub-skin, ~1 m) and the second set of columns give the same set of statistics for the SST data sets converted to foundation SST estimates.  The AATSR SST product was converted from skin to foundation SST using the Donlon et al. (2002) » skin to foundation temperature conversion algorithms and BoM LAPS forecast surface winds.  The remaining satellite and in situ sub-skin and bulk SST data streams were converted to foundation SST estimates by filtering out observations using the wind speed thresholds (6 m/s for day and 2 m/s for night) in Donlon et al. (2002). 

The BLUElink> daily, 1/12° resolution, SSTfnd analyses have been compared with the BoM existing operational regional daily, 1/4° resolution, SST analyses, the latest high-resolution global SST analyses from the UK Met Office and NOAA, and the 3 Day Composite AVHRR SST product from CSIRO for the period 1 to 30 November 2006 (e.g. Figure 1). Over small spatial scales, the high-resolution BoM, OSTIA and Reynolds daily analyses exhibited temperature differences of up to 2°C. The OSTIA and Reynolds analyses do not contain HRPT AVHRR SST observations over the Australian region unlike the BoM regional analyses. This and possible systematic errors in the AMSR-E SST data or diurnal warming effects may be the cause of the spatial differences. Further work needs to be done to improve the bias correction between all input data streams to within 0.05°C and improve the correction/filtering for diurnal effects. The new BLUElink> regional analyses exhibited significantly less bias and RMS error than the current BoM operational analyses when compared with SST measurements from buoys (Table 2) and independent observations (Table 3), through comparing the current day’s analysis observations with the previous day’s analysis (background field). In the region 60°E to 180°E, 20°N to 70°S, over the period 1-30 November 2006, there was closer agreement with the Met Office OSTIA analyses than with the NOAA Reynolds AMSR-E + AVHRR blended analyses (Table 4), although the BoM 1/12° regional SSTfnd analyses more closely resolved meso-scale ocean features than the finer resolution OSTIA 1/20° SSTfnd analyses (Figure 2).

Figure 1. An example of
(a) the BLUElink> daily regional 1/12° resolution SSTfnd analysis, and the BLUElink> 1/12° resolution SSTfnd analysis minus the
(b) BoM daily regional operational 1/4° resolution SST1m analysis,
(c) Met Office Operational SST and Sea Ice Analysis (OSTIA) daily 1/20° SSTfnd analysis,
(d) NOAA Reynolds et al. (2007) daily 1/4° AMSR-E + AVHRR blended SST analysis, and
(e) CSIRO 3 Day Composite 1/25° AVHRR SST1m product for 1 December 2006.
For comparison, the coarser resolution products have been linearly interpolated to the grid resolution of the higher resolution products. Note that the CSIRO product shown here is a composite of HRPT AVHRR SST data from 29 November to 1 December 2006.

Figure 2. An example of the
(a) BLUElink> daily regional 1/12° resolution SSTfnd analysis,
(b) BoM daily regional operational 1/4° resolution SST1m analysis,
(c) Met Office Operational SST and Sea Ice Analysis (OSTIA) daily 1/20° SSTfnd analysis,
(d) NOAA Reynolds et al. (2007) daily 1/4° AMSR-E + AVHRR blended SST analysis, and
(e) CSIRO 3 Day Composite 1/25° resolution AVHRR SST1m product for 1 December 2006 over the oceans south-west of Australia (110°E to 120°E, 30°S to 36°S), showing the southern section of the Leeuwin Current.
Note that the CSIRO product shown here is a composite of HRPT AVHRR SST data from 29 November to 1 December 2006.

Table 2. The average of the input observations minus background field SST (<OMF>) and its RMS error (<rms OMF>), and the average of the input observations minus analysed SST (<OMA>) and its RMS error (<rms OMA>) for the period 1 – 30 November 2006 for the BoM regional operational and test BLUElink> SST analysis systems. All statistics are in °C.

Table 3. Analysis SST minus Buoy foundation SST per day over the region 20°N - 70°S, 60°E - 180°E averaged over the period 1 - 30 November 2006, <m>. Data are considered “matched” if measured within same 24-hour period and buoy observations are within the analysis SST grid cell. <s> is the average of all daily standard deviations of the differences. <N> is the average number of daily match-ups.

Table 4. BLUElink> regional daily 1/12° resolution SSTfnd analysis minus other SST analyses or composite products over the region 20°N - 70°S, 60°E - 180°E averaged over the period 1 - 30 November 2006, <m>. <s> is the average of all daily standard deviations of the differences. For comparison, the coarser resolution products have been linearly interpolated to the grid resolution of the higher resolution products.

References

Donlon, C. J., P. Minnett, C. Gentemann, T. J. Nightingale, I. J. Barton, B. Ward and J. Murray (2002) Towards improved validation of satellite sea surface skin temperature measurements for Climate Research, J. Climate, 15, No. 4, 353-369.

Smith, N. S., B. Ebert and G. Warren (1999) The Bureau of Meteorology SST Analysis System », An informal paper produced as background for the OOPC/AOPC Workshop on SST Analyses for Climate, International Research Institute, LDEO, Palisades NY USA, Nov 10-12, 1998.