Fig. 1. Sample map of SST based on AVHRR data

Sea surface temperature based on AVHRR data

Sea surface temperature [°C] (SST) is determined by the remote sensing of IR radiation using an AVHRR/3 (henceforth AVHRR) in cloud-free and ice-free areas (representative of a 1 km² area). The AVHRR (Advanced Very High Resolution Radiometer) sensor is a broad-band radiometer with 5 spectral channels in the visible and infrared ranges. It is installed on TirosN/NOAA and METOP type satellites, placed in circumpolar, sun-synchronous orbits at an altitude of ca 850 km. Data from the AVHRR sensor are obtained operationally from NOAA-19, METOP-B and METOP-A satellites using the EUMETCAST system in the HRPT/AHRPT format (Level 0), and in NRT mode (delay of up to 3.5 h after recording). The SatBałtyk database also contains archival images (NOAA-15, -17, -18, -19 satellites) recorded using the University of Gdańsk’s HRPT station (from 2010 to 2012), and also acquired from the EUMETSAT archive. The spatial resolution of AVHRR data at the nadir is ca 1 km and decreases towards both edges of the band, which is ca 2500 km wide. The band width is wide enough for all or most of the Baltic Sea to be “seen” by the sensor during two consecutive orbits (at roughly 100 minute intervals). This situation recurs twice every 24 h (in the hours from 00:00 to 04:00 and from 10:00 to 14:00 for the NOAA–19 satellite and from 08:00 to 11:00 and from 10:00 to 13:00 for the METOP–A  and –B satellites). On the basis of remote-sensing data the radiation temperature relating to a very thin surface layer (thickness ca 10 µm) is measured. SST is then calculated using a semi-empirical formula, the coefficients of which are calibrated with respect to in situ measurements representative of a ca. 1 m deep layer (National Data Buoy Center, 2002).

Methodology of determining sea surface temperature

The procedure for processing AVHRR data into SST maps involves a number of important steps:

1. Calibration of the signal recorded in the various spectra channels in accordance with standard NOAA procedures, using the AAPP package.
2. Geometrical correction using an automatic correction algorithm and ASDIK image merging (Kowalewski & Krężel, 2003).
3. Detection of areas where SST cannot be measured, i.e. a sea surface hidden by clouds or covered by ice, and also of pixels with erroneous values due to technical problems with the equipment or transmission. The pixels in these areas are coded as “no data”. The MAIA3 cloud detection algorithm or other filters are used in this step. Maps on which “no data” covers more than 90% of the Baltic’s area, are treated as being of low quality and are not processed further.
4. Calculation of SST using the non-linear NLSST formula (Woźniak et al. 2008) based on the radiation temperature established in step (1) for channels 4 and 5 of the AVHRR. The NLSST equation takes account of coefficients determined empirically for particular satellites, different during the day (solar zenith angle < 80°) and at night (solar zenith angle > 90°). For transition periods (dawn/dusk) the SST is calculated as the weighted mean of both versions of the algorithm, the weight depending on the locally calculated solar zenith angle.
5. Data resampling (nearest neighbour method) to a fixed grid with a 1 km resolution.

More information on the algorithm for processing AVHRR satellite data can be found on the website of the Department of Physical Oceanography, Institute of Oceanography, University of Gdańsk: SST-AVHRR.

Fig. 2. Comparison of temperatures calculated from AVHRR data and measured in situ.

Validation (assessment of accuracy)

The accuracy of the temperature maps is assessed by analysing the differences between in situ temperatures and those obtained from remote sensing data. The in situ data for such comparisons must have been measured during a period no longer than ±3 h from the instant an image within the pixel under scrutiny was registered. For such assessments we use data from IO PAN research cruises, ICES databases and autonomous measurement buoys in the Baltic Operational Oceanographic System (BOOS). The statistical error, expressed as the standard deviation of these differences, is estimated at 0.9°C. The systematic error (mean difference) is 0.1°C.

The most common source of errors in underestimating SST from satellite data is the incorrect identification of clouds. On the other hand, SST can be overestimated when the sea surface microlayer is overheated in relation to deeper water layers (local conditions causing deviations from the statistical dependence of the radiation temperature on SST, assumed in the NLSST equation).

Interesting phenomena visible on satellite  temperature maps

Many processes taking place in the sea affect the distribution of temperature in the water. Those giving rise to temperature differences at the sea surface can be identified by remote sensing. The spatial resolution of AVHRR data is sufficient to identify such phenomena as coastal upwelling, eddies, the scattering of river plumes and thermal fronts.  More will be said about these phenomena on the Sea surface temperature page.

Link to the parameter in the SatBaltic System:

 Sea surface temperature (AVHRR) (for registered users)

Bibliografia

National Data Buoy Center, 2002, AVHRR Sea Surface Temperature Products.

Kowalewski M., Krężel A., 2003, System automatycznego dowiązania geograficznego i korekcji geometrycznej danych AVHRR, Archiwum Fotogrametrii, Kartografii i Teledetekcji, Vol. 13 B, 397-407.

Woźniak B., Krężel A., Darecki M., Woźniak S.B., Majchrowski R., Ostrowska M., Kozłowski Ł, Ficek D., Olszewski J., Dera J., 2008, Algorithm for the remote sensing of the Baltic ecosystem (DESAMBEM). Part 1: Mathematical apparatus, Oceanologia, 50(4), pp. 451-508.