Previous - 5 Minimizing Exposure to Uncertainty Index Next - 7 Concluding Remarks and Future Directions
Although the present review is principally concerned with in situ measurements of SST it is necessary to mention the important role that satellite data play in understanding SST variability and uncertainty. The advantages of satellite data are obvious; particularly the ability to measure large areas of the ocean using a single instrument, giving a more nearly global view of SST.
However, the first thing to note is that satellites monitor radiances and do not directly measure SSTs. The measured radiances are affected by the state and constituents of the atmosphere as well as variations in the state and temperature of the sea-surface. The wavelengths that are sampled are set by the design of the instrument. Retrieving SST from the radiances is a difficult inverse process and sensitive to biases and other errors [Merchant et al. 2008b]. The second thing to note is that satellite instruments are sensitive to the skin (upper few microns), or sub-skin (upper few millimeters) temperature depending on the wavelengths measured by the satellite. Because satellite instruments are sensitive to the topmost layer of the ocean, the diurnal range of retrieved SSTs is larger than for measurements made at depth. Thirdly, accurate SST retrievals from infra-red instruments are only possible when the view is not obscured by cloud. Microwave retrievals can penetrate cloud, but suffer from problems near to coastlines, and where precipitation rates are high. They also have coarser spatial resolution and higher measurement uncertainties [O'Carroll et al., 2008].
The longest records of SST from satellite are derived from the AVHRR (Advanced Very High Resolution Radiometer) instruments. These instruments make nadir measurements using two infra-red channels. The retrievals are usually calibrated relative to in situ data. More recent re-processings use optimal estimation to obtain a retrieval that is independent of the in situ record [Merchant et al., 2008b] but these have not yet been extended to calculating global averages. Furthermore, the AVHRR instrument is prone to systematic errors caused by aerosols in the atmosphere and the satellite orbits drift slowly altering the sampling of the diurnal cycle through time. Despite the numerous shortcomings of the AVHRR record, Good et al. [2007] showed that there was a long-term warming trend in SSTs as measured by AVHRR.
The Along-Track Scanning Radiometers (ATSR) [Smith et al., 2012] were designed to meet the needs of climate monitoring. The satellite is a dual view instrument, taking nadir and forward views using three infra-red channels. The dual view configuration allows for more effective screening of aerosols and the three channels allow for accurate retrievals across a wider range of conditions. Furthermore, the onboard calibration system allows the stability of the radiance measurements from the instrument to be maintained. The ATSR data have been reprocessed in the ATSR Reanalysis for Climate (ARC) project [Merchant et al., 2008a] and the resulting time series have been shown to have biases of less than 0.1 K and stability better than 5 mK/year since 1993 in the tropics where reliable long term moorings can be found [Embury et al., 2012; Merchant et al., 2012]. The ARC reprocessing is almost independent of the in situ network therefore it can be used to corroborate trends seen in the in situ network. In a comparison between global average SST anomalies (at a nominal depth of 0.2 m) calculated using the ARC data and HadSST3, the two time series agree within the estimated HadSST3 uncertainties except for parts of the ATSR1 record in the early 1990s. The ATSR1 period is believed to be of lower quality as a result of the failure of one of the IR channels, failure of the satellite cooling system as well as the high stratospheric aerosol loadings following the eruption of Mount Pinatubo in 1991.
The nearly global, high-resolution view of the world's oceans provided by satellite instruments can be used as a way of improving and testing many aspects of SST analysis. By combining the more detailed fields produced by satellites with the long records of in situ measurements, more detailed reconstructions are possible over a wider area of the Earth [Rayner et al., 2003; Smith et al., 2008; Hirahara et al., 2013]. Satellite data can also be used to assess the verisimilitude of reconstructions based on sparser in situ data.
Previous - 5 Minimizing Exposure to Uncertainty Index Next - 7 Concluding Remarks and Future Directions