14 The Validation of Sea Surface Temperature Retrievals
239
calibration of ship-board radiometers is assessed by pointing them into the cone.
The radiation emerging from the cone depends not only on its temperature, as given
by the thermometers in the water bath, but also on its emissivity. The emissivity was
determined, and hence the calibration system characterized, by the NIST Transfer
Radiometer (TXR; Rice and Johnson, 1998), which is the infrared radiometric standard for the NASA Earth Observing System program (Rice and Johnson, 1996).
The TXR was also used to characterize the laboratory blackbody calibrators used
elsewhere to check the internal calibration of the ship-deployed radiometers (Rice
et al., 2004). The comparative performance of the radiometers in conditions such as
they experience in the field has been determined by mounting them together on a
short cruise of the R/V F. G. Walton Smith (Barton et al., 2004), or on a pier. The
outcome of these exercises is that the radiometers with internal calibration and an
effective correction for the reflected sky radiance are capable of measuring the skin
SST with uncertainties <0.1 K. Thus they can be used to help generate CDRs of SST.
Additional at-sea comparisons between radiometers of different design have taken
place on a more ad-hoc basis, but with similar results (e.g. Branch et al., 2008).
14.5 Uncertainties
The uncertainties in the retrievals of SST from spacecraft radiometers are often
expressed as a mean error, or bias, and a scatter, or standard deviation about the
mean; this terminology assumes a Gaussian error distribution. However, the reduction of the uncertainty fields to a pair of numbers does not indicate the complexity
of the information that is required for some applications, such as the assimilation of the data in NWP and Ocean Forecasting models (Donlon et al., 2007).
Some of the uncertainties are caused by systematic dependences on some of the
parameters that determine the conditions at the time of the satellite measurement,
such as the satellite zenith angle (through the slant path length through the atmosphere and the surface emissivity), the atmospheric water vapor and temperature
distribution, the characteristics of aerosols, and the surface wind speed. And, of
course, the uncertainties also depend on the details of the processing algorithms,
in particular the cloud screening approaches, and the effectiveness of the atmospheric correction algorithms. These algorithms are generally optimized in the sense
of producing minimum errors on a global basis, and application of the resulting
SST fields in regionally or seasonally constrained analyses can result in different error characteristics (Eugenio et al., 2005; Kumar et al., 2003; Minnett, 1990;
Shenoi, 1999).
The following summary of accuracies of SST derived from various satellite
radiometers is determined by comparison with independent measurements and
generally the discrepancies are ascribed to the satellite measurement. Since the independent measurements have their own uncertainties, and the method of comparison
introduces additional uncertainties, the SST retrievals from the satellite-based measurements are likely to be more accurate than indicated. The study of O’Carroll
et al. (2008) involved collocated measurements of three different sources of SSTs
239
calibration of ship-board radiometers is assessed by pointing them into the cone.
The radiation emerging from the cone depends not only on its temperature, as given
by the thermometers in the water bath, but also on its emissivity. The emissivity was
determined, and hence the calibration system characterized, by the NIST Transfer
Radiometer (TXR; Rice and Johnson, 1998), which is the infrared radiometric standard for the NASA Earth Observing System program (Rice and Johnson, 1996).
The TXR was also used to characterize the laboratory blackbody calibrators used
elsewhere to check the internal calibration of the ship-deployed radiometers (Rice
et al., 2004). The comparative performance of the radiometers in conditions such as
they experience in the field has been determined by mounting them together on a
short cruise of the R/V F. G. Walton Smith (Barton et al., 2004), or on a pier. The
outcome of these exercises is that the radiometers with internal calibration and an
effective correction for the reflected sky radiance are capable of measuring the skin
SST with uncertainties <0.1 K. Thus they can be used to help generate CDRs of SST.
Additional at-sea comparisons between radiometers of different design have taken
place on a more ad-hoc basis, but with similar results (e.g. Branch et al., 2008).
14.5 Uncertainties
The uncertainties in the retrievals of SST from spacecraft radiometers are often
expressed as a mean error, or bias, and a scatter, or standard deviation about the
mean; this terminology assumes a Gaussian error distribution. However, the reduction of the uncertainty fields to a pair of numbers does not indicate the complexity
of the information that is required for some applications, such as the assimilation of the data in NWP and Ocean Forecasting models (Donlon et al., 2007).
Some of the uncertainties are caused by systematic dependences on some of the
parameters that determine the conditions at the time of the satellite measurement,
such as the satellite zenith angle (through the slant path length through the atmosphere and the surface emissivity), the atmospheric water vapor and temperature
distribution, the characteristics of aerosols, and the surface wind speed. And, of
course, the uncertainties also depend on the details of the processing algorithms,
in particular the cloud screening approaches, and the effectiveness of the atmospheric correction algorithms. These algorithms are generally optimized in the sense
of producing minimum errors on a global basis, and application of the resulting
SST fields in regionally or seasonally constrained analyses can result in different error characteristics (Eugenio et al., 2005; Kumar et al., 2003; Minnett, 1990;
Shenoi, 1999).
The following summary of accuracies of SST derived from various satellite
radiometers is determined by comparison with independent measurements and
generally the discrepancies are ascribed to the satellite measurement. Since the independent measurements have their own uncertainties, and the method of comparison
introduces additional uncertainties, the SST retrievals from the satellite-based measurements are likely to be more accurate than indicated. The study of O’Carroll
et al. (2008) involved collocated measurements of three different sources of SSTs