respect to skin temperature while AVHRR uses a buoy bulk temperature (Minnett
personal communication).
For scientific applications, TIR SST retrievals are generally used at *1 km
spatial resolution, based on heritage instruments such as the NOAA AVHRR and
MODIS. For many purposes (e.g., global SST mapping), subsamples, or binned
data, are used at lower resolution, such as 4 or 9 km pixels. Some applications of
satellite-derived SST, such as monitoring the positions and evolution of surface
expressions of thermal fronts, or meso-scale features in the ocean, require measurement precision or relative accuracy, with absolute accuracy of lesser importance. However, for many applications, the absolute accuracy of the satellitederived SST is of prime importance (Minnett 2010). The application with the most
demanding accuracy requirement is climate research, where a multi-decadal time
series of global SST is required to detect small changes against a background of
short-term variations (NRC 2000). Analysis of SST time-series data for signatures
of change over long periods requires measurement uncertainties and errors to be
less than the anticipated signal or magnitude of variation, which is likely to be
\0.2 °C per decade and thus requires consistent and accurate SST retrievals
(Donlon 2010; Minnett 2010). Given that data records of SST may span several
satellite missions, it is crucial to ensure that the validating measurements are
themselves accurate over the climate data record period. It is also critical to
remove artifacts caused by orbital-decay effects. All of the NOAA polar-orbiting
platforms lose altitude after launch, which can result in an artificial cooling trend
in the satellite-derived temperature series if not corrected. The decay correction
varies from year to year, being greatest during periods of maximum solar activity
when higher levels of solar ultraviolet radiation heat the upper atmosphere and
increase the drag on the spacecraft (Wentz and Schabel 1998). Removing the effect
of orbital decay and the derivation of trends from multiple satellites is a complex
process that requires accurate and robust climate-monitoring systems. The accuracy that can be obtained for SST derived from TIR data is now at the limit of its
capabilities with respect to the currently available operational in situ infrastructure,
with biases achieved in global SST compared to drifting-buoy measurements
typically \0.15 °C (Donlon 2010; O’Carroll et al. 2008).
12.1.3 Quality Control
SST measurements are generally provided with a separate ‘‘quality’’ field that is
often derived from the cloud-screening portion of the retrieval algorithm. The
overall quality flag associated with each SST value is a relative assignment of SST
quality based on a hierarchical suite of tests, which allows users to mask SST values
based on the quality threshold that meets their specific needs. However, data providers derive quality fields differently, requiring special attention by the user to the
different meanings listed in the relevant documentation (Donlon 2010). For example, MODIS SST quality levels are set from 0 to 4, where 0 indicates best quality and
12 Thermal Applications
317
personal communication).
For scientific applications, TIR SST retrievals are generally used at *1 km
spatial resolution, based on heritage instruments such as the NOAA AVHRR and
MODIS. For many purposes (e.g., global SST mapping), subsamples, or binned
data, are used at lower resolution, such as 4 or 9 km pixels. Some applications of
satellite-derived SST, such as monitoring the positions and evolution of surface
expressions of thermal fronts, or meso-scale features in the ocean, require measurement precision or relative accuracy, with absolute accuracy of lesser importance. However, for many applications, the absolute accuracy of the satellitederived SST is of prime importance (Minnett 2010). The application with the most
demanding accuracy requirement is climate research, where a multi-decadal time
series of global SST is required to detect small changes against a background of
short-term variations (NRC 2000). Analysis of SST time-series data for signatures
of change over long periods requires measurement uncertainties and errors to be
less than the anticipated signal or magnitude of variation, which is likely to be
\0.2 °C per decade and thus requires consistent and accurate SST retrievals
(Donlon 2010; Minnett 2010). Given that data records of SST may span several
satellite missions, it is crucial to ensure that the validating measurements are
themselves accurate over the climate data record period. It is also critical to
remove artifacts caused by orbital-decay effects. All of the NOAA polar-orbiting
platforms lose altitude after launch, which can result in an artificial cooling trend
in the satellite-derived temperature series if not corrected. The decay correction
varies from year to year, being greatest during periods of maximum solar activity
when higher levels of solar ultraviolet radiation heat the upper atmosphere and
increase the drag on the spacecraft (Wentz and Schabel 1998). Removing the effect
of orbital decay and the derivation of trends from multiple satellites is a complex
process that requires accurate and robust climate-monitoring systems. The accuracy that can be obtained for SST derived from TIR data is now at the limit of its
capabilities with respect to the currently available operational in situ infrastructure,
with biases achieved in global SST compared to drifting-buoy measurements
typically \0.15 °C (Donlon 2010; O’Carroll et al. 2008).
12.1.3 Quality Control
SST measurements are generally provided with a separate ‘‘quality’’ field that is
often derived from the cloud-screening portion of the retrieval algorithm. The
overall quality flag associated with each SST value is a relative assignment of SST
quality based on a hierarchical suite of tests, which allows users to mask SST values
based on the quality threshold that meets their specific needs. However, data providers derive quality fields differently, requiring special attention by the user to the
different meanings listed in the relevant documentation (Donlon 2010). For example, MODIS SST quality levels are set from 0 to 4, where 0 indicates best quality and
12 Thermal Applications
317