14 The Validation of Sea Surface Temperature Retrievals
231
these near-surface gradients into the error budget of the satellite retrieval and leads
to an over-estimate of the uncertainties (Kearns et al., 2000). Physical models of
the growth and decay of the diurnal thermocline (e.g. Woods and Barkmann, 1986;
Price et al., 1986; Schiller and Godfrey, 2005; Gentemann et al., 2009a) require
high temporal resolution forcing fields to produce reliable predictions, and this is
a limitation on their use in relating bulk to skin temperatures for the validation of
satellite-derived SSTs.
14.3 Required Accuracies
Some applications of satellite-derived SSTs require good precision, or relative
accuracy, and the absolute accuracy is of a lesser importance. Examples include
monitoring the positions and evolution of the surface expressions of thermal fronts
in the ocean. However for many applications, it is the accuracy of the SSTs derived
from satellite data that is of prime importance. The application with the most
demanding accuracy requirement is “climate research” where a multi-decadal time
series of global SSTs is required to detect small changes that are expected to reveal
the response of the climate to changing forcing. Analysis of a time-series of SSTs
to search for signatures of climate change will not lead to a convincing result if the
uncertainties associated with the measurements are larger than the anticipated signal, which is likely to be <0.2 K/decade which requires 15–20 years of consistent
and accurate SSTs with uncertainties <0.3 K.
Time series intended for use in Climate Research are referred to as “Climate Data
Records” (CDRs), which has been defined as “a data set designed to enable study
and assessment of long-term climate change, with ‘long-term’ meaning year-to-year
and decade-to-decade change. Climate research often involves the detection of small
changes against a background of intense, short-term variations” (NRC, 2000). To
derive CDRs from satellite data “calibration and validation should be considered
as a process that encompasses the entire system, from the sensor performance to
the derivation of the data products.” Furthermore, it is important to continue validation efforts over the lifetimes of the spacecraft sensors to ensure that the effects
of degradation of the instruments in orbit are not misinterpreted as being caused by
environmental signals (NRC, 2000). In generating time series of surface temperatures that span several satellite missions, the role of validation includes providing
the necessary continuity in the derived fields.
An important aspect of the validation exercise is sampling the full ranges of
orbital and atmospheric conditions. The orbital aspect is important as the thermal conditions on the spacecraft change markedly around the orbit, and these can
propagate to the radiometers with the consequence of a changing thermal environment in and around the instrument. A prime example is the thermal shock
experienced as the satellite enters and leaves the shadow of the earth (Brown
et al., 1985).
An example of how knowledge of the uncertainties in the SST retrieval can
be used is in the assimilation of SST retrievals in Numerical Weather Prediction
231
these near-surface gradients into the error budget of the satellite retrieval and leads
to an over-estimate of the uncertainties (Kearns et al., 2000). Physical models of
the growth and decay of the diurnal thermocline (e.g. Woods and Barkmann, 1986;
Price et al., 1986; Schiller and Godfrey, 2005; Gentemann et al., 2009a) require
high temporal resolution forcing fields to produce reliable predictions, and this is
a limitation on their use in relating bulk to skin temperatures for the validation of
satellite-derived SSTs.
14.3 Required Accuracies
Some applications of satellite-derived SSTs require good precision, or relative
accuracy, and the absolute accuracy is of a lesser importance. Examples include
monitoring the positions and evolution of the surface expressions of thermal fronts
in the ocean. However for many applications, it is the accuracy of the SSTs derived
from satellite data that is of prime importance. The application with the most
demanding accuracy requirement is “climate research” where a multi-decadal time
series of global SSTs is required to detect small changes that are expected to reveal
the response of the climate to changing forcing. Analysis of a time-series of SSTs
to search for signatures of climate change will not lead to a convincing result if the
uncertainties associated with the measurements are larger than the anticipated signal, which is likely to be <0.2 K/decade which requires 15–20 years of consistent
and accurate SSTs with uncertainties <0.3 K.
Time series intended for use in Climate Research are referred to as “Climate Data
Records” (CDRs), which has been defined as “a data set designed to enable study
and assessment of long-term climate change, with ‘long-term’ meaning year-to-year
and decade-to-decade change. Climate research often involves the detection of small
changes against a background of intense, short-term variations” (NRC, 2000). To
derive CDRs from satellite data “calibration and validation should be considered
as a process that encompasses the entire system, from the sensor performance to
the derivation of the data products.” Furthermore, it is important to continue validation efforts over the lifetimes of the spacecraft sensors to ensure that the effects
of degradation of the instruments in orbit are not misinterpreted as being caused by
environmental signals (NRC, 2000). In generating time series of surface temperatures that span several satellite missions, the role of validation includes providing
the necessary continuity in the derived fields.
An important aspect of the validation exercise is sampling the full ranges of
orbital and atmospheric conditions. The orbital aspect is important as the thermal conditions on the spacecraft change markedly around the orbit, and these can
propagate to the radiometers with the consequence of a changing thermal environment in and around the instrument. A prime example is the thermal shock
experienced as the satellite enters and leaves the shadow of the earth (Brown
et al., 1985).
An example of how knowledge of the uncertainties in the SST retrieval can
be used is in the assimilation of SST retrievals in Numerical Weather Prediction