SATELLITE MEASUREMENTS
177
atmosphere brightness temperatures recorded by the sensors identified in
Table 3 is performed by a number of different agencies around the world,
leading to a variety of SST data products. For example Table 4 lists the SST
products available for European seas and the Atlantic Ocean. These are
produced in near–real time, most are publicly available and can be served for
use by operational models. Other SST sensors such as the infrared channels
on MODIS have not been included because at present they are not processed
within an operational timeframe.
Each product can be considered to be independent of the others. Even
those derived from the same satellite source by different agencies are by no
means identical because each agency has its own protocols regarding matters
such as cloud detection, atmospheric correction algorithms, rules for
compositing, confidence flags and error statistics. However there is at
present little, if any, independent validation of most of the products,
although ESA does have a formal AATSR product validation process.
The type of SST (see section 4.2.3) also differs according to the
producer. Note for example that although AATSR and AVHRR measure
radiation emitted from the sea-surface skin, the SST products from AVHRR
are classified as either subskin or bulk, and from AATSR as skin, because of
the different ways each producer calibrates the atmospheric correction.
The wide choice and apparent redundancy offered by the different
sensors in Table 3 and SST data products in Table 4 prompts the question of
which is the best to use for assimilation into ocean forecasting models.
Because the measurement of global SST from space using polar orbiting
infra-red sensors is a well established mature observational system, having
acquired useful data for more than 20 years, it might seem reasonable to
assume that it is ready to provide data for assimilation into ocean models.
However, stringent sampling requirements and a higher degree of
accuracy are now demanded for applications in both climate monitoring and
operational oceanography (Robinson and Cromwell, 2003). On closer
inspection it seems increasingly difficult to meet these requirements using
any one of the SST data products currently produced by several different
agencies. No matter what improvements are made to sensor technology or
atmospheric correction algorithms, the problem of cloud cover imposes
unavoidable limits on the use of infra-red sensors, while microwave sensors
which can penetrate the cloud are not capable of the required spatial
resolution.
The most promising way to obtain the best SST data for input to models
is by combining data from the different sensor types of Table 3 so that each
product from Table 4 complements the others (Robinson and Donlon, 2003).
Data from the AATSR provides the best absolute accuracy through that
sensor’s dual view, but coverage suffers from the narrow swath inherent in
the viewing geometry and so it cannot achieve a revisit interval appropriate
to operational applications at all latitudes. In contrast this is achieved by the
177
atmosphere brightness temperatures recorded by the sensors identified in
Table 3 is performed by a number of different agencies around the world,
leading to a variety of SST data products. For example Table 4 lists the SST
products available for European seas and the Atlantic Ocean. These are
produced in near–real time, most are publicly available and can be served for
use by operational models. Other SST sensors such as the infrared channels
on MODIS have not been included because at present they are not processed
within an operational timeframe.
Each product can be considered to be independent of the others. Even
those derived from the same satellite source by different agencies are by no
means identical because each agency has its own protocols regarding matters
such as cloud detection, atmospheric correction algorithms, rules for
compositing, confidence flags and error statistics. However there is at
present little, if any, independent validation of most of the products,
although ESA does have a formal AATSR product validation process.
The type of SST (see section 4.2.3) also differs according to the
producer. Note for example that although AATSR and AVHRR measure
radiation emitted from the sea-surface skin, the SST products from AVHRR
are classified as either subskin or bulk, and from AATSR as skin, because of
the different ways each producer calibrates the atmospheric correction.
The wide choice and apparent redundancy offered by the different
sensors in Table 3 and SST data products in Table 4 prompts the question of
which is the best to use for assimilation into ocean forecasting models.
Because the measurement of global SST from space using polar orbiting
infra-red sensors is a well established mature observational system, having
acquired useful data for more than 20 years, it might seem reasonable to
assume that it is ready to provide data for assimilation into ocean models.
However, stringent sampling requirements and a higher degree of
accuracy are now demanded for applications in both climate monitoring and
operational oceanography (Robinson and Cromwell, 2003). On closer
inspection it seems increasingly difficult to meet these requirements using
any one of the SST data products currently produced by several different
agencies. No matter what improvements are made to sensor technology or
atmospheric correction algorithms, the problem of cloud cover imposes
unavoidable limits on the use of infra-red sensors, while microwave sensors
which can penetrate the cloud are not capable of the required spatial
resolution.
The most promising way to obtain the best SST data for input to models
is by combining data from the different sensor types of Table 3 so that each
product from Table 4 complements the others (Robinson and Donlon, 2003).
Data from the AATSR provides the best absolute accuracy through that
sensor’s dual view, but coverage suffers from the narrow swath inherent in
the viewing geometry and so it cannot achieve a revisit interval appropriate
to operational applications at all latitudes. In contrast this is achieved by the