SATELLITE MEASUREMENTS
179
4.2.2
Combining data from different sources
It is one thing to decide in concept that the data from different types of
SST sensors should be combined in order to benefit from their
complementary sampling characteristics; it is another to implement a system
in which this can be done without introducing more errors that result in a
poorer assimilation performance than using a single SST data product from a
single sensor. If each of the sensors measured precisely the same value of
SST with the same accuracy then each data value could be used with equal
weight, although the coarser spatial resolution of the microwave data would
still have to be taken into account.
In practice, there are at least three factors in addition to their different
spatial resolutions which cause the different data products to deliver
different values of SST. These are:
a) different accuracy, with errors in calibration and noise leading to a
different bias and standard deviation when compared against an in situ
“standard” of SST;
b) the fact that different methods of measuring SST result in different
values even when there are no measurement errors, because they sample
different parts of the near surface thermal microstructure; and
c) there are circumstances where the measured SST varies considerably
throughout the day because of a strong diurnal variability signal.
Factor (a) is unavoidable when different sensor measurement techniques
are being used. It is best treated by obtaining a confident knowledge of the
errors (mean bias and standard deviation) associated with each measurement
type. However, it is important when estimating the errors that factors (b)
and (c) are also taken into account and do not create additional bias or
variance. Factors (b) and (c) both relate to the physical behaviour of the
upper layer of the ocean and are discussed in the next paragraph.
A particular problem that faces the use of SST measurements is that of
the difference between precisely which part of the sea surface is being
measured, as mentioned already in Section 3.3.4. Figure 8 identifies the
difference between the skin SST which is measured by an infrared
radiometer and the sub-skin SST a short distance below the surface (of order
tenths of a millimetre). They are separated by the thermal skin layer where
heat transport is restricted to molecular conductivity because of the
suppression of turbulence close to the surface. The sub-skin is typically a
few tenths of a degree warmer than the actual skin. Microwave radiometers
measure the temperature at approximately the same depth as the sub-skin.
Thus when microwave and infrared SST measurements are compared the
thermal skin layer difference must be modelled and allowed for.
179
4.2.2
Combining data from different sources
It is one thing to decide in concept that the data from different types of
SST sensors should be combined in order to benefit from their
complementary sampling characteristics; it is another to implement a system
in which this can be done without introducing more errors that result in a
poorer assimilation performance than using a single SST data product from a
single sensor. If each of the sensors measured precisely the same value of
SST with the same accuracy then each data value could be used with equal
weight, although the coarser spatial resolution of the microwave data would
still have to be taken into account.
In practice, there are at least three factors in addition to their different
spatial resolutions which cause the different data products to deliver
different values of SST. These are:
a) different accuracy, with errors in calibration and noise leading to a
different bias and standard deviation when compared against an in situ
“standard” of SST;
b) the fact that different methods of measuring SST result in different
values even when there are no measurement errors, because they sample
different parts of the near surface thermal microstructure; and
c) there are circumstances where the measured SST varies considerably
throughout the day because of a strong diurnal variability signal.
Factor (a) is unavoidable when different sensor measurement techniques
are being used. It is best treated by obtaining a confident knowledge of the
errors (mean bias and standard deviation) associated with each measurement
type. However, it is important when estimating the errors that factors (b)
and (c) are also taken into account and do not create additional bias or
variance. Factors (b) and (c) both relate to the physical behaviour of the
upper layer of the ocean and are discussed in the next paragraph.
A particular problem that faces the use of SST measurements is that of
the difference between precisely which part of the sea surface is being
measured, as mentioned already in Section 3.3.4. Figure 8 identifies the
difference between the skin SST which is measured by an infrared
radiometer and the sub-skin SST a short distance below the surface (of order
tenths of a millimetre). They are separated by the thermal skin layer where
heat transport is restricted to molecular conductivity because of the
suppression of turbulence close to the surface. The sub-skin is typically a
few tenths of a degree warmer than the actual skin. Microwave radiometers
measure the temperature at approximately the same depth as the sub-skin.
Thus when microwave and infrared SST measurements are compared the
thermal skin layer difference must be modelled and allowed for.