Estimates of Surface Heat Fluxes
71
In the 1990s, a group of researchers at Southampton Oceanography Centre
(SOC), England developed a new global air–sea flux climatology. The new SOC
climatology utilised results from a careful examination of the raw data obtained by
merchant ships in the North Atlantic. The data contained standard meteorological
reports of air temperature, humidity, wind speed and direction, SST, and cloud cover,
together with additional information regarding the sensors used to make these measurements. The sensor information allowed corrections to be developed for various
observational biases, for example a warm bias in measured air temperatures due to
solar heating of the ship superstructure. Subsequently, these corrections were applied
where possible at an individual level (for the first time) to each of the 30 million
meteorological reports contained in the Comprehensive Ocean–Atmosphere Dataset
(COADS) which were used to obtain estimates of the various fluxes.
In short, the SOC researchers were doing for the merchant ship data the same
kind of hard work that the COARE researchers did on their raw data (see Josey
et al., 1998, for a full description of the SOC climatology). When the SOC climatology
was compared with fluxes from IMET buoys in the subduction region of the eastern
North Atlantic, agreement to within 10 W/m
2 was achieved at 3 of 5 sites and to within
20 W/m
2 at the other sites (Josey et al., 1999). However, a significant bias of order
50 W/m
2 was still evident in a comparison with a short-period buoy deployment
in the western North Atlantic. Little high-quality comparison data were available
elsewhere, but in the Arabian Sea, the SOC annual mean net heat flux differed from
that obtained from an IMET mooring by 18 W/m
2 (e.g., Weller et al., 1998); and
with the IMET mooring in the COARE region by 10 W/m
2 . All this looked very
encouraging. However, although the ship corrections had a significant impact at a
regional level, the global net heat flux with the SOC climatology still averaged to
30 W/m
2 —i.e., the global bias evident in earlier products had not been solved; perhaps
because there are vast areas in the Southern Hemisphere with virtually no merchant
ship traffic, and hence no surface data. Ongoing research continues to target the
various causes of this bias. An alternative approach, which uses the mathematical
technique of inverse analysis with hydrographic estimates of the ocean heat transport
as constraints, has recently led to a globally balanced version of the SOC climatology
(Grist and Josey, 2003).
Meanwhile, satellite observations improved rapidly through the 1990s and early
2000s. These have the huge advantage, compared to merchant ship data, of coming
from single sensors, many of which scan the globe at frequent intervals: the “data gap”
problem is thus immediately much reduced. There are certainly difficult problems
with radiometer drifts, and with the exact interpretation of what one is sensing; but
salt cannot encrust satellite sensors, and no birds can sit on them. They offered the
promise—given ingenuity, attention to detail, and luck—of providing a global raw
data set of more homogeneous data quality than that from merchant ships.
One example comes with the use of scatterometers for measuring wind stresses.
The improved quality of satellite-based wind stress estimates has been evident for
some years to ocean modellers, through the improvements that had been achieved in
71
In the 1990s, a group of researchers at Southampton Oceanography Centre
(SOC), England developed a new global air–sea flux climatology. The new SOC
climatology utilised results from a careful examination of the raw data obtained by
merchant ships in the North Atlantic. The data contained standard meteorological
reports of air temperature, humidity, wind speed and direction, SST, and cloud cover,
together with additional information regarding the sensors used to make these measurements. The sensor information allowed corrections to be developed for various
observational biases, for example a warm bias in measured air temperatures due to
solar heating of the ship superstructure. Subsequently, these corrections were applied
where possible at an individual level (for the first time) to each of the 30 million
meteorological reports contained in the Comprehensive Ocean–Atmosphere Dataset
(COADS) which were used to obtain estimates of the various fluxes.
In short, the SOC researchers were doing for the merchant ship data the same
kind of hard work that the COARE researchers did on their raw data (see Josey
et al., 1998, for a full description of the SOC climatology). When the SOC climatology
was compared with fluxes from IMET buoys in the subduction region of the eastern
North Atlantic, agreement to within 10 W/m
2 was achieved at 3 of 5 sites and to within
20 W/m
2 at the other sites (Josey et al., 1999). However, a significant bias of order
50 W/m
2 was still evident in a comparison with a short-period buoy deployment
in the western North Atlantic. Little high-quality comparison data were available
elsewhere, but in the Arabian Sea, the SOC annual mean net heat flux differed from
that obtained from an IMET mooring by 18 W/m
2 (e.g., Weller et al., 1998); and
with the IMET mooring in the COARE region by 10 W/m
2 . All this looked very
encouraging. However, although the ship corrections had a significant impact at a
regional level, the global net heat flux with the SOC climatology still averaged to
30 W/m
2 —i.e., the global bias evident in earlier products had not been solved; perhaps
because there are vast areas in the Southern Hemisphere with virtually no merchant
ship traffic, and hence no surface data. Ongoing research continues to target the
various causes of this bias. An alternative approach, which uses the mathematical
technique of inverse analysis with hydrographic estimates of the ocean heat transport
as constraints, has recently led to a globally balanced version of the SOC climatology
(Grist and Josey, 2003).
Meanwhile, satellite observations improved rapidly through the 1990s and early
2000s. These have the huge advantage, compared to merchant ship data, of coming
from single sensors, many of which scan the globe at frequent intervals: the “data gap”
problem is thus immediately much reduced. There are certainly difficult problems
with radiometer drifts, and with the exact interpretation of what one is sensing; but
salt cannot encrust satellite sensors, and no birds can sit on them. They offered the
promise—given ingenuity, attention to detail, and luck—of providing a global raw
data set of more homogeneous data quality than that from merchant ships.
One example comes with the use of scatterometers for measuring wind stresses.
The improved quality of satellite-based wind stress estimates has been evident for
some years to ocean modellers, through the improvements that had been achieved in