2.

1
OBSERVATIONS AND THE PARAMETERISATION OF AIR-SEA FLUXES DURING DIAMET

Peter A. Cook * and Ian A. Renfrew

School of Environmental Sciences, University of East Anglia, Norwich, UK

1. INTRODUCTION In DIAMET we are using the FAAM BAe-146 aircraft in

low-level flying to measure the flux of momentum,
1.1 The DIAMET project sensible heat and latent heat from the sea surface
around the UK.
DIAMET, the DIAbatic influences on Mesoscale
structures in Extra-Tropical storms, is part of the UK 1.2 Air-sea fluxes
Natural Environment Research Council (NERC)
funded Storm Risk Mitigation (SRM) program. The horizontal wind at the sea surface results in
Although storms are well forecast on the synoptic- shear, turbulence and vertical flux (figure 1). Early
scale, the precise timing, location and morphology of studies of air-sea interactions due to wind shear and
the mesoscale and convective-scale structures such turbulence at the sea surface, leading to fluxes of
as strong winds and intense precipitation within these momentum, heat and moisture, were carried out by
cyclones remain uncertain. So we need to Large and Pond (1981, 1982) and by Donelan (1990).
parameterise key processes, such as the turbulence More recently Petersen and Renfrew (2009) have
from the wind over the sea surface leading to an demonstrated the ability of the FAAM BAe-146 aircraft
upwards flux of momentum, heat and moisture which to produce high-quality turbulence observations (using
adds to the storms, to better understand and forecast the eddy covariance technique) at ~40 m above the
the mesoscale structure of severe storms over the UK. sea surface during high wind speed conditions.

Figure 1: Turbulent fluxes due to wind shear at the sea surface, and the formulae for the eddy co-variance fluxes
and the bulk fluxes.

2. THEORY

2.1 The eddy covariance technique

__________________________________________
Aircraft are flown at low level over the sea surface
* Corresponding author address: Peter A. Cook, in long straight legs. The turbulent fluxes of
School of Environmental Sciences, University of East momentum, sensible heat and latent heat from the
Anglia, Norwich Research Park, Norwich, NR4 7TJ, sea surface to the boundary layer are then determined
UK; email: [email protected] from high frequency measurements of wind speed (u
v), updraft (w), air temperature (Ɵ) and humidity (q). Atmospheric Measurements (FAAM) uses a specially
The flight legs are divided into two minute runs (~12 built BAe-146 to provide a very versatile platform for
km length) and within each run the products of the instruments, it is operated by the UK Met Office and
perturbations from the mean values are calculated the National Centre for Atmospheric Science (NCAS)
(u’w’, v’w’, w’Ɵ’ and w’q’). These products are used in on behalf of NERC, and has been used in many field
eddy covariance to determine the fluxes (figure 1). campaigns around the world since 2004 (figure 2).
The aircraft is able to fly down to only 30-40 m above
2.2 Bulk fluxes the sea surface, and here measure the low level
winds, updrafts, air pressure, temperature and
However large-scale numerical weather prediction humidity, and also the sea surface temperature.
(NWP) models require parameterised bulk fluxes
between the sea surface and air, which just depend 3.2 Why an aircraft?
on the wind speed at the 10 m neutral reference
height (U10N) and on the differences in temperature But why use an aircraft to measure low level
(Ɵs – Ɵa) and humidity (qs – qa). The formulae for the winds, sea surface temperature and air-sea fluxes
bulk fluxes contain three neutral exchange coefficients rather than a ship? With an aircraft we have a
(CDN, CHN and CEN) which need to be determined platform that is independent of the sea surface, giving
(figure 1). less distortion of the airflow, able to cover larger
areas, and can quickly reach and investigate areas
3. AIRCRAFT MEASUREMENTS with interesting meteorology. Although with an aircraft
we obtain lower resolution data, require more quality
3.1 The BAe-146 control of the data and have limited time of operation.

In DIAMET we are using aircraft measurements to

determine the air-sea fluxes. The Facility for Airborne

Figure 2: The FAAM BAe-146, and the five flight paths with low-level legs during DIAMET. Though B653 is
counted as a Met Office EXMIX flight.

4. THE DATA COLLECTED AND USED Low level measurements have been obtained from 5
flights (figure 2), divided into a total of 151 straight line
4.1 The 5 DIAMET flights runs of 2 minute length (~12 km). In each low-level
run the fluxes of momentum (wind stress), sensible
For DIAMET the BAe-146 has so far been used in heat and latent heat can be calculated from the high
three flying campaigns: from Cranfield, Bedfordshire, frequency measurements by using the eddy
in September 2011, from Exeter, Devon, in November covariance technique. Although many of the runs
and December 2011, and from Cranfield in May 2012. contain mesoscale features and not pure turbulence,
and these need to be left aside as poor quality data However the sea surface temperature (SST) has
since any calculations would lead to inaccurate to be measured remotely which is more difficult.
turbulent fluxes. During each flight a Heimann Radiometer is used (at 4
Hz), though at present there are some questions over
4.2 The 19 further flights its calibration, the corrected values can be less
accurate than the uncorrected values. So many of the
Measurements from a further 19 flights with low- flights (but not all) also used an interferometer
level legs over the sea have also been used in this (ARIES) to make limited SST measurements and so
study, mostly from around the UK, and the same check the Heimann.
calculations and quality control were applied.
5.3 Use of OSTIA satellite data
5. INSTRUMENTATION
Because of this difficulty for each flight we use
5.1 In-situ instruments independent SST measurements from OSTIA – the
Operational Sea Surface Temperature and Sea Ice
Wind speed (u, v), updraft (w) and air pressure (p) Analysis, Donlon et al. (2011). The OSTIA values at
are measured by a Five-Port Pressure Measurement noon on the date of the flight at many regularly spaced
System at 32 Hz (giving a spatial resolution of ~3 m). locations are interpolated to the aircraft position for
Air temperature (Ɵ) is measured by a Rosemount every second of low level flight. For the calculations in
Sensor at 32 Hz, and air humidity (q) by a Lyman this study we use uncorrected Heimann values which
Alpha Hygrometer at 64 Hz. are offset so that they coincide with OSTIA values at
low levels. The humidity of the air at sea surface is
5.2 Remote sensing of surface temperature calculated from the temperature by assuming 98%
saturated relative humidity (for salt water).

Figure 3: Scatter plots comparing the 3 sets of sea surface temperature values, from 13 flights.

OSTIA website: http://ghrsst-pp.metoffice.com/pages/latest_analysis/ostia.html

5.4 Comparing the temperature measurements the aircraft measurements and the eddy covariances,
are detailed in Petersen and Renfrew (2009). These
We need to determine the best sea-surface include a correction for the friction velocity at the
temperature (SST) measurements during each flight surface due to the Coriolis force, Donelan (1990), and
because the values of CHN and CEN strongly depend empirical stability functions to account for the
on these. The 3 sets of SST measurements are in boundary layer stability, Paulson (1970) for unstable
good agreement for some flights (e.g. B650 and conditions.
B652), for others the ARIES and OSTIA are in good
agreement but uncorrected Heimann values are lower 7. RESULTS
by around 1.0 K (e.g. B656 and B695). This gives
confidence in the ARIES and OSTIA values. The SST 7.1 The calculated values
measurements from 13 flights are compared in figure
3 (there were no ARIES measurements during flight The preliminary results from the measurements
B653). Since the OSTIA values are typically larger are over a range of wind speeds 7 – 24 m/s (U10N at
than the uncorrected Heimann, by more than 1.0 K for 10 m), and have a range of calculated wind stresses 0
2 2
some flights, using the OSTIA values in the – 3 N/m , sensible heat fluxes -40 – 280 W/m , and
2
calculations results in significantly smaller values for latent heat fluxes -60 – 570 W/m . The
the CHN and CEN neutral exchange coefficients. instrumentation produces measurement errors in the
wind speed, updraft, temperature and humidity.
6. USING THE DATA These errors are combined according to the bulk
formulae to determine the errors on the neutral
6.1 Quality control exchange coefficients.

The data and co-variances are put through a 7.2 Plotting the fluxes
complex quality control to find the runs which contain
2
pure turbulence, and any runs which also contain The plots of wind stress (τ) vs U10N , sensible heat
larger mesoscale features (100’s of metres wide) are flux (SH) vs U10N*(Ɵs – Ɵa), and latent heat flux (LH)
left aside. First basic power spectra are taken of both vs U10N*(qs – qa) from the five flights show roughly
the data (u, v, w, Ɵ, q) and the covariances (u’w’, v’w’, linear relationships, supporting the bulk flux equations
w’Ɵ’ and w’q’) within each run, and these should have (figures 4-6). The dashed lines on the figures show
a -5/3 gradient if there is just turbulence. constant coefficient values of 0.001 and 0.002.
Then cumulative summations of the covariances
over each run are examined, and these will be close 7.3 Plotting the coefficients
to a straight line if there are no large mesoscale
features. However the neutral exchange coefficients
Finally the co-spectra of the covariances are actually depend on the wind speed U10N and plotting
calculated, and these should have almost zero the calculated values of CDN, CHN and CEN reveals the
variation at low frequencies (~1 Hz), and cumulative different dependences. Here the values from the five
summations of the co-spectra should show a smooth DIAMET flights are plotted with the values calculated
curve. The quality control process is detailed in from nineteen other flights which had low level legs
Petersen and Renfrew (2009). (figures 7-9). CDN is seen to increase sharply with
U10N, but CHN and CEN show only small increases with
6.2 Calculation of the coefficients U10N.

All the calculations used to find the wind speed

(U10N) and the three exchange coefficients (CDN, CHN
and CEN) at the 10 m neutral reference height, from
Figure 4: The momentum flux (wind stress) values from the low-level runs calculated by eddy co-variance.

Figure 5: The sensible heat flux values from the low-level runs calculated by eddy co-variance.
Figure 6: The latent heat flux values from the low-level runs calculated by eddy co-variance.

Figure 7: Momentum flux coefficient vs wind speed at the neutral reference height, low-level runs from many
flights, with the previously developed algorithms marked on. Values marked by O are from flights where the
boundary layer was very unstable, such as B656.
Figure 8: Sensible heat flux coefficient vs wind speed at the neutral reference height, low-level runs from many
flights, with the previously developed algorithms marked on. The values from flights where Ɵs and Ɵa are very
similar, such as B652, have very large errors and so are excluded here.

Figure 9: Latent heat flux coefficient vs wind speed at the neutral reference height, low-level runs from many
flights, with the previously developed algorithms marked on. The values from flights where qs and qa are very
similar, such as B652, have very large errors and so are excluded here.
7.4 Comparison to previous algorithms Hence figures 7-9 also show the algorithms for
CDN, CHN and CEN that have been developed and used
There have been a number of previous studies of by the European Centre for medium range weather
air-sea fluxes and different groups have developed forecasts (ECMWF), Smith (Smith 1988), COARE 3.0
their own bulk flux algorithms for how the three neutral (Fairall 2003), and the UK Met Office (UKMO).
exchange coefficients depend on wind speed. The new coefficient values calculated from the
However the measurements of fluxes are still limited DIAMET measurements are clearly greater at large
in number, particularly at large wind speeds where the wind speeds (U10N > 10 m/s) than the previously
coefficients increase so that the fluxes have a non- developed algorithms, many CDN values by around
linear response to the wind. So there remains 50% and many CHN and CEN values by around 20%.
considerable uncertainty in the parameterisations of Although using the OSTIA SST values instead of the
the air-sea fluxes, and the good quality fluxes uncorrected Heimann measurements has reduced the
calculated here can be used to develop and validate CHN and CEN values.
new bulk flux parameterisation algorithms.

Figure 10: Measured theta profiles from the 5 DIAMET flights.

Figure 11: Momentum flux coefficient vs stability, low-level runs from many flights.

Figure 12: Friction velocity (from covariances u’w’ and v’w’) vs stability, low-level runs from many flights.
Figure 13: Friction velocity vs mean wind speed, low-level runs from many flights.

7.5 Stability of the boundary layer

8. Conclusions
However some of the CDN values are very large
(figure 7), particularly from flight B656 and one of the The recent DIAMET flights have gone well and
other flights, and this seems to be due to unstable some good quality measurements have been obtained
boundary layers, even though empirical stability at low levels in a variety of wind speeds. These
functions have been applied in the calculations. results, along with the data from many other flights,
Figure 10 shows the measured theta profiles from the will be useful for the development and validation of
five DIAMET flights, B650 and B652 had stable bulk flux algorithms and potentially testing coupled
boundary layers were theta increases with altitude, Numerical Weather Prediction models. Although
while the other flights were in unstable boundary quality control is required to find the low level runs
layers. Figure 11 plots CDN against the air stability at containing pure turbulence and no mesoscale
the flight level (z / L), and the values are clearly features.
greater within unstable air. There is still a We are still determining the best sea surface
dependence on air stability after the empirical stability temperature (SST) measurements to use in the
functions have been applied. calculations, since for many of the flights the OSTIA
Unstable boundary layers probably set up SST values are significantly greater than the
additional turbulence which adds to the turbulence uncorrected Heimann radiometer measurements.
from the wind shear at the sea surface. Figure 12 However the limited SST measurements from the
plots the friction velocity U* (calculated from the ARIES interferometer support the OSTIA values.
covariances u’w’ and v’w’) against the air stability at Using the OSTIA SST values the coefficients CHN and
flight level (z / L), and this is also clearly greater within CEN are reduced, but these are still ~20% greater than
unstable air. Though since the friction velocity also in the algorithms developed in previous studies at
depends on the mean wind speed, the friction velocity large wind speeds (U10N > 10 m/s).
is also plotted against wind speed (figure 13). The Many of the CDN coefficients are ~50% greater at
flights within unstable boundary layers measure large wind speeds (U10N > 10 m/s) than in the
greater friction velocities at similar wind speeds than previously developed algorithms. However the flights
the other flights, following a steeper gradient and which have produced the largest values took place in
implying that there was additional turbulence. unstable boundary layers, and here vertical mixing
seems to have added further turbulence to that from
the wind shear at the sea surface. An increase in the
friction velocity U* of 15% would lead to a 50%
increase in the calculated values of CDN.
We will go onto collaborate with the DIAMET
partners on validation of the Met Office Unified Model,
and in further investigation of particular cases.

9. Acknowledgements

We wish to thank NERC for funding DIAMET

through the Storm Risk Mitigation program, and FAAM
for flying the BAe-146 and providing the detailed
measurements.

10. References

Donelan, M. A., 1990: Air-sea interaction. The Sea,

Wiley-Interscience, New York, 239-292.

Donlon, C. J., et al., 2011: The Operational Sea

Surface Temperature and Sea Ice Analysis (OSTIA).
Remote Sensing of the Environment, doi:
10.1016/j.res.2010.10.017.

Fairall, C. W., et al., 2003: Bulk parameterization of

air-sea fluxes: Updates and verification for the
COARE algorithm. J. Climate, 16, 571-591.

Large, W. G., and Pond, S., 1981: Open ocean

momentum flux measurements in moderate to strong
winds. J. Phys. Oceanogr., 11, 324-336.

Large, W. G., and Pond, S., 1982: Sensible and latent

heat flux measurements over the ocean. J. Phys.
Oceanogr., 12, 464-482.

Paulson, C. A., 1970: The mathematical

representation of wind speed and temperature in the
unstable atmospheric boundary layer. J. Appl.
Meteorol., 9, 857-861.

Petersen, G. N., and Renfrew, I. A., 2009: Aircraft-

based observations of air-sea fluxes over Denmark
Strait and the Irminger Sea during high wind speed
conditions. Q. J. R. Meteorol. Soc., 135, 2030-2045.

Smith, S. D., 1988: Coefficients for sea surface wind

stress, heat flux and wind profiles as a function of
wind speed and temperature. J. Geophys. Res., 93,
15467-15472.