Southern Ocean AirSea Flux Climatologies and Uncertainties

1
Southern Ocean Air-Sea Flux Climatologies and
Uncertainties

• Outline
• Background Global Datasets /The Southern Ocean
  Sampling Problem

• 2) Current Understanding Evaluation of
  Available Fields
• 3) Future Improvements Routes to More
  Accurate Estimates of the Heat Exchange

S. A. Josey , CLIVAR Southern Ocean panel,
Sydney, Feb 16th - 19th 2009
2
Components of the Air-Sea Heat Exchange

• Example values from the NCEP/NCAR reanalysis

NCEP Climatological Annual mean (50 S, 90 E)
values - Wm-2

• Annual mean heat exchange close to zero -
  unclear whether ocean gains or loses heat over
  much of region (dataset dependent).
• Strong seasonal variation, significant spatial
  asymmetry.

Longwave -59
Latent -52
Shortwave 122
Sensible -17
Net heat flux -6 Wm-2
3
The Southern Ocean Sampling Problem

• Attempts to estimate air-sea exchanges in
  Southern Ocean are plagued by lack of
  observations.
• SST coverage from ships / floats reasonable (?).

Jan SST 2000-2004
Jul SST 2000-2004
4
The Southern Ocean Sampling Problem

• Attempts to estimate air-sea exchanges in
  Southern Ocean are plagued by lack of
  observations.
• SST coverage from ships / floats reasonable
  (?).
• However, latent heat flux requires wind speed,
  air temperature and humidity as well. Some
  coverage in summer but virtually nothing in
  winter.

Jan Latent 2000-2004
Jul Latent 2000-2004

• All ship based fields (e.g. NOC , formerly SOC)
  have significant sampling problems as do
  reanalyses.

5
Atmospheric Model Reanalyses

• Atmospheric reanalyses
• Assimilate various data types and potentially
  useful if model physics reliable. Employed in
  studies of variability (e.g. SAM, Ciasto and
  Thompson, 2007).
• Only dataset class suitable for forcing ocean
  models.
• However, not fully evaluated, suffer from biases
  in flux algorithm (Dong et al., 2007) and may
  have sampling inflicted regional biases.
• Pressing need for reanalysis evaluations

6
Annual Mean Heat Exchange

• If NCEP provides some measure of reality, what
  can we learn?
• Annual mean field shows pattern of heat gain in
  Atlantic-Indian sector and heat loss in Pacific
  sector.

80
NCEP Annual Mean Net Air-Sea Heat Flux (Red -
ocean heat gain, W m-2)
-80
7
Driving Factors for Heat Exchange

• Heat flux asymmetry initially puzzling as
  uniform wind field (arrows).

80
-80
NCEP Annual Mean Net Air-Sea Heat Flux (Red -
ocean heat gain, W m-2)
8
Driving Factors for Heat Exchange

• Heat flux asymmetry initially puzzling as
  uniform wind field (arrows) set by variations in
  sea-air temp difference.

1.5
80
-80
-1.5
NCEP Annual Mean Net Air-Sea Heat Flux (Red -
ocean heat gain, W m-2)
NCEP Annual Mean Sea-Air Temperature (Red - ocean
warmer than atmosphere, deg C)
9
Seasonal Cycle Extremes
Of course, annual mean field never actually
occurs. Instead we have strong seasonal
variation Balance between latent / sensible /
longwave heat loss and shortwave gain.
200
-200
NCEP - January Mean Net Heat Flux (W m-2)
NCEP - July Mean Net Heat Flux (W m-2)
10
Seasonal Cycle Mid-Points
Transition months (March and October) show heat
loss in most regions except Atlantic
sectorimplications for buoyancy gain of
different water masses?
200
-200
NCEP - April Mean Net Heat Flux (W m-2)
NCEP - September Mean Net Heat Flux (W m-2)
11
Comparison with ECMWF Reanalysis
Similar net heat flux asymmetry found for ECMWF
in Southern Ocean
NCEP / NCAR Net Air-Sea Heat Flux (W m-2)
ECMWF Net Air-Sea Heat Flux (W m-2)
12
Comparison with ECMWF Reanalysis
Similar net heat flux asymmetry found for ECMWF
in Southern Ocean Poor agreement in sub-tropics.
NCEP / NCAR Net Air-Sea Heat Flux (W m-2)
ECMWF Net Air-Sea Heat Flux (W m-2)
13
Coupled Model Insights
NCEP / NCAR Net Heat Flux (W m-2)

• Coupled models unaffected by sampling variations
  may indicate whether strong zonal variations in
  NCEP are realistic.
• HadCM3 coupled model comparison shows more
  zonally uniform net heat loss than
  NCEPparticularly at 50-60 S.

HadCM3 Net Heat Flux (W m-2)
14
Impact of Sampling Problem

• Comparison of NCEP net heat flux field with
  distribution of surface observations required to
  generate turbulent flux estimates.
• Suggests that patchy nature of heat loss may
  reflect variations in amount of data assimilated
  by reanalysis.
• Further work required to fully assess this.

Note data hole in Pacific sector, this is one of
the best months for coverage - January!
NCEP Mean Net Heat Flux and All January Surface
Obs for 2000-2004
15
Density Flux Field
Accurate determination of heat flux field
critical for assessing buoyancy gain of wind
driven northward flow. NCEP fields suggest
buoyancy gain in some regions and buoyancy loss
in others
NCEP Total Surface Density Flux (Purple -surface
buoyancy gain, kg m-2 s-1)
16
Thermal and Haline Density Flux Fields
Thermal term sets large scale pattern, haline
term leads to weak buoyancy gain over much of
region.
NCEP Thermal Surface Density Flux
NCEP Haline Surface Density Flux
17
Routes Forward.

• Reanalyses evaluations against high quality ship
  meteorological measurements (e.g.
  Hobart-Antarctic line - Astrolabe, S of Australia
  -Southern Surveyor - Eric Schulz, data contact).
• Further evaluations against moored surface flux
  buoys (WHOI built). One to be deployed at 47
  S.April 2010?

NCEP / NCAR Net Air-Sea Heat Flux (W m-2)
18
Interlude Can We Estimate North Atlantic MOC
Variability from Surface Fluxes Alone?

• Separate study Recently employed Walin (1982)
  water mass transformation to estimate mid-high
  latitude MOC variability from surface forced
  overturning circulation (SFOC) in several coupled
  models.
• Used 400 yrs of HadCM3 control run. Averaged
  SFOC over a 10 year interval prior to MOC
  estimate (Josey, Grist and Marsh, 2009,
  submitted).

MOC (black) and SFOC (green)
Transport Anomaly at 48 oN (Sv)

• Able to estimate significant portion of North
  Atlantic MOC variability from surface density
  fluxes alone (works over 35-65 oN range).
• Southern Hemisphere analogues?

19
Air-Sea Heat Flux Datasets

• Synthesis of satellite and reanalysis fields
  OAFLUX.
• Ship based flux datasets NOC1.1, NOC1.1a, NOC2.

• Atmospheric model reanalyses NCEP, ERA40. And
  corrected (ad hoc adjusted) versions CORE -
  Yeager and Large.
• Ocean syntheses GECCO, SOSE etc

NCEP Mean Net Heat Flux and All January Surface
Obs for 2000-2004
20
An Inconvenient Reminder.
Climatological December mean pseudo-latent heat
flux ( wind speed x sea-air humidity difference)
from all available COADS obs. 1960-1993.
Woodruff, S. et al., Phys. Chem. Earth, Vol.23,
No. 5-6, pp. 517-526, 1998
21
Objectively Analysed Air-Sea Heat Fluxes Project
(OAFlux)

• Led by Lisan Yu. Current Status Version 3 (Yu,
  Jin and Weller, WHOI Tech. Report, 2008).

• Air-sea fluxes determined following synthesis of
  satellite reanalysis meteorological variables
  (wind speed, near surface atmospheric temperature
  and humidity, SST).
• Latent and sensible heat fluxes from synthesized
  met. fields using COARE bulk flux algorithm.
• Global, 1x1 grid, monthly fields for 1958 - 2006.
  Daily for 1985 onwards. Available at
  http//oaflux.whoi.edu/.

22
OAFlux - Data Source

• Combination of reanalysis fields (black lines)
  and satellite data (magenta). Table 1 from Yu, et
  al., (2008).

1949
1958
1979
1985
2007
Wind Speed
SST
Humidity
Air Temperature
23
OAFlux - Method

• Minimise an objective function which is a
  weighted sum of the difference between each input
  dataset (NCEP, ECMWF, satellite) and the analysed
  field.
• Weighting matrices are inversely proportional to
  the errors for each dataset.
• Errors determined by reference to NOC dataset
  (assumed error free) and research buoy
  measurements.

• Problem resulting met. fields and derived
  fluxes not independent of NOC or buoy
  measurements.

24
OAFlux - Example Field Latent Heat Flux
25
OAFlux - Buoy Evaluations

• Partly taken up recommendations of Josey and
  Smith (2006) GSOP White Paper on flux dataset
  evaluation.
• Butbuoys used to tune OAFlux weighting matrices,
  not possible to determine whether OAFlux is an
  improvement over the reanalyses by this method.

• Comparison against wide range of research buoy
  measurements (107 in situ flux time series).
• Turbulent heat flux. Mean absolute difference
  OAFlux 7.4 Wm-2, 11.4 Wm-2 ERA40 , 17.3 Wm-2
  NCEP1.

26
Summary

• What do we know?
• Balance between large ocean heat gain
  (shortwave) and wind-driven heat loss (latent)
  components.
• Major seasonal cycle.
• What dont we know?
• - Sign of annual mean heat exchange over much of
  the region.
• - Key air temp and humidity fields. Particularly
  in Pacific sector.
• - Very little about freshwater budget, with
  consequent errors for the buoyancy exchange
• Potential for improvements
• - Research ship / surface flux buoy evaluations
  of reanalyses.