1. Introduction
Air–sea fluxes of energy establish the link between ocean surface temperature change and atmospheric circulation variability, and provide mechanisms by which ocean variability is forced by the atmosphere. Accurate knowledge of air–sea flux variability is extremely important for understanding and simulating variations in the coupled ocean–atmosphere system and feedbacks in the climate system (e.g., Curry et al. 2004; Frankignoul et al. 2004). Nowhere is this knowledge more critical than the convectively active Southern Ocean.
The Southern Ocean is the crossroad of the global ocean’s water mass, connecting the Atlantic, Pacific, and Indian Oceans as well as the deep ocean to the surface (e.g., Gordon 1988; White and Peterson 1996; Lumpkin and Speer 2007; Mayewski et al. 2009). The Southern Ocean hosts the climatologically strongest sea surface winds in the world, which drive the deep and vigorous Antarctic Circumpolar Current (ACC; e.g., Rintoul et al. 2001). The maximum flow of the current is normally concentrated in the Subantarctic Front and the Polar Front. These winds push the surface water away from the Antarctic continent through Ekman transport, creating massive divergence-driven upwelling south of the current and strongly tilting the isopycnal surfaces along the path of the ACC. The wind-driven upwelling of the Circumpolar Deep Water, which exposes the deep-water masses to surface buoyancy fluxes, produces Subantarctic Mode Water, Antarctic Intermediate Water, and Antarctic Bottom Water (e.g., Schmitz 1996; Speer et al. 2000; Sloyan and Rintoul 2001; Karsten and Marshall 2002; Talley 2003).
The unique conditions in the Southern Ocean suggest that lessons learned from tropical and subtropical oceans are not necessarily translated into improvements in the air–sea fluxes in the Southern Ocean (U.S. CLIVAR 2009). High winds over the Southern Ocean in magnitude and frequency of occurrence can exceed the speed for which scatterometer wind retrieval algorithms have been tested and the range of validity for standard drag coefficients. Ocean and atmospheric stratification in the Southern Ocean can be extremely weak, resulting in deep mixed layers, pushing the limits of existing stability parameterizations. Also, sea ice coverage in the Southern Ocean undergoes a large seasonal cycle, with a nearly fivefold increase in ice area/extent from the minimum in February to the maximum in September, which adds additional complexity to estimate air–sea fluxes.
For many of the simulations of the Atmospheric Model Intercomparison Project (available online at http://www-pcmdi.llnl.gov/projects/amip/index.php), the implied ocean heat transport in the Southern Hemisphere is equatorward. This unrealistic transport might result from deficiencies in the simulated air–sea fluxes in the Southern Ocean (e.g., Randall and Gleckler 1996; Gleckler and Weare 1997). Although the need for flux correction in coupled climate models has declined in the past several years as many modeling groups show increasing success in reducing climate drift, the tuning required to minimize this drift remains a sensitive and unfortunate necessity (e.g., Covey et al. 2006). Identifying and understanding the biases in the simulated air–sea fluxes is an important step toward resolving some potential problems in coupled climate models, leading to correct representations of the energy and water cycle.
Some studies of ocean observations have shown that much of the oceanic heat storage in response to changes in radiative forcing associated with increased loading of greenhouse gases in the atmosphere occurs in the Southern Ocean (e.g., Levitus et al. 2000; Gille 2002). Recent studies suggest that the poleward intensification of westerly winds might slow the rate of stratification in the Southern Ocean, thereby slowing the decrease in the oceanic storage of heat and anthropogenic carbon dioxide due to ocean warming and stratification (e.g., Russell et al. 2006). Since the Southern Ocean is undergoing significant climate change, the air–sea fluxes in the Southern Ocean are expected to change, where the characteristics of either the overlying atmosphere or the upper ocean are changing. These fluxes are expected to have a large influence on both atmospheric and oceanic circulation and meridional energy transport that will impact global climate in fundamental ways.
Although the air–sea fluxes in the Southern Ocean are critical to the climate system, severe undersampling has limited our observation-based knowledge of the heat exchange; thus, it is not clear whether the ocean loses or gains heat over much of this region. To date, there has been comparatively little effort on evaluating the current state of knowledge for the air–sea fluxes in the Southern Ocean as compared to the tropical and subtropical oceans. Thus, it is necessary to assess the degree of consistency and discrepancy of current-day air–sea flux products that include the Southern Ocean and disseminate the evaluation to the broader scientific community. Here, we examine the latent and sensible heat fluxes (LHF and SHF, respectively) in the Southern Ocean for five recently released air–sea flux products. We expect that the intercomparisons coming out of this study will be useful for 1) the air–sea flux community to improve bulk flux algorithms and flux estimates, 2) the modeling community to select suitable flux products for evaluating model simulations and forcing ocean models, and 3) the observational community to plan and deploy in situ flux measurements in key regions where the flux products differ substantially.
This paper includes a description of the five flux products used in this study (section 2). Section 3 evaluates similarities and differences of the heat fluxes among the flux products from the climatological perspective, including the annual mean, meridional profile of the zonal average, and seasonal cycle. Section 4 presents the variability of the heat fluxes on interannual-to-decadal time scales as well as their associations with dominant modes of climate variability in the Southern Hemisphere. Discussions and summary are given in section 5. Also, we discuss possible attributions of the identified consistency and discrepancy for some flux products based on the availability of the following input meteorological state variables: sea–air specific humidity and temperature differences, and wind speeds (note that this cannot determine which of these products uses better input meteorological state variables).
2. Description of datasets
Five different air–sea surface flux products that include the Southern Ocean are analyzed in this study (Table 1). A brief description of each product is given below.
The National Oceanography Centre, Southampton flux dataset version 2 (NOCS2) is a major update of the NOCS flux dataset version 1.1 (NOCS1.1, often referred to as the SOC flux climatology; Josey et al. 1999). The NOCS2 is based on Voluntary Observing Ship (VOS) Project from the International Comprehensive Ocean–Atmosphere Data Set (ICOADS; Worley et al. 2005), which has a spatial resolution of 1° for the period 1973–2006. Major improvements of the NOCS2 as compared to the NOCS1.1 include updated bias adjustments for air temperature, humidity, and wind speed; an improved dataset construction method using optimal interpolation; use of new ice datasets; estimates of random and bias uncertainty for fluxes; calculation of daily-mean fluxes from daily meteorological fields to retain the correlations in synoptic variability between the variables; and reduce bias due to the nonlinearity of the bulk formulas (see Berry and Kent 2009 for details).
The National Centers for Environmental Prediction (NCEP) and European Centre for Medium-Range Weather Forecasts (ECMWF) numerical weather prediction (NWP) reanalyses use improved forecast model and data assimilation systems, which are kept unchanged over the reanalysis period, to produce a consistent climate record without discontinuities due to changes in forecast models and data assimilation techniques, although the observation network itself can change (see Kanamitsu et al. 2002; Uppala et al. 2005 for details). The NCEP/Department of Energy Global Reanalysis 2 (NCEP-2) provides data at a spatial resolution of T62 (∼1.9°) from 1979 to the present. The 40-yr ECMWF Re-Analysis (ERA-40) provides data at a spatial resolution of 2.5° from September 1957 to October 2002.
The new version of the Hamburg Ocean Atmosphere Parameters and Fluxes from Satellite Data version 3 (HOAPS-3; see Andersson et al. 2007 for details) contains updated precipitation and evaporation over the global ocean and all basic-state variables needed for the derivation of the fluxes. Except for the Advanced Very High Resolution Radiometer (AVHRR) Pathfinder sea surface temperature, the other variables are derived from the Special Sensor Microwave Imager (SSM/I) over the ice-free ocean. The HOAPS-3 covers the period from July 1987 to December 2005, with 0.5° spatial resolution.
The objectively analyzed air–sea fluxes (OAFlux; see Yu and Weller 2007 for details) are constructed from a blending of surface meteorological variables from three major sources: VOS marine surface weather reports, satellite retrievals, and NWP reanalyses (Table 1). Fluxes are computed from the optimally estimated variables using version 3 of the Coupled Ocean–Atmosphere Response Experiment bulk flux algorithm (Fairall et al. 2003). The OAFlux provides data at a spatial resolution of 1° for the period 1958–2008.
Here, the NOCS2 is denoted as the analyzed product, since it is not generated from the NWP model outputs and satellite-based products. The NCEP-2 and ERA-40 are denoted as the NWP products, the HOAPS-3 is denoted as the satellite-based product, and the OAFlux is denoted as the blended product. Positive (negative) values indicate sea-to-air (air to sea) fluxes. The domain selected for this study encompasses all the oceans south of 35°S (hereafter the ACC indicates the regions between 45° and 60°S). The same “ice mask” for all the flux products is generated based on their sea ice cover or missing values to take into account the same area for calculations. The common period for all the datasets is 1989–2005, with the exception of the ERA-40, which is based on the period 1989–2001.
3. Climatology
a. Annual mean
The annual-mean LHF averaged over the Southern Ocean for 1989–2005 varies from 54.7 (NOCS2) to 69.4 W m−2 (NCEP-2) across the datasets (Fig. 1), with an average of 63.6 W m−2. Encouragingly, the spatial distribution of the LHF shows qualitative agreement for all the datasets (Fig. 2, top row). The LHF gradually decreases poleward. The strongest LHF gradient occurs in the western Indian sector at latitude ∼45°S. Despite differences in magnitude, large LHF values are located in four regions: the Agulhas retroflection region (the southern tip of Africa), the west Brazil–Malvinas confluence (just off the coast of Argentina and Uruguay), and the southeast and southwest Australian Current. Within the ACC, the LHF value in the Atlantic and Indian sectors is smaller than that in the Pacific sector. Compared to the other datasets, the HOAPS-3 has larger LHF values (ocean heat loss) in the ACC, and the difference can be as large as ∼20–30 W m−2. As reflected by the spatial distribution of the across-data standard deviation (Fig. 3a), the largest LHF discrepancy is found in the central Indian sector of the ACC, south of New Zealand, and the aforementioned four regions having large LHF values, where the value exceeds 12 W m−2. The large LHF discrepancy in the central Indian sector of the ACC and south of New Zealand might be mostly associated with the large scatter of the sea–air specific humidity difference among the datasets (not shown), whereas the large LHF discrepancies over the western boundary currents might be primarily associated with the large scatter of the wind speed in addition to the contribution from the scatter of the sea–air specific humidity difference among the datasets (not shown).
Compared to the annual-mean LHF, the annual-mean SHF is much smaller, but it varies widely across the datasets. The annual-mean SHF averaged over the Southern Ocean for 1989–2005 varies greatly from −0.2 (NCEP-2) to 20.6 W m−2 (HOAPS-3) across the datasets (Fig. 1), with an average of 10 W m−2. As shown in Fig. 2 (bottom row), the spatial pattern of the NOCS2, ERA-40, and OAFlux is similar. The Southern Ocean is dominated by positive values, but the ACC has some small-scale structures of negative values (ocean heat gain), mainly in the Atlantic and Indian sectors. By contrast, the NCEP-2 shows that almost the entire ACC is a heat gain region, and that the magnitude of negative values is significantly larger than that of the NOCS2, ERA-40, and OAFlux, with the maximum difference of ∼20 W m−2. Different from the other datasets, the HOAPS-3 exhibits significantly large positive values in the entire ACC, even larger than the values in the subtropics. The across-data standard deviation (Fig. 3b) indicates that the largest SHF discrepancy is found in the ACC, particularly in the eastern Atlantic and Indian sectors, where the value exceeds 12 W m−2. The large SHF discrepancy in the ACC might be mostly associated with the large scatter of the sea–air temperature difference among the datasets (not shown).
b. Meridonal profile of zonal mean
Figure 4a shows the meridional profile of the zonally averaged LHF over the Southern Ocean for 1989–2005. All the datasets show similar latitudinal variations, which resemble latitude variations of the sea–air specific humidity difference for the datasets (Fig. 13). The LHF monotonically decreases poleward from ∼90 to 110 to ∼25 to 40 W m−2. The NOCS2 and OAFlux tend to follow each other for north of 50°S, whereas the NWP reanalyses (the NCEP-2 and ERA-40) tend to follow each other between 40° and 60°S. The HOAPS-3 is biased high relative to the other datasets in the ACC. The spread among the datasets is the smallest near 43°S as compared to the regions to the north and south. This seems to be primarily due to the smallest sea–air specific humidity differences among the datasets there, although the wind speed of the NWP reanalyses is ∼3 m s−1 weaker than that of the other three datasets there (Fig. 13).
As shown in Fig. 4b, the meridional profile of the zonal-mean SHF for each dataset is unique, which resembles latitudinal variations of the sea–air temperature difference for the datasets (Fig. 13). Although the NOCS2, NCEP-2, ERA-40, and OAFlux show one minimum around 50°S, there is a large difference in magnitude. By contrast, the HOAPS-3 shows a maximum there. The difference between the two extremes (the NCEP-2 and HOAPS-3) can be as large as ∼35 W m−2, since the NCEP-2 has a different sign (heat gain) than the other products. The latitudinal variations vary widely south of 56°S, since there are no two datasets alike.
c. Seasonal cycle
Figure 5a shows the seasonal cycle of the LHF averaged over the Southern Ocean for 1989–2005. All the datasets suggest that the LHF reaches a maximum in late fall or winter of the Southern Hemisphere and a minimum in summer of the Southern Hemisphere. There is a lag of two months between the NOCS2 (peak in May, 64.9 W m−2) and HOAPS-3 (peaks in July, 85.3 W m−2), which is consistent with the lag (two months) in the peak sea–air specific humidity difference between the NOCS2 and HOAPS-3 (Fig. 13). The spread among the datasets in the cold season can be as large as ∼25 W m−2, a factor of 2 larger than that in the warm season. The NOCS2 shows reduced seasonal variation as compared to the other datasets.
The structure of the seasonal cycle of the SHF (Fig. 5b) is similar to that of the LHF. Again, the largest phase difference occurs between the NOCS2 and HOAPS-3. The NOCS2 shows a peak in May [13.9 W m−2, consistent with its peak for the sea–air specific humidity difference (Fig. 13)], whereas the HOAPS-3 presents a maximum in August (33.2 W m−2). The spread among the datasets in the cold season (∼20–25 W m−2) is larger than that of the warm season (∼15–20 W m−2). The NCEP-2 exhibits enhanced seasonal variation as compared to the other datasets. In addition, the NCEP-2 is the only dataset showing that the ACC is a heat gain region in late spring and summer.
4. Variability on interannual-to-decadal time scales
a. Year-to-year variability and standard deviation
Figure 6a shows the year-to-year variability of the LHF averaged over the Southern Ocean. First, the NCEP-2, ERA-40, and HOAPS-3 values are larger than the NOCS2 and OAFlux. Second, the NCEP-2, ERA-40, HOAPS-3, and OAFlux show an increase of the LHF from 1991 to 1995/96, which is primarily due to an increase of sea–air specific humidity difference (Fig. 13). Following that, the LHF of the NCEP-2 and HOAPS-3 tends to fluctuate around 70 W m−2, whereas the LHF of the ERA-40 and OAFlux shows a decreasing tendency. Different from the other datasets, the NOCS2 exhibits a small persistent increasing tendency for the entire period. As suggested by Table 2, only 1) the NCEP-2 and HOAPS-3 and 2) the reanalysis (NCEP-2 and ERA-40) and OAFlux have statistically significant correlations at 95% confidence level.
Substantial discrepancy also exists in the SHF time series (Fig. 6b). The HOAPS-3 (NCEP-2) value is significantly biased high (low) relative to the other datasets. As shown in Table 2, statistically significant correlations at 95% confidence level are only found between the reanalysis (NCEP-2 and ERA-40) and OAFlux.
We notice that for the seasonal cycle and interannual variability, the magnitude of the LHF for the NCEP-2 and HOAPS-3 is in relatively good agreement, whereas the magnitude of the SHF for the NCEP-2 is much smaller than that of the HOAPS-3. The former might be attributed to the larger turbulent exchange coefficient used in the NCEP-2 bulk flux algorithm as compared to the HOAPS-3 bulk flux algorithm, since both the sea–air specific humidity difference and wind speed of the NCEP-2 are smaller than that of the HOAPS-3 (Fig. 13). The latter might be mainly due to a combination of the much smaller sea–air temperature difference (even negative from late spring to early fall) and weaker wind speed of the NCEP-2 as compared to the HOAPS-3 (Fig. 13).
The standard deviation of the monthly anomalies (which are calculated relative to the respective monthly means) of the LHF for 1989–2005 is shown in Fig. 7 (top). The variability of the NOCS2 in the regions north of 45°S is systematically too large (∼30–60 W m−2), a factor of 2 larger than the other datasets. By contrast, the variability of the NOCS2 in the ACC is comparable to the other datasets. Despite large differences in magnitude, all the datasets indicate that the variability of the LHF decreases poleward and is generally less than 15 W m−2 in the ACC. Within the ACC, all the datasets (except the NOCS2) suggest that the Atlantic and Indian sectors seem to have weak variance relative to that in the Pacific sector.
In contrast to the poleward decrease of the LHF variability, the standard deviation of the SHF shows large variability in the ACC (∼12 W m−2) and small variability in the regions to the north (Fig. 7, bottom). Again, the variability of the NOCS2 is biased high relative to the other datasets. The NCEP-2, ERA-40, and OAFlux indicate that the locations of the largest SHF variability are in the western Indian sector (near 45°S) and the Pacific sector (near 60°S), while the HOAPS-3 shows large SHF variability in relatively similar amplitude for much of the ACC.
b. Dominant EOF mode
To determine if the primary characteristics of spatiotemporal variability of the heat flux in the Southern Ocean are consistent across the datasets for 1989–2005, we perform an empirical orthogonal function analysis on the monthly anomalies (which are calculated relative to the respective monthly means) of the LHF and SHF. Figure 8 shows the first leading EOF mode of the LHF (top row) and SHF (bottom row). Except for the NOCS2, despite differences in the magnitude of variability, the other datasets show a distinct wavenumber 3 pattern in the Southern Ocean for both the LHF and SHF, explaining ∼10%–14% of the spatial variance and reaching maximum amplitudes in the ACC. The largest negative (positive) anomalies are located over the eastern Pacific sector (the western Atlantic sector), south of Africa (the central Indian sector), and southwest of Australia (southeast of New Zealand). This reflects the meridional component of the large-scale atmospheric circulation in southern high latitudes (e.g., Raphael 2004). In contrast to the other datasets, the NOCS2 shows scattered positive and negative values over the Southern Ocean, like a noise pattern. Also, the principle component of the NOCS2 exhibits an extremely large excursion after 2000 (not shown).
c. Associations with the AAO and ENSO
The large-scale atmospheric variability in the Southern Hemisphere is dominated by two patterns of climate variability: the Antarctic Oscillation (AAO) and the remote response to the El Niño-Southern Oscillation (ENSO). The AAO is characterized by nearly zonally symmetric north–south vacillations in the latitude of the midlatitude atmospheric westerly jet, which is believed to reflect a positive feedback between the zonal flow and transient eddies in the Southern Hemisphere storm track. In contrast to the zonally symmetric structure of the AAO, the Southern Hemisphere atmospheric response to the ENSO is predominantly wavelike.
As a first step to investigate physical representations of the heat flux variability in the Southern Ocean in association with some of the most prominent large-scale modes of climate variability in the Southern Hemisphere, we calculate the correlations between the monthly anomalies of the LHF and SHF, and the AAO index (available online at http://www.esrl.noaa.gov/psd/data/climateindices) for the datasets. The threshold correlation coefficient at 90% confidence level is 0.12. As shown in Fig. 9, except for the NOCS2, despite differences in the magnitude, the other datasets show similar correlation pattern for both the LHF and SHF.
Associated with the AAO, the NOCS2 shows scattered positive and negative correlations over the Southern Ocean, which is different from the other datasets. By contrast, the other datasets show similar spatial structure and magnitude of the correlations. In response to the positive phases of the AAO, there is a more zonal structure with a decrease of the heat flux lying between 40° and 60°S throughout the far eastern Pacific, Atlantic, and western Indian Oceans and a more meridional structure with an increase of the heat flux in the central/eastern Pacific (100°–150°W) and eastern Indian Oceans, and a decrease of the heat flux south of New Zealand.
Figure 10 is analogous to Fig. 9 but shows the correlations based on the ENSO index (available online at http://www.esrl.noaa.gov/psd/data/climateindices). The most notable difference between the AAO and ENSO correlation patterns lie in their magnitude. The ENSO-related flux correlation is about half the magnitude of the AAO-related flux correlation. Again, the NOCS2 shows scattered positive and negative correlations over the Southern Ocean. By contrast, the other datasets show a wavenumber 2 pattern. Associated with positive phases of the ENSO, there is an increase of the heat flux in the central Pacific and western Atlantic sectors, and a decrease of the heat flux in the eastern Pacific and central/eastern Indian sectors.
d. Trend
Trend analysis for the monthly anomalies of the LHF is performed for 1989–2005 using the linear least squares fit regression. The spatial distribution of the LHF trends is shown in Fig. 11 (top row). The NCEP-2 and HOAPS-3, which are similar, exhibit an increase of the LHF in much of the Southern Ocean, with statistically significant increasing trends in the western Atlantic and western/central Pacific sectors more toward the subtropics. By contrast, the ERA-40 and OAFlux, which are in qualitative agreement, exhibit a decrease of the LHF in much of the ACC, with pronounced decreasing trends in the eastern Pacific, central Atlantic, and western Indian sectors.
As shown in Fig. 11 (bottom row), the SHF trends of the NCEP-2, ERA-40, and OAFlux resemble each other, showing a scattered statistically significant increase of the SHF in the subtropics and a broad-scale decrease of the SHF in the ACC, with statistically significant decreasing trends in the eastern Pacific, central Atlantic, and western Indian sectors. By contrast, for the HOAPS-3, the SHF trends are mainly positive in the Southern Ocean.
Remarkably different from the other datasets, the NOCS2 shows small-scale structures of positive and negative trends in the Southern Ocean.
5. Discussion and summary
This assessment provides a snapshot of the consistencies/discrepancies of the current-day analyzed (NOCS2), NWP modeled (the NCEP-2 and ERA-40), satellite-based (HOAPS-3), and blended (OAFlux) air–sea heat flux products over the Southern Ocean.
Our intercomparison results suggest that despite complicated atmosphere–ocean interactions in the Southern Ocean, the majority of the flux products do show encouraging qualitative agreements in the following aspects. (Note that without the observational references, the agreement of the datasets does not equally mean the accuracy of the datasets.) 1) The spatial variations of the annual-mean LHF and the meridional variations of the zonal-averaged LHF for the flux products are quite similar. 2) The dominant EOF modes of the LHF and SHF of the flux products resemble each other, characterized by a wavenumber 3 pattern. The possible reason for this good consistency is that the zonal wavenumber 3 pattern is self-sustained by a coupling between the atmosphere and the ocean: a spatially fixed atmospheric pattern and the ACC-advected oceanic anomalies. Previous studies have demonstrated that the zonal wavenumber 3 is quasi-stationary and barotropic, which is associated with the blockings in the Southern Hemisphere (Trenberth and Mo 1985). Moreover, the zonal wavenumber 3 is a dominant feature of the Southern Hemisphere atmospheric circulation (reflecting the meridional component), which contributes significantly to variability on daily, seasonal, and interannual time scales in the latitudes of 45°–55°S (Mo and White 1985; Raphael 2004; Venegas 2003). 3) From the large-scale perspective, the responses of the LHF and SHF to the AAO and ENSO are in good agreements among the flux products. This is generally consistent with the observed relationships between variability in the Southern Hemisphere extratropical atmospheric circulation, sea surface temperature, and sea ice, and the AAO and ENSO as illustrated in previous studies (e.g., Liu et al. 2004; Ciasto and Thompson 2008). These consistencies give us some confidence for utilizing the flux products to investigate associations between the heat flux and dominant climate modes of variability in the Southern Hemisphere and its feedback to the large-scale atmosphere and the ocean.
However, substantial temporal and spatial discrepancies are noteworthy, which can be attributed to the determination of the air–sea heat flux through the use of different sources of the input meteorological state variables (e.g., Curry et al. 2004) and different bulk flux algorithms (e.g., Brunke et al. 2003). On the annual basis, the HOAPS-3 shows much larger heat loss through the SHF in the ACC than the other datasets, which might be partly the effect of larger wind speed in the ACC in the HOAPS-3 as compared to that in the other datasets (Figs. 12 and 13). By contrast, the NCEP-2 suggests the entire ACC as a region of heat gain through the SHF as compared to the NOCS2, ERA-40, and OAFlux, because the ACC is dominated by negative sea–air temperature differences in the NCEP-2 as compared to that in the other datasets (Figs. 12 and 13). The zonal mean of the SHF varies widely across the datasets, and the difference can be ∼35 W m−2 in the central ACC. The largest across-data scatter is found in the central Indian sector of the ACC for the LHF and in the Atlantic and Indian sectors of the ACC for the SHF, which is comparable to and even larger than their corresponding interannual variability, highlighting a large uncertainty for estimating climate change and climate sensitivity in the Southern Ocean. Dependent on seasons, the spread of heat flux across the flux products can vary from ∼25 to ∼10 W m−2.
The flux products have interannual variability of various amplitudes and decadal trends of different signs. The ERA-40 and OAFlux show a decrease of the LHF in much of the ACC, while the opposite is the case for the NCEP-2 and HOAPS-3. The NCEP-2, ERA-40, and OAFlux show a decrease of the SHF in much of the ACC, whereas the HOAPS-3 exhibits an increase of the SHF. Additionally, the spatial pattern of the decadal trends of the LHF resembles that of the SHF for the ERA-40, HOAPS-3, and OAFlux; however, it is not the case for the NCEP-2. As shown in Fig. 14, the increase of the LHF in the Southern Ocean for the HOAPS-3 results from positive trends in both sea–air specific humidity differences and wind speed. For the OAFlux, large negative trends of sea–air specific humidity (temperature) differences contribute to the decrease of the LHF (SHF) in the ACC, which offset weak positive trends in wind speed. For the NCEP-2, it is not clear which input meteorological state variable has a larger contribution to the identified heat flux trends, particularly the LHF. The LHF and SHF variability of the NOCS2, including the dominant EOF mode, relationships with the AAO and ENSO, and decadal trends is substantially different from the other flux products, implying that the NOCS2 hardly reproduces accurate variability in the data-sparse Southern Ocean.
Further investigations are needed to better understand the reasons leading to the discrepancies among these flux products and to reduce the uncertainty, since a range of climate applications in the Southern Ocean require the air–sea fluxes to have an accuracy within ∼10 W m−2 (U.S. CLIVAR 2009). Given that the existing flux products differ substantially and in some cases there is no clear consensus, it is not entirely clear to what extent the results of this study can be generalized to infer which flux product is superior to others and the most suitable dataset for forcing ocean models. Follow-up model studies testing impacts of these temporal and spatial discrepancies in different flux products on basin-scale ocean simulations will give us additional information for assessing which dataset (or which combination of datasets) is optimal.
To date, no moored air–sea flux buoys have been deployed in the Southern Ocean. Ship observations of direct air–sea fluxes, being difficult to measure, are rarely available outside of a few dedicated campaigns, primarily during the International Polar Year (i.e., the Southern Ocean Gas Exchange Experiment). This paucity of observations is particularly troublesome, since air–sea heat fluxes are strongly dependent on wind speed and the Southern Ocean has the strongest sea surface winds in the world, exceeding 20 m s−1 in contrast with tropical and subtropical winds that are typically less than 14 m s−1 (U.S. CLIVAR 2009). In addition, high winds drive high wave and sea spray that can modulate air–sea fluxes in ways that might not be predictable on the basis of tropical and subtropical measurements. Thus, ongoing in situ monitoring from buoys and ships in the Southern Ocean, particularly in the Indian sector of the ACC (where the largest across-data scatter is found for both the LHF and SHF), are needed to validate flux products (including input meteorological state variables) and support continuing algorithm improvements.
Acknowledgments
We thank the anonymous reviewers for their helpful comments. This research was supported by the NSF OPP Antarctic Program (Grant 0838920), NASA NEWS, NSFC (Grant 40876099), 973 program (Grant 2011CB309704), and CAS “Hundred Talent Program.”
REFERENCES
Andersson, A., S. Bakan, K. Fennig, H. Grassl, C. P. Klepp, and J. Schulz, cited. 2007: Hamburg ocean atmosphere parameters and fluxes from satellite data—HOAPS-3—monthly mean. World Data Center for Climate. [Available online at http://cera-www.dkrz.de/WDCC/ui/Compact.jsp?acronym=HOAPS3_MONTHLY].
Berry, D. I., and E. C. Kent, 2009: A new air–sea interaction gridded dataset from ICOADS with uncertainty estimates. Bull. Amer. Meteor. Soc., 90 , 645–656.
Brunke, M. A., C. W. Fairall, X. Zeng, L. Eymard, and J. A. Curry, 2003: Which bulk aerodynamic algorithms are least problematic in computing ocean surface turbulent fluxes? J. Climate, 16 , 619–635.
Ciasto, L. M., and D. W. J. Thompson, 2008: Observations of large-scale ocean–atmosphere interaction in the Southern Hemisphere. J. Climate, 21 , 1244–1259.
Covey, C., P. J. Gleckler, T. J. Phillips, and D. C. Bader, 2006: Secular trends and climate drift in coupled ocean-atmosphere general circulation models. J. Geophys. Res., 111 , D03107. doi:10.1029/2005JD006009.
Curry, J. A., and Coauthors, 2004: SEAFLUX. Bull. Amer. Meteor. Soc., 85 , 409–424.
Fairall, C. W., E. F. Bradley, J. E. Hare, A. A. Grachev, and J. B. Edson, 2003: Bulk parameterization of air–sea fluxes: Updates and verification for the COARE algorithm. J. Climate, 16 , 571–591.
Frankignoul, C., E. Kestenare, M. Botzet, A. F. Carril, H. Drange, A. Pardaens, L. Terray, and R. Sutton, 2004: An intercomparison between the surface heat flux feedback in five coupled models, COADS, and the NCEP reanalysis. Climate Dyn., 22 , 373–388.
Gille, S. T., 2002: Warming of the Southern Ocean since the 1950s. Science, 295 , 1275–1277.
Gleckler, P. J., and B. C. Weare, 1997: Uncertainties in global ocean surface heat flux climatologies derived from ship observations. J. Climate, 10 , 2764–2781.
Gordon, A. L., 1988: The southern-ocean and global climate. Oceanus, 31 , 39–46.
Josey, S. A., E. C. Kent, and P. K. Taylor, 1999: New insights into the ocean heat budget closure problem from analysis of the SOC air–sea flux climatology. J. Climate, 12 , 2856–2880.
Kanamitsu, M., W. Ebisuzaki, J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino, and G. L. Potter, 2002: NCEP–DOE AMIP-II Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83 , 1631–1643.
Karsten, R., and J. Marshall, 2002: Testing theories of the vertical stratification of the ACC against observations. Dyn. Atmos. Oceans, 36 , 233–246.
Levitus, S., J. L. Antonov, T. P. Boyer, and C. Stephens, 2000: Warming of the World Ocean. Science, 287 , 2225–2229.
Liu, J., J. A. Curry, and D. G. Martinson, 2004: Interpretation of recent Antarctic sea ice variability. Geophys. Res. Lett., 31 , L02205. doi:10.1029/2003GL018732.
Lumpkin, R., and K. Speer, 2007: Global ocean meridional overturning. J. Phys. Oceanogr., 37 , 2550–2562.
Mayewski, P. A., and Coauthors, 2009: State of the Antarctic and Southern Ocean climate system. Rev. Geophys., 47 , RG1003. doi:10.1029/2007RG000231.
Mo, K. C., and G. H. White, 1985: Teleconnections in the Southern Hemisphere. Mon. Wea. Rev., 113 , 22–37.
Randall, D. A., and P. J. Gleckler, 1996: Systematic biases in AGCM ocean surface heat fluxes. WCRP workshop on air–sea flux fields for forcing ocean models and validating GCMS. WCRP-95, WMO/TD-762.
Raphael, M. N., 2004: A zonal wave 3 index for the Southern Hemisphere. Geophys. Res. Lett., 31 , L23212. doi:10.1029/2004GL020365.
Rintoul, S. R., C. W. Hughes, and D. Olbers, 2001: The Antarctic circumpolar current system. Ocean Circulation and Climate, G. Siedler, J. Church, and J. Gould, Eds., Academic Press, 271–302.
Russell, J. L., K. W. Dixon, A. Gnanadesikan, R. J. Stouffer, and J. R. Toggweiler, 2006: The Southern Hemisphere westerlies in a warming world: Propping open the door to the deep ocean. J. Climate, 19 , 6382–6390.
Schmitz, W. J. Jr., 1996: Some global features/North Atlantic circulation. Vol. 1, On the World Ocean circulation. WHOI Tech. Rep. WHOI-96-03, 148 pp. [Available from Woods Hole Oceanographic Institution, Woods Hole, MA 02543].
Sloyan, B. M., and R. S. Rintoul, 2001: Circulation renewal and modification of Antarctic mode and intermediate water. J. Phys. Oceanogr., 31 , 1005–1030.
Speer, K., S. R. Rintoul, and B. Sloyan, 2000: The diabatic Deacon cell. J. Phys. Oceanogr., 30 , 3212–3222.
Talley, L. D., 2003: Shallow, intermediate, and deep overturning components of the global heat budget. J. Phys. Oceanogr., 33 , 530–560.
Trenberth, K. E., and K. C. Mo, 1985: Blocking in the Southern Hemisphere. Mon. Wea. Rev., 113 , 38–53.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131 , 2961–3012.
U.S. CLIVAR cited. 2009: Working group on high latitude surface fluxes. [Available online at http://www.usclivar.org/Organization/HighLatitudeWG/hlatwg_USCLIVAR_report.pdf].
Venegas, S., 2003: The Antarctic circumpolar wave: A combination of two signals? J. Climate, 16 , 2509–2525.
White, W. B., and R. G. Peterson, 1996: An Antarctic circumpolar wave in surface pressure, wind, temperature, and sea ice extent. Nature, 380 , 699–702.
Worley, S. J., S. D. Woodruff, R. W. Reynolds, S. J. Lubker, and N. Lott, 2005: ICOADS release 2.1 data and products. Int. J. Climatol., 25 , 823–842.
Yu, L., and R. A. Weller, 2007: Objectively analyzed air–sea heat fluxes for the global ice-free oceans (1981–2005). Bull. Amer. Meteor. Soc., 88 , 527–539.
Annual-mean LHF and SHF (W m−2) averaged over the Southern Ocean during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Annual-mean LHF and SHF (W m−2) averaged over the Southern Ocean during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Annual-mean LHF and SHF (W m−2) averaged over the Southern Ocean during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the annual-mean (top) LHF (W m−2) and (bottom) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the annual-mean (top) LHF (W m−2) and (bottom) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the annual-mean (top) LHF (W m−2) and (bottom) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the (a) LHF and (b) SHF (W m−2) across the five flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the (a) LHF and (b) SHF (W m−2) across the five flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the (a) LHF and (b) SHF (W m−2) across the five flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Meridional profile of the zonal-mean (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Meridional profile of the zonal-mean (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Meridional profile of the zonal-mean (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Seasonal cycle of the (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Seasonal cycle of the (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Seasonal cycle of the (a) LHF and (b) SHF (W m−2) during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Time series of the annual-mean (a) LHF and (b) SHF (W m−2) for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Time series of the annual-mean (a) LHF and (b) SHF (W m−2) for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Time series of the annual-mean (a) LHF and (b) SHF (W m−2) for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the monthly (top) LHF (W m−2) and (bottom) SHF (W m−2) anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the monthly (top) LHF (W m−2) and (bottom) SHF (W m−2) anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Spatial distribution of the standard deviation of the monthly (top) LHF (W m−2) and (bottom) SHF (W m−2) anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
The first EOF mode of the monthly (top) LHF and (bottom) SHF anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
The first EOF mode of the monthly (top) LHF and (bottom) SHF anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
The first EOF mode of the monthly (top) LHF and (bottom) SHF anomalies during 1989–2005 for each individual flux product.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Correlations between the monthly anomalies of the (top) LHF and (bottom) SHF, and AAO during 1989–2005 for each individual flux product. Contours give the correlations at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Correlations between the monthly anomalies of the (top) LHF and (bottom) SHF, and AAO during 1989–2005 for each individual flux product. Contours give the correlations at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Correlations between the monthly anomalies of the (top) LHF and (bottom) SHF, and AAO during 1989–2005 for each individual flux product. Contours give the correlations at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends [W m−2 (decade)−1] of the (top) LHF and (bottom) SHF during 1989–2005 for each individual flux product. Contours give the trends at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends [W m−2 (decade)−1] of the (top) LHF and (bottom) SHF during 1989–2005 for each individual flux product. Contours give the trends at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends [W m−2 (decade)−1] of the (top) LHF and (bottom) SHF during 1989–2005 for each individual flux product. Contours give the trends at 90% confidence level.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Annual-mean sea–air (top) specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Annual-mean sea–air (top) specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Annual-mean sea–air (top) specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
(left) Meridional profile of the zonal mean, (middle) seasonal cycle, and (right) time series of the annual-mean for (top) sea–air specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
(left) Meridional profile of the zonal mean, (middle) seasonal cycle, and (right) time series of the annual-mean for (top) sea–air specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
(left) Meridional profile of the zonal mean, (middle) seasonal cycle, and (right) time series of the annual-mean for (top) sea–air specific humidity difference (g kg−1), (middle) temperature difference (K), and (bottom) wind speed (m s−1) during 1989–2005 for some flux products.
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends of (top) sea–air specific humidity difference (g kg−1 decade−1), (middle) temperature difference (K decade−1), and (bottom) wind speed (m s−1 decade−1) during 1989–2005 for some flux products (contours give the trends at 90% confidence level).
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends of (top) sea–air specific humidity difference (g kg−1 decade−1), (middle) temperature difference (K decade−1), and (bottom) wind speed (m s−1 decade−1) during 1989–2005 for some flux products (contours give the trends at 90% confidence level).
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Trends of (top) sea–air specific humidity difference (g kg−1 decade−1), (middle) temperature difference (K decade−1), and (bottom) wind speed (m s−1 decade−1) during 1989–2005 for some flux products (contours give the trends at 90% confidence level).
Citation: Journal of Climate 24, 4; 10.1175/2010JCLI3699.1
Summary of the resolution and input variables of the flux products. Values in parenthesis are the weighting assigned to different sources of input variables. NCEP-1 refers to NCEP–National Center for Atmospheric Research (NCAR) Global Reanalysis 1. GSSTF refers to Goddard Satellite-Based Surface Turbulent Fluxes. OI refers to optimum interpolation. QSCAT refers to Quick Scatterometer. AMSR-E refers to Advanced Microwave Scanning Radiometer for Earth Observing System. Qa is air specific humidity, Ta is air temperature, SST is sea surface temperature, and U is wind speed.
Correlations of the annual-mean LHF/SHF between the datasets. Correlations at 95% confidence level are in boldface.
