1. Introduction
Transfer of energy between the atmosphere and the ocean is a fundamental process within the global climate system. Attempts to quantify this transfer, specifically the fields of air–sea heat and momentum exchange, have primarily relied on meteorological reports from merchant ships at sea (e.g., Bunker 1976; da Silva et al. 1994). The need for accurate climatological descriptions of the heat and momentum transfer is well recognized (e.g., WCRP 1986). In addition to the primary insight they provide into the physics of the ocean–atmosphere system, fields of these quantities are vital for modeling studies both as boundary conditions for ocean models and for verification of flux fields obtained from coupled ocean–atmosphere models. Climatological flux estimates also have a role to play in the integration within a consistent framework of the increasing number of hydrographic estimates of the ocean heat and freshwater transport (Macdonald and Wunsch 1996).
In this paper, we present results from the analysis of a new global air–sea heat flux climatology generated at the Southampton Oceanography Centre (SOC) from an enhanced version of the Comprehensive Ocean Atmosphere Dataset 1a (COADS1a; Woodruff et al. 1993), which consists of marine meteorological reports spanning the period 1980–93. The new fields will be referred to as the SOC climatology hereafter. The global distribution of ship reports is spatially inhomogeneous (Woodruff et al. 1993) with a relatively large number of observations in the midlatitude North Atlantic and North Pacific but very few in the Southern Hemisphere. Despite the limitations imposed by this distribution we believe that climatological analysis of fluxes derived from ship observations can provide useful information. Satellite and numerical weather prediction (NWP) model-derived fluxes potentially offer full global coverage but suffer from other disadvantages with continuing uncertainty over retrieval methods used with satellite observations (e.g., WCRP/GEWEX 1996; Schulz et al. 1997) and significant differences between the output from the various model reanalysis programs. Ship-based climatological fields have been produced in many earlier analyses (e.g., Bunker 1976; Esbensen and Kushnir 1981; Oberhuber 1988). Most recently, a global climatological flux dataset has been generated from reports contained in the version of COADS covering the period 1945–89 by da Silva et al. (1994); that climatology will be referred to as UWM/COADS hereafter.
We have made several advances both in the generation and the analysis of the SOC climatology relative to past studies. For the first time, the individual ship meteorological reports have been corrected for biases arising from variations in observing procedure (Kent et al. 1993a,b). Further, the results of several supporting studies have been used to justify our choice of the visual wind estimate conversion scale (Kent and Taylor 1997), the longwave flux formula (Josey et al. 1997), and the method employed to average the flux fields (Josey et al. 1995). The accuracy of the individual monthly flux fields has been assessed using high quality research buoy measurements at several climatically different locations; the first time that such an approach has been carried out. On a larger scale, we have compared box mean surface heat fluxes derived from hydrographic section data with the corresponding SOC values giving further insight into regional biases within the climatology.
The global climatological mean net heat flux should average to zero; however, mean heat gains of 30 W m−2 or greater have been found in many previous studies. The common remedy for this heat budget closure problem has been a scale adjustment of the flux components using inverse analysis (e.g., Isemer et al. 1989). This implicitly assumes that the problem arises from fundamental biases in the bulk formula estimates; it does not allow for the possibility that regional biases are significant. Such regional biases might arise, for example, from undersampling of extreme conditions in regions such as the Southern Ocean and through variations in the applicability of the flux parameterizations between climatic regions. We will show that, although we are also unable to close the global heat budget with the SOC climatology, we do not believe that global adjustments to the flux components are the correct solution. Comparison with high quality independent measurements made by Woods Hole Oceanographic Institute research buoys (e.g., Moyer and Weller 1997) shows that, at least in some regions, our unadjusted flux set provides a more accurate description of the heat exchange than the corresponding fluxes adjusted following da Silva et al. (1994). While we are not arguing against the necessity for corrections to the climatological fields in order to close the global heat budget, we suggest that these corrections should be made at a regional level.
In the following section we describe the primary datasets that form the basis for the climatology. The method by which the individual heat flux estimates have been obtained and the procedure used to generate the climatological mean fields are described in section 3. In section 4, we discuss the main characteristics of the heat flux fields, the impact of the corrections to the ship reports, and the main differences between the SOC fields and the UWM/COADS climatology. We do not attempt to provide a comprehensive description of the SOC climatological monthly mean fields, for which purpose an atlas has been prepared (Josey et al. 1998). Evaluation studies using buoy and hydrographic data are presented in section 5. In the final section we summarize our findings and discuss possible future improvements to the climatology.
2. Datasets
The primary dataset used to produce the climatology was the COADS release 1a (Woodruff et al. 1993), which contains some 30 million in situ meteorological reports covering the period 1980–93. Its composition changes significantly throughout this period with reports from the Coastal Marine Automated Network (C-MAN), drifting and moored buoys dominating toward the end. The reliability of the C-MAN and drifting buoy reports has been questioned and prior to our analysis we have removed these reports together with moored buoy reports that occur at intervals shorter than 3 h.
In order to apply the bias corrections described in the following section, additional metadata regarding the observing procedures used on specific ships have been merged onto the COADS1a from the International List of Selected, Supplementary and Auxiliary Ships (WMO 1993), which is published annually by the World Meteorological Organisation (WMO), specifically, the height at which the anemometer was deployed, the height of the observing platform (the latter being used as a proxy for the temperature and humidity sensor heights), and the methods by which the sea surface temperature, the air temperature, and the humidity were measured. The COADS1a and WMO47 were merged using the ship call sign. For each report within the COADS1a, the WMO47 file for the corresponding year was searched for the listed call sign. If the call sign was present in the WMO47 file, the metadata for that particular ship were appended to the COADS1a report. If not, the procedure was repeated for the WMO47 files for the following two years and finally for the preceding year. If no matching call sign was found, absent data were appended to the report.
Time series of the total number of reports, the number made by ships, and the number that have been matched in the merged dataset are shown in Fig. 1. The proportion of reports for which matching was successful increased throughout the time period reflecting an increase in the proportion of valid call signs in COADS. However, there is also an increasing number of buoy reports. No attempt has been made to include additional metadata for the buoys since no comprehensive dataset describing their characteristics is available.
3. Flux calculation method
a. Corrections for systematic errors
The existence of systematic biases in surface meteorological observations made by voluntary observing ships (VOS), which form the core of the reports contained in COADS1a, was identified during the VOS Special Observing Project for the North Atlantic (VSOP-NA; Kent and Taylor 1991; Kent et al. 1991; Kent et al. 1993a). In that study, detailed information regarding observing procedure was collected for a subset of the VOS fleet reporting in the North Atlantic and corrections developed for various observational biases by using NWP model analysis output as a comparison standard. Note that the problem of possible systematic biases in the absolute values of the model fields was avoided as only differences with respect to the model values were considered. We are aware that the corrections were developed using a sample of ships from the North Atlantic and that their global applicability has not been tested. However, we note that the corrections are based on instrument type and, thus, provided that the information regarding measurement method supplied in the WMO47 list is accurate, we expect them to be equally applicable in other regions of the global ocean.
Corrected values of the reported meteorological observations were obtained before calculating the heat fluxes. The corrections are as follows:
(i) Sea surface temperature (SST) measurements made with a thermometer situated in the engine cooling system are reduced by 0.35°C. If no information was available about the method of SST measurement a default correction of 0.2°C was applied since the WMO47 list indicated that the proportion of ships making engine inlet measurements was about 60%.
(iv) The reported values of wind speed (if based on anemometer readings), air temperature, and humidity have been height corrected using the merged metadata from WMO47 for the cases where matching was successful. Default heights were applied where no information was available as follows: ships (anemometer height = 20 m/temperature and humidity sensor height = 18 m), buoys (8/7 m), and platforms (80/60 m).
Many of the wind speed estimates reported in COADS1a were obtained from visual observations of the sea state. These visual estimates are based on the WMO1100 Beaufort equivalent scale (WMO 1970). The accuracy of this scale has often been questioned and many alternative scales suggested. Kent and Taylor (1997) compared ship-based monthly mean wind speeds on a 1° × 1° grid. The Beaufort scale suggested recently by Lindau (1995) was found to give the best correspondence between winds from anemometers at known heights and the visual mean wind speeds; it has therefore been used in the present study.
b. Flux parameterizations
Estimates of the heat flux components have been obtained for each report in the merged COADS1a–WMO47 dataset for which the relevant meteorological fields are available using the parameterizations described below.
1) Turbulent heat fluxes
2) Longwave flux
3) Shortwave flux
c. Averaging procedure and objective analysis
If any of the variables on which an individual flux estimate depended lay outside the 4.5 σ limits that are set within COADS1a, it was removed from the flux dataset. The other flux estimates derived from the same report were retained provided the variables on which they depended lay within the limits. Raw mean flux fields were generated from the resulting trimmed dataset by averaging over individual reports for each month in the period January 1980–December 1993 on a 1° × 1° grid from 85°N to 85°S. This approach (termed the sampling approach) has been taken rather than calculating fluxes from monthly averaged meteorological variables (the classical approach) as significant errors are known to be possible with the latter method. Esbensen and Reynolds (1981) found errors of up to 10% in classical estimates of the latent and sensible January and July mean heat fluxes relative to the corresponding sampling values from an analysis of weather ship data in the North Atlantic and North Pacific. More recently, Josey et al. (1995) analyzed the seasonal variation using data from a weather ship at Station Lima in the North Atlantic and found that the estimated heat loss may be biased high by of order 15 W m−2 if classical means are used due to the neglect of correlations between meteorological variables, particularly wind speed and humidity. The sign and magnitude of such biases are likely to vary with region depending on variations in the correlation fields of the different meteorological variables (Hasse and Smith 1997).
It is important to remember that the distribution of ship reports, and hence of the estimated fluxes, shows strong spatial variations. As an example, the number of latent heat flux estimates per 1° × 1° square used to form the climatological monthly mean, averaged over all months of the year, is shown in Fig. 2. In the midlatitude Northern Hemisphere oceans there were typically greater than 10 observations per 1° × 1° square with greater than 100 in the heavily sampled regions. However, large areas with very few or no observations are evident in the Southern Hemisphere particularly in the austral winter.
The raw fields have been smoothed and estimates obtained in the unsampled regions of the grid using a successive correction method similar to that employed by da Silva et al. (1994) in which an analyzed mean is produced from the raw mean estimates within a certain radius of influence; see Josey et al. (1998) for full details. Each individual monthly raw mean field has been analyzed with four passes of the successive correction method at influence radii of 1541, 1211, 771, and 331 km. We chose to use 331 km as the minimum influence radius as opposed to the value of 771 km adopted by da Silva et al. (1994) as test studies indicated that it provides a better definition of frontal features, for example, the Gulf Stream, in well-sampled areas. Climatologically ice-covered regions were excluded from the analysis using the ice mask of Alexander and Mobley (1976). Finally, climatological monthly mean fields were generated from the individual monthly analyzed fields for 1980–93.
d. Estimation of random errors
Estimation of the random errors in ship-based climatological flux fields requires knowledge of the errors inherent in the original meteorological observations, of the strength of the spatial and temporal correlations between the variables in the flux formulas, and of the propagation of errors during the objective analysis. Gleckler and Weare (1997) have recently attempted to estimate the errors in the Oberhuber (1988) flux climatology using estimates for the random errors and standard sampling theory. Following the method of Lindau (1995), Kent et al. (1998) have determined the random errors in the meteorological observations used in calculating the SOC climatology. However, at present we do not consider our knowledge of the many meteorological variable correlation terms sufficient to use their results to derive error fields for the climatology. Thus, in the present study our main assessment of the overall accuracy of the calculated flux values will use comparisons with flux estimates from buoy deployments (section 5b). We shall also provide comparisons with hydrographic estimates (section 5c).
4. Field characteristics
In this section we briefly describe the main characteristics of the flux fields, a detailed description is given in the SOC flux atlas (Josey et al. 1998). As might be expected, the main characteristics of the flux fields are similar to those found in previous studies. A difference in this study is the attempt to correct for systematic errors in the observations. We shall therefore describe the size and spatial variation of the corrections for observational bias described in section 3a. In addition, we discuss the main differences between the SOC fields and the UWM/COADS climatology.
a. SOC climatological mean fields
Climatological January and July mean fields of the different components and net heat flux are shown in Figs. 3a–j. The sign convention is for positive fluxes to represent heat gain by the ocean. We are aware that although we have produced a dataset on a globally complete grid, the fields cannot be regarded as truly global given the very sparse data coverage in certain areas, particularly the Southern Ocean. In several past studies (e.g., Harrison 1989) high-latitude extrema have been excluded from global plots. We have decided to present the full global fields rather than impose a restriction based on latitude. Examination of Fig. 2 shows that the number of observations does not follow a simple zonal variation and application of a boundary at say 40°S would lead both to the exclusion of data in some relatively well sampled areas while at the same time giving the misleading impression that the fields are as reliable in the southeast Pacific as they are in the North Atlantic. We stress that caution must be exercised in attaching significance to the fields in data-sparse regions and Fig. 2 should be used for reference when making such judgments.
For the turbulent heat flux components (i.e., the sensible and latent terms), the areas of strongest loss are in the winter hemisphere with latent heat losses of order 250 W m−2 over the Gulf Stream and Kuroshio. Enhanced heat loss is also seen over the Agulhas and east Australian currents in the Southern Hemisphere winter but these regions are somewhat weaker than their Northern Hemisphere counterparts. Broader areas of less intense heat loss are observed under the main trade wind belts with particularly strong loss seen in the southern Indian Ocean in July. Strong sensible heat fluxes occur in regions where very cold air is advected over the ocean from neighboring landmasses particularly the Labrador and Norwegian Seas. Note, that the true northward extent of these regions is probably underestimated in our analysis as we have not been able to include heat loss through partially ice-covered regions of the ocean and the adjacent ice-free regions are badly undersampled.
The global variation in the net longwave flux is relatively small, typical values ranging from 30 to 70 W m−2. However, within this range there is a degree of structure that reflects the balance between the sea–air temperature difference, the cloud cover, and the amount of water vapor in determining the net longwave flux. Noticeable features include the bands of reduced longwave loss under the intertropical and South Pacific convergence zones (ITCZ and SPCZ), while maxima occur over the western boundary currents in winter where the sea–air temperature difference is high. In contrast, the shortwave field has a primarily meridional variation determined by the mean solar elevation with peak summer values of order 350 W m−2. The main departures from this variation occur under regions of increased cloud cover such as the ITCZ and SPCZ. Finally, the net heat flux field is seen to be dominated by the contributions from the shortwave and latent heat fluxes with heat gain primarily confined to the summer hemisphere and heat loss to the winter.
b. Estimation of systematic errors due to observing procedure
We now quantify the impact of the corrections described in section 3a, which is equivalent to the systematic error in an uncorrected set of flux fields due to the observing procedure. Mean fields have been calculated with and without each of the corrections for alternate months in 1990. Results are presented for January and July of that year; we expect these to be similar to those for the climatological mean fields. Note that the corrections do not affect the shortwave radiation.
1) Effect of individual corrections
The reduction in engine intake measured SST acted to lower both the sensible and latent heat loss, the latter as a result of the reduction in specific humidity at the sea surface. The effects were largest in January with the sensible heat loss reduced by up to 4 W m−2 in the northern North Pacific and the latent heat by 10 W m−2 in the midlatitude and equatorial Pacific. The SST correction also acted to decrease the net longwave cooling due to the reduction in the upwelling component. However, changes were generally small reaching a maximum of 2 W m−2 in the western North Pacific.
In contrast, the reduction of the air temperature due to the radiation correction acted to increase the sensible heat and net longwave loss. The effect on each of these components was strongly seasonal and zonal with changes up to 5 W m−2. This correction also had an influence, via the stability, on the estimated latent heat flux. Such changes were small over much of the ocean;however, the heat loss was increased locally by up to 3 W m−2 in the Indian Ocean and equatorial regions.
The reduction applied to screen-measured dewpoint temperatures served to increase the latent heat loss. This correction had the greatest effect over the Atlantic where there is a higher proportion of ships using screens than in the Pacific. Over most of the Atlantic the increase was greater than 4 W m−2 with peak values of about 10 W m−2. The effect of the changes on the latent heat flux is greater in winter than in the summer. Changes in the net longwave and sensible heat flux due to the dewpoint correction were negligible.
The use of the Lindau (1995) scale and individual anemometer heights affected the latent and sensible heat fluxes through changes in the estimated wind speed. With the Lindau scale, visual wind speed estimates of less than 11 m s−1 were increased relative to the WMO1100 scale values by about 0.3–0.4 m s−1. At higher wind speeds the scale acted to progressively reduce the original values, for example, by 1 m s−1 at 16 m s−1 and by 3 m s−1 at 26 m s−1. The combined effect on the turbulent fluxes was complex, depending on the ratio of visual to measured wind speed estimates. In the Atlantic, visual wind estimates dominate with the consequence that the turbulent fluxes were reduced in regions of high wind speed, for example, the central North Atlantic in January, and increased in regions of low wind speed, for example, much of the Atlantic in July, the magnitude of the changes being up to 2 and 8 W m−2, respectively. In the Pacific, anemometers rather than visual observations were the dominant source of wind speed values. Mean anemometer heights were found to be higher than in the Atlantic, typically greater than 30 m, and the major change in the turbulent fluxes in the North Pacific was a reduction of order 10 W m−2 in the latent heat flux due to the reduced wind speed.
2) Combined effect of the corrections
We now consider the combined effects of the corrections on each of the flux components and the net heat flux. Plots of the difference between the fully corrected and uncorrected flux fields for January and July 1990 are shown in Figs. 4a–h, while summary statistics for certain key regions are listed in Table 1. Positive values indicate an increased heat loss by the ocean as a result of the corrections.
The combined effect of the corrections on the sensible heat flux was similar in both the Atlantic and the Pacific. In January, the SST correction dominated in the northern latitudes and the sensible heat loss was reduced by up to 4 W m−2. Farther south the radiation correction became important and in the Tropics the heat loss was increased by a similar amount. In July increases in sensible heat loss of order 2 W m−2 were observed in most regions due to the radiation correction. However, toward the southern limit, roughly in the range 35°–45°S, the effect of the SST correction still dominated and small reductions were found. The latent heat loss in January was increased by up to 8 W m−2 in the Southern Hemisphere, the Indian Ocean, and tropical North Atlantic. In contrast over most of the North Pacific and the Gulf Stream region reductions of up to 10 W m−2 were found as the effects of the anemometer height corrections and use of the Lindau scale dominated. The area of strong reduction noted in the North Pacific was significantly smaller in July. For the longwave flux, the January distribution of flux change showed a mainly zonal pattern due to the radiation correction with the largest increases, above 4 W m−2, in the Southern Hemisphere midlatitudes. The situation was reversed in July with the effects of the radiation correction shifted toward the Northern Hemisphere.
The combined effect of the corrections on the net heat flux, or equivalently the systematic error in an uncorrected heat flux field due to observational bias, has a complex spatial variation (Figs. 4g,h). In January increases in the heat loss of up to 15 W m−2 are seen in the Southern Hemisphere (driven mainly by changes in the latent heat flux) while reduced loss is seen in the Northern Hemisphere. In July the situation is reversed in the northern oceans with increased heat loss occurring; farther south the heat loss also tends to increase as a result of the corrections but by a smaller amount than in January. Finally note that there was reduced heat loss at the limit of data coverage in the south where the SST correction starts to dominate.
c. Comparison with the UWM/COADS climatology
Finally, we briefly discuss the main differences between the SOC and UWM/COADS (da Silva et al. 1994) climatological heat flux fields. Differences between the two climatologies may result from the effects of the observational bias corrections that were included in the SOC analysis but not in UWM/COADS, the use of different formulas to estimate the various flux components, the difference in the time periods sampled, and the different scales used in the objective analysis. The main variations are found to occur in the sensible and latent heat flux fields. Difference plots of the SOC–UWM/COADS annual mean values for these components and the net heat flux are shown in Figs. 5a–c. The radiative flux fields (not shown) are typically in agreement to within 10 W m−2. The major differences in the latent heat flux occur over the Gulf Stream and the subtropical North Pacific. In UWM/COADS, the Gulf Stream region of strong latent heat loss is broader and appears to be displaced farther south than in the SOC analysis. The shift in position appears as adjacent bands of positive and negative heat flux difference in Fig. 5a. The narrower region of strong heat loss found in the SOC analysis may be due to our choice of a smaller objective analysis scale relative to that adopted for UWM/COADS. Similar behavior is seen over the Kuroshio although the differences are smaller in magnitude. Over a broad area in the subtropical central North Pacific, the UWM/COADS heat loss is stronger by up to 30 W m−2. It is unlikely that this feature is a result of the ship corrections as these are only expected to result in differences of order 5–10 W m−2 in this region; see Figs. 4c,d. A further possibility is that it is due to changes in the instrumentation used by ships in this area between the periods used to generate the two climatologies. Alternatively, the difference may reflect an actual climatic change between these two periods.
Differences between the two climatologies are also seen in the sensible heat flux field for which the UWM/COADS analysis has losses of order 10–20 W m−2 greater in the annual mean over the Gulf Stream and Kuroshio and localized areas of the high-latitude North Atlantic and North Pacific. The effect of the ship corrections, which is typically less than 5 W m−2, does not appear to be large enough to explain these differences. The difference in objective analysis scales may again be partly responsible. The net heat flux difference field, Fig. 5c, is dominated by the features in the latent and sensible heat flux fields; agreement between the two datasets is typically found to within 15 W m−2 with regions of stronger loss in the UWM/COADS fields over the Gulf Stream and subtropical central North Pacific of up to 50 W m−2. Finally, we note in advance from the discussion in section 5a and the values in Table 2 that although strong regional differences are observed, the global mean values in each climatology agree to within 1 W m−2 for the latent and 2 W m−2 for the sensible heat flux.
5. Analysis results
a. The global ocean heat budget
Over a sufficiently long timescale the net heat flux into the ocean must average to close to zero otherwise there would be significant warming or cooling trends. A simple calculation shows that a global mean net heat gain of 1 W m−2 maintained over the period of the climatology would, if mixed into the upper 1000 m of the ocean, lead to a mean temperature increase in that layer of 0.1°C. Observational limits on decadal temperature changes are at most a few tenths of a degree (e.g., Parrilla et al. 1994). Hence, the climatological global mean heat budget must be closed to within 2–3 W m−2.
As noted earlier, a common problem in past climatological studies has been the inability to close the ocean heat budget within the above limits. We encounter the same problem, the global mean net heat flux is an ocean gain of 30 W m−2 in our analysis, which is very close to the value of 31 W m−2 obtained from the UWM/COADS fields. The latter study did not include corrections for systematic errors at the level of individual ships. This suggests that, although errors due to observing procedure are important at the regional level, they are not responsible for the global heat budget imbalance. Very similar values for the global means of the heat flux components are obtained in each case; see Table 2. The magnitude of the shortwave gain is greater by 3 W m−2 in UWM/COADS as is the sensible heat loss, by 2 W m−2. It is not clear whether these differences are due to variations in flux calculation method or in the period sampled in each analysis.
Is the heat budget imbalance significant given its likely random error? The standard deviation of the sample of 14 individual yearly mean values in our dataset is 7 W m−2. This represents an upper limit for the random error since it also includes interannual variability. Even so, it indicates that the bias of 30 W m−2 is indeed significant.
Attempts to remedy the closure problem in past analyses have rested on the assumption that there are errors in the various flux formulas that can be corrected by proportional adjustment. Each component of the heat flux has been allowed to vary within an assumed error range and correction factors have been sought using linear inverse analysis. Closure of the heat budget and various hydrographic heat transport estimates have been employed as constraints. For UWM/COADS, da Silva et al. (1994) suggested that closure should be obtained by reducing the shortwave flux by 8%, increasing the latent heat by 13% and making smaller adjustments to the longwave and sensible heat fluxes. However, this approach does not allow for the possibility that regional biases, for example as a result of undersampling at high latitudes, may play a significant role. We carry out regional comparisons with hydrography in section 5c to explore this possibility and discuss sources of bias in section 5d. First we consider in section 5b how the SOC fluxes, without adjustment, compare with high quality meteorological buoy measurements.
b. Comparison with research buoy measurements
1) Introduction
Recent years have seen the deployment of Woods Hole Oceanographic Institute (WHOI) surface buoys equipped with research quality meteorological instrumentation in several different ocean regions for periods ranging from four months to two years. The buoy instrumentation has been described by Moyer and Weller (1997); measurements of the radiative fluxes were made with radiometers while the turbulent fluxes were determined from 1-min mean meteorological measurements using the algorithm of Fairall et al. (1996). We have compared the monthly mean fluxes from the buoys with collocated values from the SOC flux dataset calculated for the buoy deployment period (where we use the term “flux dataset” to refer to the monthly means from individual years as opposed to the flux climatology, which refers to the monthly means averaged over the period 1980–93). The positions of the buoys are indicated by crosses on Fig. 2; summary statistics for the buoy and SOC estimates of the fluxes are listed in Tables 3 and 4. We also present comparisons with a version of the dataset that has been adjusted to close the global heat budget following da Silva et al. (1994). For this purpose we have employed a simplified version of the da Silva scheme in which the shortwave is reduced by 8% and the latent heat increased by 13%. With these adjustments our global heat budget would be closed to within 4 W m−2. We note that as data from these buoys were not included in the merged COADS1a–WMO47 dataset used to generate the SOC fluxes, the buoys represent an independent source of flux estimates.
2) The comparison datasets
(i) The subduction buoy array
During the Subduction Experiment five buoys were deployed between 18° and 33°N in the eastern North Atlantic from July 1991 to June 1993 (Moyer and Weller 1997). The data recorded by the buoys varied in quality; Moyer and Weller (1997) focus on the measurements made by the northeast buoy as it was the only one to remain operational throughout the planned deployment period and provided the most reliable data. The measurement records of the other four were less complete. In our comparison we have used the SOC monthly flux estimates averaged over 2° × 2° boxes centered on each of the buoy locations for those periods for which reliable data was recorded by the buoys as specified by Moyer and Weller (1997).
(ii) The FASINEX buoy array
The Frontal Air–Sea Interaction Experiment (FASINEX) moored buoy array consisted of five buoys deployed within 50 km of each other in a region of strong SST fronts southwest of Bermuda (Weller et al. 1995). Monthly mean estimates are only available for February 1986–May 1986 so we are unable to examine the full annual cycle. We compare array-averaged buoy estimates of the fluxes with values from the SOC flux dataset averaged over the box (26°–28°N, 69°–71°W), which is centered on the array. Note that for the current study revised estimates of the air–sea fluxes were obtained from the buoy meteorological variable measurements using the algorithm of Fairall et al. (1996). Unfortunately the buoys were not instrumented with longwave sensors and so we are unable to make any evaluation of this component.
(iii) The TOGA COARE buoy
A WHOI mooring was deployed during the Tropical Ocean Global Atmosphere Couple Ocean–Atmosphere Research Experiment (TOGA COARE) experiment in the western equatorial Pacific warm pool at (1.75°S, 156°E). Complete monthly records are available from the buoy for November 1992–February 1993 and these are compared with concurrent means from the SOC flux dataset for the box (1°–2°S, 155°–157°E).
(iv) The Arabian Sea buoy
The most recently completed WHOI buoy deployment consisted of a single buoy moored in the Arabian Sea at (15.5°N, 61.5°E) from October 1994 to October 1995. Measurements made on this buoy are compared with an extension of the SOC flux dataset for the additional years 1994–95 prepared from a preliminary COADS dataset (S. Worley 1997, personal communication) using the same method as described in section 3. The detailed results of this study, which suggest that contrary to past analyses the Arabian Sea gains rather than loses heat during the southwest monsoon, are described in detail in a separate paper (Weller et al. 1998). For our comparison we consider only the mean fluxes over the deployment period.
We expect the monthly means determined from the buoy measurements to be more accurate than the SOC flux dataset estimates as a result of the greater temporal sampling and direct measurement of the radiative fluxes, rather than empirical estimation. However, it is important to be aware of the limitations on the buoy values. The buoy measurements are made at specific locations while the comparison values from the SOC dataset are spatial averages over boxes as large as 2° × 2°, as detailed above. The question then arises of whether the buoy values are representative of the mean air–sea interaction at these scales given the potential for spatial variation in the magnitude of the basic variables, particularly sea surface temperature. We note that strong gradients in SST were a feature of the FASINEX site but these were not found to significantly modify the air–sea fluxes in the monthly mean (Weller et al. 1995). As a precaution, we have restricted our comparison at this site to the array-averaged values and note that the other sites are not in regions with strong fronts in SST.
Estimates of the accuracy of the buoy-measured fluxes have been made by the WHOI group for each of the buoy deployments. Comparisons of the buoy-measured fluxes with values obtained from collocated research ships during TOGA COARE suggest that the individual flux components are accurate in the monthly mean to within 5 W m−2 while the net heat flux is obtained to better than 10 W m−2 (Weller and Anderson 1996). Similar accuracy is expected for the Arabian Sea buoy (Weller et al. 1998). For the subduction buoy array Moyer and Weller (1997) estimate an extreme error for the case where the errors in the individual fluxes sum constructively to be 50 W m−2 but note that in practice some of the biases are likely to be counteractive resulting in errors closer to those obtained for the TOGA COARE deployment. The error in the shortwave flux measurements from the FASINEX array was estimated to be similar in magnitude to that for the later deployments (Weller et al. 1995); while the errors in the turbulent fluxes may be somewhat larger given the stronger air–sea interaction in this region.
3) Comparison for the radiative heat fluxes
Time series of the measured fluxes at the northeast buoy of the Subduction Array are compared with the SOC estimates in Fig. 6. Fairly good agreement is found for the shortwave flux although there is a tendency for the SOC values to underestimate the gain relative to the buoy measurements, this difference being greatest in the summer. In the mean over the deployment period, the SOC estimate is biased low by 13 W m−2 relative to the buoy-measured value of 195 W m−2. This bias is partially offset by an underestimation of the net longwave loss with the result that the net radiative flux from the two sources agrees to within 4 W m−2. A summary plot of the differences between the SOC and buoy mean shortwave and longwave fluxes for all the sites is shown in Fig. 7a. Partial cancellation of the shortwave and longwave components is also seen at most of the other buoy sites, which suggests that the ships’ cloud amount estimate is different from that implied by the buoy data. The SOC–buoy net radiative flux difference is only significantly different from zero at the TOGA site where both the shortwave and longwave differences are significant and of the same sign. A simple calculation with the Reed formula shows that the SOC–buoy mean shortwave flux difference of 23 ± 4 W m−2 could be explained if the reported mean cloud cover, 5.2 octas for the period considered, had been underestimated from a true value of 6.2. Note that we do not find evidence that climatological estimates of the shortwave flux are generally overestimated, as has been suggested in many past analyses in order to account for the difficulties in closing the global heat budget. Rather, the results of our comparison for five of the eight buoys suggests that we have underestimated the net shortwave gain.
4) Comparison for the turbulent heat fluxes
The time series for the SOC estimates of the turbulent flux components at the northeast buoy site are noticeably less smooth than the buoy values, see Fig. 6, probably reflecting the lower sampling rate. For this area, the number of ship observations on which the SOC monthly raw mean estimates are based is of order 10 while the buoy values typically represent an average over a continuous 1-min mean time series. The time series for the other buoy sites (not shown) exhibit the same characteristic. The comparisons indicate that the relatively low sample size on which the ship estimates are based does not lead to biases in the estimated fluxes. However, as the buoy fluxes are also determined from bulk formulas we cannot rule out the possibility of residual biases due to inadequate representation of the heat exchange in the parameterizations. The differences between the SOC and buoy mean latent and sensible heat fluxes for all the sites are summarized in Fig. 7b, agreement was typically found to within 10 W m−2. The exception to this rule is the FASINEX site at which the SOC estimates of the turbulent loss terms are persistently weaker than the values obtained from the buoy measurements, the differences being 45 ± 3 W m−2 and 7 ± 1 W m−2 for the latent and sensible heat fluxes, respectively. Comparison of the meteorological variable measurements indicates that the bias in the latent heat flux is primarily due to overestimation of the atmospheric humidity and underestimation of the SST in the SOC analysis relative to the buoy measurements. Although the application of the da Silva et al. (1994) corrections to the latent heat flux of 13% would improve the comparison with the buoy data at the FASINEX site (where a significant bias remains), and also at the Arabian Sea site, we find no evidence that the SOC latent heat fluxes are generally biased low. At the subduction and TOGA sites such adjustment would degrade the comparison.
5) Comparison for the net heat fluxes
At the northeast subduction buoy site the various offsets in the flux components tend to cancel giving similar estimates of the mean net heat exchange, the SOC and buoy net heat flux values being 28 and 25 W m−2, respectively, an insignificant difference. Despite the additional uncertainty in the data from the remaining subduction buoys, they provide further indication that the SOC flux estimates are not strongly biased; see Fig. 7c. At the southeast buoy, the overall mean net heat exchange estimates are identical at 54 W m−2 although the standard error of the difference is 9 W m−2, somewhat larger than at the northeast buoy. For the other three subduction buoys the differences lie slightly outside one standard error from zero. Note that in the flux time series, Fig. 6, the stronger heat loss measured by the northeast buoy in December 1992 relative to December 1991 is also reflected in the SOC values, suggesting that the SOC flux dataset will prove useful for studies of interannual variability.
For the TOGA COARE buoy the difference in net heat flux is dominated by the difference in the shortwave flux, the buoy estimate of the net heat gain being 21 W m−2 over the 4-month period and the SOC value 44 W m−2. The SOC values also suggest greater heat gain by the ocean at the Arabian Sea buoy site. However, here it is the underestimation of the heat loss by the longwave and latent heat terms, moderated only slightly by the shortwave difference, that leads to a value for the SOC net heat gain of 78 W m−2, which is 17 W m−2 greater than the value obtained by the buoy. Weller et al. (1998) note that this discrepancy is reduced if an alternative longwave formula suggested by Bignami et al. (1995) is employed for this region.
Calculation of the net heat flux for the FASINEX array is prevented by the lack of longwave data from the buoys. However, we note that the sum of the shortwave, latent, and sensible terms is 35 W m−2 for the buoy and 85 W m−2 for the SOC analysis. Given that the longwave values would be unlikely to differ by more than 10 W m−2, this suggests that the SOC estimate of the net heat flux is likely to overestimate the buoy value by 40–60 W m−2. Combining the SOC estimate for the longwave loss with the buoy-measured values for the other components gives a net heat loss of −28 W m−2 as opposed to the value obtained with SOC estimates for all of the components, which is a net heat gain of 22 W m−2.
The effects of a scale adjustment to the SOC flux dataset similar to that proposed by da Silva et al. (1994) are also shown in Fig. 7c. The result is a poorer picture of the heat exchange over the subduction buoy array, no improvement at the Arabian Sea site, and values that are in better agreement with the buoy measurements for the TOGA buoy (because the shortwave heating is reduced) and, by inference, the FASINEX buoy site (because the latent heat cooling is increased).
6) Summary
The results of these comparisons suggest that the level of accuracy of the SOC flux fields is regionally dependent. At the northeast and southeast subduction buoys, the SOC and buoy mean net heat flux estimates are not significantly different. Agreement to within slightly greater than one standard error was obtained at the remaining subduction buoys while at the Arabian Sea buoy the difference was within 20 W m−2. The TOGA buoy had an offset of 23 W m−2, mainly due to the shortwave flux estimate, while the greatest difference was found at the FASINEX site, stemming primarily from the latent heat flux estimate. We note that the periods for which data were available from the latter two sites were just 4 months long and it is not clear how large these differences would be in the annual mean. Overall there does not appear to be a bias with respect to the buoy measurements that is consistent with scale adjustments to the bulk formulas. Although we have focused on the adjustment scheme suggested by da Silva et al. (1994), any scheme in which the ocean heat budget is closed by a combination of reduced shortwave and increased latent heat flux (e.g., Isemer et al. 1989) will lead to larger discrepancies with respect to the buoy measurements at the subduction buoy site as each of these terms is significant in magnitude; modifying either of them serves to reduce the net heat gain. We also note that the air–sea exchange fields in this region of trade wind forcing are typical of a significant portion of the World Ocean.
c. Comparison of climatological area mean heat exchange with hydrographic estimates
Climatological heat flux fields have been evaluated (and adjusted) in many analyses by comparison of the implied ocean heat transport with a limited number of hydrographic estimates (e.g., Isemer et al. 1989; da Silva et al. 1994). The disadvantage with this approach is that errors accumulate in the transport calculation with the potential to generate differences between the climatological and hydrographic heat transport estimates in regions where the surface fields are reliable. With the advent of an increasing number of hydrographic estimates of the ocean heat transport, a better approach is to compare implied mean surface net heat fluxes in regions bound by hydrographic sections with the corresponding climatological area means. Regional imbalances in the climatological fluxes linked to processes that are not well represented in the flux calculations may then be identified.
We have calculated implied surface heat fluxes for a number of boxes bound by hydrographic sections in the Atlantic (Hall and Bryden 1982; Rago and Rossby 1987;Klein et al. 1995; Saunders and King 1995; Bacon 1997;Holfort and Siedler 1997; Speer et al. 1999, manuscript submitted to J. Phys. Oceanogr.) and North Pacific (Roemmich and McCallister 1989; Bryden et al. 1991;Wijffels et al. 1996) and compared them with the corresponding climatological mean values, see Figs. 8 and 9 and Table 5. The variation of the two sets of estimates with latitude is shown for the Atlantic in Fig. 10, in which the box mean values are plotted against the midlatitude for each box. Note that error values for the climatological means are the standard deviation of the sample of 14 individual yearly mean values, which is taken as an upper limit for the random error. In the Atlantic, we find agreement within the error range in the majority of regions considered although there is a tendency for the climatological heat gain to be stronger than the hydrographic values. The main exception is the area bound by 32°N and the U.K. Control Volume Experiment (CONVEX) section for which the climatological heat loss is some 70 W m−2 weaker than the hydrographic value. Disagreement is also found in the southernmost box although the 30°S heat transport estimate is a preliminary value. Assessing the accuracy of the climatology with reference to hydrography in the Tropics is hampered by uncertainty over the contribution of the Ekman component of the heat transport, which is primarily responsible for the large error estimate of order 200 W m−2 on the heat flux in the region between 8° and 14°N. Fillenbaum et al. (1997) have revised the value for the heat transport across 24°N using measurements that resolve the seasonal variability of the boundary current contribution, to 1.44 ± 0.33 PW, which is 0.2 PW greater than the value of Hall and Bryden (1982). Using this estimate we obtain a reduction in the exchange for the 24°–32°N box to −9 ± 56 W m−2 and an increase in that for 14°–24°N to 23 ± 56 W m−2; the climatological estimates for each box remain within the error bounds of these revised hydrographic values.
Fewer hydrographic estimates are available for the Pacific; see Fig. 9. In the northern midlatitudes, comparison with the results of Roemmich and McCallister (1989) suggest that a significant component of the heat loss over the Kuroshio is not included in the climatological analysis; for the westernmost box in this region the climatological mean heat loss is −32 W m−2 while that from hydrography is −256 W m−2. We note that the FASINEX buoy comparison indicated that the turbulent heat loss was underestimated toward the southern edge of the region of strong winter losses that is centered on the Gulf Stream. The hydrographic comparison suggests that a similar bias may exist in the Kuroshio area. In contrast, in the central box the two estimates agree to within 1 W m−2. In the eastern box they differ by 25 W m−2; it is not clear whether this difference is significant as errors are not available for the hydrographic estimate. Farther south, the climatological heat gain is 40 W m−2 greater than the hydrographic value found by Wijffels et al. (1996) for the area between 10° and 24°N. A similar result was obtained for the 14°–24°N box in the Atlantic although in that case the climatological value remained within the hydrographic error bounds.
d. Possible mechanisms for regional bias in the climatological fields
1) Introduction
We now consider various processes that may cause regional biases in the heat flux fields and hence contribute to the global heat budget imbalance. The problem lies in determining reliable estimates for their relative contributions. We are not yet at the stage where we can claim to be able to make such estimates and hence suggest regional corrections to the SOC flux climatology that would close the heat budget. Indeed it is likely that adequate corrections will require additional information from satellites in the regions that are poorly sampled by ships. However, we do attempt to provide some reasoned discussion of which processes are likely to be most dominant.
2) Heat loss over the ice-covered ocean
Predominantly ice-covered regions of the world ocean have been excluded from our analysis. However, heat is transported by the ocean to these regions and lost to the atmosphere directly through openings in the ice (both small-scale leads and larger-scale polynyas) and indirectly by conductive processes through the ice layer. Large heat losses are known to occur over leads, turbulent fluxes of order 300 W m−2 being typical with the sensible heat dominating due to the strong sea–air temperature difference (Peixoto and Oort 1992). Estimating the mean heat loss over the ice-covered ocean is very difficult as the fractional area of the polar seas that are directly exposed to the atmosphere is not well known. Indirect estimates are possible based on atmospheric budget studies, in which the annual mean surface heat flux is obtained as a residual, but must also be treated with caution given the sparse nature of the atmospheric observing network at high latitudes (Nakamura and Oort 1988). Such studies give upper limits for the mean heat loss of order 50 W m−2, which would reduce the global imbalance by no more than 3 W m−2 as the proportion of the global ocean covered by ice is relatively small, just 7% in the annual mean.
3) Heat loss in the Southern Ocean
The Southern Ocean covers a relatively large proportion, nearly 20%, of the global ocean and is an area in which there are very few ship reports (see Fig. 2). Hence, it has the potential to be a major source of error in the global heat budget calculation. In particular, given the high winds that are typical of the Southern Ocean one might expect the true turbulent heat loss to be somewhat stronger than the analyzed values that we have obtained (the latter being strongly influenced by the raw estimates obtained in more heavily sampled areas farther north). The picture is further complicated by uncertainty over whether the turbulent heat loss is moderated because of the ocean and atmosphere being in a state of near equilibrium, as one might expect given the persistent westerly wind field, or whether it is enhanced by outbreaks of cold dry air from the Antarctic landmass as suggested by Trenberth and Solomon (1994).
Satellite-based estimates of the latent heat flux (Jourdan and Gautier 1995; Chou et al. 1997; Schulz et al. 1997) in the Southern Ocean tend to indicate stronger losses than we have obtained. However, there remain doubts over their accuracy because the techniques used to retrieve the near-surface humidity have not been properly validated at high latitudes. Preliminary comparisons with the various satellite fields indicate that the SOC estimates of the latent heat loss are lower by of order 20–60 W m−2. We note that if this difference represents an underestimate in the SOC analysis of 40 W m−2 maintained over the Southern Ocean, it would account for 8 W m−2 of the imbalance in the global heat budget. In addition to the latent heat flux, we might expect an underestimate of the sensible heat loss in the Southern Ocean but unfortunately we cannot appeal to satellite measurements to estimate the size of any bias.
Estimates of the heat loss over the Southern Ocean based on hydrographic measurements will be important in ascertaining whether this region plays a significant role in closing the global heat budget. A recent analysis of hydrographic measurements in the southeastern Pacific sector of the Southern Ocean indicates that the Antarctic Circumpolar Current loses large amounts of heat in this region (Gille and Donohue 1998). However, it remains unclear as to whether this heat is lost primarily to the atmosphere or through meridional exchanges with the subtropics.
4) Reduction in heat gain due to aerosol loading
The foregoing biases are caused by regional undersampling; it is also possible that the imbalance in the global heat budget is partly driven by processes that are not well represented in the flux formulas. Gilman and Garrett (1994) noted that attenuation of shortwave radiation by aerosols may cause a reduction in the surface insolation relative to the value obtained with the Reed formula in regions of strong aerosol loading due to reductions in the transmission factor for clear-sky radiation. The problem with attempting to quantify this effect is that reliable climatological estimates of the aerosol loading are not yet available (Kaufman et al. 1997). Using idealized representations of the marine aerosol layer, Gilman and Garrett (1994) found that the shortwave flux was attenuated by between 2% and 4%. Taking the latter number as an upper limit one might expect the SOC shortwave estimates in heavily aerosol laden regions of the Tropics to be biased high by up to 10 W m−2. The impact of this bias on the global heat budget will clearly depend on the spatial distribution of aerosols, which as noted above is not well known; given that aerosol loading is relatively light away from the Tropics, it is unlikely to be more than 2–3 W m−2 in the global mean.
5) Heat loss over boundary currents
The results of the hydrographic analyses in the Northern Hemisphere midlatitudes and the FASINEX buoy comparisons indicate that a significant portion of the turbulent heat loss is being missed in the SOC analysis in areas containing strong surface losses that are associated with the major western boundary currents. The FASINEX results suggest that this is in part due to an overestimate of the atmospheric humidity in our analysis. We have attempted to remove positive biases in the reported dewpoint temperatures that arise from observing procedure. However, it is possible that the application of Eq. (2) does not fully correct the reported values. Some evidence for an additional bias was noted in the VSOP-NA analysis (Kent et al. 1993a) from comparisons with weather ship measurements; however, the large spatial separation between the weather ship and the main shipping routes caused uncertainty in this conclusion. The result that the latent heat flux is strongly underestimated in our analysis at the FASINEX site but not over the subduction array might be explained if there was an additional bias in the reported dewpoint temperature which only becomes significant at low relative humidities. In that case, one would expect the latent heat flux to be underestimated in regions where the strongest exchanges occur during cold, dry air outbreaks. In contrast a reasonable estimate of the flux should be obtained in regions influenced by trade wind forcing, such as the subduction array, for which the loss process is steady and occurs at moderate relative humidities. We stress that this is a rather speculative suggestion, we plan to do a detailed comparison of the buoy and SOC analysis meteorological variables, at the level of individual ship reports in the latter, in order to resolve this issue.
In summary, we note that a combination of the various regionally dependent processes outlined above may be responsible for a significant proportion of the global imbalance in the net heat flux that we have found in the analysis of the SOC climatological fields. The processes discussed arise from undersampling [sections 5d(2) and 5d(3)], inadequate representation of specific processes in the flux formulas [section 5d(4)], and uncorrected ship errors [section 5d(5)]. We do not rule out the possibility that part of the imbalance may be due to residual scale biases in the bulk formulas that should be corrected globally. However, we note that such biases should not be regarded as the only solution to the heat budget problem and stress that the comparisons with the buoy measurements show that simple global adjustments of the bulk formulas can lead to greater biases in the heat flux estimates than exist in the original fields.
6. Discussion and summary
We have presented results from an analysis of the Southampton Oceanography Centre (SOC) heat flux climatology that show that a more sophisticated approach to the problem of closure of the global ocean heat budget is required than the simple scale adjustment of fluxes employed in many past studies. The climatology has been calculated using empirical flux formulas applied to in situ reports in the COADS1a onto which additional metadata from the WMO47 list of ships have been merged. The additional metadata have allowed corrections to be made to the meteorological variables for biases arising from observational procedure. The effect of the corrections is complex and we have attempted to quantify it by calculating repeat fields for sample months in which various subsets of the corrections are made. Regional changes of up to 15 W m−2 are found with significant seasonal variations depending primarily on the spatial distribution of different measurement methods and incident solar radiation as well as the magnitude of the flux component being corrected. We note that our use of in situ reports is likely to impose spatial variations in the accuracy of the calculated fluxes with the reduction in sampling frequency in the Southern Hemisphere leading to larger errors.
A common problem in global climatological analyses of the heat exchange has been the inability to close the global ocean heat budget. We encounter the same problem in our analysis; the global mean net heat flux is a gain of 30 W m−2 by the ocean. A similar value was obtained from the UWM/COADS fields, which did not include corrections for systematic errors at the level of individual ships. Hence, it seems likely that although systematic errors due to observing procedure are important at the regional level they do not have a significant impact on the global mean net heat flux. It is worth noting that the closure problem is not confined to analyses based on in situ reports. In a recent study, in which the best available combination of turbulent heat flux fields from the National Centers for Environmental Prediction atmospheric model reanalysis program and radiative fluxes from the International Satellite Cloud Climatology Project were selected to force an ocean model, the unadjusted global mean net heat gain by the ocean was 43 W m−2 (Large et al. 1997).
Proportional adjustments of the various flux components have been made in order to balance the global heat budget in many previous studies. In order to assess the validity of this approach, we have compared both our original flux estimates, and adjusted values obtained using a scheme similar to that suggested by da Silva et al. (1994), with high quality research buoy measurements. Close agreement of the unadjusted net heat fluxes with the measured values is found for several buoys deployed in the Subduction Array off the coast of northwest Africa. However, the adjusted values differ from the buoy measurements by of order 30 W m−2. Although we have focused on the scheme suggested by da Silva et al. (1994), any scheme in which the ocean heat budget is closed by a combination of reduced shortwave and increased latent heat fluxes will lead to differences with respect to the buoy measurements as each of these terms is significant in magnitude at this site and modifying either of them reduces the net heat gain. We note that the atmospheric forcing in the area spanned by the subduction buoy array is typical of that over a significant proportion of the global ocean, that is, the trade wind belts. In contrast, at other buoy deployment sites in the western equatorial Pacific warm pool and south of Bermuda in the North Atlantic, the flux adjustment was found to improve the estimate of the net heat exchange. However, even at these sites the results for one of the flux components were made worse in making the adjustment. Thus, the results of the buoy comparisons indicate that simple scale adjustments to the flux components applied in a global sense are not appropriate. Instead more detailed consideration needs to be given to the processes responsible for regional biases in the climatology.
The amount by which the latent heat flux may be increased in order to close the heat budget can potentially be constrained using the ocean freshwater budget in which this term appears as the evaporation rate (e.g., Schmitt 1995). The problem with appealing to this constraint is that precipitation rates over the ocean are still poorly known. We have obtained precipitation estimates within the SOC climatology project using the empirical relation of Dorman and Bourke (1978) applied to present weather reports. However, the validity of this relation has been called into question (Elliot and Reed 1979) and we feel that before the SOC estimates can reliably be employed as part of a freshwater budget study they must be evaluated using satellite datasets that are now becoming available. Hence, we have not attempted to use the freshwater budget as a constraint in this paper.
We have made a first attempt at identifying some of those regions in which biases in the heat fluxes are important by using hydrographic results. Values of the SOC climatological mean surface heat flux in various boxes bound by hydrographic sections have been compared with the corresponding figures derived from the hydrographic heat transport estimates across the box boundaries. This method is preferred to comparing the climatologically implied ocean heat transport with hydrography as it avoids the accumulation of errors. The climatological and hydrographic estimates agree within the error limits in most of the boxes considered with some evidence being found for regional biases in several areas. In particular, significant underestimates of the surface heat loss are found in the North Atlantic and North Pacific boxes, which contain the strongest surface flux expression of the major western boundary currents. We note that in the tropical Atlantic, the errors on the hydrographic estimates are uncomfortably large and thus it is possible that significant biases exist in the climatological fluxes in this region that have not been revealed by the comparison with hydrography.
Various regional biases that may contribute to the heat budget closure problem have been discussed. The lack of observations in the Southern Ocean coupled with the influence of more heavily sampled regions farther to the north may have led us to underestimate the turbulent heat loss at high southern latitudes. Heat loss over ice-covered regions of the World Ocean has not been included in our analysis but the small fractional area that this occupies suggests that this term would only play a minor role in closing the heat budget. Further observational biases that we have not been able to correct for at present may prove important, in particular a residual bias in the reported dewpoint temperature at low relative humidities may be partly responsible for the underestimated heat loss in the western North Atlantic and North Pacific. Finally, we have noted that the neglect of the attenuating effect of aerosol loading on the estimated shortwave flux in our analysis may also contribute to the heat budget problem at least for certain regions.
In conclusion, we restate the new insights into the heat budget closure problem provided by our study. The first of these is that systematic errors due to observing procedure are important at the regional level but do not have a significant impact on the global mean net heat flux. The second is that buoy collocation comparisons show that the ocean heat budget must be closed by regional not global adjustments. Comparisons with hydrography potentially provide a method to correct for the regional biases. At present we are limited by the area of ocean bound by reliable hydrographic estimates of the heat transport. However, as more estimates become available in the future, particularly as a result of the World Ocean Circulation Experiment program, we envisage being able to use the implied surface heat fluxes to produce a regionally adjusted version of the SOC climatology in which the global heat budget is closed without the introduction of biases in regions that do not require adjustment.
Acknowledgments
The work described in this paper was partially funded by a commissioned research project for the Hadley Centre, U.K. Meteorological Office. The authors would like to thank Steven Worley at the Data Support Section, NCAR, and Scott Woodruff at the NOAA/ERL Climate Diagnostics Center for supplying the COADS 1a; Bob Weller of WHOI for making available the various research buoy datasets; Arlindo da Silva at the Data Assimilation Office, NASA/GSFC, for advice and providing the objective analysis code; and Daniel Oakley for assistance in preparing the merged COADS1a–WMO47 dataset. We also wish to acknowledge the useful comments of the two anonymous referees.
Note on availability of fields: The SOC flux climatology is freely available to interested users for noncommercial scientific research, for details of how to access the fields see http://www.soc.soton.ac.uk/JRD/MET/fluxclimatology.html.
REFERENCES
Alexander, R., and R. Mobley, 1976: Monthly average sea-surface temperature and ice-pack limits on a 1° global grid. Mon. Wea. Rev.,104, 143–148.
Bacon, S., 1997: Circulation and fluxes in the North Atlantic between Greenland and Ireland. J. Phys. Oceanogr.,27, 1420–1435.
Bignami, F., S. Marullo, R. Santoleri, and M. E. Schiano, 1995: Longwave radiation budget in the Mediterranean Sea. J. Geophys. Res.,100 (C2), 2501–2514.
Bryden, H. L., D. H. Roemmich, and J. A. Church, 1991: Ocean heat transport across 24°N in the Pacific. Deep-Sea Res.,38, 297–324.
Bunker, A. F., 1976: Computations of surface energy flux and annual air–sea interaction cycles of the North Atlantic Ocean. Mon. Wea. Rev.,104, 1122–1140.
Chou, S.-H., C-L. Shie, R. M. Atlas, and J. Ardizzone, 1997: Air–sea fluxes retrieved from special Sensor Microwave Imager data. J. Geophys. Res.,102 (C6), 12 705–12 726.
Clark, N. E., L. Eber, R. M. Laurs, J. A. Renner, and J. F. T. Saur, 1974: Heat exchange between ocean and atmosphere in the eastern North Pacific for 1961–71. NOAA Tech. Rep. NMFS SSRF-682, U.S. Dept. of Commerce, Washington, DC, 108 pp.
da Silva, A. M., C. C. Young, and S. Levitus, 1994: Atlas of Surface Marine Data. Vol. 1, Algorithms and Procedures. NOAA Atlas Series, NOAA, 74 pp.
Dobson, F. W., and S. D. Smith, 1988: Bulk models of solar radiation at sea. Quart. J. Roy. Meteor. Soc.,114, 165–182.
Dorman, C. E., and R. H. Bourke, 1978: A temperature correction for Tucker’s ocean rainfall estimates. Quart. J. Roy. Meteor. Soc.,104, 765–773.
Elliot, W. P., and R. K. Reed, 1979: Comments on “A temperature correction for Tucker’s ocean rainfall estimates.” Quart. J. Roy. Meteor. Soc.,105, 1082–1083.
Esbensen, S. K., and Y. Kushnir, 1981: The heat budget of the global ocean: An atlas based on estimates from surface marine observations. Climate Research Institute Rep. 29, Oregon State University, 27 pp. [Available from Department of Atmospheric Sciences, OSU, Corvallis, OR 97531.].
——, and R. W. Reynolds, 1981: Estimating monthly averaged air–sea transfers of heat and momentum using the bulk aerodynamic method. J. Phys. Oceanogr.,11, 457–465.
Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: Bulk parameterization of air–sea fluxes for TOGA-COARE. J. Geophys. Res.,101, 3747–3764.
Fillenbaum, E. R., T. N. Lee, W. E. Johns, and R. J. Zantopp, 1997:Meridional heat transport variability at 26.5°N in the North Atlantic. J. Phys. Oceanogr.,27, 153–174.
Gille, S. T., and K. A. Donohue, 1998: Alongstream changes in Antarctic Circumpolar Current heat transport: A case study in the southeastern Pacific. J. Geophys. Res.,104 (C3), 5191–5210.
Gilman, C., and C. Garrett, 1994: Heat flux parameterizations for the Mediterranean Sea: The role of atmospheric aerosols and constraints from the water budget. J. Geophys. Res.,99 (C3), 5119–5134.
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.
Hall, M. M., and H. L. Bryden, 1982: Direct estimates and mechanisms of ocean heat transport. Deep-Sea Res.,29 (3A), 339–359.
Harrison, D. E., 1989: On climatological monthly mean wind stress and wind stress curl fields over the World Ocean. J. Climate,2, 57–70.
Hasse, L., and S. D. Smith, 1997: Local sea surface wind, wind stress, and sensible and latent heat fluxes. J. Climate,10, 2711–2724.
Holfort, J., and G. Siedler, 1997: The heat and mass transport in the South Atlantic. Int. WOCE News.,28, 3–5.
Isemer, H.-J., J. Willebrand, and L. Hasse, 1989: Fine adjustment of large scale air–sea energy flux parameterizations by direct estimates of ocean heat transport. J. Climate,2, 1173–1184.
Josey, S. A., E. C. Kent, and P. K. Taylor, 1995: Seasonal variations between sampling and classical mean turbulent heat flux estimates in the eastern North Atlantic. Ann. Geophys.,13, 1054–1064.
——, D. Oakley, and R. W. Pascal, 1997: On estimating the atmospheric longwave flux at the ocean surface from ship meteorological reports. J. Geophys. Res.,102 (C13), 27 961–27 972.
——, E. C. Kent, and P. K. Taylor, 1998: The Southampton Oceanography Centre (SOC) Ocean–Atmosphere Heat, Momentum and Freshwater Flux Atlas. Southampton Oceanography Centre Rep. 6, Southampton, United Kingdom, 30 pp. + figures.
Jourdan, D., and C. Gautier, 1995: Comparison between global latent heat flux computed from multisensor (SSM/I and AVHRR) and from in situ data. J. Atmos. Oceanic Technol.,12, 46–72.
Kaufman, Y. J., and Coauthors, 1997: Passive remote sensing of tropospheric aerosol and atmospheric correction for the aerosol effect. J. Geophys. Res.102 (D14), 16 815–16 830.
Kent, E. C., and P. K. Taylor, 1991: Ships observing marine climate:A catalogue of the voluntary observing ships participating in the VSOP-NA. WMO Marine Meteorology and Related Oceanographic Activities Rep. 25, 123 pp.
——, and ——, 1996: Accuracy of humidity measurements on ships:Consideration of solar radiation effects. J. Atmos. Oceanic Technol.,13, 1317–1321.
——, and ——, 1997: Choice of a Beaufort equivalent scale. J. Atmos. Oceanic Technol.,14, 228–242.
——, ——, B. S. Truscott, and J. A. Hopkins, 1991: The accuracy of ship’s meteorological observations—Results of the VSOP-NA. WMO Marine Meteorology and Related Oceanographic Activities Rep. 26, 86 pp.
——, ——, ——, and J. S. Hopkins, 1993a: The accuracy of voluntary observing ships meteorological observations—Results of the VSOP–NA. J. Atmos. Oceanic Technol.,10, 591–608.
——, R. J. Tiddy, and P. K. Taylor, 1993b: Correction of marine air temperature observations for solar radiation effects. J. Atmos. Oceanic Technol.,10, 900–906.
——, P. G. Challenor, and P. K. Taylor, 1999: A statistical determination of the random observational errors present in voluntary observing ships’ meteorological reports. J. Atmos. Oceanic Technol.,16, 920–929.
Klein, B., R. L. Molinari, T. J. Mueller, and G. Siedler, 1995: A transatlantic section at 14.5°N: Meridional volume and heat fluxes. J. Mar. Res.,53, 929–957.
Large, W. G., G. Danabasoglu, S. C. Doney, and J. C. McWilliams, 1997: Sensitivity to surface forcing and boundary layer mixing in a global ocean model: Annual mean climatology. J. Phys. Oceanogr.,27, 2418–2447.
Lindau, R., 1995: A new Beaufort equivalent scale. Proc. Int. COADS Winds Workshop, Kiel, Germany, Institüt für Meerskunde, 232–252.
Macdonald, A. M., and C. Wunsch, 1996: An estimate of global ocean circulation and heat fluxes. Nature,382, 436–439.
Moyer, K. A., and R. A. Weller, 1997: Observations of surface forcing from the Subduction Experiment: A comparison with global model products and climatological datasets. J. Climate,10, 2725–2742.
Nakamura, N., and A. H. Oort, 1988: Atmospheric heat budgets of the polar regions. J. Geophys. Res.,93, 9510–9524.
Oberhuber, J. M., 1988: An atlas based on the COADS data set: The budgets of heat, buoyancy and turbulent kinetic energy at the surface of the global ocean. MPI Rep. 15, 19 pp. [Available from Max-Plank-Institut für Meteorologie, Bundestrasse 55, D-2000, Hamburg 13, Germany.].
Parrilla, G., A. Lavin, H. Bryden, M. Garcia, and R. Millard, 1994:Rising temperatures in the subtropical North Atlantic ocean over the past 35 years. Nature,369, 48–51.
Payne, R. E., 1972: Albedo of the sea surface. J. Atmos. Sci.,29, 959–970.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.
Rago, T. A., and H. T. Rossby, 1987: Heat transport into the North Atlantic Ocean north of 32°N latitude. J. Phys. Oceanogr.,17, 854–871.
Reed, R. K., 1977: On estimating insolation over the ocean. J. Phys. Oceanogr.,7, 482–485.
Roemmich, D., and T. McCallister, 1989: Large scale circulation of the North Pacific Ocean. Progress in Oceanography, Vol. 22, Pergamon Press, 171–204.
Saunders, P. M., and B. A. King, 1995: Oceanic fluxes on the WOCE A11 section. J. Phys. Oceanogr.,25, 1942–1958.
Schmitt, R. W., 1995: The ocean component of the global water cycle. Rev. Geophys.,33 (Suppl.), 1395–1409.
Schulz, J., J. Meywerk, S. Ewald, and P. Schlussel, 1997: Evaluation of satellite derived latent heat fluxes. J. Climate,10, 2782–2795.
Seckel, G. R., and F. H. Beaudry, 1973: The radiation from sun and sky over the Pacific Ocean (abstract). Trans. Amer. Geophys. Union,54, 1114.
Smith, S. D., 1980: Wind stress and heat flux over the ocean in gale force winds. J. Phys. Oceanogr.,10, 709–726.
——, 1988: Coefficients for sea surface wind stress, heat flux and wind profiles as a function of wind speed and temperature. J. Geophys. Res.,93, 15 467–15 474.
Trenberth, K. E., and A. Solomon, 1994: The global ocean heat balance: Heat transports in the atmosphere and ocean. Climate Dyn.,10, 107–134.
WCRP, 1986: Scientific plan for the World Ocean Experiment. WCRP Publ. Series 6, WMO/TD 122, 83 pp.
WCRP/GEWEX, 1996: Report of the Eighth Session of the Working Group on Radiative Fluxes. WCRP Informal Rep. 13/1996, Killiney Bay, Dublin, Ireland, 26 pp.
Weller, R. A., and S. P. Anderson, 1996: Surface meteorology and air–sea fluxes in the western equatorial warm pool during the TOGA Coupled Ocean–Atmosphere Response Experiment. J. Climate,9, 1959–1990.
——, D. L. Rudnick, and N. J. Brink, 1995: Meteorological variability and air–sea fluxes at a closely spaced array of surface moorings. J. Geophys. Res.,100 (C3), 4867–4883.
——, M. F. Baumgartner, S. A. Josey, A. S. Fischer, and J. Kindle, 1998: Atmospheric forcing in the Arabian Sea during 1994–1995: Observations and comparisons with climatology and model. Deep-Sea Res.,45, 1961–1999.
Wijffels, S. E., J. M. Toole, H. L. Bryden, R. A. Fine, W. J. Jenkins, and J. L. Bullister, 1996: The water masses and circulation at 10°N in the Pacific. Deep-Sea Res.,43, 501–544.
WMO, 1970: The Beaufort scale of wind force (technical and operational aspects). World Meteorological Organisation Reports on Marine Science Affairs, Rep. 3, 22 pp.
——, 1993: International list of selected, supplementary and auxiliary ships. WMO Rep. 47, 264 pp.
Woodruff, S. D., S. J. Lubker, K. Wolter, S. J. Worley, and J. D. Elms, 1993: Comprehensive Ocean–Atmosphere Data Set (COADS) release 1a: 1980–92. Earth Syst. Monit.,4, 4–8.
Yelland, M. J., B. I. Moat, P. K. Taylor, R. W. Pascal, J. Hutchings, and V. C. Cornell, 1998: Wind stress measurements from the open ocean corrected for air flow disturbance by the ship. J. Phys. Oceanogr.,28, 1511–1526.
Time series of the total number of reports, N(reports), for each individual month over the period Jan 1980–Dec 1993 in the filtered version of COADS1a, that is, that in which reports with the source exclusion flag set have been removed; the number of those that are ship reports, N(ships), and the number of reports for which matching with an entry in the WMO47 list of ships has been possible, N(matched).
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Time series of the total number of reports, N(reports), for each individual month over the period Jan 1980–Dec 1993 in the filtered version of COADS1a, that is, that in which reports with the source exclusion flag set have been removed; the number of those that are ship reports, N(ships), and the number of reports for which matching with an entry in the WMO47 list of ships has been possible, N(matched).
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Time series of the total number of reports, N(reports), for each individual month over the period Jan 1980–Dec 1993 in the filtered version of COADS1a, that is, that in which reports with the source exclusion flag set have been removed; the number of those that are ship reports, N(ships), and the number of reports for which matching with an entry in the WMO47 list of ships has been possible, N(matched).
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the number of latent heat flux estimates per 1° × 1° square used to form the average climatological monthly mean for the period 1980–93. White indicates that there were no observations over the entire period considered. Black crosses indicate the positions of the Woods Hole buoy deployments used in the comparison described in section 5b.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the number of latent heat flux estimates per 1° × 1° square used to form the average climatological monthly mean for the period 1980–93. White indicates that there were no observations over the entire period considered. Black crosses indicate the positions of the Woods Hole buoy deployments used in the comparison described in section 5b.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the number of latent heat flux estimates per 1° × 1° square used to form the average climatological monthly mean for the period 1980–93. White indicates that there were no observations over the entire period considered. Black crosses indicate the positions of the Woods Hole buoy deployments used in the comparison described in section 5b.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Objectively analyzed climatological mean fields for Jan and Jul of (a) and (b) the latent heat flux, (c) and (d) the sensible heat flux, (e) and (f) the longwave flux, (g) and (h) the shortwave flux, and (i) and (j) the net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Objectively analyzed climatological mean fields for Jan and Jul of (a) and (b) the latent heat flux, (c) and (d) the sensible heat flux, (e) and (f) the longwave flux, (g) and (h) the shortwave flux, and (i) and (j) the net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Objectively analyzed climatological mean fields for Jan and Jul of (a) and (b) the latent heat flux, (c) and (d) the sensible heat flux, (e) and (f) the longwave flux, (g) and (h) the shortwave flux, and (i) and (j) the net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the fully corrected and uncorrected flux fields for Jan and Jul 1990 (a) and (b) sensible, (c) and (d) latent, (e) and (f) longwave, and (g) and (h) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the fully corrected and uncorrected flux fields for Jan and Jul 1990 (a) and (b) sensible, (c) and (d) latent, (e) and (f) longwave, and (g) and (h) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the fully corrected and uncorrected flux fields for Jan and Jul 1990 (a) and (b) sensible, (c) and (d) latent, (e) and (f) longwave, and (g) and (h) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
(Continued)
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the SOC and UWM/COADS annual mean fields for (a) latent heat flux, (b) sensible heat flux, and (c) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the SOC and UWM/COADS annual mean fields for (a) latent heat flux, (b) sensible heat flux, and (c) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Global distribution of the difference between the SOC and UWM/COADS annual mean fields for (a) latent heat flux, (b) sensible heat flux, and (c) net heat flux; units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Time series of monthly mean flux measurements from the NE buoy (solid line) and estimates from the SOC flux dataset (dash–dot line) for the individual heat flux components (QH, sensible; QE, latent; QLW, longwave; QSW, shortwave), the turbulent (QTUR) and radiative (QRAD) fluxes, and the net heat flux (QNET); units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Time series of monthly mean flux measurements from the NE buoy (solid line) and estimates from the SOC flux dataset (dash–dot line) for the individual heat flux components (QH, sensible; QE, latent; QLW, longwave; QSW, shortwave), the turbulent (QTUR) and radiative (QRAD) fluxes, and the net heat flux (QNET); units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Time series of monthly mean flux measurements from the NE buoy (solid line) and estimates from the SOC flux dataset (dash–dot line) for the individual heat flux components (QH, sensible; QE, latent; QLW, longwave; QSW, shortwave), the turbulent (QTUR) and radiative (QRAD) fluxes, and the net heat flux (QNET); units W m−2.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Bar chart representation of the difference between the SOC and buoy mean heat fluxes at each deployment site for (a) the longwave (striped bar) and shortwave (shaded bar) fluxes, (b) the sensible (striped) and latent (shaded) fluxes, and (c) the unadjusted (striped) and adjusted (shaded) net heat fluxes (see text for description of adjustment); units W m−2. Note that the values for the net heat flux difference shown for the FASINEX site are intended to be indicative. They have been calculated under the assumption that if the FASINEX buoys had been instrumented with pyrgeometers the measured longwave flux would have been the same as the SOC estimate.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Bar chart representation of the difference between the SOC and buoy mean heat fluxes at each deployment site for (a) the longwave (striped bar) and shortwave (shaded bar) fluxes, (b) the sensible (striped) and latent (shaded) fluxes, and (c) the unadjusted (striped) and adjusted (shaded) net heat fluxes (see text for description of adjustment); units W m−2. Note that the values for the net heat flux difference shown for the FASINEX site are intended to be indicative. They have been calculated under the assumption that if the FASINEX buoys had been instrumented with pyrgeometers the measured longwave flux would have been the same as the SOC estimate.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Bar chart representation of the difference between the SOC and buoy mean heat fluxes at each deployment site for (a) the longwave (striped bar) and shortwave (shaded bar) fluxes, (b) the sensible (striped) and latent (shaded) fluxes, and (c) the unadjusted (striped) and adjusted (shaded) net heat fluxes (see text for description of adjustment); units W m−2. Note that the values for the net heat flux difference shown for the FASINEX site are intended to be indicative. They have been calculated under the assumption that if the FASINEX buoys had been instrumented with pyrgeometers the measured longwave flux would have been the same as the SOC estimate.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the Atlantic; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the Atlantic; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the Atlantic; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the North Pacific; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap, in the absence of error bars on the hydrographic estimate a value of 30 W m−2 is assumed.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the North Pacific; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap, in the absence of error bars on the hydrographic estimate a value of 30 W m−2 is assumed.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Comparison of the climatological and implied hydrographic mean heat gain for various boxes bound by hydrographic sections in the North Pacific; units W m−2. Normal text indicates the hydrographic value, italic text the SOC climatological mean. Shading indicates that the SOC and hydrographic error ranges do not overlap, in the absence of error bars on the hydrographic estimate a value of 30 W m−2 is assumed.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Variation in climatological (dash–dot line) and implied hydrographic (solid line) mean heat gain with latitude in the Atlantic; units W m−2. Error values for the climatological means are the standard deviation of the sample of fourteen individual yearly mean values, which are taken as an upper limit for the random error.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Variation in climatological (dash–dot line) and implied hydrographic (solid line) mean heat gain with latitude in the Atlantic; units W m−2. Error values for the climatological means are the standard deviation of the sample of fourteen individual yearly mean values, which are taken as an upper limit for the random error.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Variation in climatological (dash–dot line) and implied hydrographic (solid line) mean heat gain with latitude in the Atlantic; units W m−2. Error values for the climatological means are the standard deviation of the sample of fourteen individual yearly mean values, which are taken as an upper limit for the random error.
Citation: Journal of Climate 12, 9; 10.1175/1520-0442(1999)012<2856:NIITOH>2.0.CO;2
Effect of the VSOP-NA corrections on mean heat fluxes for the specified regions in Jan and Jul of 1990, positive values indicate an increase in heat loss by the ocean following the corrections (units, W m−2). For the sensible, latent, and longwave fluxes the dominant corrections are indicated as follows: “S” for the SST correction, “I” for the combined effects of the individual height correction and the Lindau scale, “R” for the radiation correction to air temperature, and “H” for the screen dewpoint correction.
Values for the global mean air–sea heat flux components and the net heat flux from the SOC climatology over the period 1980–93 and the UWM/COADS climatology over 1945–89 (units, W m−2).
Comparison of mean heat fluxes from the Subduction Array buoys and the SOC flux dataset averaged over the period of each buoy deployment for which reliable observations were available according to Moyer and Weller (1997) (units, W m−2). The number of months covered in each comparison is given by Nm. In addition, the mean monthly difference (SOC–buoy) for each field is listed with its standard error. The value in the last column,
Comparison of mean heat fluxes from various WHOI buoy deployments and the SOC flux dataset averaged over the deployment period (units, W m−2). The number of months covered in each comparison is given by Nm. In addition, the mean monthly difference (SOC–buoy) for each field is listed with its standard error. The value in the last column,
Table 5a. Comparison of climatological and implied hydrographic mean heat flux for various boxes in the Atlantic (units, W m−2). Sources for hydrographic heat transport estimates: CONVEX section [southern Greenland–Ireland; Bacon (1997)]; 32°N (Rago and Rossby 1987); 24°N (Hall and Bryden 1982); 14°N, 8°N (Klein et al. 1995);11°S (Speer et al. 1999, manuscript submitted to J. Phys. Oceanogr.);30°S (Holfort and Siedler 1997); A11 section [South Atlantic dogleg, Saunders and King (1995)]. Error values for the climatological means are the standard deviation of the sample of 14 individual yearly mean values, which is taken as an upper limit for the random error.
Table 5b. Comparison of climatological and implied hydrographic mean heat flux for various boxes in the North Pacific (units, W m−2). Sources for hydrographic heat transport estimates W, C, and E boxes (Roemmich and McCallister 1989); 24°N (Bryden et al. 1991); 10°N (Wijffels et al. 1996). Error values for the climatological means are the standard deviation of the sample of 14 individual yearly mean values which is taken as an upper limit for the random error.
