• Alados-Arboledas, L., J. Vida, and J. I. Jimenez, 1988: Effects of solar radiation on the performance of pyrgeometers with silicon domes. J. Atmos. Oceanic Technol.,5, 666–670.

  • Albrecht, B., and S. K. Cox, 1977: Procedures for improving pyrgeometer performance. J. Appl. Meteor.,16, 188–197.

  • ——, C. S. Bretherton, D. Johnson, W. H. Schubert, and A. S. Frisch, 1995: The Atlantic Stratocumulus Transition Experiment—ASTEX. Bull. Amer. Meteor. Soc.,76, 889–904.

  • Berliand, M. E., and T. G. Berliand, 1952: Determining the net longwave radiation of the Earth with consideration of the effect of cloudiness. Isv. Akad. Nauk. SSSR Ser. Geofis,1..

  • Bignami, F., S. Marullo, R. Santoleri, and M. E. Schiano, 1995: Longwave radiation budget in the Mediterranean Sea. J. Geophys. Res.,100, 2501–2514.

  • Bolton, D., 1980: The computation of equivalent potential temperature. Mon. Wea. Rev.,108, 1046–1053.

  • Bradley, E. F., and R. A. Weller, 1995: Proceedings of the Joint Workshop of the TOGA COARE Flux and Atmospheric Working Groups, Rep. TOGA COARE International Project Office, UCAR, 34 pp. [Available from UCAR, P. O. Box 3000, Boulder, CO 80307.].

  • Brink, N. J., K. A. Moyer, R. P. Trask, and R. A. Weller, 1995: The Subduction Experiment: Mooring field program and data summary. Tech. Rep. WHOI-95-08, 113 pp. [Available from WHOI, Woods Hole, MA 02543.].

  • Budyko, M. I., 1974: Climate and Life. International Geophysical Series, Vol. 18, Academic Press, 508 pp.

  • 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.

  • Cardone, V. J., J. G. Greenwood, and M. Cane, 1990: On trends in historical marine wind data. J. Climate,3, 113–127.

  • Charnock, H., 1955: Wind stress on a water surface. Quart. J. Roy. Meteor. Soc.,81, 639–640.

  • Crescenti, G. H., and R. A. Weller, 1989: Improved meteorological measurements from buoys and ships: Preliminary comparison of precipitation sensors. Tech. Rep. WHOI-89-44, IMET-TR-89-01, 32 pp. [Available from WHOI, Woods Hole, MA 02543.].

  • da Silva, A. M., C. C. Young, and S. Levitus, 1994: Atlas of Surface Marine Data 1994. Vol. 1, Algorithms and Procedures, U.S. Department of Commerce, 74 pp.

  • Dickey, T. D., D. V. Manov, R. A. Weller, and D. A. Siegel, 1994: Determination of longwave heat flux at the air–sea interface using measurements from buoy platforms. J. Atmos. Oceanic Technol.,11, 1057–1078.

  • ECMWF Technical Attachment, 1994: The description of the ECMWF/WCRP level III-A global atmospheric data archive. Rep. European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 72 pp. [Available from ECMWF, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Efimova, N. A., 1961: On methods of calculating monthly values of net longwave radiation. Meteor. Gidrol.,10, 28–33.

  • Fairall, C. W., E. F. Bradley, J. S. Godfrey, G. A. Wick, J. B. Edson, and G. S. Young, 1996a: Cool-skin and warm-layer effects on sea surface temperature. J. Geophys. Res.,101, 1295–1308.

  • ——, ——, D. P. Rogers, J. B. Edson, and G. S. Young, 1996b: Bulk parameterization of air–sea fluxes for TOGA COARE. J. Geophys. Res.,101, 3747–3764.

  • Gill, G. C., 1983: Comparison testing of selected naturally ventilated solar radiation shields. Rep. NOAA/National Data Buoy Center, Bay St. Louis, MS, 38 pp.

  • 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, 5119–5134.

  • Godfrey, J. S., and A. C. M. Beljaars, 1991: On the turbulent fluxes of buoyancy, heat, and moisture at the air–sea interface at low wind speeds. J. Geophys. Res.,96, 22th043–22048.

  • Hosom, D. S., R. A. Weller, R. E. Payne, and K. E. Prada, 1995: The IMET (Improved Meteorology) ship and buoy systems. J. Atmos. Oceanic Technol.,12, 527–540.

  • Isemer, H.-J., and L. Hasse, 1985: The Bunker Climate Atlas of theNorth Atlantic Ocean. Vol. 1, Observations, Springer-Verlag, 218 pp.

  • ——, and ——, 1987: The Bunker Climate Atlas of the North Atlantic Ocean. Vol. 2, Air-Sea Interactions, Springer-Verlag, 252 pp.

  • Kanamitsu, M., 1989: Description of the NMC global data assimilation and forecast system. Wea. Forecasting,4, 335–342.

  • ——, J. C. Alpert, K. A. Campana, P. M. Caplan, D. G. Deaven, M. Iredell, B. Katz, H.-L. Pan, J. Sela, and G. H. White, 1991: Recent changes implemented into the global forecast system at NMC. Wea. Forecasting,6, 425–435.

  • Kaufeld, L., 1981: The development of a new Beaufort equivalent scale. Meteor. Rundsch.,34, 17–23.

  • Large, W. G., and S. Pond, 1981: Open ocean momentum flux measurements in moderate to strong winds. J. Phys. Oceanogr.,11, 324–336.

  • ——, and ——, 1982: Sensible and latent heat fluxes over the ocean. J. Phys. Oceanogr.,12, 464–482.

  • Liu, W. T., K. B. Katsaros, and J. A. Businger, 1979: Bulk parameterization of air-sea exchanges of heat and water vapor including the molecular constraints at the interface. J Atmos. Sci.,36, 1722–1735.

  • Luyten, J. R., J. Pedlosky, and H. Stommel, 1983: The ventilated thermocline. J. Phys. Oceanogr.,13, 292–309.

  • Marshall, J. C., A. J. G. Nurser, and R. G. Williams, 1993: Inferring the subduction rate and period over the North Atlantic. J. Phys. Oceanogr.,23, 1315–1329.

  • Miller, M. J., A. C. M. Beljaars, and T. N. Palmer, 1992: The sensitivity of the ECMWF model to the parameterization of evaporation from the tropical oceans. J. Climate,5, 418–434.

  • 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. Max-Planck-Institut fur Meteorologie Rep. 15. [Available from Max-Planck-Institut für Meteorologie, Bundesstrasse 55, D-20146, Hamburg, Germany.].

  • Ockert-Bell, M., and D. Hartmann, 1992: The effect of cloud type on earth’s energy balance: Results for selected regions, J. Climate,5, 1157–1171.

  • Payne, R. E., 1972: Albedo of the sea surface. J. Atmos. Sci.,29, 959–970.

  • Quayle, R. G., 1980: Climatic comparisons of estimated and measured wind from ships. J. Appl. Meteor.,19, 142–156.

  • Reed, R. K., 1977: On estimating insolation over the ocean. J. Phys. Oceanogr.,7, 482–485.

  • Reynolds, R. W., 1991: A new global sea surface temperature analysis. Tech. Procedure Bulletin 393, 31 pp. [Available from National Weather Service, Office of Meteorology, Silver Spring, MD 20910.].

  • Smith, S. D., 1988: Coefficients for sea surface wind stress, heat flux, and wind profiles as a function of wind speed and temperature. J. Geophys. Res.,93, 15467–15472.

  • Stommel, H. M., 1979: Determination of the water mass properties of water pumped down from the Ekman layer to the geostrophic flow below. Proc. Natl. Acad. Sci. U.S.A.,76, 3051–3055.

  • Weller, R. A., and R. E. Davis, 1980: A vector measuring current meter. Deep-Sea Res.,27A, 565–582.

  • ——, D. L. Rudnick, R. E. Payne, J. P. Dean, N. J. Pennington, and R. P. Trask, 1990: Measuring near-surface meteorology over the ocean from an array of surface moorings in the subtropical convergence zone. J. Atmos. Oceanic Technol.,7, 85–103.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A Comprehensive Atmosphere–Ocean Data Set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

  • Wright, P. B., 1988: An atlas based on the “COADS” data set: Fields of mean wind, cloudiness and humidity at the surface of the global ocean. Max-Planck-Institut für Meteorologie Rep. 14, 73 pp. [Available from Max-Planck-Institut für Meteorologie, Bundesstrasse 55, D-20146, Hamburg, Germany.].

  • Young-Molling, C. C., A. M. da Silva, and S. Levitus, 1995: Atlas of Surface Marine Data 1994 Supplement A: Anomalies of Directly Observed Quantities and Surface Marine Fluxes for 19901993. U.S. Department of Commerce, 41 pp.

  • View in gallery

    The locations of the five Subduction Experiment buoys juxtaposed with the average Isemer and Hasse climatological sea level pressure in millibars for (a) the winter months of December–February and (b) the summer months of June–August.

  • View in gallery

    A time line for the Subduction Experiment buoys. A (+) indicates that IMET instrumentation was used and a (−) that VAWR instrumentation was used. The absence of any symbols indicates that no reliable data were available from either system.

  • View in gallery

    Seasonal means of (a) sea–air temperature, (b) sea–air specific humidity, (c) wind speed, (d) wind stress, (e) sensible heat flux, (f) latent heat flux, (g) net longwave flux, (h) net shortwave flux, and (i) net heat flux from the Isemer and Hasse (*), Oberhuber (○), ECMWF (×), NCEP (+), and buoy (—) datasets at the NE location. The 95% confidence limits for the buoy data are represented by the error bars. Note that the buoy and Isemer and Hasse wind speeds have been corrected to a height of 10 m.

  • View in gallery

    (Continued)

  • View in gallery

    Four-day running means of (a) wind stress, (b) sensible heat flux, (c) latent heat flux, (d) net longwave flux, (e) net shortwave flux, and (f) net heat flux from 1 January 1992–31 December 1992 at the NE location. The buoy data are represented by a solid line (—), while the ECMWF data are denoted by a dotted line (····).

  • View in gallery

    Time series of sea temperature from 15 October 1992 to 5 November 1992 at the NE location. The buoy measurements at a depth of 1 m are represented by a dashed line (– – –), the skin temperature estimated from the buoy data by a dotted line (····), and the ECMWF sea surface temperature analyses by a solid line (—).

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 172 34 3
PDF Downloads 13 6 1

Observations of Surface Forcing from the Subduction Experiment: A Comparison with Global Model Products and Climatological Datasets

View More View Less
  • 1 Woods Hole Oceanographic Institution, Woods Hole, Massachusetts
Full access

Abstract

Reliable estimates of the exchange of heat, moisture, and momentum across the air–sea interface are essential in assessing the local “representativeness” of the surface forcing fields depicted by global model and climatological datasets. The reliability and extended length of the in situ data collected by a large-scale array of buoys deployed during the Subduction Experiment make this dataset particularly well suited to providing such an assessment. The Subduction Experiment was designed to explore the process by which the mixed layer waters of the eastern subtropical North Atlantic are incorporated into the top of the main thermocline. To this end, an array of five buoys was maintained between 18°–33°N and 22°–34°W from June 1991 to June 1993. In situ dynamic, thermodynamic, and radiometric measurements are utilized along with a state-of-the-art bulk flux algorithm to estimate the time-dependent surface forcing at each of the buoys. The resulting air–sea fluxes are compared to similar quantities offered by the Isemer and Hasse (Bunker) and the Wright–Oberhuber Comprehensive Ocean–Atmosphere Data Set climatological datasets, and global model forecasts from the European Centre for Medium-Range Weather Forecasts and the National Centers for Environmental Prediction.

Some substantial differences are exhibited between the surface forcing components garnered from the Subduction Experiment buoys and those of the climatological and model products. The mean net heat flux from the Subduction Experiment buoys exhibits a qualitatively similar spatial gradient to that of the climatological and model products across the array, but generally reflects a greater oceanic heat gain in summer and a smaller oceanic heat loss in winter. On shorter timescales, the models’ inability to replicate the heat and radiometric fluxes of the buoys is reflected in large mean standard deviations of the differences between the buoy and model fluxes at 6-h intervals. Some of the observed differences are attributed to differences in bulk formulas and/or differences in the mean variables from which the bulk air–sea fluxes are derived, while others are simply an artifact of the spatial and temporal filtering inherent within the climatological and model products.

Corresponding author address: Dr. Kerry A. Moyer, Geoscience Dept., Indiana University of Pennsylvania, 111 Walsh Hall, Indiana, PA 15705.

Email: KMOYER@grove.iup.edu

Abstract

Reliable estimates of the exchange of heat, moisture, and momentum across the air–sea interface are essential in assessing the local “representativeness” of the surface forcing fields depicted by global model and climatological datasets. The reliability and extended length of the in situ data collected by a large-scale array of buoys deployed during the Subduction Experiment make this dataset particularly well suited to providing such an assessment. The Subduction Experiment was designed to explore the process by which the mixed layer waters of the eastern subtropical North Atlantic are incorporated into the top of the main thermocline. To this end, an array of five buoys was maintained between 18°–33°N and 22°–34°W from June 1991 to June 1993. In situ dynamic, thermodynamic, and radiometric measurements are utilized along with a state-of-the-art bulk flux algorithm to estimate the time-dependent surface forcing at each of the buoys. The resulting air–sea fluxes are compared to similar quantities offered by the Isemer and Hasse (Bunker) and the Wright–Oberhuber Comprehensive Ocean–Atmosphere Data Set climatological datasets, and global model forecasts from the European Centre for Medium-Range Weather Forecasts and the National Centers for Environmental Prediction.

Some substantial differences are exhibited between the surface forcing components garnered from the Subduction Experiment buoys and those of the climatological and model products. The mean net heat flux from the Subduction Experiment buoys exhibits a qualitatively similar spatial gradient to that of the climatological and model products across the array, but generally reflects a greater oceanic heat gain in summer and a smaller oceanic heat loss in winter. On shorter timescales, the models’ inability to replicate the heat and radiometric fluxes of the buoys is reflected in large mean standard deviations of the differences between the buoy and model fluxes at 6-h intervals. Some of the observed differences are attributed to differences in bulk formulas and/or differences in the mean variables from which the bulk air–sea fluxes are derived, while others are simply an artifact of the spatial and temporal filtering inherent within the climatological and model products.

Corresponding author address: Dr. Kerry A. Moyer, Geoscience Dept., Indiana University of Pennsylvania, 111 Walsh Hall, Indiana, PA 15705.

Email: KMOYER@grove.iup.edu

1. Introduction

Reliable estimates of the exchange of heat, moisture, and momentum between the atmosphere and ocean are of great interest to both atmospheric scientists and oceanographers alike. Such estimates are needed to verify the local “representativeness” of the surface forcing fields offered by global atmospheric model and climatological datasets. Modelers from both communities rely on these gridded atmospheric forcing fields to provide the proper boundary conditions with which to drive their respective models. Unfortunately, high-quality, long- term, in situ flux estimates are scarce over marine areas. The buoy data collected during the Subduction Experiment, however, meet all of these requirements, making this dataset ideally suited for validation purposes.

Oceanic subduction is a process by which mixed layer water is incorporated into the main thermocline (Stommel 1979; Luyten et al. 1983; Marshall et al. 1993). In an effort to more fully understand the sequence of events leading to subduction, the Subduction Experiment was undertaken in the eastern subtropical North Atlantic. Since subduction is inherently a nonlocal process, one of the primary goals of the Subduction Experiment was to directly observe large-scale horizontal gradients in atmospheric forcing. Such gradients in sensible and latent heat flux, wind stress, and net shortwave and longwave radiation are thought to play a significant role in the subduction process. To this end, dynamic, thermodynamic, and radiometric data were collected by a large- scale buoy array over a period of 2 years. The accuracy of these measurements was assured through intercomparisons and pre- and postdeployment instrument calibrations.

The harsh environment and remote location of the Subduction Experiment mandated that the local estimation of the heat, moisture, and momentum fluxes beaccomplished via the bulk method. Thus, basic in situ measurements were utilized in conjunction with bulk formulas to estimate the time-dependent surface forcing at each of the buoys. The state-of-the-art bulk flux algorithm used in this endeavor was originally developed to meet the stringent accuracy requirements of the Tropical Ocean Global Atmosphere (TOGA) Coupled Ocean–Atmosphere Response Experiment (COARE) and has been verified by several recent tropical and extratropical measurement programs (Fairall et al. 1996b).

The purpose of this investigation is to utilize direct measurements of the basic meteorological observables and bulk estimates of the heat, moisture, and momentum fluxes garnered from the Subduction Experiment buoys to locally evaluate the performance of climatological and global model representations of the surface forcing field. The ultimate motivation behind this comparative study lies in the authors’ desire to simulate the subduction process numerically using a three-dimensional ocean model driven by realistic atmospheric forcing. If the subduction process is to be simulated numerically in this manner, then the gridded surface forcing fields from either global atmospheric models or climatological datasets must be used in conjunction with the in situ estimates from the Subduction Experiment, with the latter serving to identify persistent differences in the gridded forcing fields that are not simply the product of subgrid-scale variability.

An overview of the datasets employed within this comparative study is provided in section 2. Section 3 details the subsequent processing of these datasets. The mean accuracy of the buoy instrumentation and the potential effect of instrumental bias upon the mean accuracy of the buoy fluxes are addressed in section 4. Section 5 contains a discussion of the climatology of the Subduction Experiment region. In section 6, the basic observables and bulk fluxes from the Subduction Experiment buoys are compared to similar quantities offered by the Isemer and Hasse (Bunker) and the Wright–Oberhuber COADS (Comprehensive Ocean–Atmosphere Data Set) climatological datasets, and operational products from the data assimilation and global forecast systems of the European Centre for Medium- Range Weather Forecasts (ECMWF) and the National Centers for Environmental Prediction (NCEP, formerly the National Meteorological Center). Differences in the buoy, model, and climatological surface forcing are explained in terms of temporal and spatial averaging, differences in the mean variables utilized within the bulk formulas, and differences in the bulk parameterization schemes themselves.

2. An overview of the datasets

a. The Subduction Experiment buoy dataset

The mooring component of the Subduction Experiment took place in the eastern subtropical North Atlanticbetween June 1991 and June 1993 (Brink et al. 1995). A large-scale, five-buoy array was maintained across the eastern flank of the Azores high over two annual cycles. As shown in Figs. 1a and 1b, the buoys were located at 33°N, 34°W; 33°N, 22°W; 18°N, 34°W; 18°N, 22°W; and 25.5°N, 29°W and are referred to by their relative positions (NW, NE, SW, SE, and C) within the framework of the array.

The mooring component of the Subduction Experiment was separated into three distinct 8-month settings in order to reduce the deleterious effect that a prolonged exposure to the harsh oceanic environment would have upon the instrumented buoys. Thus, at 8-month intervals, the buoys were systematically retrieved and refurbished. Despite this careful attention, several buoys did not remain on station for their entire 8-month tenure. This was especially true on the first deployment, as threeof the five moorings parted at one time or another during this initial setting. However, subsequent deployments were not as susceptible to mooring failure, and the overall scope and quality of the Subduction Experiment buoy data are exceptionally good. The specific times when each of the five buoys were on station are illustrated in Fig. 2.

The five Subduction Experiment buoys were outfitted with a full compliment of meteorological instrumentation. Many of the buoys carried two independent meteorological instrument systems: a Vector Averaging Wind Recorder (VAWR, Weller et al. 1990) and an Improved Meteorological recorder (IMET; Hosom et al. 1995). Both of these systems measured barometric pressure, wind speed and direction, air temperature, sea temperature, relative humidity, and incoming shortwave and longwave radiation. The IMET measured precipitation as well. The IMET recorded data once every 1 min, while the VAWR recorded data once every 15 min. While all of the IMET observables represent 1-min averages, the averaging intervals of the VAWR observables were variable dependent. Unlike the wind and radiation measurements, which represent 15-min averages, the remaining VAWR observables were averaged over a subset of the recording interval. For example, sea surface temperature was averaged over the initial 7.5 min, while air temperature was averaged over the final 7.5 min. Barometric pressure and relative humidity were sampled for 2.5 s and 3.5 s, respectively, midway through the 15-min period. During postprocessing, however, all of the IMET data were subsequently averaged over 15 min to match the recording rate of the VAWR.

The two instrument systems were mounted on the deck of a 2-m-high white aluminum tower, which in turn was secured to the upper face of either a 3-m- diameter discus or a 2.3-m-diameter toroid buoy. The former buoys were provided by the Woods Hole Oceanographic Institution (WHOI), while the latter were provided by Scripps Institution of Oceanography (SIO). A vane was attached to one side of the tower in order to maintain the buoy’s orientation relative to the wind. Special care was taken to ensure that the meteorological instrumentation was configured in an optimal manner. For example, the radiometers, which require an unobstructed hemispheric view of the sky, occupied the uppermost position on the downwind side of the buoy, while the temperature, humidity, and wind instrumentation were mounted at slightly lower levels on the buoy’s upwind side. The upwind positioning of these latter sensors was designed to reduce any inadvertent temperature modification or flow distortion associated with the buoy’s superstructure. The exact height at which the sensors were mounted was a function of instrument system and buoy type, but ranged between 2.4 m and 3.5 m above a nominal water line.

Wind speed on the VAWR was measured by an R. M. Young model 12170C three-cup anemometer, while the vector components of the wind were measured witha WHOI—EG&G vane, vane follower, and compass system. IMET winds were measured with an R. M. Young model 5103 propeller and vane unit. This unit uses a propeller to measure wind speed, and a vane and a KVH Industries model C100 flux gate compass to determine wind direction.

The VAWR used thermistors to measure air and sea temperatures, while the IMET employed platinum resistance thermometers to make these measurements. The air temperature sensors were mounted inside multiplate radiation shields (Gill 1983). These plates were designed to allow air to flow freely past the sensors, yet protect the sensors from direct and diffuse solar radiation. Below the surface, the sea temperature sensors were positioned within a pressure-protected enclosure and affixed to the buoy bridle at a depth of 1 m.

Barometric pressure was measured by the VAWR system using a Paroscientific model 215 quartz crystal transducer. The IMET system measured barometric pressure with an Atmospheric Instrumentation Research model DB-1A sensor. A Gill static pressure port was employed to reduce errors caused by winds blowing over the exposed sensor port. Relative humidity was measured by a Vaisala humicap element on the VAWR and a Rotronic MP-100F relative humidity sensor on the IMET system. The humidity sensors were protected from solar heating by multiplate radiation shields, and from salt spray and precipitation by Gore-Tex and Teflon shields.

Measurements of incoming shortwave radiation from the VAWR and IMET systems were made with an Eppley model 8-48 pyranometer and an Eppley Precision Spectral Pyranometer, respectively. The voltage output from the former instrument is proportional to the temperature difference between adjacent black and white surfaces, while that of the latter is proportional to the temperature difference between a surface exposed to the solar radiation and a thermistor enclosed within its supporting case. Both the VAWR and IMET systems used Eppley precision infrared (PIR) pyrgeometers to measure incoming longwave radiation. The dome of the pyrgeometer allows for the transmission of longwave radiation, but is opaque to shortwave radiation. The measurements of incoming longwave radiation are then proportional to the temperature difference between a receiving surface beneath the dome and its supporting case, which serves as a reference mass.

The IMET system measured rainfall using an R. M. Young model 50201 self-siphoning rain gauge. This instrument relates measurable changes in the capacitance of its collection chamber to the volume of water contained within its depth, hence providing a cumulative record of rainfall or, more precisely, rainfall minus evaporation. The instrument is accurate to within 2 mm and self-siphons at a depth of 50 mm.

In addition to the meteorological instrumentation, the five moorings were also outfitted with subsurface instrumentation. The only subsurface instrumentation relevant to this investigation, however, are the 1-m sea temperature sensors discussed above and the Vector Measuring Current Meters (VMCMs) deployed at a depth of 10 m below the surface. VMCMs use two propeller sensors and a compass to measure the east and north components of horizontal velocity (Weller and Davis 1980).

b. The climatological products

The climatological data used in this investigation are from the Bunker Climate Atlas of the North Atlantic Ocean (Isemer and Hasse 1985, 1987) and An Atlas Based on the “COADS” Data Set (Wright 1988; Oberhuber 1988). These frequently referenced climatological databases contain long-term monthly and annual averages of basic observables and air–sea fluxes. The Bunker climatological data were derived from observations collected by ships of the Voluntary Observing Fleet (VOF) during the period 1941–72 (Bunker 1976). In recent years, Isemer and Hasse (1985, 1987) have modified the original Bunker dataset by interpolating the data onto a 1° grid and correcting some of the known biases in both the basic observables and the parameterizations used to compute the air–sea fluxes.

The second climatological dataset employed in this investigation is derived from COADS and, thus, is also a compilation of ship weather observations. Specific details concerning COADS can be found in Woodruff et al. (1988). Wright (1988) constructed the mean fields of the basic observables, while Oberhuber (1988) derived the complementary flux fields. This dataset encompasses the period from 1950 to 1979 and is available on a 2° grid. The Oberhuber fluxes were computed using the “classical method,” whereby the mean flux fields are computed from the mean fields of the basic observables. In contrast, the Isemer and Hasse fluxes were computed using the “sampling method,” whereby fluxes are computed from each individual set of observations and subsequently averaged over time.

c. The global assimilation products

The global assimilation products used in this investigation are from the global operational numerical weather prediction analysis–forecast systems of the ECMWF (ECMWF Technical Attachment 1994) and the NCEP (Kanamitsu, 1989; Kanamitsu et al. 1991). The Subduction Experiment buoy data are compared to ECMWF products generated by their operational T106/L19 model prior to 17 September 1991 and their operational T213/L31 model at all times thereafter. This latter model has twice the horizontal resolution as its predecessor. The Subduction Experiment data are also compared to products generated by the operational version of the T126/L28 NCEP model. Unfortunately, a total of 44 days of data are unavailable from the NCEP archive over the 2-year period. These gaps in the NCEParchive occur sporadically throughout the 2 years and are typically only 6–12 hours in duration, although a few persisted for several days.

The model fluxes are archived as accumulated values over an initial 6-h forecast from which 6-h averages were subsequently computed. The basic observables utilized within this investigation are available at 0000, 6000, 1200, and 1800 UTC, and represent uninitialized analyses produced by the data assimilation systems of each center. IMET data from the Subduction Experiment buoys were accessible in real time by these data assimilation systems via the Global Telecommunications System. Unfortunately all of the buoys were not equipped with an IMET system, and the extent to which this IMET data were utilized when it was available is unknown.

3. Processing of the datasets

Redundant measurements from the VAWR and IMET systems often allowed for deficiencies in either system to be readily exposed in the field. In addition, shipboard meteorological observations were collected by hand- held and bridge-mounted sensors both prior to the retrieval of the buoys and immediately after their redeployment. These periods of intensive meteorological observations typically lasted for several hours and were utilized as yet another field check on the accuracy of the buoy data. More subtle deficiencies in the data were brought to light and corrected during postdeployment calibrations of the instrumentation.

Although the Subduction Experiment buoys were typically outfitted with both a VAWR and an IMET system, there were several occasions when a VAWR was singly deployed. Given that the VAWR was used more frequently and the quality of its data was comparable to that of the IMET system, the VAWR was selected as the primary supplier of meteorological data from the Subduction Experiment buoys. However, data from the IMET system were utilized on those occasions when the VAWR data were either unavailable or deemed to be unreliable. As shown in Fig. 2, IMET supplied all of the basic observables on the northeast (18 June 1991–14 February 1992) and northwest (24 February 1992–16 October 1992) buoys, relative humidity (9 February 1992–12 September 1992) and barometric pressure (10 November 1992–19 June 1993) on the southeast buoy, barometric pressure (23 June 1991–11 February 1992) and incoming longwave radiation (23 September 1991–11 February 1992) on the central buoy, and incoming longwave radiation (24 August 1991–2 November 1991) on the southwest buoy.

Also indicated in Fig. 2 are those occasions when the moorings experienced a structural failure and, therefore, the buoys were off station for an extended period of time. There were several additional occasions when the buoys were on station, but neither the VAWR nor the IMET systems provided an accurate measure of a specific variable. Such instances relate to relative humidityon the southeast (12 September 1992–6 October 1992) and central (28 May 1992–14 October 1992) buoys, barometric pressure on the southwest buoy (28 March 1992–3 June 1992), winds on the southeast (29 June 1991–9 October 1991) and northwest (3 July 1991–3 August 1991) buoys, and incoming longwave radiation on the southeast (26 August 1991–9 October 1991) and northwest (24 February 1992–16 October 1992) buoys.

The incoming longwave radiation measured by a stock Eppley model PIR pyrgeometer contains an additional output equivalent to 3.6% of incoming solar radiation (Alados-Arboledas et al. 1988). For some time now, investigators have attributed the inflated measurements to a heating of the pyrgeometer’s silicon dome (Albrecht and Cox 1977). It has been suggested that some of this heating may be caused by the inadvertent transmission of shortwave radiation through the dome (Dickey et al. 1994). Further investigation has revealed that the predominant cause of dome heating is a previously unaccounted for emittance from a cover resident beneath the dome (S. Anderson 1995, personal communication). This removable cover shields the upper face of the pyrgeometer’s case. The cover on the VAWR pyrgeometer is constructed of stainless steel, while the IMET cover is aluminum. The larger emissivity of the stainless steel significantly enhances the difference in temperature between the dome and the case, and is now thought to be the primary source of error, although some residual heating effects are still evident when the aluminum case is used as well. Thus, the VAWR incoming longwave radiation was reduced by the empirically determined value of 3.6% of incoming shortwave radiation, while the IMET longwave radiation was left unaltered. This correction of less than 7 W m−2 over an annual cycle, but up to 40 W m−2 instantaneously in summer under full sun, significantly reduced, but did not completely eliminate, the daytime enhancement of incoming longwave radiation caused by solar heating and produced a much better agreement between the VAWR and IMET pyrgeometers.

Although incoming shortwave and longwave radiation were measured by radiometers aboard the buoys, their outgoing components were not measured and had to be estimated. The surface albedo formulation of Payne (1972), which expresses the solar albedo as a function of both solar altitude and atmospheric transmittance, was employed in the calculation of outgoing shortwave radiation. Outgoing longwave radiation, on the other hand, consists of both the longwave radiation emitted from the surface and the reflected portion of the incoming radiation and was computed using the Stefan–Boltmann law and an infrared emissivity of 0.97.

The nonradiometric fluxes were estimated using 15-min averages of the basic observables and a bulk flux algorithm developed in conjunction with TOGA COARE (Fairall et al. 1996b). The bulk transfer coefficients used in the algorithm vary with both wind speed and stability and are based on the transfer coefficientsemployed within the Liu, Katsaros, and Businger (LKB) model, with some modifications based on observations from recent measurement programs (Liu et al. 1979). The neutral transfer coefficients for momentum, heat, and moisture are functions of their respective roughness lengths. The LKB model parameterizes the more difficult-to-measure scalar roughness lengths in terms of a roughness Reynolds number. This roughness Reynolds number is in turn a function of the velocity roughness length, which is simply expressed as the sum of the Charnock relation (Charnock 1955) and a smooth flow limit (Smith 1988).

In their strictest sense, the bulk formulas require knowledge of the wind speed relative to the sea surface current. Since the current at the interface was not known, the current at 10 m was utilized to approximate the surface current. On those occasions when the 10-m current was unavailable, the surface current was arbitrarily set to zero. This action has a negligible effect on the long-term mean heat and momentum fluxes, as the mean wind speed is roughly two orders of magnitude larger than the mean sea surface velocity.

Cool-skin and warm-layer corrections were also incorporated into the algorithm, as sea temperatures on the buoys were measured at a depth of 1 m (Fairall et al. 1996a). Since it is not the sea temperature at depth, but rather the interfacial temperature, that is required in the computation of the fluxes of sensible and latent heat and outgoing longwave radiation, the skin temperature was estimated from the measured temperature. The cool- skin correction accounts for the fact that the sensible, latent, and longwave fluxes are actually realized at the air–sea interface. This effect is relatively persistent and effectively lowers the skin temperature by an average of 0.2°C. The warm-layer correction, on the other hand, accounts for the diurnal temperature variations that can occur as a result of the absorption of solar radiation within the upper few meters of the ocean. The precise profile shape and magnitude of this near-surface warming are functions of the optical properties of the water and the extent to which winds are acting to diffuse this heating through mixing. However, in light winds, the temperature of the water above the sensor can warm several degrees during the course of the day, while the water temperature at a depth of 1 m remains relatively unchanged.

Light wind regimes, in general, represent a unique challenge in terms of accurately estimating the bulk air–sea exchanges, for it is not the magnitude of the mean wind vector, but rather the average wind speed, that should be used in the computation of the transfer coefficients. Godfrey and Beljaars (1991) have addressed this issue by augmenting the measured wind speed with a “gustiness velocity,” which they relate to the convective velocity scale. The inclusion of this gustiness velocity within the COARE algorithm accounts for the enhanced turbulent exchange generated by the passage of convective eddies. Such an enhancement in the degree of air–sea coupling during periods of light winds has been shown to produce more realistic simulations of atmospheric phenomena within the ECMWF global model (Miller et al. 1992).

With the air–sea fluxes from the buoys in hand, time series of the global assimilation products were extracted from the grid points nearest to each of the buoys. The nearest grid points were in all cases within 75 km of the Subduction Experiment buoys. Monthly and annual means were extracted from the nearest climatological grid points as well, with the additional caveat that if the buoy lay midway between two grid points, then adjacent grid points were averaged to arrive at a more representative value.

The 2-m height at which temperature and humidity data are available from the global analyses compares favorably with the 2.4–2.8-m height at which these variables were measured at the buoys, and thus, no height corrections were applied. Specific humidity was derived from the relative humidity measured by the buoys and the dewpoint temperature available from the ECMWF analyses using a variation of Teten’s formula (Bolton 1980). In addition, a stability-dependent height correction was applied to the buoy winds in order to adjust these 3.0–3.5-m winds to the 10-m height of the ECMWF and NCEP wind analyses. The precise measurement heights for temperature, humidity, and winds within the climatological datasets are uncertain. However, shipboard temperature and humidity were typically measured at a height of ∼10 m, while winds, on those rare occasions when they were measured and not estimated, were measured at a height of ∼20 m (Cardone et al. 1990).

Monthly and seasonal means of the basic observables and the air–sea fluxes were then computed at the Subduction Experiment buoy locations. Monthly means were computed by averaging the buoy and assimilation products over 30.5-day intervals beginning on 1 July 1991. Seasonal means representing June–August (July–August in 1991), September–November, December–February, and March–May were subsequently computed directly from these monthly means.

In an effort to quantify the ability of the assimilation products to replicate the basic meteorological observables from the buoys on timescales as short as 6 h, those 15-min block averages from the buoys that spanned the 0000, 6000, 1200, and 1800 UTC analyses times wereextracted and compared to their counterparts from ECMWF and NCEP. A similar comparison was undertaken with the heat, momentum, and radiometric fluxes as well. In the case of the fluxes, however, 24 15-min block averages from the buoys were averaged over 6-h periods that were coincident in time with the 6-h average fluxes from the models. Mean differences between the buoy and assimilation products were then computed whenever the buoys were on station and recording reliable data. The standard deviation of these differences were also calculated, providing insight into the spread of the observed differences about any existing biases. Lastly, correlation coefficients between the buoy data and the assimilation products were computed in order to ascertain the degree to which the assimilation products were capturing the temporal trends exhibited by the buoy data.

Strictly speaking, the model analyses represent basic observables collected from a window of ±3 h around each of the analyses times. However, comparisons of 6-h averages of the buoy data centered around the analyses times were only slightly different from those made using a single 15-min average coincident with the analyses times. Specifically, the mean differences and correlation coefficients of sea temperature, air temperature, and relative humidity were virtually identical, while the mean difference and correlation coefficient of the winds differed by only ∼0.1 m s−1 and ∼0.02, respectively.

4. Accuracy of the buoy data

The accuracy of the surface forcing components garnered from the buoys depends largely on the accuracy of the basic observables from which they are derived. The long-term accuracy of the basic observables and the estimated accuracy of the mean air–sea fluxes are provided in Table 1.

The cup anemometer employed by the VAWR is likely to overspeed at times, resulting in a mean wind speed that is up to 6% too high, while the propeller used by the IMET system is expected to underpredict the mean winds by up to 3% (Weller et al. 1990). Although both the VAWR and IMET systems were relied upon to supply wind data at various times throughout the course of the experiment, the former system was utilized far more frequently. Thus, one might anticipate that the mean wind speeds measured during the Subduction Experiment would be a bit too high on average. A 6% uncertainty in wind speed results in potential biases of 0.3–0.5 m s−1 and 0.007–0.010 N m−2 in wind speed and wind stress, respectively, at the typical wind speeds observed across the Subduction Experiment array.

The VAWR and IMET air temperature sensors are susceptible to solar heating errors, particularly during periods of low winds and high solar radiation. The utilization of Gill radiation shields significantly reduced, but did not entirely eliminate, these radiative effects. Differences in daytime versus nighttime air temperature were computed for the northeast and central buoys for the 2 years of the Subduction Experiment. The daytime measurements from these two buoys were 0.4°C greater on average than those collected at night over the duration of the experiment. However, a diurnal variation of 0.2°–0.3°C was also observed in sea skin temperature, from which it can be inferred that a significant fraction of this observed 0.4°C difference in air temperature may represent a true warming of the lower boundary layer. Thus, long-term mean air temperatures are thought to be accurate to 0.2°C.

Most of the uncertainty concerning the sensible heat flux revolves around this potential error of 0.2°C in air temperature. Since the mean sea–air temperature observed across the buoys is only 0.5°–0.9°C, this uncertainty in air temperature can potentially alter the sensible heat flux by 2.0–2.5 W m−2. A 6% error in wind speed potentially alters the sensible heat flux by an additional 0.3–0.5 W m−2. It is anticipated, however, that the error associated with the radiative warming of the temperature sensor is counteracted to some degree by the error associated with the overspeeding problem, thus reducing the true error in the sensible heat flux term below what might have been expected from solar heating alone.

The relative humidities measured by the buoys are accurate to 3%. A 3% uncertainty in relative humidity produces a potential error of 12–17 W m−2 in the mean latent heat flux term plus an additional 6% or 5–7 W m−2 arising from the uncertainty in wind speed. Like the air temperature sensors, the relative humidity sensors were also subject to solar heating errors. However, the mean daytime latent heat fluxes estimated at the northeast and central buoys differed from their nighttime counterparts by only 1–2 W m−2 over the 2-year period.

The pyranometers employed by the VAWR and IMET systems are accurate to within 3%. This represents a potential error of 5–7 W m−2 annually in the incoming shortwave radiation measured across the array. A somewhat larger error is associated with the VAWR and IMET pyrgeometers. The pyrgeometers are accurate to within 5% or 15–20 W m−2. As previously mentioned, the incoming longwave radiation measured by the VAWR system was corrected to account for daytime radiative heating effects. The success of this correction is exemplified by the negligible difference in daytime and nighttime incoming longwave measurements exhibited by the northeast and central buoys over the duration of the experiment.

If all of these potential biases in the individual surface forcing components were to sum constructively, then the mean net heat flux could potentially be in error by up to 50 W m−2, solely as a result of instrumental biases. Fortunately, the buoy instrumentation appear to have functioned well within the limits cited here. In addition, some of the biases that may be present inherently act so as to partially counteract each other. This was observed to be the case during TOGA COARE, where a similarly outfitted buoy and bulk flux algorithm produced mean monthly net heat fluxes that were within 10 W m−2 of shipboard measurements (R. Weller and S. Anderson 1997, manuscript submitted to J. Climate). Although the opportunities for intercomparison were more limited during the Subduction Experiment, the results from COARE are encouraging.

5. Climatology of the Subduction Experiment region

The dominant atmospheric feature within the Subduction Experiment region is the Azores high. Much of the spatial and temporal variability observed across the Subduction Experiment array is directly linked to the strength and position of this subtropical high pressure system. Climatologically speaking, the Azores high gradually evolves from a relatively weak feature possessing a broad zonal 1020-mb center located near 28°N in January to a more virulent 1025-mb cellular entity whose center is resident near 35°N and 35°W by July (Isemer and Hasse 1985).

A clockwise atmospheric circulation around the Azores high dictates that the southern portion of the array remain under the influence of northeasterly trades for most of the year. Although these trades are extremely persistent, they reach their peak intensity during the summer months in conjunction with the intensification of the high and its associated pressure gradient. Thus, the southern buoys experience fair weather on most days throughout the year, interrupted only by an occasional wave imbedded within the easterly trades during the summer or the infrequent passage of an extratropical front in winter. Although the northern portion of the array is also within the northeasterly trade regime during the warm season, the northern buoys periodically fall under the influence of the prevailing westerlies during the winter season as the center of the Azores high moves equatorward. During these winter months, the northern portion of the array experiences frequent frontal passages associated with extratropical cyclones that commonly pass by to the north. Such frontal passages are characterized by clockwise changes in wind direction and brief precipitation events. Occasionally, however, one of these eastward propagating cyclones will follow a more southerly track, allowing the center of the cyclone to pass directly through the array and causingsustained high winds and precipitation within its radius of influence.

Seasonal anomalies for the Subduction Experiment period have recently become available (Young-Molling et al. 1995). These anomalies are derived from climatological values compiled by da Silva et al. (1994). The position and strength of the Azores high during the 2 years of the Subduction Experiment agree quite well with climatology during the summer of both years, but the high tended to be several millibars stronger, and its zonal axis was often situated north of the northernmost buoy locations during the winter months. The northward displacement of the high resulted in an easterly wind anomaly across the northern portion of the array during winter and spring of the first year and again during the fall and winter of the second year. Mean wind speeds were up to 1 m s−1 greater than normal during the spring and fall of 1992 over the eastern portion of the array. Sea and air temperatures were in excess of 0.5°C warmer than normal across the northern and central portions of the array during the summer and fall of 1991, while specific humidities were up to 0.5 g kg−1 higher than normal. Fractional cloud cover anomalies did not exceed 0.1 in any season.

6. Data comparisons

Some substantial differences are exhibited between the surface forcing components garnered from the Subduction Experiment buoys and those offered by the climatological and model products. Many of the long-term mean differences are of a similar sign throughout the array and differ only in a quantitative sense. Thus, in the interest of brevity, the authors provide data in both tabular and graphical form for the NE location only, as this buoy was essentially on station collecting reliable data continuously for 2 years. Two-year means of the buoy data, the global assimilation products, and the climatological products at the NE location are provided in Table 2. Similarly, seasonal means are presented in Fig. 3. Time series of the surface forcing components from the buoy and the ECMWF model are illustrated in Fig. 4, where 6-h averages of the air–sea fluxes have been filtered using a 4-day running mean in order to enhancereadability. The reader should be advised, however, that the statistics cited within the forthcoming text are by no means limited to the NE location, but rather reflect the range of values computed across all five buoys whenever the buoys were on station and recording reliable data. For the reader’s convenience, the ranges of these statistics are summarized in Table 3.

a. Wind forcing

More often than not, the shipboard wind speeds utilized within the climatological datasets were not directly measured, but rather were estimated from the sea state using Beaufort scale equivalents. In fact, Quayle (1980) found that the wind reports from VOF ships in the North Atlantic represented actual measurements less than 15% of the time between 1950 and 1971. Isemer and Hasse utilize the revised Beaufort equivalent scale of Kaufeld (1981) in the computation of their wind speeds. The Kaufeld scale, however, systematically overestimates wind speed (da Silva et al. 1994). Wright (1988), on the other hand, employs the original WMO code 1100 Beaufort scale, which underestimates wind speeds for Beaufort numbers less than 6 and overestimates wind speeds for Beaufort numbers greater than 6 (da Silva et al. 1994). At the 6–9 m s−1 wind speeds typically encountered across the Subduction Experiment array, Kaufeld’s scale produces wind speeds that are 1.4–1.8 m s−1 larger than those of the original Beaufort scale. Not surprisingly then, the Isemer and Hasse wind speeds, corrected to a height of 10 m, are still 1.3–1.9 m s−1 larger than those of Wright across the array.

The 3.0–3.5-m buoy winds, corrected to a height of 10 m, are consistently smaller than the 10-m winds of Isemer and Hasse at all of the buoys and during every month that the buoys remained on station. Specifically, the monthly differences between the buoy winds and those of Isemer and Hasse often exceed 1.0 m s−1. In reality, the tendency of the buoy anemometers to overspeed make the differences potentially even larger. In contrast, the monthly buoy winds were usually, but not always, larger than those of Wright, with differences generally being less than 0.5 m s−1.

Mean differences between the ECMWF and NCEPwind analyses are less than 0.2 m s−1 across the array. The buoy winds, corrected to a height of 10 m, are only slightly stronger than the 10-m winds from the ECMWF and NCEP analyses on average, with differences of up to 0.8 m s−1 noted across the southern buoys. Mean standard deviations of these differences are 1.3–2.0 m s−1, with correlation coefficients of 0.71–0.83. These relatively large standard deviations are not unexpected in light of the large high-frequency variability associated with wind speed measurements.

Mean wind stress from the Isemer and Hasse climatology ranges from 0.05 to 0.13 N m−2 across the Subduction Experiment array. Despite the fact that the Isemer and Hasse wind speeds are consistently greater than the buoy wind speeds throughout the array, the same is not true of wind stress. Although the monthly wind stress offered by Isemer and Hasse generally exceeds that of the southern buoys by ∼0.03–0.05 N m−2, the stress from the northern buoys actually exceeds that of Isemer and Hasse by ∼0.01–0.03 N m−2. This dichotomy occurs because the Isemer and Hasse wind stress was computed from its monthly average vector components. The computation of the climatological wind stress in this manner, coupled with a poleward decrease in the directional persistence of the winds across the array, results in an underestimation of climatological wind stress across the northern buoys.

The mean wind stresses forecast by the ECMWF and NCEP models are extremely similar to one another, differing by less than 0.006 N m−2. Similarly, the mean buoy wind stress differs from the model forecasts by less than 0.005 N m−2 (8%) across the northern portion of the array. However, the mean wind stress estimated from the southern buoys is 0.012–0.020 N m−2 (17%–22%) larger than the stress depicted by the models. These larger differences are primarily the result of the greater wind speeds exhibited by the southern buoys relative to their model counterparts. Mean standard deviations of the differences between the buoy and model wind stress are 0.04–0.05 N m−2, with correlation coefficients of 0.74–0.89.

b. Sensible and latent heat fluxes

The air–sea exchange of sensible heat is a function of both wind speed and the thermal gradient across the interface. Climatologically speaking, sea–air temperatures are typically small and positive across the array. The thermal gradient achieves its peak amplitude of 1.0°–1.5°C across the northern buoys during the winter months in response to increasingly frequent intrusions of cooler air from the north and approaches zero during the summer months, particularly over the southeastern portion of the array, in response to the intensification and westward propagation of coastal upwelling. Although the precise reference levels for the climatological sea and air temperatures are uncertain, the monthly sea–air temperatures measured at the buoys are slightly larger than those depicted by either climatology for the better part of the year.

The mean interfacial thermal gradients from the ECMWF and the buoys are quite similar, differing by less than 0.15°C. However, sea–air temperatures from the NCEP analyses are an average of 1.1°–1.8°C greater than that from the buoys. Since both the ECMWF and NCEP use identical sea surface temperature analyses and these analyses differ from the mean skin temperatures of the buoys by less than 0.2°C, the source of this large discrepancy must lie in the air temperature analyses of the NCEP. Indeed, further investigation reveals a 1.0°–1.8°C cool bias in the 2-m NCEP air temperature relative to the buoy observations. In contrast, the 2-m ECMWF air temperature differs from the mean buoy air temperatures by less than 0.2°C.

On timescales of 6 h or less, both global assimilation products have a great deal of difficulty replicating the observed interfacial temperature gradient. This difficulty is reflected in the large mean standard deviations of the differences between the buoy and model sea–air temperatures of 0.63°–1.0°C and low correlation coefficients of 0.34–0.64. Mean standard deviations of the differences between the buoy air temperatures and those offered by the ECMWF and NCEP range between 0.78° and 1.10°C, with correlation coefficients of 0.84–0.96. Mean standard deviations of the differences between the sea surface temperatures of the buoy and the ECMWF and NCEP are not much better at 0.48°–0.75°C, with correlation coefficients of 0.92–0.97. Although the correlation coefficients associated with the air and sea temperatures are fairly high, this is largely due to the success of the assimilation products in capturing the seasonal variations in these variables, rather than their variability at shorter timescales.

The inability of the global assimilation products to replicate the sea surface temperatures at the buoys on shorter timescales is not unexpected given the temporal limitations imposed upon the sea surface temperature analyses (Reynolds 1991). These analyses are produced daily at the NCEP, and yet the sea surface temperatures depicted in these analyses are the product of a 7-day running mean. The sea surface temperature analyses not only lags the buoy measurements, but also inherently fails to capture the diurnal changes in sea surface temperature as well. The inability of the analyses to capture day-to-day variations in sea surface temperature is illustrated in Fig. 5, which tracks the sea surface temperature from the buoy and the analyses during a 3-week period in the fall of 1992. The differences between the sea temperature measured at a depth of 1 m, the skin temperature estimated from the 1-m measurement, and the sea surface temperature analyses during this 3-week period are typical of the differences observed throughout the 2-yr record at the buoys. A persistent ∼4-day lag in the sea surface temperature analyses is a direct product of using a 7-day running mean, while the spike in the measured sea temperature and skin temperature on 28 October represents the warming that occurred in the upper few meters of the ocean that day in response to strong insolation and very light winds.

In response to seasonal variations in wind speed and sea–air temperature, the sensible heat flux across the Subduction Experiment array exhibits a distinct annual cycle, with greater losses of oceanic sensible heat during the winter months and smaller losses during the summer. Annual losses offered by Isemer and Hasse range between 9 and 14 W m−2 across the array, while the Oberhuber losses are somewhat smaller at 3–6 W m−2. The monthly bulk estimates of sensible heat loss from the buoys are typically less than those of Isemer and Hasse, but greater than those of Oberhuber. The largest discrepancies occur during the winter months, when the Isemer and Hasse loss often exceeds the observed loss by more than 10 W m−2. The larger wind speeds of Isemer and Hasse certainly promote larger sensible heat fluxes. However, it should be noted that Isemer and Hasse also augmented their sea–air temperature by 0.07°C and their temperature transfer coefficient by 4.4% prior to the calculation of their sensible heat fluxes. These increases were guided by the results of inverse modeling studies (Isemer and Hasse 1987).

The mean sensible heat fluxes forecast by the ECMWF and NCEP models are also greater than those estimated at the buoys. Specifically, the mean sensible heat loss of the ECMWF model exceeds that of the buoy estimates by 2.4–6.1 W m−2 (26%–127%). Substantially greater differences of 5.0–13.9 W m−2 (70%–280%) were observed when the NCEP sensible heat fluxes were compared to those of the buoys. The larger sensible heat losses of the NCEP model are not surprising in light of the roughly two to three times greater sea–air temperatures found within the NCEP analyses. At shorter timescales, the inability of either model to replicate the buoys’ thermal gradient across the interface contributed to substantial mean standard deviations of the differences between the buoy and model sensible heat fluxes of 8.0–11.3 W m−2, with correlation coefficients of 0.40–0.80.

The latent heat flux is largely governed by wind speed and potential evaporation from the sea surface. A measure of this potential evaporation is the difference between the saturation specific humidity at the sea surfaceand the specific humidity at some reference level above the air–sea interface. The climatological interfacial humidity gradient reaches a maximum of 5 g kg−1 in the early fall across the northern buoys and somewhat later in the calendar year across the southern buoys. The observed interfacial humidity gradient generally agrees with the climatological products to within 1 g kg−1. The observations and the ECMWF and NCEP analyses also agree quite well with respect to potential evaporation, at least on average, as mean differences are less than 0.4 g kg−1. However, as was the case with the thermal gradient, the global assimilation products do not replicate the observed interfacial moisture gradient very well at shorter timescales. Mean standard deviations between the buoy and assimilation products’ depictions of sea–air specific humidity are 0.9–1.3 g kg−1, with correlations coefficients of 0.56–0.77. The sizable magnitude of these deviations is attributable to the inability of the assimilation products to capture the short-term variability in specific humidity, both at the interface and within the lower levels of the atmosphere. The latter is reflected by mean standard deviations in the differences between the buoy and the ECMWF and NCEP 2-m specific humidities of 1.0–1.3 g kg−1. However, the spatial and temporal limitations imposed upon the global sea surface temperature analyses play a role here as well, on account of the proportionality of the interfacial specific humidity to the sea surface temperature.

Like the sensible heat flux, the latent heat flux across the array is also larger in winter than in summer. Isemer and Hasse portray oceanic latent heat losses of 129–168 W m−2 annually across the buoys, while the losses depicted by Oberhuber range between 97 and 126 W m−2. The monthly latent heat losses at all five buoys are consistently smaller than those of Isemer and Hasse during every month that the buoys remained on station. The largest deficits are again observed during the winter months, when the latent losses from the buoys are consistently 25–75 W m−2 lower than those of Isemer and Hasse. The buoy estimates are in much better agreement with those of Oberhuber, differing by less than 25 W m−2 during each month. Once again, the larger wind speeds of Isemer and Hasse promote larger latent heat losses. However, Isemer and Hasse also increased their sea–air dewpoint temperature by 0.07°C and their moisture transfer coefficient by 5.7% in response to inverse modeling studies. These increases served to promote larger moisture fluxes in their dataset as well.

The models’ mean latent heat losses are in fairly good agreement with one another. However, unlike sensible heat flux, where the NCEP model predicted greater losses than those of the ECMWF, the latent heat losses forecast by the ECMWF model generally exceed those of NCEP. Both models, however, forecast latent heat losses that are greater on average than the losses estimated at the buoys. Specifically, the mean latent heat losses from the models were up to 19 W m−2 (20%) larger than those from the buoys. Mean standard deviations of these differences were 33–47 W m−2, with mean correlation coefficients between the buoy and model data of 0.69–0.84. These large standard deviations are, at least in part, a reflection of the inability of the assimilation products to replicate the short-term interfacial moisture gradient observed at the buoys.

c. Radiometric fluxes

Intrannual variations in net shortwave radiation across the array are predominantly a reflection of seasonal changes in the sun’s zenith angle. However, seasonal variations in cloud cover brought on by the stabilizing effects of West African coastal upwelling and the migration of the subsidence field associated with the Azores high are also reflected in the time series of net shortwave radiation. Annual variations in net shortwave radiation typically range in magnitude from 175 W m−2 across the northern buoys to 100 W m−2 across their southern counterparts. The monthly net shortwave radiation observed at the buoys generally exceeds that depicted by both the Isemer and Hasse and the Oberhuber datasets, except at the southeast location, where the climatological net shortwave radiation is typically larger or more in line with what was observed. The mean annual net shortwave radiation offered by Isemer and Hasse is ∼20 W m−2 greater than that depicted by the Oberhuber dataset across the northern buoys and ∼30 W m−2 greater across the southern buoys. The largest discrepancies between the two climatologies occur during the summer months, when monthly differences of 50 W m−2 are not uncommon. The largest differences between the observations and the Oberhuber dataset also occur during the summer with monthly means often differing in excess of 50 W m−2.

Departures in the observations from climatology are not that surprising, if for no other reason than that climatological shortwave radiation is computed from estimates of clear-sky shortwave radiation and cloud cover. Both climatologies utilize the cloud cover correction of Reed (1977) in order to compute the incident solar radiation at the surface. Subsequent results of inverse modeling studies prompted Isemer and Hasse to reduce their clear-sky radiation estimates by 1.4% and augment Reed’s cloud cover coefficient by 0.017. Similar studies prompted Oberhuber to reduce his incident solar radiation by 10%. This 10% reduction in incident shortwave radiation causes the net shortwave radiation of Oberhuber to be consistently less than that of the buoys and Isemer and Hasse, with the greatest discrepancies occurring during the summer months, when the shortwave radiation reaches its peak values.

The fact that the climatological shortwave values either exceed or better approximate the buoy measurements at the southeast location may be attributed to several factors. Both climatologies correct their clear- sky shortwave values by a cloudiness factor given by Reed (1977). This factor overestimates net shortwaveradiation at very low cloud values (Gilman and Garrett 1994). The magnitude of this error is a function of the number of cloud-free or nearly cloud-free days encountered at a given location over time. The SE location is climatologically prone to have a greater number of these cloud-free days than the other array locations on account of the stabilizing effect of coastal upwelling. Thus, one would expect that the climatologies might overestimate the net shortwave radiation at the SE location. The SE location is also downwind of the Saharan region and closer to this source of mineral aerosols than the other buoys. The attenuation of incoming solar radiation by these aerosols can have a significant effect on the shortwave budget (Gilman and Garrett 1994) and is not accounted for by the climatologies. Traces of red sand were found on the instrumentation aboard the southeast buoy during its retrieval. The deposition of the Saharan particles upon the radiometers might also reduce the insolation values measured at the southeast buoy.

The ECMWF model employs a constant sea surface albedo of 0.07, while the authors and the NCEP model utilize a surface albedo that is a function of solar angle (Payne 1972). The use of a solar angle–dependent surface albedo rather than a constant albedo of 0.07 results in slightly less reflection in summer and slightly greater reflection in winter, with differences of less than 2 W m−2 in the mean annual net shortwave flux. The mean difference in net shortwave radiation between the buoy data and the ECMWF model is not of a consistent sign, but is always less than 9 W m−2 (4%) across the array. The NCEP model also differs by less than 7% at three of the five buoys; however, it depicts mean net shortwave fluxes that are 27 W m−2 and 37 W m−2 lower at the C and SW locations, respectively. These latter two differences represent 12% and 17% of the annual net shortwave flux observed at the central and southwestern buoys, respectively. Mean standard deviations of the differences between the net shortwave data from the buoys and that of the ECMWF and NCEP models are 75–108 W m−2 and 85–120 W m−2, respectively. Such large standard deviations are not unexpected given the high subgrid-scale variability of this quantity, coupled with the fact that the effect of clouds on the model shortwave fluxes is not considered at every time step (i.e., 10–15 min) and grid point, but rather at 3-h increments and then over a coarser grid. This cost saving measure further hampers the already difficult task of correctly diagnosing cloud cover within the models.

The Oberhuber dataset exhibits net longwave radiation losses of 54–59 W m−2 at the buoys, while Isemer and Hasse offer more conservative values of 43–47 W m−2. Differences in net longwave radiation between the two climatologies arise from their disparate treatment of the effects of humidity and cloud cover on the longwave flux. Oberhuber bases his treatment of humidity on the formulas of Berliand and Berliand (1952), while Isemer and Hasse’s parameterization of this effect isbased upon the updated formula of Efimova (1961). At the high humidities typical of marine environments, the difference between these two parameterizations are fairly minimal. A larger difference stems from their disparate treatment of the effects of cloud cover. Although both climatologies use identical latitude-dependent cloud cover coefficients, as suggested by Budyko (1974), Isemer and Hasse use a cloud cover exponent of 1.1, while Oberhuber uses an exponent of 2. The use of a larger exponent effectively reduces the impact of cloud upon the longwave flux, thus promoting larger net longwave losses.

The monthly longwave radiation losses from the buoys generally exceed those offered by both climatologies. This is not unexpected, as the bulk formulas used by both climatologies to estimate the incoming longwave radiation have been shown to overestimate this quantity, even in clear-sky conditions (Bignami et al. 1995). The coefficients utilized within the bulk formulas have been empirically determined through land- based measurements. These coefficients overestimate the effect of water vapor upon the incoming longwave radiation at sea. Interestingly, the sign of the differences between the buoy and climatological longwave flux becomes seasonally dependent at the SE location. The observed net longwave losses typically exceed those portrayed by both climatologies during the winter months, but are typically less than the climatological norms during the summer. As boundary layer clouds are more prevalent during the summer season at the SE location, one wonders whether this dichotomous behavior is not merely a reflection of the inability of the bulk formulas to distinguish between cloud types. Ockert-Bell and Hartmann (1992) found that the variance in the longwave flux attributed to differences in cloud type can be 50% greater than that of total cloud cover alone. Although the presence of aerosols is not thought to play a significant role in the longwave budget, as their opacity decreases at longer wavelengths (Gilman and Garrett 1994), the deposition of Saharan particles upon the pyrgeometer may introduce some error in its measurements. Unfortunately, the precise nature of this error is unknown.

Although the net longwave losses from the buoys are generally larger than those of either climatology, both models exhibit even greater mean net longwave radiation losses than those of the buoys. Specifically, the mean longwave losses from the models exceed those of the buoys by up to 16 W m−2 (35%), with the greatest differences, once again, observed at the southeast location. Mean standard deviations of the differences between the buoy and model longwave fluxes are 20–26 W m−2, with low correlation coefficients of 0.32–0.49. The high standard deviations and low correlation coefficients reflect the inability of the models to reproduce the high-frequency variability in cloudiness that is observed at the buoys.

d. Net heat flux

The net heat flux represents the sum of the sensible, latent, shortwave, and longwave fluxes. Climatology suggests that all five buoy locations experience a net oceanic heat gain during the summer months and a net oceanic heat loss during the winter months. Intrannual variations in the climatological net heat flux range from 200 W m−2 across the southern buoys to 275 W m−2 across the northern buoys. Isemer and Hasse depict small annual oceanic net heat losses of 4–15 W m−2 across the northern and western buoys, zero net heat flux at the central buoy, and a heat gain of 34 W m−2 at the southeast buoy. The Oberhuber data portray annual oceanic net heat losses of 0–5 W m−2 at the northern and western buoys and gains of 6 and 30 W m−2 at the central and southeastern buoys, respectively. The buoy estimates of net heat flux exhibit a qualitatively similar spatial gradient across the array to that of the climatological datasets, but indicate an annual oceanic net gain of heat at all five buoys. In general, the buoy data reflect a greater oceanic heat gain in the summer and a smaller oceanic heat loss in the winter relative to the climatological datasets.

The Isemer and Hasse dataset exhibits greater annual net heat losses relative to the buoys primarily on the strength of smaller net shortwave gains and larger sensible and latent heat losses. Even greater net annual heat losses would have been exhibited by the Isemer and Hasse dataset but for the conservative nature of its net longwave losses. The greatest discrepancies between the net heat flux of the buoys and that of Isemer and Hasse occur during the winter months, with monthly differences of up to 100 W m−2. In contrast, the Oberhuber dataset arrives at greater net annual heat losses principally on the strength of smaller net shortwave fluxes, as both its sensible and longwave losses are typically less than those of the buoys and its latent heat losses are generally of a comparable magnitude. Differences of up to 75 W m−2 are observed between the buoy estimates and those of Oberhuber during both the summer and winter seasons.

Both models also display greater mean net heat losses than the buoys, with the NCEP model exhibiting the greatest net heat losses overall. Mean differences between the net heat fluxes of the buoys and those of the ECMWF model are 14–37 W m−2, with standard deviations of these differences of 80–104 W m−2. A similar comparison between the buoy and the NCEP model reveals mean differences of 13–56 W m−2 and standard deviations of 91–122 W m−2. The net heat losses predicted by the ECMWF model are larger than those estimated at the buoys, primarily on the strength of greater sensible, latent, and longwave losses, as the differences in the net shortwave fluxes are fairly small and not of a consistent sign across the buoys. The sensible, latent, and longwave losses of the NCEP model are also typically greater than those from the buoys and, hence,contribute to overall greater annual net heat losses. However, the smaller net shortwave fluxes offered by the NCEP model at the C and southwest locations acts to significantly augment the heat losses above and beyond what they otherwise might be at these two locations.

e. Freshwater flux

The R. M. Young sensor, which measured rainfall on the Subduction Experiment buoys, has performed well in side-by-side tests with other conventional tipping bucket guages on land (Crescenti and Weller 1989). It also performed well at sea during TOGA COARE, as its rainfall measurements agreed to within 10% of those from other types of guages on nearby platforms (Bradley and Weller 1995). During the Subduction Experiment, precipitation was only measured by the IMET system. Unfortunately, each of the Subduction Experiment buoys was not always outfitted with an IMET system. The periodic absence of the IMET system, coupled with instrumental problems associated with the rain gauges themselves, severely limited the temporal and spatial scope of reliable precipitation measurements during the Subduction Experiment. Even if the data return had been perfect, however, the intrinsically high spatial variability of precipitation would have made any quantitative assessment of the freshwater flux from five widely spaced buoys difficult at best.

Nevertheless, a number of qualitative points can be garnered from the precipitation data that are available. For example, the NCEP model has more instances of measurable precipitation falling at the buoy locations than are indicated by the rain gauge data. Most of these instances are associated with minor rainfall events. On the other hand, most of the heavier precipitation events at the rain gauges are qualitatively, at least, depicted well by the NCEP model. As far as satellite-derived precipitation estimates are concerned, the IMET measurements support the idea that the daily data from the Microwave Sounding Unit tends to overestimate rainfall in the fall while underestimating rainfall in winter (G. Reverdin 1995, personal communication).

7. Discussion and conclusions

Considerable differences have been noted in the atmospheric forcing terms garnered from the Subduction Experiment buoy data and those of commonly referenced model and climatological products. Of course, one intrinsic difference between the buoy data and data from the model and climatological products is that the former represents a single point measurement, while the latter’s domain extends over some specified grid box. Thus, some of the differences between the buoy data and the gridded products are a direct consequence of spatial averaging. Nevertheless, the mean differences in many of the surface forcing terms offered by the buoyand the gridded datasets are generally of a consistent sign across the entire array. This suggests that these mean differences are not simply a product of spatial averaging, but rather are a reflection of differences in bulk formulas and in the mean variables utilized within these formulas.

The COARE bulk flux algorithm represents a state- of-the art bulk parameterization scheme. It has been verified in the field by observations in both tropical and extratropical environments (Fairall et al. 1996b). In fact, several intercomparisons were made in the Subduction Experiment region with the aid of direct flux measurements collected aboard the RV Malcolm Baldrige during the Atlantic Stratocumulus Transition Experiment (ASTEX). The field phase of ASTEX took place in the Azorean region during June 1992 and, thus, coincided with the Subduction Experiment both in space and time (Albrecht et al. 1995). During those time periods conducive to making accurate shipboard covariance flux measurements, the bulk heat and momentum fluxes derived from the COARE algorithm and shipboard measurements of the basic observables compared quite favorably with the direct measurements of these fluxes. Specifically, the mean bulk sensible and latent heat flux estimates from the COARE algorithm exceeded the direct flux measurements by 0.8 and 7.0 W m−2, respectively, while the measured wind stress exceeded its bulk counterpart by only 0.006 N m−2.

In an effort to assess the relative sensitivity of the bulk flux estimates cited within this investigation to the bulk formulas used in their derivation, alternative heat and momentum fluxes were computed using the mean observables from the buoys and the wind speed and stability-dependent bulk formulas of Large and Pond (1981, 1982). The mean sensible heat loss generated by the bulk formulas of Large and Pond exceeded the sensible heat loss produced by the COARE algorithm by 3.5–4.0 W m−2. Roughly half of this discrepancy, however, can be directly attributed to the application of the warm-layer–cool-skin correction within the COARE algorithm. On the other hand, larger humidity transfer coefficients in the COARE algorithm produced mean latent heat losses that were 8–12 W m−2 greater than those generated by the bulk formulas of Large and Pond. Finally, the mean wind stress derived from the COARE algorithm was found to agree on average with that offered by the bulk formulas of Large and Pond to within 0.003 N m−2.

In summary, the results from intercomparisons involving both direct measurements and alternate bulk formulas suggest that the bulk fluxes generated by the COARE algorithm are a reliable representation of the true fluxes in the area. If any inherent biases are revealed from these intercomparisons, it is that the bulk latent heat losses from version 2.0 of the COARE algorithm may be a bit too large. In fact, plans are currently underway to reduce the scalar transfer coefficients by 6% in a subsequent version of this algorithm (C. Fairall,personal communication), which, in turn, would reduce the mean sensible and latent heat fluxes cited here by 1%–2%. Nevertheless, since the mean sensible and latent heat losses of Isemer and Hasse and both global models are larger than those from the buoys, a further reduction in the buoy sensible and latent heat losses will only act to accentuate these climatological and model biases still further.

It is difficult to draw many firm conclusions concerning the existence of local bias within the climatological datasets from only 2 consecutive years of data. Certainly, climatic variations on timescales greater than 2 yr have been observed in the North Atlantic. Nevertheless, no persistent climatic anomalies have been reported that would explain the significant differences found here. Chief among these persistent differences are the larger wind speeds and latent heat fluxes offered by the Isemer and Hasse dataset and the smaller net longwave fluxes typically offered by both climatologies. Both climatological datasets, but particularly that of Oberhuber, also generally exhibited smaller net shortwave fluxes. With the exception of the net longwave flux of the Oberhuber dataset, all of these differences between the buoy and climatological means exceed the uncertainties in the buoy data, which arise on account of potential instrumental biases. Furthermore, the seasonal means of all of these climatological quantities generally lie well outside of the 95% confidence interval computed from the buoy data. It should be noted, however, that the number of ship observations on which these climatological values are based is fairly low in this region of the North Atlantic. Therefore, one must exercise some degree of caution in extending these conclusions to other more heavily sampled oceanic regions.

Significant differences have also been revealed relative to the global model products. These differences assume the form of both persistent differences of a consistent sign and short-term differences of variable sign at timesscales of 6 h or less. For example, both models typically forecast greater mean sensible, latent, and longwave losses than do the buoys. Once again, the seasonal averages of these model fluxes tend to lie well outside the 95% confidence limits established by the buoy data. The persistence of these mean differences in sensible and latent heat flux imply that they are a product of either persistent differences in the mean variables from which the heat fluxes are derived or differences in bulk formulas. While the cold bias in the NCEP 2-m air temperature may explain the large sensible heat losses of this model, the basic observables from the ECMWF analyses are in fairly good agreement with those measured at the buoys. Furthermore, when the COARE algorithm was applied to the variables from the ECMWF surface analyses, the resulting mean sensible and latent losses were considerably less than those forecast by the models and more in line with the bulk fluxes from the buoys. Thus, some of the observed discrepancies are an artifact of the utilization of dissimilar transfer coefficients within the bulk formulas.

On timescales of 6 h and less, the models’ inability to replicate the heat and radiometric fluxes of the buoys is reflected in large mean standard deviations of the differences between the buoy and model fluxes. Subgrid-scale variability undoubtedly plays a large role in creating these differences at shorter timescales. Nevertheless, the imperfect parameterizations that the models must use to simulate subgrid-scale processes that are captured directly within the buoy time series inevitably contribute to some of these differences as well. So although the mere existence of differences at shorter timescales is not surprising in itself, the utility of a study such as this lies in its ability to quantify the models’ limitations in this regard.

Changes in the data assimilation systems and operational forecast models of both the ECMWF and NCEP occurred sporadically during the 2 years of the Subduction Experiment. Most of these changes were minor, but a few significant modifications occurred as well. In an effort to include a greater amount of data and produce a more consistent product overall, both the NCEP and ECMWF have developed reanalyses programs. Reanalyses for the Subduction Experiment time period have just recently become available from the NCEP. A cursory look at the NCEP reanalyses reveals that the excessive sensible heat losses of their operational model have been reduced somewhat. Net longwave losses have also been reduced. Unfortunately, the NCEP reanalyses still exhibit significantly less incoming shortwave radiation than measured at the SW location. Better agreement in the incoming shortwave flux, however, is observed at the central buoy. Overall, it is comforting to find that the individual surface forcing terms within the NCEP reanalyses are, for the most part, in better agreement with those of the buoys than were their counterparts from NCEP’s operational predecessor.

Acknowledgments

The Subduction Experiment mooring array was deployed by the members of the Upper Ocean Processes Group (Woods Hole Oceanographic Institution) and maintained in cooperation with R. Davis (Scripps Institution of Oceanography). Special thanks are extended to S. Anderson, N. Brink, and R. Goldsmith (Woods Hole Oceanographic Institution) for their integral contributions to this investigation. The authors would also like to recognize the Data Support Section at NCAR for providing the ECMWF and NCEP model data cited here, and C. Fairall and A. White (NOAA Environmental Technology Laboratory) for providing the data from the RV Malcolm Baldrige. This work is supported by the Office of Naval Research under Contract N00014-90-J-1490.

REFERENCES

  • Alados-Arboledas, L., J. Vida, and J. I. Jimenez, 1988: Effects of solar radiation on the performance of pyrgeometers with silicon domes. J. Atmos. Oceanic Technol.,5, 666–670.

  • Albrecht, B., and S. K. Cox, 1977: Procedures for improving pyrgeometer performance. J. Appl. Meteor.,16, 188–197.

  • ——, C. S. Bretherton, D. Johnson, W. H. Schubert, and A. S. Frisch, 1995: The Atlantic Stratocumulus Transition Experiment—ASTEX. Bull. Amer. Meteor. Soc.,76, 889–904.

  • Berliand, M. E., and T. G. Berliand, 1952: Determining the net longwave radiation of the Earth with consideration of the effect of cloudiness. Isv. Akad. Nauk. SSSR Ser. Geofis,1..

  • Bignami, F., S. Marullo, R. Santoleri, and M. E. Schiano, 1995: Longwave radiation budget in the Mediterranean Sea. J. Geophys. Res.,100, 2501–2514.

  • Bolton, D., 1980: The computation of equivalent potential temperature. Mon. Wea. Rev.,108, 1046–1053.

  • Bradley, E. F., and R. A. Weller, 1995: Proceedings of the Joint Workshop of the TOGA COARE Flux and Atmospheric Working Groups, Rep. TOGA COARE International Project Office, UCAR, 34 pp. [Available from UCAR, P. O. Box 3000, Boulder, CO 80307.].

  • Brink, N. J., K. A. Moyer, R. P. Trask, and R. A. Weller, 1995: The Subduction Experiment: Mooring field program and data summary. Tech. Rep. WHOI-95-08, 113 pp. [Available from WHOI, Woods Hole, MA 02543.].

  • Budyko, M. I., 1974: Climate and Life. International Geophysical Series, Vol. 18, Academic Press, 508 pp.

  • 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.

  • Cardone, V. J., J. G. Greenwood, and M. Cane, 1990: On trends in historical marine wind data. J. Climate,3, 113–127.

  • Charnock, H., 1955: Wind stress on a water surface. Quart. J. Roy. Meteor. Soc.,81, 639–640.

  • Crescenti, G. H., and R. A. Weller, 1989: Improved meteorological measurements from buoys and ships: Preliminary comparison of precipitation sensors. Tech. Rep. WHOI-89-44, IMET-TR-89-01, 32 pp. [Available from WHOI, Woods Hole, MA 02543.].

  • da Silva, A. M., C. C. Young, and S. Levitus, 1994: Atlas of Surface Marine Data 1994. Vol. 1, Algorithms and Procedures, U.S. Department of Commerce, 74 pp.

  • Dickey, T. D., D. V. Manov, R. A. Weller, and D. A. Siegel, 1994: Determination of longwave heat flux at the air–sea interface using measurements from buoy platforms. J. Atmos. Oceanic Technol.,11, 1057–1078.

  • ECMWF Technical Attachment, 1994: The description of the ECMWF/WCRP level III-A global atmospheric data archive. Rep. European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 72 pp. [Available from ECMWF, Shinfield Park, Reading RG2 9AX, United Kingdom.].

  • Efimova, N. A., 1961: On methods of calculating monthly values of net longwave radiation. Meteor. Gidrol.,10, 28–33.

  • Fairall, C. W., E. F. Bradley, J. S. Godfrey, G. A. Wick, J. B. Edson, and G. S. Young, 1996a: Cool-skin and warm-layer effects on sea surface temperature. J. Geophys. Res.,101, 1295–1308.

  • ——, ——, D. P. Rogers, J. B. Edson, and G. S. Young, 1996b: Bulk parameterization of air–sea fluxes for TOGA COARE. J. Geophys. Res.,101, 3747–3764.

  • Gill, G. C., 1983: Comparison testing of selected naturally ventilated solar radiation shields. Rep. NOAA/National Data Buoy Center, Bay St. Louis, MS, 38 pp.

  • 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, 5119–5134.

  • Godfrey, J. S., and A. C. M. Beljaars, 1991: On the turbulent fluxes of buoyancy, heat, and moisture at the air–sea interface at low wind speeds. J. Geophys. Res.,96, 22th043–22048.

  • Hosom, D. S., R. A. Weller, R. E. Payne, and K. E. Prada, 1995: The IMET (Improved Meteorology) ship and buoy systems. J. Atmos. Oceanic Technol.,12, 527–540.

  • Isemer, H.-J., and L. Hasse, 1985: The Bunker Climate Atlas of theNorth Atlantic Ocean. Vol. 1, Observations, Springer-Verlag, 218 pp.

  • ——, and ——, 1987: The Bunker Climate Atlas of the North Atlantic Ocean. Vol. 2, Air-Sea Interactions, Springer-Verlag, 252 pp.

  • Kanamitsu, M., 1989: Description of the NMC global data assimilation and forecast system. Wea. Forecasting,4, 335–342.

  • ——, J. C. Alpert, K. A. Campana, P. M. Caplan, D. G. Deaven, M. Iredell, B. Katz, H.-L. Pan, J. Sela, and G. H. White, 1991: Recent changes implemented into the global forecast system at NMC. Wea. Forecasting,6, 425–435.

  • Kaufeld, L., 1981: The development of a new Beaufort equivalent scale. Meteor. Rundsch.,34, 17–23.

  • Large, W. G., and S. Pond, 1981: Open ocean momentum flux measurements in moderate to strong winds. J. Phys. Oceanogr.,11, 324–336.

  • ——, and ——, 1982: Sensible and latent heat fluxes over the ocean. J. Phys. Oceanogr.,12, 464–482.

  • Liu, W. T., K. B. Katsaros, and J. A. Businger, 1979: Bulk parameterization of air-sea exchanges of heat and water vapor including the molecular constraints at the interface. J Atmos. Sci.,36, 1722–1735.

  • Luyten, J. R., J. Pedlosky, and H. Stommel, 1983: The ventilated thermocline. J. Phys. Oceanogr.,13, 292–309.

  • Marshall, J. C., A. J. G. Nurser, and R. G. Williams, 1993: Inferring the subduction rate and period over the North Atlantic. J. Phys. Oceanogr.,23, 1315–1329.

  • Miller, M. J., A. C. M. Beljaars, and T. N. Palmer, 1992: The sensitivity of the ECMWF model to the parameterization of evaporation from the tropical oceans. J. Climate,5, 418–434.

  • 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. Max-Planck-Institut fur Meteorologie Rep. 15. [Available from Max-Planck-Institut für Meteorologie, Bundesstrasse 55, D-20146, Hamburg, Germany.].

  • Ockert-Bell, M., and D. Hartmann, 1992: The effect of cloud type on earth’s energy balance: Results for selected regions, J. Climate,5, 1157–1171.

  • Payne, R. E., 1972: Albedo of the sea surface. J. Atmos. Sci.,29, 959–970.

  • Quayle, R. G., 1980: Climatic comparisons of estimated and measured wind from ships. J. Appl. Meteor.,19, 142–156.

  • Reed, R. K., 1977: On estimating insolation over the ocean. J. Phys. Oceanogr.,7, 482–485.

  • Reynolds, R. W., 1991: A new global sea surface temperature analysis. Tech. Procedure Bulletin 393, 31 pp. [Available from National Weather Service, Office of Meteorology, Silver Spring, MD 20910.].

  • Smith, S. D., 1988: Coefficients for sea surface wind stress, heat flux, and wind profiles as a function of wind speed and temperature. J. Geophys. Res.,93, 15467–15472.

  • Stommel, H. M., 1979: Determination of the water mass properties of water pumped down from the Ekman layer to the geostrophic flow below. Proc. Natl. Acad. Sci. U.S.A.,76, 3051–3055.

  • Weller, R. A., and R. E. Davis, 1980: A vector measuring current meter. Deep-Sea Res.,27A, 565–582.

  • ——, D. L. Rudnick, R. E. Payne, J. P. Dean, N. J. Pennington, and R. P. Trask, 1990: Measuring near-surface meteorology over the ocean from an array of surface moorings in the subtropical convergence zone. J. Atmos. Oceanic Technol.,7, 85–103.

  • Woodruff, S. D., R. J. Slutz, R. L. Jenne, and P. M. Steurer, 1987: A Comprehensive Atmosphere–Ocean Data Set. Bull. Amer. Meteor. Soc.,68, 1239–1250.

  • Wright, P. B., 1988: An atlas based on the “COADS” data set: Fields of mean wind, cloudiness and humidity at the surface of the global ocean. Max-Planck-Institut für Meteorologie Rep. 14, 73 pp. [Available from Max-Planck-Institut für Meteorologie, Bundesstrasse 55, D-20146, Hamburg, Germany.].

  • Young-Molling, C. C., A. M. da Silva, and S. Levitus, 1995: Atlas of Surface Marine Data 1994 Supplement A: Anomalies of Directly Observed Quantities and Surface Marine Fluxes for 19901993. U.S. Department of Commerce, 41 pp.

Fig. 1.
Fig. 1.

The locations of the five Subduction Experiment buoys juxtaposed with the average Isemer and Hasse climatological sea level pressure in millibars for (a) the winter months of December–February and (b) the summer months of June–August.

Citation: Journal of Climate 10, 11; 10.1175/1520-0442(1997)010<2725:OOSFFT>2.0.CO;2

Fig. 2.
Fig. 2.

A time line for the Subduction Experiment buoys. A (+) indicates that IMET instrumentation was used and a (−) that VAWR instrumentation was used. The absence of any symbols indicates that no reliable data were available from either system.

Citation: Journal of Climate 10, 11; 10.1175/1520-0442(1997)010<2725:OOSFFT>2.0.CO;2

Fig. 3.
Fig. 3.

Seasonal means of (a) sea–air temperature, (b) sea–air specific humidity, (c) wind speed, (d) wind stress, (e) sensible heat flux, (f) latent heat flux, (g) net longwave flux, (h) net shortwave flux, and (i) net heat flux from the Isemer and Hasse (*), Oberhuber (○), ECMWF (×), NCEP (+), and buoy (—) datasets at the NE location. The 95% confidence limits for the buoy data are represented by the error bars. Note that the buoy and Isemer and Hasse wind speeds have been corrected to a height of 10 m.

Citation: Journal of Climate 10, 11; 10.1175/1520-0442(1997)010<2725:OOSFFT>2.0.CO;2

Fig. 4.
Fig. 4.

Four-day running means of (a) wind stress, (b) sensible heat flux, (c) latent heat flux, (d) net longwave flux, (e) net shortwave flux, and (f) net heat flux from 1 January 1992–31 December 1992 at the NE location. The buoy data are represented by a solid line (—), while the ECMWF data are denoted by a dotted line (····).

Citation: Journal of Climate 10, 11; 10.1175/1520-0442(1997)010<2725:OOSFFT>2.0.CO;2

Fig. 5.
Fig. 5.

Time series of sea temperature from 15 October 1992 to 5 November 1992 at the NE location. The buoy measurements at a depth of 1 m are represented by a dashed line (– – –), the skin temperature estimated from the buoy data by a dotted line (····), and the ECMWF sea surface temperature analyses by a solid line (—).

Citation: Journal of Climate 10, 11; 10.1175/1520-0442(1997)010<2725:OOSFFT>2.0.CO;2

Table 1.

Long-term accuracy of the basic observables from the Subduction Experiment buoys and the uncertainties that instrumental bias places upon the resulting mean air–sea fluxes.

Table 1.
Table 2.

Two-year means from the buoy, the global assimilation products, and the climatological products at the NE location. Sea temperature from the buoy is represented by the estimated skin temperature, and the buoy and Isemer and Hasse wind speeds have both been corrected to a height of 10 m.

Table 2.
Table 3.

Ranges of the mean differences, standard deviations of these differences, and correlation coefficients across the five buoy locations. The differences are defined as buoy values minus model values. Therefore, positive differences in temperature, humidity, and wind speed, and wind stress are indicative of larger buoy values; positive differences in sensible, latent, and net longwave flux reflect smaller oceanic heat losses in the buoy data; and positive differences in net shortwave and net heat flux reflect larger oceanic heat gains in the buoy data.

Table 3.

|Woods Hole Oceanographic Institution Contribution Number 9133.

Save