• Berliand, M. E., and T. G. Berliand, 1952: Determining the net longwave radiation of the earth with consideration of the effect of cloudiness (in Russian). Izv. Akad. Nauk. SSSR Ser. Geofiz., No. 1.

  • Brunt, D., 1932: Notes on the radiation in the atmosphere. Quart. J. Roy. Meteor. Soc.,58, 389–420.

  • Budyko, M. I., 1974: Climate and Life. Academic Press, 508 pp.

  • Fairall, C., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: The TOGA COARE bulk flux algorithm. J. Geophys. Res.,101, 3747–3764.

  • Godfrey, J. S., M. Nunez, E. F. Bradley, P. A. Coppin, and E. J. Lindstrom, 1991: On the net surface flux into the western equatorial Pacific. J. Geophys. Res.,96, 3391–3400.

  • Gruber, A., and A. F. Krueger, 1984: The status of the NOAA outgoing longwave radiation data set. Bull. Amer. Meteor. Soc.,65, 958–962.

  • Gutzler, D. S., G. N. Kiladis, G. A. Meehl, K. M. Weickmann, and M. Wheeler, 1994: The global climate of December 1992–February 1993. Part II: Large-scale variability across the tropical western Pacific during TOGA COARE. J. Climate,7, 1606–1622.

  • Hayes, S. P., L. J. Mangum, J. Picaut, A. Sumi, and K. Takeuchi, 1991: TOGA-TAO: A moored array for real-time measurements in the tropical Pacific Ocean. Bull. Amer. Meteor. Soc.,72, 339–347.

  • Hendon, H. H., and B. Liebmann, 1990: The intraseasonal (30–50 day) oscillation of the Australian summer monsoon. J. Atmos. Sci.,47, 2909–2923.

  • ——, and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci.,51, 2225–2237.

  • ——, and J. Glick, 1997: Intraseasonal air–sea interaction in the tropical Indian and Pacific Oceans. J. Climate,10, 647–661.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc.,77, 437–471.

  • Knutson, T. R., and K. M. Weickmann, 1987: 30–60 day atmospheric oscillations: Composite life cycles of convection and circulation anomalies. Mon. Wea. Rev.,115, 1407–1436.

  • Krishnamurti, T. N., D. K. Osterhof, and A. V. Mehta, 1988: Air–sea interaction on the timescale of 30 to 50 days. J. Atmos. Sci.,45, 1304–1322.

  • Lau, K.-M., and C.-H. Sui, 1997: Mechanisms of short-term sea surface temperature regulation: Observations from TOGA COARE. J. Climate,10, 465–472.

  • Li, Z., C. H. Whitlock, and T. P. Charlock, 1995: Assessment of the global monthly mean surface insolation estimate from satellite measurements using global energy balance archive data. J. Climate,8, 315–328.

  • Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci.,28, 702–708.

  • ——, and ——, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Atmos. Sci.,29, 1109–1123.

  • Pinker, R., and I. Laszlo, 1992: Modeling of surface solar irradiance for satellite applications on a global scale. J. Appl. Meteor.,31, 194–211.

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

  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate,7, 929–948.

  • Rossow, W. B., and R. A. Schiffer, 1991: ISCCP cloud data products. Bull. Amer. Meteor. Soc.,72, 2–20.

  • Rui, H., and B. Wang, 1990: Developement characteristics and dynamic structure of tropical intraseasonal convection anomalies. J. Atmos. Sci.,47, 357–379.

  • Salby, M. L., and H. H. Hendon, 1994: Intraseasonal behavior of clouds, temperature, and motion in the Tropics. J. Atmos. Sci.,51, 2207–2224.

  • Shinoda, T., and H. H. Hendon, 1998: Mixed layer modeling of intraseasonal variability in the tropical western Pacific and Indian Oceans. J. Climate, in press.

  • Spencer, R. W., 1993: Global oceanic precipitation from the MSU during 1979–91 and comparisons to other climatologies. J. Climate,6, 1301–1326.

  • Trenberth, K. E., 1992: Global analyses from ECMWF and atlas of 1000 to 10 mb circulation statistics. NCAR Tech. Note NCAR/TN-373+STR, 191 pp. [Available from National Center for Atmospheric Research, Boulder, CO 80307.].

  • Waliser, D. E., and N. E. Graham, 1993: Convective cloud systems and warm-pool sea surface temperatures: Coupled interactions and self-regulation. J. Geophys. Res.,98 (D7), 12 881–12 893.

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Weller, R. A., and S. P. Anderson, 1996: Surface meteorology and air–sea fluxes in the western equatorial Pacific warm pool during the TOGA Coupled Ocean–Atmosphere Response Experiment. J. Climate,9, 1959–1990.

  • Whitlock, C. H., and Coauthors, 1995: First Global WCRP Shortwave Surface Radiation Budget Dataset. Bull. Amer. Meteor. Soc.,76, 905–922.

  • Zhang, C., 1996: Atmospheric intraseasonal variability at the surface in the tropical western Pacific Ocean. J. Atmos. Sci.,53, 739–758.

  • ——, and H. H. Hendon, 1997: On propagating and standing components of the intraseasonal oscillation in tropical convection. J. Atmos. Sci.,54, 741–752.

  • Zhang, G. J., and M. J. McPhaden, 1995: The relationship between sea surface temperature and latent heat flux in the equatorial Pacific. J. Climate,8, 589–605.

  • View in gallery
    Fig. 1.

    Point-by-point correlation coefficient between daily OLR and shortwave radiation from SRB data during 1985–88. The rectangular area in the map indicates (10°N–10°S, 60°E–180°). The contour interval is 0.1. Mean OLR less than 250 W m−2 is shaded (interval 10 W m−2).

  • View in gallery
    Fig. 2.

    (a) Same as Fig. 1 except correlation coefficient between SRB and shortwave radiation calculated from empirical formula by Reed (1977). Shading shows mean total cloudiness from the ISSCP C2 data for the period 1985–88 (shading intervals at 55%, 65%, and 75%). (b) Same as in (a) except for the correlation coefficient between SRB and net shortwave radiation from NCEP reanalyses. Shading shows net shortwave radiation from the NCEP data for the period 1985–88 (shading intervals at 250, 240, and 230 W m−2).

  • View in gallery
    Fig. 3.

    Relation between daily OLR and percent cloud from ISCCP data during 1986–90 for the area (10°N–10°S, 60°E–180°). The circles indicate means of percent cloud in 1 W m−2 bins of OLR. The dotted line indicates plus and minus one standard deviation of total cloud in each bin. The dashed line indicates the number of observations in each bin. The thick line is a third-order polynomial obtained by least squares fit to the binned data.

  • View in gallery
    Fig. 4.

    First (a) and second (b) eigenvectors of intraseasonally filtered OLR data for the domain (15°N–15°S, 60°E–90°W). The eigenvectors have been scaled for a one standard deviation anomaly of each principal component (contour interval 3 W m−2). Shading indicates correlation coefficient between intraseasonally filtered SST and the respective principal components of OLR (shading interval is 0.1 with first level at 0.2).

  • View in gallery
    Fig. 5.

    Daily amplitude, during TOGA COARE, 22 October 1992 through 2 March 1993, of first (horizontal axis) and second (vertical axis) principal components of the EOF analysis of OLR. Numbers on the right side of the points indicate the month, day, and year. The circle indicates the first date (22 October).

  • View in gallery
    Fig. 6.

    Time series of (a) wind stress, (b) latent heat flux, (c) shortwave radiation, (d) longwave radiation, (e) net surface heat flux, (f) precipitation, and (g) SST for TOGA COARE 22 October 1992 through 2 March 1993. Dashed lines in (a)–(f) indicate daily mean flux estimates from the IMET mooring observations (Weller and Anderson 1996) sited at 1°45′S, 156°E. The dashed line in (g) indicates the weekly mean SST from the IMET mooring observations. Thick lines indicate estimates from gridded data centered at 2.5°S, 155°E. The dotted line in (c) indicates shortwave flux estimated from the formula by Reed (1977). The dotted line in (d) indicates net longwave flux estimated from the Berliand and Berliand (1952) formula using ISCCP total cloudiness. The dotted line in (g) indicates the daily mean SST from the IMET mooring observations.

  • View in gallery
    Fig. 6.

    (Continued)

  • View in gallery
    Fig. 7.

    Time series of SST from the gridded analyses and from the TAO mooring observations. Data from three TAO moorings along 156°E at 2°N, 0°, and 2°S are averaged. The thick line indicates the gridded SST analyses at 0°, 155°E. The dashed and thin lines indicate the daily mean and weekly mean TAO data, respectively.

  • View in gallery
    Fig. 8.

    Composite (a) shortwave radiation anomaly, (b) latent heat flux anomaly, (c) net surface heat flux anomaly, and (d) SST anomaly for 10 MJO events along 5°S. The vertical axis indicates the phase based on the EOFs of OLR.

  • View in gallery
    Fig. 8.

    (Continued)

  • View in gallery
    Fig. 9.

    Composite (a) SST tendency, net surface heat flux, shortwave radiation, and latent heat flux; and (b) wind stress and precipitation for 10 MJO events at 2.5°–7.5°S, 160°–170°E.

  • View in gallery
    Fig. 9.

    (Continued)

  • View in gallery
    Fig. 10.

    Composite (a) SST tendency, net surface flux, shortwave radiation, and latent heat flux; and (b) wind stress and precipitation for 10 MJO events at 2.5°–7.5°S, 100°–110°E.

  • View in gallery
    Fig. 10.

    (Continued)

  • View in gallery
    Fig. 11.

    Composite difference between sum of latent and sensible heat flux estimates based on weekly SST and low-pass-filtered SST along 5°S. Vertical axis is as in Fig. 7.

  • View in gallery
    Fig. 12.

    Composite latent and sensible heat flux based on weekly SST (solid line) and low-pass-filtered SST (dashed line) at 2.5°–7.5°S, 160°–170°E (upper panel). The difference between the two composites is shown in the lower panel.

  • View in gallery
    Fig. 13.

    Composite difference along 5°S between net surface heat flux based on weekly SST and low-pass-filtered SST. Vertical axis is as in Fig. 7.

  • View in gallery
    Fig. 14.

    Schematic diagram showing magnitude and phase relationship relative to the convective anomaly of the surface fluxes and SST variations produced by the canonical MJO. The asymmetric zonal scale of the cloudy-windy and suppressed-calm phases and eastward phase speed (4 m s−1) of the joint atmosphere–ocean disturbance across the warm pool are indicated. Typical extrema of surface fluxes and SST over life cycle of MJO are shown for western Pacific.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 538 200 32
PDF Downloads 353 175 21

Intraseasonal Variability of Surface Fluxes and Sea Surface Temperature in the Tropical Western Pacific and Indian Oceans

Toshiaki ShinodaClimate Diagnostics Center, University of Colorado, Boulder, Colorado

Search for other papers by Toshiaki Shinoda in
Current site
Google Scholar
PubMed
Close
,
Harry H. HendonClimate Diagnostics Center, University of Colorado, Boulder, Colorado

Search for other papers by Harry H. Hendon in
Current site
Google Scholar
PubMed
Close
, and
John GlickClimate Diagnostics Center, University of Colorado, Boulder, Colorado

Search for other papers by John Glick in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Composites of sea surface temperature (SST), surface heat, momentum, and freshwater flux anomalies associated with intraseasonal oscillations of convection are developed for the warm pool of the western Pacific and Indian Oceans during 1986–93. The composites are based on empirical orthogonal function analysis of intraseasonally filtered outgoing longwave radiation (OLR), which efficiently extracts the Madden–Julian oscillation (MJO) in convection. Surface fluxes are estimated using gridded analyses from the European Centre for Medium-Range Weather Forecasts, weekly SST, OLR, microwave sounding unit precipitation, and the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) bulk flux algorithm. At intraseasonal timescales, these surface flux estimates agree reasonably well with estimates based on mooring observations collected during TOGA COARE.

The amplitude of the composite SST variation produced by the MJO is about 0.25°C in the western Pacific, 0.35°C in the Indonesian region, and 0.15°C in the Indian Ocean. The intraseasonal anomalies of SST and net surface heat flux propagate eastward at about 4 m s−1 along with the convective anomaly. The amplitude of the net surface heat flux variation is 50–70 W m−2 in the western Pacific, with anomalous insolation and latent heat flux making similar contributions. Across the Indian Ocean, the net surface heat flux anomaly is weaker (30–40 W m−2), and anomalous insolation appears to make a greater contribution than anomalous latent heat flux. Across the entire warm pool, the net surface heat flux leads the SST variation by about one-quarter cycle, which is consistent with the notion that surface heat flux variations are driving the SST variations at these intraseasonal timescales. The intraseasonal SST variation, however, is estimated to significantly reduce the amplitude of the latent and sensible heat fluxes produced by the MJO.

Corresponding author address: Dr. Toshiaki Shinoda, Climate Diagnostics Center, University of Colorado, Campus Box 449, Boulder, CO 80309.

Email: ts@cdc.noaa.gov

Abstract

Composites of sea surface temperature (SST), surface heat, momentum, and freshwater flux anomalies associated with intraseasonal oscillations of convection are developed for the warm pool of the western Pacific and Indian Oceans during 1986–93. The composites are based on empirical orthogonal function analysis of intraseasonally filtered outgoing longwave radiation (OLR), which efficiently extracts the Madden–Julian oscillation (MJO) in convection. Surface fluxes are estimated using gridded analyses from the European Centre for Medium-Range Weather Forecasts, weekly SST, OLR, microwave sounding unit precipitation, and the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment (TOGA COARE) bulk flux algorithm. At intraseasonal timescales, these surface flux estimates agree reasonably well with estimates based on mooring observations collected during TOGA COARE.

The amplitude of the composite SST variation produced by the MJO is about 0.25°C in the western Pacific, 0.35°C in the Indonesian region, and 0.15°C in the Indian Ocean. The intraseasonal anomalies of SST and net surface heat flux propagate eastward at about 4 m s−1 along with the convective anomaly. The amplitude of the net surface heat flux variation is 50–70 W m−2 in the western Pacific, with anomalous insolation and latent heat flux making similar contributions. Across the Indian Ocean, the net surface heat flux anomaly is weaker (30–40 W m−2), and anomalous insolation appears to make a greater contribution than anomalous latent heat flux. Across the entire warm pool, the net surface heat flux leads the SST variation by about one-quarter cycle, which is consistent with the notion that surface heat flux variations are driving the SST variations at these intraseasonal timescales. The intraseasonal SST variation, however, is estimated to significantly reduce the amplitude of the latent and sensible heat fluxes produced by the MJO.

Corresponding author address: Dr. Toshiaki Shinoda, Climate Diagnostics Center, University of Colorado, Campus Box 449, Boulder, CO 80309.

Email: ts@cdc.noaa.gov

1. Introduction

The Coupled Ocean–Atmosphere Response Experiment (COARE) was conducted during October 1992 through March 1993 with the goal of understanding the mechanisms that maintain and perturb the warm pool of the equatorial western Pacific Ocean, which is the region of warmest sea surface temperature (SST) and largest annual precipitation over the open ocean (Webster and Lukas 1992). SST in the intensive flux array, centered at 2°S and 156°E, was observed to fluctuate at two disparate timescales: diurnal and intraseasonal (Weller and Anderson 1996). The diurnal variability was strongly modulated by the intraseasonal variability, whose dominant period (40–50 days) was similar to that of the atmospheric Madden–Julian oscillation (MJO; Madden and Julian 1971, 1972). Pronounced intraseasonal variability of convection, with a similar 40–50-day period, was also observed during COARE (e.g., Gutzler et al. 1994). Hendon and Glick (1997, hereafter HG97) indicated that the intraseasonal SST variation observed during COARE can be traced westward back into the Indian Ocean and was associated with eastward propagating, intraseasonal convective anomalies, which were taken to be manifestations of the atmospheric MJO.

Since the atmospheric intraseasonal oscillation was first described by Madden and Julian (1971, 1972), its evolution and structure have been detailed by many composite studies (e.g., Knutson and Weickmann 1987;Rui and Wang 1990; Hendon and Salby 1994). However, until recently very little information has been provided about the associated variability in the ocean (particularly SST) and the concomitant surface fluxes of moisture, heat, momentum, and radiation. Krishnamurti et al. (1988) examined the global distribution of the latent and sensible heat fluxes and SST variations on the timescale of 30–50 days. They showed that the region of pronounced intraseasonal variance in SST lies over the tropical western Pacific and eastern Indian Oceans. Zhang and McPhaden (1995) and Zhang (1996) investigated the intraseasonal variability of surface fluxes and SST using extended records of observations from the Tropical Ocean Global Atmosphere Tropical Atmosphere–Ocean (TOGA TAO) buoys in the equatorial western Pacific (Hayes et al. 1991). They found that intraseasonal SST anomalies generally lagged reduced evaporation by about one-quarter cycle and thus concluded that latent heat flux anomalies were primarily responsible for driving SST anomalies on the intraseasonal timescale. HG97 explored the relationship between the MJO-induced surface fluxes of moisture and shortwave radiation and SST variability across the equatorial Indian and western Pacific Oceans, which is where intraseasonal variability is dominated by the MJO. HG97 showed that SST anomalies with an amplitude of approximately 0.3°C move coherently, but with ≈one-quarter cycle lag, along with the eastward propagating convective anomaly across the Indian Ocean out past the date line. They further deduced that both anomalous shortwave radiation and latent heat flux played important roles in driving the observed SST anomalies (see also Lau and Sui 1997).

The goal of the present study is to quantify the magnitude of the surface fluxes of heat, momentum, radiation, and moisture produced by the MJO such that the impact on the warm pool of the Indian and western Pacific can be better understood. To this end, a composite evolution of the MJO-induced anomalies in SST and surface fluxes will be developed based on well-defined occurrences of the MJO during 1986–93. The surface stress and latent and sensible heat fluxes will be estimated using bulk formulas with analyzed winds, whereas the freshwater flux will be estimated from satellite-observed oceanic rainfall. The fluxes of shortwave and longwave radiation will be estimated by empirical formulas and regression relationships based on, among other parameters, outgoing longwave radiation (OLR). The datasets, bulk formulas, and empirical and regression relationships, along with the compositing technique, are described in section 2. Comparison of these surface fluxes with those measured and derived from detailed observations collected during COARE is provided in section 3. The composite evolution of SST and these surface fluxes for the life cycle of the MJO is described in section 4. The effect of the SST variation associated with MJO on the surface fluxes is discussed in section 5, and conclusions are provided in section 6.

2. Data, surface fluxes, and compositing method

a. Data

To understand how the MJO interacts with the warm pool, estimates of all components of the air–sea heat flux, surface stress, and freshwater flux (precipitation minus evaporation) are required. The air–sea heat flux includes latent and sensible heat and longwave and shortwave radiation. Surface fluxes of latent and sensible heat and surface wind stress are estimated from daily uninitialized analyses of surface winds and sea level pressure produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), along with empirical estimates of surface humidity and air temperature (described below) and weekly SST analyses (Reynolds and Smith 1994). The weekly SST data were first linearly interpolated to daily resolution for compatibility with the daily meteorological analyses. Daily average precipitation over the ocean is available from the Microwave Sounding Unit (MSU; Spencer 1993). Daily average OLR from the Advanced Very High Resolution Radiometer on the National Oceanic and Atmospheric Administration’s (NOAA) operational polar-orbiting satellites (Gruber and Krueger 1984) is used as a proxy for deep convection, from which net surface shortwave radiation and total cloudiness (for use in an empirical formula for net longwave radiation) are estimated (described below). All of these datasets are available for the 7-yr period July 1986–June 1993 with 2.5° horizontal resolution (except for the SST, which is provided on a 1° grid and was subsequently interpolated onto a 2.5° grid). These gridded data are also averaged onto a 5° lat × 10° long grid for the composite analyses in order to emphasize the large scales associated with intraseasonal variations. Data collected on the Improved Meteorological Instrument (IMET) mooring (156°E, 1°45′S) during TOGA COARE (22 October 1992–2 March 1993; Weller and Anderson 1996) are used as an independent sample with which to compare the flux estimates from the gridded analyses.

b. Surface flux calculation

The TOGA COARE bulk flux algorithm (Fairall et al. 1996) is employed to estimate the surface wind stress and latent and sensible heat fluxes from the gridded analyses. As inputs we use the ECMWF analyzed winds at 10 m, weekly SST analyses, and empirical estimates of surface specific humidity and air temperature. These empirical estimates are used rather than the analyses from ECMWF since, according to Trenberth (1992), a questionable amount of observed humidity data actually gets into the analyses. Furthermore, continual changes to the analysis system have introduced spurious jumps in the humidity record. The empirical estimate of specific humidity and air temperature is that of Waliser and Graham (1993):
i1520-0442-11-7-1685-e1
where qa is the specific humidity, Ta is the air temperature, qs(Ta) is the saturation specific humidity at Ta, and the relative humidity (RH) is assigned a value of 80%. These formulas fit satellite-inferred values of specific humidity in the western equatorial Pacific (Waliser and Graham 1993).

Accurate estimates of surface insolation have been produced using retrievals from satellites for a number of years (e.g., Li et al. 1995). However, no such daily estimates are currently available for the entire record considered here. Hence, we develop an algorithm to estimate surface insolation using daily observations of the readily available OLR. The rationale is that variations of surface insolation over the warm pool are produced primarily by variations in cloudiness and that variations in cloudiness are primarily produced by variations of deep convection, for which OLR is a good proxy. The algorithm is based on linear regression of OLR onto a gridded record of daily net surface insolation for the period 1985–88. This record of net insolation is derived from daily estimates of downwelling shortwave radiation produced by Pinker and Laszlo (1992) as part of the Surface Radiation Budget (SRB) project (Whitlock et al. 1995). Daily net insolation is estimated from the daily downwelling shortwave radiation by use of the monthly mean surface albedo provided by Pinker and Laszlo (1992), which we interpolate to daily resolution. Hereafter we will refer to this estimate of the daily net insolation, which has 2.5° horizontal resolution, as SRB. Li et al. (1995) compared the monthly mean downwelling shortwave radiation from Pinker and Laszlo (1992) to a worldwide (but mostly from Northern Hemisphere land stations) archive of surface observations. They found a mean correlation of 0.95 and estimated the rms error to be 5 W m−2. However, a global-mean positive bias of 11 W m−2 was diagnosed. Li et al. (1995) conclude that this bias results from too little clear-sky absorption by water vapor in the radiative transfer calculation used in the Pinker and Laszlo algorithm. This positive bias is removed here from the SRB data, without regard to any possible spatial dependence, prior to developing the OLR-based algorithm.

To provide some justification for developing a linear prediction of daily net surface insolation based on OLR, we show in Fig. 1 the point-by-point correlation between daily OLR and SRB for the period 1985–88. Prior to computing the correlations, the data were averaged onto a 5° lat × 10° long grid. Also shown is the mean OLR with values less than 250 W m−2 shaded, which emphasizes the regions of time-mean convection. The correlation between OLR and SRB is largest (>0.8) where the mean OLR is the lowest (i.e., over the warm pool), which is consistent with our assumption that variations of convection captured by the OLR are indicative of variations of surface insolation.

The algorithm to predict net surface insolation from OLR is based on linear regression of OLR onto SRB using time series at each 2.5° grid box in the highlighted domain in Fig. 1, which encompasses the warm pool region of interest in this study. The following regression relation is obtained:
QsQo
where Qs is the net surface insolation in watts per square meter, and Qo is the OLR in watts per square meter. This regression relationship captures approximately 50% of the daily variance of SRB in the boxed domain. For larger grids and longer time averages, the explained variance increases accordingly. For instance, 62% of the variance for weekly data averaged onto 5° lat × 10° long is explained. The standard error of this surface insolation estimate is about 19 W m−2, whereas the weekly standard deviation of the SRB data is about 32 W m−2.

This empirical formula based on OLR appears to perform as well over the warm pool as other empirical formulas, which typically use geographical location and cloud amount as predictors. For instance, Fig. 2a shows the correlation between daily SRB and the net insolation calculated from the popular formula of Reed (1977) for the same 1985–88 period. Godfrey et al. (1991) showed that this formula performed well for the equatorial western Pacific. Daily total cloudiness from International Satellite Cloud Climatology Project (ISCCP) C2 data (Rossow and Schifer 1991) was used in Reed’s formula. In contrast, the correlation of daily estimates from Reed’s formula with SRB increases away from the Tropics and is a minimum in regions of maximum cloudiness in the Tropics. In fact, the correlation in the boxed region over the warm pool is lower (0.66) than that with OLR (0.7). The empirical estimate based on OLR also outperforms the estimate of insolation from the National Centers for Environmental Prediction (NCEP) reanalyses (Kalnay et al. 1996). Figure 2b shows the correlation between SRB and net insolation from the NCEP data. As with Reed’s formula, the correlation increases away from the Tropics and is a minimum in regions of minimum insolation (i.e., maximum cloudiness). The correlation coefficient in the highlighted region over the warm pool is, in fact, only 0.38. In other words, less than 15% of the daily variance of SRB is captured by the NCEP insolation. Hence, we conclude that our regression relationship based on OLR provides daily estimates of net insolation over the warm pool that are as accurate as those from the popular empirical formula of Reed (1977) and are superior to the insolation from NCEP reanalyses.

Net surface longwave radiation is estimated using the formula originally suggested by Brunt (1932) and modified by Berliand and Berliand (1952) (cited by Budyko 1974). Godfrey et al. (1991) showed this to be the most accurate for the warm pool of three popular empirical formulas. The Berliand and Berliand formula requires surface air temperature and humidity, sea level pressure, and SST, as well as total cloud amount. Unfortunately, daily total cloud amounts from ISCCP are not available for the entire record considered here. Hence, we turn again to OLR in order to develop an empirical prediction of total cloudiness. The basic justification for predicting total clouds from OLR over the warm pool region is the same as for predicting net surface insolation. The empirical prediction of total cloudiness from OLR is based on the observed relationship between daily OLR and ISCCP cloudiness for the period 1986–90. We first bin all simultaneous observations of C2 total cloudiness into 1 W m−2 bins of OLR (open circles in Fig. 3) for the highlighted region across the warm pool in Fig. 1. The long dashed line in Fig. 3 is the total number of observations in each bin: few occurrences of OLR below 150 W m−2 and above 290 W m−2 are indicated. In this range a highly nonlinear relationship is observed with cloudiness dramatically increasing for decreasing OLR between 290 and 200 W m−2, whereas a much less dramatic rise occurs for OLR less than 200 W m−2. A large spread about the mean estimates is also indicated by the plus and minus one standard deviation in each bin. Nonetheless, a third-order polynomial fit to this binned data,
Qo−3Q2o−6Q3o,
where C is the cloud amount in percent and Qo is the OLR in watts per square meter, is able to capture 41% of the daily variance of ISCCP cloudiness in the boxed region in Fig. 1. This explained variance increases to 57% for daily data averaged onto a 5° lat × 10° long grid and to 67% for weekly averaged data. Hence, the larger space and time scales of cloudiness associated with intraseasonal activity are better captured than the higher frequency synoptic scales. Furthermore, a positive impact of including this predicted cloudiness in the empirical estimate of net longwave radiation is shown in section 3.

c. Compositing technique

Surface fluxes, precipitation, OLR, and SST are composited based on identification of individual occurrences of the MJO in convection. The MJO in convection is identified by empirical orthogonal function (EOF) analysis of OLR for the period 1986–93 in the domain (60°E–90°W, 15°N–15°S). The OLR data are first intraseasonally filtered (via spectral transform) to periods of 30–90 days, which is where the signal of the MJO is concentrated (e.g., Salby and Hendon 1994). Figure 4 shows the leading two EOFs, which capture 17% and 14% of the intraseasonally filtered variance, respectively. These two modes form a pair (i.e., their eigenvalues are not significantly separated and their principal components are predominantly in quadrature) and together they describe a zonally propagating convective disturbance with near 50-day period. The maximum amplitude is found along the equator (though shifted slightly into the Southern Hemisphere) in the Indian and western Pacific Oceans. The explained variance of the two modes combined exhibits a pronounced seasonality (not shown) with local values across the warm pool often exceeding 60% during late austral spring and summer (see HG97). Hereafter, these two EOFs combined are taken to capture the MJO in convection [see also Zhang and Hendon (1997) and references therein].

Also shown in Fig. 4 is the correlation of intraseasonally filtered SST with the principal components of each EOF. For both EOFs, anomalous SST is approximately in quadrature with the OLR. Taking the two EOFs together indicates that warm SST lags anomalously high OLR (reduced convection) by about one-quarter cycle and that the SST anomalies accompany the OLR anomalies as they propagate eastward from the Indian Ocean out past the date line. The SST correlations have a similar horizontal scale as the OLR anomalies and are also shifted southward into the Southern Hemisphere in the western Pacific. This latter feature is consistent with the tendency for the MJO to produce the largest convective and circulation anomalies there during late austral spring and summer, for which the EOF analyses of OLR discriminates to (see also Salby and Hendon 1994; Zhang and Hendon 1997).

Individual occurrences of the MJO are identified by examining the amplitude of the two leading principal components. Figure 5 shows an example of the daily amplitude of the leading two principal components during the TOGA COARE Intensive Operating Period (IOP), when intraseasonal activity was prominent. At this time, the leading two EOFs capture in excess of 45% of the intraseasonal variance of OLR (see also HG97). The trajectory of PC-1 and PC-2 lies on a circle, which indicates systematic eastward propagation of a convective disturbance. Well-defined intraseasonal events, which exhibit systematic eastward propagation, are thus selected by the following criteria:

  1. The minimum of PC-1 exceeds −1.5 standard deviation and

  2. within 25 days after PC-1 exceeds −1.5 standard deviation, the minimum of PC-2 exceeds −1.5 standard deviation and PC-1 is positive.

The second criterion assures that some continuity of propagation exists with each event. We identify 10 events for the period 1986–93. The 120-day time series for each of these events are extracted. The central day of the time series is the day when the trajectory in the (PC-1, PC-2) plane crosses the negative PC-1 axis. Table 1 shows the middle dates of the 10 events, which is when enhanced convection is centered near the date line and suppressed convection covers the Indian Ocean (Fig. 4). Note that most of the events occur during late austral spring and summer. For each event, time is converted to phase defined by tan−1(−a1/a2), where a1 and a2 are the amplitude of PC-1 and PC-2 of the EOF analysis of the OLR data. Composites are formed by averaging into phase bins that span one-eighteenth cycle. Compositing according to phase bin is adopted to account for the varying period of the individual events. The final step is to remove the linear trend at each location from the composite in order to eliminate spurious longer period timescales, which might otherwise obscure intraseasonal variations. The maximum trend of the SST composite is 0.29°C cycle−1. No such trends are seen in composites of surface fluxes.

3. Comparison of surface fluxes

The accuracy of our estimates of the various surface fluxes is judged by comparison to estimates based on observations from the IMET mooring deployed at 1°45′S, 156°E from 22 October 1992 through 2 March 1993 for TOGA COARE (Weller and Anderson 1996). Our estimates from the grid point centered at 2.5°S, 155°E are used for the comparison. Three intraseasonal convective events traversed the western Pacific warm pool during the experiment (e.g., Gutzler et al. 1994). The active convective phase of these intraseasonal events passed over the IMET mooring during early November, late December, and late January.

Surface winds, air temperature, humidity, SST, rainfall, and downwelling shortwave and longwave radiation were measured on the IMET mooring [see Weller and Anderson (1996) for a complete description of the measurements]. Hourly surface fluxes of sensible and latent heat and surface stress were estimated from the winds, temperatures, and humidities using the COARE bulk flux algorithm (Fairall et al. 1996). Net surface insolation was estimated by assuming a constant ocean albedo of 0.055. Net longwave radiation was estimated by subtracting the measured downwelling longwave radiation from the sum of the estimated emitted surface longwave radiation and the reflected downwelling radiation (a graybody approximation was used such that the surface longwave albedo was taken to be one minus the surface emittance). The accuracy of these estimates is discussed in Weller and Anderson (1996). Daily averages were formed for the comparison with the estimates based on the gridded analyses.

Time series of the magnitude of the surface stress from the IMET and gridded analyses are shown in Fig. 6a. The magnitude of the daily wind stress is calculated from daily averages of the individual zonal and meridional stresses. Three episodes of increased stress are associated with the three active phases of intraseasonal convection during early November, late December, and late January. The December event was particularly strong. These intraseasonal variations (≈0.05 N m−2 amplitude) appear to be well captured in our estimate based on the gridded analyses. Figure 6b displays the two estimates of the latent heat flux. The increased winds during the active convective phases are seen to also be associated with enhanced latent heat flux. Again, the intraseasonal variations (≈35 W m−2 amplitude) appear to be faithfully depicted in our estimate based on gridded analyses.

Table 2 summarizes the means, correlations, and rms differences between the estimates based on the IMET mooring and the gridded analyses. The two daily stress estimates correlate at 0.81 and the two latent heat fluxes correlate at 0.69, suggesting that a significant portion of the daily variability is captured in our estimates. With weekly data, these correlations increase to 0.91 and 0.86, respectively, suggesting that the intraseasonal variations are faithfully captured in our estimate based on the ECMWF gridded analyses.

A major assumption of our estimation of latent heat flux is the use of Eq. (1), which estimates the surface specific humidity based on a constant relative humidity and the surface air temperature based on the SST. The use of a constant relative humidity is supported by Hendon and Leibman (1990), who diagnosed an amplitude of the relative humidity variation of about 1.5% during the MJO cycle at Darwin, Australia. Similarly, Zhang (1996) diagnosed an intraseasonal amplitude of about 2% using near-equatorial moorings in the western Pacific. A 1.5% change of relative humidity corresponds to about a 0.4 g kg−1 change of specific humidity if typical air temperature and pressure for the warm pool are used. This causes approximately 6 W m−2 latent heat flux change using the simple bulk formula (see also section 5), which is about 15%–20% of the amplitude on the intraseasonal variation (40 W m−2; see section 4). However, most of the variability of relative humidity appears to be caused by temperature (cooling due to convection), not by changing specific water vapor content (C. Zhang 1997, personal communication). In this case, the impact on our estimate of latent heat flux is smaller since the specific humidity does not change much.

On the other hand, the specific humidity calculated by Eq. (1) does not agree well with the IMET observation (the correlation coefficient is 0.38 for daily data). We also calculated the latent heat flux using the ECMWF wind, weekly SST analyses, and specific humidity from the IMET measurements (not shown). The correlation coefficient between this calculation and the estimate from the IMET observation (weekly mean) is 0.92, which is significantly higher than the value in Table 2, which uses Eq. (1). However, the latent heat flux estimate using the ECMWF specific humidity does not improve the correlation (the correlation coefficient between this calculation and IMET estimate is 0.76 for the weekly data). The above calculations do suggest that while the wind effect is dominant for intraseasonal variations of latent heat flux, accurate measurements of surface humidity would certainly further improve our estimates.

Prominent intraseasonal variations of net shortwave radiation are evident in both IMET observation and OLR-based estimates (Fig. 6c). Minima coincide with the active convective phases and maxima occur during the periods of weak winds. The intraseasonal variation of the shortwave radiation (≈50 W m−2 amplitude) appears well captured in the estimate based on OLR. The means are particularly similar (Table 2). The poorer correlation for daily data (0.66) than for weekly averaged data (0.79) may reflect the small-scale variability occurring at a point that is not coherent over a 2.5° grid box.

Daily net shortwave radiation estimated from the formula by Reed (1977) is also shown in Fig. 6c. Total cloudiness from ISCCP analyses for TOGA COARE (Rossow and Schifer 1991) was used in Reed’s empirical formula. Similar intraseasonal variations are also seen in this empirical estimate. However, the OLR-based estimate agrees better with the IMET observation (Table 2).

Daily net longwave radiation estimates are shown in Fig. 6d. Again, a prominent intraseasonal variation (≈10 W m−2 amplitude) is evident in both the estimates based on the IMET and gridded analyses. Minimum net longwave radiation occurs during the convectively active phases and maxima during the calm-clear phases. Daily estimates using the formula by Berliand and Berliand (1952) with ISCCP cloudiness are also shown in Fig. 6d. The intraseasonal variation is quite similar in all three estimates. Table 2 indicates that no accuracy is lost by estimating the net longwave radiation using cloudiness empirically estimated from OLR rather than from ISCCP.

Figure 6e shows the net surface heat flux estimated from the gridded data and the IMET. Again, a similar intraseasonal variation (≈75 W m−2 amplitude) is evident in both estimates with net cooling (≈−75 W m−2) during the convectively active phases and net warming (≈+75 W m−2) during the calm-suppressed phases. The IOP mean from the gridded analyses is slightly lower than the IMET estimate (+7 versus +18 W m−2), which stems from less longwave cooling and latent heat flux in the IMET estimates. Whether this difference is predominantly a sampling problem is not known.

Figure 6f shows the precipitation estimates from the MSU and IMET data. The IMET estimate is based on measurements from the mooring along with measurements by rain gauges on research vessels when they were in close proximity to the mooring (Weller and Anderson 1996). Both estimates depict a well-defined intraseasonal variation (≈40 mm day−1 amplitude). The mean from the MSU data is lower than that from the IMET (7.8 mm day−1 vs 11.2 mm day−1), but the good correlation (particularly for weekly averaged data, Table 2) indicates that the MSU estimates are probably reliable.

Figure 6g shows the gridded SST analyses and the SST from the IMET mooring observation. The maximum temperature of the daily IMET data during early December is about 30.5°C and decreases to a minimum temperature of about 28.5°C after the passage of the MJO event in late December. This approximately 2°C temperature variation during the MJO cycle is largely reduced by weekly averaging: the amplitude in weekly averaged IMET data is about 0.6°C, which is only slightly larger than that depicted by the gridded analyses. However, the warming during early January is not as well captured by the gridded weekly analyses. This warming could be a smaller-scale phenomenon, which cannot be resolved in the weekly gridded SST.

The ability of the gridded SST analyses to accurately capture intraseasonal variations in the warm pool is further demonstrated by considering the SST time series created by averaging observations from three TOGA TAO moorings (Hayes et al. 1991) along 156°E at 2°N, 0°, and 2°S (Fig. 7). As with the daily IMET observations, weekly averaging of the TAO time series results in much better agreement with the gridded analyses at 0°, 155°E, especially for the large intraseasonal variations in early October 1992 through early January 1993. Furthermore, the warming in early January observed at the IMET mooring is much smaller in the TAO data, which suggests that it may have been a relatively small-scale phenomenon. However, during February–March 1992, the SST from the TAO mooring is higher than the gridded SST, which is consistent with the comparison with the IMET mooring data (Fig. 6g). It seems that the gridded SST has some problems in this particular period. Nevertheless, the good agreement during the period of large intraseasonal events suggests that the weekly gridded analyses are reliable for this study.

In summary, the surface fluxes estimated from gridded analyses and OLR and the MSU precipitation capture the variability observed at the IMET mooring reasonably well. While the differences are greatest for daily data, the improved correlations and rms differences for weekly data (Table 2) suggest that the intraseasonal variations, which are the focus of this study, are realistically depicted in the estimates based on the gridded analyses. The differences at higher frequency may mostly reflect variability at small spatial scales that is not resolved in the 2.5° gridded data. The use of OLR to empirically predict shortwave radiation and total cloudiness (for subsequent use in the empirical estimate of longwave radiation) does not appear to introduce a significant error, particularly for the intraseasonal scales of interest here.

4. Composite evolution of intraseasonal variation of SST and surface fluxes

The evolution of surface fluxes and SST for the 10-event composite is now considered. The emphasis here is on the phasing, amplitude, and zonal propagation of the fluxes in concert with the convection and SST anomalies. Hence, we display the composite fluxes and SST as Hovmöller diagrams averaged between 2.5°S and 7.5°S, which is the latitude of maximum SST and OLR anomalies associated with the MJO across the warm pool during austral summer (Fig. 4; see also HG97). Insolation and latent heat flux, the dominant components of the surface heat budget, and the net surface heat flux (sum of sensible, latent, net shortwave and net longwave radiation) are displayed in Figs. 8a, 8b, and 8c, respectively. The evolution of OLR and hence convection can be deduced from that of insolation (Fig. 8a), with low OLR coinciding with low insolation and vice versa. Continuous eastward propagation of enhanced convection and decreased insolation is evident from the Indian Ocean to the date line, indicating that our strict selection of events for which to composite was successful in isolating eastward-propagating convective events. The phase speed of propagation can be estimated from the rate of change of location with respect to phase bin. The phase changes about 240° during the propagation from 75°E to 175°E, which for a near 50-day period indicates a speed of about 4 m s−1. Such a phase speed is typical of the MJO across the warm pool during austral summer (e.g., Rui and Wang 1990; Hendon and Salby 1994).

Amplitude of the insolation anomaly associated with the canonical MJO is about 25 W m−2. Amplitude of the latent heat flux is about 10 W m−2 in the Indian Ocean. The amplitude is larger in the western Pacific (about 40 W m−2). There, maximum latent heat flux is seen to lag minimum insolation (maximum convection) by about 20° (or 3 days for a typical 50-day period). The larger amplitude of the latent heat flux in the western Pacific, and the phase lag of the latent heat flux relative to the convective anomaly are consistent with previous analyses of the MJO (e.g., HG97).

Amplitude of the net surface heat flux variation is 30–70 W m−2. The net surface heat flux closely follows the evolution of insolation in the Indian Ocean, while a slight lag is evident in the western Pacific. There the relative amplitude of the latent heat flux anomaly is greater and hence shifts the net flux slightly behind the insolation.

The composite SST anomaly is shown in Fig. 8d. The amplitude of the intraseasonal variation is about 0.2°C in the Indian Ocean, 0.25°C in the western Pacific Ocean, and 0.35°C in the seas surrounding Indonesia. The spatial variation of the amplitude of the intraseasonal SST variation shown here is further discussed in a companion paper (Shinoda and Hendon 1998). Note that the MJO event during TOGA COARE (Fig. 6) is large compared to other events and hence the composite. The longitudinal scale (8000 km) and eastward phase propagation of the SST perturbations are similar to that of the net surface heat flux. However, a phase lag of about 90° is evident, consistent with the notion that the intraseasonal SST variations across the warm pool are driven by surface heat flux variations, which propagate systematically eastward along with the anomalous convection.

Also apparent in Fig. 8 is that the anomalous fluxes during the cloudy-windy phase of the MJO are more concentrated and of shorter duration than the fluxes during the calm-clear phase. This asymmetry is particularly apparent for the latent heat flux anomalies, such that the increased latent heat flux associated with anomalous surface westerlies is stronger but less extensive than the decreased latent heat flux associated with anomalous surface easterlies. Such longitudinal asymmetry of the surface zonal wind relative to the convective anomaly has also been previously reported (e.g., Hendon and Salby 1994).

The amplitudes and phase relationship of the SST and surface fluxes are examined in more detail at 165°E, representative of open ocean conditions in the western Pacific, and at 105°E, representative of the Indian Ocean. The SST tendency (composite of centered difference for each event), net surface heat flux, and its two dominant components (insolation and latent heat flux) are shown in Figs. 9a and 10a for 165°E and 105°E, respectively. At both locations, the SST tendency closely tracks the net surface heat flux. At 165°E the relative contribution by insolation and latent heat flux to the net heat flux are similar, while at 105°E the dominance of the insolation is evident. Insolation is also seen to lead the latent heat flux by about 20° (3 days) at 165°E, consistent with (but slightly less than) the analyses of Hendon and Glick (1997). At 105°E, this phase relationship is difficult to judge due to the smaller amplitude of the latent heat flux variation.

The magnitude of the wind stress and precipitation at each location are shown in Figs. 9b and 10b. At 165°E, the maximum precipitation coincides with minimum insolation (Fig. 9a). Strong winds (westerly) occur 3–15 days after the maximum precipitation, consistent with previous analyses of surface winds and OLR in the western Pacific (e.g., Hendon and Salby 1994). The amplitude of the precipitation variation is about 7 mm day−1, whereas that of stress is 0.025 N m−2. Similar phase lags are seen at 105°E; however, the amplitude of the stress fluctuations is weaker, which is consistent with the weaker latent heat flux variation there.

5. Effect of SST variation on surface fluxes

The composite analyses indicate that the typical swing in SST associated with the passage of the MJO is about 0.5°C. One important question is whether such an intraseasonal SST variation significantly affects the intraseasonally varying fluxes of moisture and heat, and, hence, possibly the convection associated with the MJO. The moisture flux due to the SST variation may significantly change the moisture content in the atmospheric boundary layer, and this may initiate the deep convection. The impact of a 0.5°C change in SST on the latent and sensible heat fluxes can be estimated using a standard bulk formula, assuming the surface humidity, wind, and air temperature remain constant as the SST increases. These changes are estimated as
i1520-0442-11-7-1685-eq3
where δQe is the latent heat flux change, δQh is the sensible heat flux change, ρa is the air density, cp is the specific heat of moist air, Ce and Ch are the exchange coefficients, L is the latent heat of vaporization, δqs is the saturation specific humidity change due just to the SST change, δTs is the SST change, and |U| is the wind speed. An SST change from 29.0°C to 29.5°C corresponds to a positive δqs of 0.76 g kg−1. For a wind speed equal to 4 m s−1, the latent heat flux increases by 11.5 W m−2, whereas the sensible heat flux increases by 3.2 W m−2. Here we have used Ce = Ch = 1.3 × 10−3, ρa = 1.2 kg m−3, cp = 1.02 × 10−3 J (kg °C)−1, and L = 2441 W s g−1. These increases are significant considering the amplitude of the composite sensible and latent heat flux variation associated with MJO is about 40 W m−2 in the western Pacific.

A more accurate estimate of the impact of the intraseasonally varying SST on the sensible and latent heat flux associated with the MJO is determined by recomputing these fluxes as in section 2 but using SST from which the intraseasonal variations have been removed. We do this by low-pass filtering the SST (via spectral transforms) to periods longer than 120 days. Hence, the seasonal cycle and longer timescales are retained, whereas the intraseasonal variations are removed. To isolate the impact of just the intraseasonally varying SST, the air temperature and specific humidity are calculated from the original weekly SST. Hence, only the sea surface temperature and saturation humidity are changed by using the low-pass-filtered SST.

The difference between the latent and sensible heat flux estimates from weekly SST and low-pass-filtered SST for the composite MJO are shown in Fig. 11. The amplitude of the composite latent and sensible heat flux variation increases uniformly by about 10 W m−2 when the low-pass-filtered SST is used in place of the weekly SST. Figure 12 shows the latent and sensible heat flux estimates using the low-pass-filtered SST and the weekly SST at 165°E. The intraseasonal SST variation tends to reduce the amplitude of the latent and sensible heat flux variation because the maximum intraseasonally varying wind speed, which lags the maximum convection by about 3 days, is shifted slightly toward the cold SST anomaly. Hence, the correlation between the intraseasonally varying wind speed and SST is negative. Thus, inclusion of the intraseasonally varying SST acts to reduce the amplitude of the latent and sensible heat fluxes over that computed with “climatological” SST. Although a 10 W m−2 change of the latent and sensible heat flux variation appears to be significant, the actual impact on the evolution of the MJO awaits appropriate modeling studies that faithfully account for this effect.

Longwave cooling from the surface also changes due to the intraseasonal SST variation. We again recompute the net longwave radiation as in section 2 but with the low-pass-filtered SST. The resulting change is to increase the net longwave anomalies by about 1–2 W m−2 (not shown).

Figure 13 shows the difference between the estimates of net surface heat flux using weekly SST and low-pass-filtered SST. The intraseasonal SST variation acts to reduce the amplitude of the net surface heat flux over that estimated using low-pass-filtered SST. This reduction is dominated by the latent heat flux difference. Whether this reduction in the net and latent heat fluxes impacts the evolution of the MJO awaits further study.

6. Conclusions

A composite study of SST and surface flux variations produced by the MJO was performed for the period 1986–93. The composite is based on EOF analysis of intraseasonally filtered OLR across the warm pool, where the signal of the MJO is greatest. The first two EOFs, which together capture the MJO in convection (e.g., Zhang and Hendon 1997), are used to determine the occurrence and phase of individual, well-defined events. Ten such events from 1986 to 1993 are selected from which the composite evolution of SST and surface fluxes is formed.

Surface fluxes of heat and momentum are estimated using ECMWF surface wind analyses, empirical estimates of temperature and humidity, and weekly SST analyses along with the TOGA COARE bulk flux algorithm. Longwave and shortwave radiation are also estimated empirically. These surface fluxes are compared with independent estimates from data collected on the IMET mooring, which was deployed during TOGA COARE (Weller and Anderson 1996). The intraseasonal variation of the surface fluxes computed from the gridded analyses and empirical estimates agree reasonably well with the estimates based on the IMET data.

The amplitude of the SST variation produced by the MJO in the western Pacific and Indian Ocean is about 0.25°C, while a maximum amplitude of about 0.35°C is found in the Indonesian region. Coherent surface flux anomalies are associated with the SST variations. The net surface heat flux anomaly is found to lead SST about one-quarter cycle, which is consistent with the notion that surface heat flux variations drive the SST variations on these timescales.

The magnitude of the surface flux variations produced by the MJO are quantified by the present analyses and are summarized schematically in Fig. 14. Typical mean values of the heat flux and SST anomalies in the western Pacific from the composite analyses have been included. Maximum SST is found between the cloudy and clear-sky regions. Latent heat flux is lowest in the clear-sky region since the wind is weak there and highest just after (≈7 days) the most convective region since the winds are strongest (most westerly) there. Maximum precipitation, which closely follows OLR, occurs just prior to and during the windiest conditions. The impact of the phasing of this large-scale variation of freshwater flux on the evolution of the thermal structure of the mixed layer is investigated further in the companion paper.

The net surface heat flux is most positive (60 W m−2 enters the ocean) in the clear-sky region and most negative (80 W m−2 exits the ocean) in the cloudy-convective region of the MJO. Insolation and latent heat flux make similar contributions, particularly over the western Pacific, to this variation of net surface heat flux over the life cycle of the MJO. Note that the negative net surface heat flux anomaly is slightly stronger and of shorter duration than the positive anomaly. The more extensive duration of the suppressed-calm phase (e.g., Fig. 8) results in a weak positive mean net surface heat flux (≈5 W m−2) despite the locally stronger negative anomaly. The entire complex of anomalous fluxes, convection, and SST propagates coherently eastward at approximately 4 m s−1 from the western Indian Ocean to the date line, where the convective disturbance diminishes.

The SST variation of 0.5°C associated with the MJO was estimated to reduce the amplitude of the latent and sensible heat fluxes produced by the MJO by about 15% from that if the SST were to remain constant. This suggests the possibility that the time-varying SST produced by the MJO may significantly affect the evolution of convection associated with the MJO, and, hence, may play an important role in determining the characteristics of the MJO. To fully examine this problem requires an atmospheric model capable of simulating convective processes at these intraseasonal scales, coupled to an active ocean mixed layer. Such a model is currently being developed.

Acknowledgments

Robert Weller kindly provided the data from the WHOI IMET mooring. The SRB data were obtained from the NASA/Langley Research Center EOSDIS Distributed Active Archive Center. Constructive comments by two reviewers are gratefully acknowledged. This work was supported by a TOGA COARE grant from NOAA’s Office of Global Programs.

REFERENCES

  • Berliand, M. E., and T. G. Berliand, 1952: Determining the net longwave radiation of the earth with consideration of the effect of cloudiness (in Russian). Izv. Akad. Nauk. SSSR Ser. Geofiz., No. 1.

  • Brunt, D., 1932: Notes on the radiation in the atmosphere. Quart. J. Roy. Meteor. Soc.,58, 389–420.

  • Budyko, M. I., 1974: Climate and Life. Academic Press, 508 pp.

  • Fairall, C., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young, 1996: The TOGA COARE bulk flux algorithm. J. Geophys. Res.,101, 3747–3764.

  • Godfrey, J. S., M. Nunez, E. F. Bradley, P. A. Coppin, and E. J. Lindstrom, 1991: On the net surface flux into the western equatorial Pacific. J. Geophys. Res.,96, 3391–3400.

  • Gruber, A., and A. F. Krueger, 1984: The status of the NOAA outgoing longwave radiation data set. Bull. Amer. Meteor. Soc.,65, 958–962.

  • Gutzler, D. S., G. N. Kiladis, G. A. Meehl, K. M. Weickmann, and M. Wheeler, 1994: The global climate of December 1992–February 1993. Part II: Large-scale variability across the tropical western Pacific during TOGA COARE. J. Climate,7, 1606–1622.

  • Hayes, S. P., L. J. Mangum, J. Picaut, A. Sumi, and K. Takeuchi, 1991: TOGA-TAO: A moored array for real-time measurements in the tropical Pacific Ocean. Bull. Amer. Meteor. Soc.,72, 339–347.

  • Hendon, H. H., and B. Liebmann, 1990: The intraseasonal (30–50 day) oscillation of the Australian summer monsoon. J. Atmos. Sci.,47, 2909–2923.

  • ——, and M. L. Salby, 1994: The life cycle of the Madden–Julian oscillation. J. Atmos. Sci.,51, 2225–2237.

  • ——, and J. Glick, 1997: Intraseasonal air–sea interaction in the tropical Indian and Pacific Oceans. J. Climate,10, 647–661.

  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Amer. Meteor. Soc.,77, 437–471.

  • Knutson, T. R., and K. M. Weickmann, 1987: 30–60 day atmospheric oscillations: Composite life cycles of convection and circulation anomalies. Mon. Wea. Rev.,115, 1407–1436.

  • Krishnamurti, T. N., D. K. Osterhof, and A. V. Mehta, 1988: Air–sea interaction on the timescale of 30 to 50 days. J. Atmos. Sci.,45, 1304–1322.

  • Lau, K.-M., and C.-H. Sui, 1997: Mechanisms of short-term sea surface temperature regulation: Observations from TOGA COARE. J. Climate,10, 465–472.

  • Li, Z., C. H. Whitlock, and T. P. Charlock, 1995: Assessment of the global monthly mean surface insolation estimate from satellite measurements using global energy balance archive data. J. Climate,8, 315–328.

  • Madden, R. A., and P. R. Julian, 1971: Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. J. Atmos. Sci.,28, 702–708.

  • ——, and ——, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Atmos. Sci.,29, 1109–1123.

  • Pinker, R., and I. Laszlo, 1992: Modeling of surface solar irradiance for satellite applications on a global scale. J. Appl. Meteor.,31, 194–211.

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

  • Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate,7, 929–948.

  • Rossow, W. B., and R. A. Schiffer, 1991: ISCCP cloud data products. Bull. Amer. Meteor. Soc.,72, 2–20.

  • Rui, H., and B. Wang, 1990: Developement characteristics and dynamic structure of tropical intraseasonal convection anomalies. J. Atmos. Sci.,47, 357–379.

  • Salby, M. L., and H. H. Hendon, 1994: Intraseasonal behavior of clouds, temperature, and motion in the Tropics. J. Atmos. Sci.,51, 2207–2224.

  • Shinoda, T., and H. H. Hendon, 1998: Mixed layer modeling of intraseasonal variability in the tropical western Pacific and Indian Oceans. J. Climate, in press.

  • Spencer, R. W., 1993: Global oceanic precipitation from the MSU during 1979–91 and comparisons to other climatologies. J. Climate,6, 1301–1326.

  • Trenberth, K. E., 1992: Global analyses from ECMWF and atlas of 1000 to 10 mb circulation statistics. NCAR Tech. Note NCAR/TN-373+STR, 191 pp. [Available from National Center for Atmospheric Research, Boulder, CO 80307.].

  • Waliser, D. E., and N. E. Graham, 1993: Convective cloud systems and warm-pool sea surface temperatures: Coupled interactions and self-regulation. J. Geophys. Res.,98 (D7), 12 881–12 893.

  • Webster, P. J., and R. Lukas, 1992: TOGA COARE: Coupled Ocean–Atmosphere Response Experiment. Bull. Amer. Meteor. Soc.,73, 1377–1416.

  • Weller, R. A., and S. P. Anderson, 1996: Surface meteorology and air–sea fluxes in the western equatorial Pacific warm pool during the TOGA Coupled Ocean–Atmosphere Response Experiment. J. Climate,9, 1959–1990.

  • Whitlock, C. H., and Coauthors, 1995: First Global WCRP Shortwave Surface Radiation Budget Dataset. Bull. Amer. Meteor. Soc.,76, 905–922.

  • Zhang, C., 1996: Atmospheric intraseasonal variability at the surface in the tropical western Pacific Ocean. J. Atmos. Sci.,53, 739–758.

  • ——, and H. H. Hendon, 1997: On propagating and standing components of the intraseasonal oscillation in tropical convection. J. Atmos. Sci.,54, 741–752.

  • Zhang, G. J., and M. J. McPhaden, 1995: The relationship between sea surface temperature and latent heat flux in the equatorial Pacific. J. Climate,8, 589–605.

Fig. 1.
Fig. 1.

Point-by-point correlation coefficient between daily OLR and shortwave radiation from SRB data during 1985–88. The rectangular area in the map indicates (10°N–10°S, 60°E–180°). The contour interval is 0.1. Mean OLR less than 250 W m−2 is shaded (interval 10 W m−2).

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 2.
Fig. 2.

(a) Same as Fig. 1 except correlation coefficient between SRB and shortwave radiation calculated from empirical formula by Reed (1977). Shading shows mean total cloudiness from the ISSCP C2 data for the period 1985–88 (shading intervals at 55%, 65%, and 75%). (b) Same as in (a) except for the correlation coefficient between SRB and net shortwave radiation from NCEP reanalyses. Shading shows net shortwave radiation from the NCEP data for the period 1985–88 (shading intervals at 250, 240, and 230 W m−2).

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 3.
Fig. 3.

Relation between daily OLR and percent cloud from ISCCP data during 1986–90 for the area (10°N–10°S, 60°E–180°). The circles indicate means of percent cloud in 1 W m−2 bins of OLR. The dotted line indicates plus and minus one standard deviation of total cloud in each bin. The dashed line indicates the number of observations in each bin. The thick line is a third-order polynomial obtained by least squares fit to the binned data.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 4.
Fig. 4.

First (a) and second (b) eigenvectors of intraseasonally filtered OLR data for the domain (15°N–15°S, 60°E–90°W). The eigenvectors have been scaled for a one standard deviation anomaly of each principal component (contour interval 3 W m−2). Shading indicates correlation coefficient between intraseasonally filtered SST and the respective principal components of OLR (shading interval is 0.1 with first level at 0.2).

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 5.
Fig. 5.

Daily amplitude, during TOGA COARE, 22 October 1992 through 2 March 1993, of first (horizontal axis) and second (vertical axis) principal components of the EOF analysis of OLR. Numbers on the right side of the points indicate the month, day, and year. The circle indicates the first date (22 October).

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 6.
Fig. 6.

Time series of (a) wind stress, (b) latent heat flux, (c) shortwave radiation, (d) longwave radiation, (e) net surface heat flux, (f) precipitation, and (g) SST for TOGA COARE 22 October 1992 through 2 March 1993. Dashed lines in (a)–(f) indicate daily mean flux estimates from the IMET mooring observations (Weller and Anderson 1996) sited at 1°45′S, 156°E. The dashed line in (g) indicates the weekly mean SST from the IMET mooring observations. Thick lines indicate estimates from gridded data centered at 2.5°S, 155°E. The dotted line in (c) indicates shortwave flux estimated from the formula by Reed (1977). The dotted line in (d) indicates net longwave flux estimated from the Berliand and Berliand (1952) formula using ISCCP total cloudiness. The dotted line in (g) indicates the daily mean SST from the IMET mooring observations.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 7.
Fig. 7.

Time series of SST from the gridded analyses and from the TAO mooring observations. Data from three TAO moorings along 156°E at 2°N, 0°, and 2°S are averaged. The thick line indicates the gridded SST analyses at 0°, 155°E. The dashed and thin lines indicate the daily mean and weekly mean TAO data, respectively.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 8.
Fig. 8.

Composite (a) shortwave radiation anomaly, (b) latent heat flux anomaly, (c) net surface heat flux anomaly, and (d) SST anomaly for 10 MJO events along 5°S. The vertical axis indicates the phase based on the EOFs of OLR.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 9.
Fig. 9.

Composite (a) SST tendency, net surface heat flux, shortwave radiation, and latent heat flux; and (b) wind stress and precipitation for 10 MJO events at 2.5°–7.5°S, 160°–170°E.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 10.
Fig. 10.

Composite (a) SST tendency, net surface flux, shortwave radiation, and latent heat flux; and (b) wind stress and precipitation for 10 MJO events at 2.5°–7.5°S, 100°–110°E.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 11.
Fig. 11.

Composite difference between sum of latent and sensible heat flux estimates based on weekly SST and low-pass-filtered SST along 5°S. Vertical axis is as in Fig. 7.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 12.
Fig. 12.

Composite latent and sensible heat flux based on weekly SST (solid line) and low-pass-filtered SST (dashed line) at 2.5°–7.5°S, 160°–170°E (upper panel). The difference between the two composites is shown in the lower panel.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 13.
Fig. 13.

Composite difference along 5°S between net surface heat flux based on weekly SST and low-pass-filtered SST. Vertical axis is as in Fig. 7.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Fig. 14.
Fig. 14.

Schematic diagram showing magnitude and phase relationship relative to the convective anomaly of the surface fluxes and SST variations produced by the canonical MJO. The asymmetric zonal scale of the cloudy-windy and suppressed-calm phases and eastward phase speed (4 m s−1) of the joint atmosphere–ocean disturbance across the warm pool are indicated. Typical extrema of surface fluxes and SST over life cycle of MJO are shown for western Pacific.

Citation: Journal of Climate 11, 7; 10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2

Table 1.

Dates of 10 intraseasonal events (time of minimum of PC-1).

Table 1.
Table 2.

Mean, rms difference, and correlation coefficients between surface fluxes estimated from IMET mooring data by Weller and Anderson (1996) and gridded data. Values obtained from daily mean data and weekly mean data are shown. Values obtained from the shortwave radiation calculation by Reed’s formula are shown in parentheses. Values obtained from the longwave radiation calculation using ISCCP cloud data are shown in parentheses.

Table 2.
Save