1. Introduction and motivation
a. The northward-propagating intraseasonal oscillation of the Indian monsoon
Intraseasonal variability (ISV) in summer monsoon rainfall affects the lives and livelihoods of hundreds of millions of people on the Indian subcontinent each season [June–September (JJAS); all seasons are for the Northern Hemisphere]. These variations span temporal scales, but the spectrum of daily monsoon rainfall shows pronounced peaks at 10–20 and 30–50 days (Annamalai and Slingo 2001). This study is concerned with the latter, which is associated with a northward-propagating intraseasonal oscillation (NPISO) between “active” and “break” conditions of enhanced and reduced rainfall, respectively, over India (e.g., Hartmann and Michelsen 1989; Lawrence and Webster 2002).
Active and break events form in the equatorial Indian Ocean and then move north toward India at approximately 1° latitude day−1 (Gadgil 1990; Webster and Hoyos 2004). They show a large-scale organized structure in fields such as 850-hPa zonal wind and outgoing longwave radiation (OLR; Annamalai et al. 1999). OLR over India is strongly anticorrelated with OLR over the eastern equatorial Indian Ocean: active (break) phases over India lead oceanic break (active) phases by approximately six days (Klingaman et al. 2008b). Using reanalysis data, Kemball-Cook and Wang (2001) and Lawrence and Webster (2002) found that approximately three-quarters of the NPISO events propagated northward from and eastward along the equator, suggesting that the NPISO was the summer manifestation of the Madden–Julian oscillation (MJO; Madden and Julian 1971).
In situ and remotely sensed observations in the Indian Ocean have demonstrated that active–break cycles were associated with sea surface temperature (SST) variations of around 1°C (e.g., Bhat et al. 2001; Webster et al. 2002; Joseph and Sabin 2008; Klingaman et al. 2008b). Rainfall and SST anomalies are out of phase: warm (cold) SSTs precede (follow) heavy rainfall by 7–10 days (Woolnough et al. 2001; Vecchi and Harrison 2002; Fu et al. 2003).
The magnitude of daily SST variations may be critical to general circulation model (GCM) representations of the MJO and the NPISO. Kim et al. (2008) and Klingaman et al. (2008a) each forced two ensembles of atmospheric GCM (AGCM) simulations with high-resolution-observed SSTs: one with daily-mean and the other with monthly-mean SSTs. The former (latter) study focused on winter (summer) ISV. The AGCMs displayed weak ISV in convection in the monthly SST case, while the daily SST forcing produced ISV consistent with observations.
b. The effect of air–sea coupling on organized variability in tropical convection
Many studies using coupled atmosphere–ocean GCMs (AOGCMs) have found that coupling improved the magnitude and propagation of the MJO and NPISO over AGCMs alone (e.g., Waliser et al. 1999; Inness and Slingo 2003; Rajendran and Kitoh 2006; Fu et al. 2007; Woolnough et al. 2007; Wang et al. 2009; Kim et al. 2010). Fu et al. (2007) and Wang et al. (2009) concluded that air–sea coupling extended NPISO predictability by one week. AOGCMs have also correctly simulated the near-quadrature SST–rainfall phase relationship (e.g., Fu and Wang 2004), unlike AGCMs, which typically collocate the strongest convection with the warmest SSTs (e.g., Rajendran and Kitoh 2006; Klingaman et al. 2008a).
AOGCMs still contain considerable biases in their representations of the MJO and the NPISO. Nearly all models from the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4) failed to simulate the observed amplitude or propagation speed of the intraseasonal oscillation in winter or summer (Lin et al. 2006, 2008). The seven AOGCMs from the Development of a European Multimodel Ensemble System for Seasonal-to-Interannual Prediction project also generated deficient summertime ISV in convection and SSTs (Xavier et al. 2008).
The coarse spatial and temporal resolutions of air–sea interactions in AOGCMs may limit SST ISV (e.g., Bernie et al. 2005, 2007; Woolnough et al. 2007). Bernie et al. (2005) concluded that a 1-m near-surface ocean vertical resolution and a 3-h air–sea coupling frequency were necessary to resolve 90% (95%) of the diurnal (intraseasonal) SST amplitude in the west Pacific; yet most contemporary AOGCMs use 10-m resolution and 24-h coupling. Neglecting the diurnal cycle of surface forcing can reduce the intraseasonal SST response to the MJO by 20% (Bernie et al. 2007).
Woolnough et al. (2007) coupled the K-Profile Parameterization (KPP) ocean boundary layer model within the European Centre for Medium-Range Weather Forecasts (ECMWF) monthly forecasting system to conduct MJO predictability experiments. Simulations with 1-m ocean vertical resolution and 3-h coupling showed improved forecast skill against simulations with 10-m resolution and 24-h coupling, as well as against forecasts using the three-dimensional Hamburg Ocean Primitive Equation ocean GCM with 10-m resolution and 24-h coupling.
The benefits of well-resolved air–sea interactions in AOGCMs likely extend beyond the MJO. Bernie et al. (2008) demonstrated that subdaily coupling improved the seasonal cycle of the Pacific Ocean: diurnal warming rectified onto the seasonal mean equatorial Pacific SST, driving stronger trade winds and increased summer equatorial upwelling. Ham et al. (2010) found that 2-h coupling in the Seoul National University AOGCM reduced tropical SST biases and improved the amplitude of the El Niño–Southern Oscillation.
In AOGCMs with full ocean dynamics, the computational cost of improving the spatial and temporal resolutions of air–sea interactions may be prohibitively expensive, because of additional ocean vertical grid points, a shorter ocean time step, and more frequent coupling. Ocean thermodynamics—not dynamics—cause the majority of subseasonal tropical upper-ocean variability (e.g., Waliser et al. 2004). Thus, replacing the three-dimensional ocean GCM with many columns of a one-dimensional ocean boundary layer model would reduce the cost of upper-ocean resolution while retaining ISV, enabling large-ensemble process studies of the intraseasonal oscillation. This study introduces and employs such an AOGCM: HadKPP.
c. The purpose of the present study
The authors use HadKPP to investigate the impact on the simulated NPISO of refined resolution in air–sea coupled processes and ocean vertical mixing. Four ensembles of HadKPP integrations are conducted to test the sensitivity of the modeled NPISO intensity and propagation to upper-ocean vertical resolution and air–sea coupling frequency, combined and separately. This study complements and extends previous studies of the effect on the winter MJO of varying these parameters in other AOGCMs (e.g., Woolnough et al. 2007; Bernie et al. 2008). This study also supplements past work on the importance of high-frequency, high-resolution SSTs simulating ISV (e.g., Kim et al. 2008; Klingaman et al. 2008a).
Section 2 introduces HadKPP, the depth-varying heat-flux-correction technique used to constrain mean SSTs, and the four ensembles. Section 3 describes the observational datasets used in this study, the Lanczos bandpass filter used to compute ISV, and the technique used to construct composite NPISO active events. In section 4, the HadKPP ensembles are compared to determine whether finely resolved atmosphere–ocean interactions affect the simulated NPISO. The authors discuss and summarize their key conclusions in section 5.
2. Model description and experiment design
a. The HadKPP
HadKPP comprises the Hadley Centre AGCM [the Hadley Centre Atmospheric Model, version 3 (HadAM3); Pope et al. 2000] coupled to many columns of the KPP ocean boundary layer model (Large et al. 1994) via the Ocean–Atmosphere–Sea Ice–Soil (OASIS; Terray et al. 1995) coupler. The chief objective in developing HadKPP was to produce an AOGCM with an improved representation of ocean vertical mixing at reduced computational cost, to enable process studies of phenomena for which subdaily atmosphere–ocean coupling is believed to be important (e.g., the MJO). KPP contains no ocean dynamics—only vertical mixing—and simulates only the top 200 m of the ocean. Depth-varying heat-flux corrections (section 2b) give HadKPP a more accurate basic-state SST than most AOGCMs, which can exhibit large biases and drift in ocean temperatures. Inness et al. (2003) demonstrated that flux corrections in the third climate configuration of the Met Office Unified Model (HadCM3) considerably improved MJO strength and propagation.
Within HadKPP, HadAM3 is configured as in Pope et al. (2000), except that the horizontal, vertical, and temporal resolutions are increased to 1.25° longitude × 0.83° latitude, 30 levels, and 10 min, respectively, as in Klingaman et al. (2008a). Stratton (1999) found that this horizontal resolution improved ISV in HadAM2b against a 3.75° × 2.5° resolution. Inness et al. (2001) showed that refining HadAM3 from 19 to 30 levels greatly improved MJO amplitude.
One KPP column runs underneath each HadAM3 horizontal grid point. Vertical diffusion in KPP is based on a nonlocal scheme, to represent the transport of heat and salt by eddies equal to the depth of the mixed layer and to account for countergradient transports under unstable forcing (Large et al. 1994). The KPP vertical grid is stretched, with more levels close to the ocean surface; the vertical resolution varies among the ensembles (section 2c).
HadAM3 and KPP are coupled within 30°S–30°N and 20°E–180°. Outside this region, HadAM3 is forced by daily climatological (2002–08) SSTs from the Forecast Ocean Assimilation Model analyses (FOAM; section 3a). At the coupling-region boundaries, a five-gridpoint linear interpolation blends the HadKPP and FOAM SSTs to remove sharp discontinuities.
b. Heat-flux corrections in HadKPP
Heat-flux corrections are required to correct for the lack of ocean advection in KPP and biases in HadAM3 surface forcing. This section introduces a depth-varying heat-correction technique, based on separate integrations of the atmospheric and oceanic components of the coupled model (Sausen et al. 1988). The technique removes the monthly-mean ocean temperature bias throughout the vertical, relative to the 2002–08 FOAM climatology (section 3a).
The 10-member KPP ensemble is then reintegrated for the same month, applying the corrections as per (2). The ensemble then continues through the next month without correction, to obtain the drift in ocean temperatures for that month. The process is repeated for each month of the year or for each month of the season of interest. The term
Klingaman (2008) compared this correction method to two other techniques: one in which depth-varying corrections were computed from uncorrected coupled HadKPP simulations and the other in which a surface-only correction was calculated from uncorrected HadKPP simulations, based on the SST bias and the ocean mixed layer depth. The depth-varying technique using coupled simulations underestimated the magnitude of the correction required, because of coupled feedbacks that damped ocean temperature biases in the uncorrected integrations. This highlighted the need to drive the ocean model with atmospheric fluxes that were correct for the SST climatology to which the ocean model was constrained.
The surface-only technique produced an erroneous shoaling of the ocean mixed layer depth where the corrections warmed the ocean, because of the large input of heat into the shallow 1-m KPP top layer. The interaction of the shoaling mixed layer and the warming corrections generated a positive feedback that resulted in a severe overcorrection of cold biases. Klingaman (2008) concluded that depth-varying corrections were required for ocean models with fine vertical resolution. Across the coupling domain, the technique described in (1) and (2) removed 90% of the monthly-mean mixed layer temperature bias; the depth-varying and surface-only techniques using coupled integrations removed 30% and 20%, respectively.
c. Ensemble configurations
Four 30-member HadKPP ensembles are conducted to assess the effect on the NPISO of subdaily coupling and improved upper-ocean mixing (Table 1). The four ensembles use the same set of 30 1 May atmospheric initial conditions, restarted from a 30-member, 1-yr HadAM3 ensemble driven by FOAM climatological SSTs. In all members, KPP is initialized with the climatological 1 May FOAM ocean temperature and salinity. As the mixed layer ocean is essentially a slave to the atmosphere, this should not affect intraensemble variability. All simulations ran for 6 months, May–October. May is discarded as spinup; October is included only to minimize the edge effects of the Lanczos filter (section 3e) on the monsoon season. Only JJAS is analyzed further.
A summary of the four ensembles conducted in this study, giving the name of the ensemble, the number of points in the vertical in the KPP boundary layer ocean model, the depth of the near-surface layer in KPP, and the atmosphere–ocean coupling frequency.
In “10M-24H,” HadKPP is configured similarly to most contemporary AOGCMs, with 10-m near-surface ocean vertical resolution and 24-h coupling frequency. This is the control integration for the sensitivity experiments. In 10M-3H, the coupling frequency is shortened to 3 h; however, the near-surface layer remains at 10 m. In 1M-24H, the near-surface ocean resolution is refined to 1 m; however, the coupling frequency remains at 24 h. In 1M-3H, both the vertical resolution and the coupling frequency are enhanced.
1M-3H and 1M-24H have finer resolution than 10M-3H and 10M-24H throughout the 200-m KPP domain, not only near the surface (Table 1). Bernie (2006) found that, given a 1-m-deep near-surface layer, some refinement to a 10-m resolution was necessary below the near-surface layer to capture SST ISV; however, a 1-m spacing was not required. We have elected to keep the fine resolution below the surface in the 1M ensembles, as HadAM3—not KPP—limits the computational efficiency of HadKPP.
Separate monthly-mean, ensemble mean, depth-varying heat corrections (section 2b) were computed for each ensemble, because the ocean vertical resolution and the coupling frequency affect the KPP temperature biases.
3. Datasets and methods
Unless noted otherwise, all datasets were interpolated from their native resolutions to the HadKPP horizontal grid using an area-weighted method, prior to any analysis.
a. FOAM ocean temperature analyses
Daily, three-dimensional, global, 1° × 1° FOAM ocean temperature analyses for 2002–08 were obtained from the U.K. National Centre for Ocean Forecasting. These are used to compute heat-flux corrections (section 2b), as well as to drive HadAM3 outside the air–sea coupling region. FOAM assimilates SSTs from the National Oceanic and Atmospheric Administration–National Aeronautics and Space Administration Advanced Very High Resolution Radiometer, as well as SSTs and subsurface temperatures from ships, Argo profiling floats, and Tropical Atmosphere Ocean–Triangle Trans-Ocean Buoy Network (TRITON) moorings (Bell et al. 2000).
b. TMI sea surface temperatures
ISV in SSTs from HadKPP is compared against that from the Tropical Rainfall Measuring Mission Microwave Imager. Unlike infrared instruments, TMI allows reliable readings through clouds (Wentz 1998, 2000), which is particularly important during the summer monsoon. Tropical ISV in TMI SSTs has been shown to agree closely with observations, while infrared-only products typically contain smaller ISV (Harrison and Vecchi 2001; Senan et al. 2001). Daily, microwave-based, optimally interpolated 0.25° × 0.25° analyses for 1998–2008 were obtained from Remote Sensing Systems (RSS). Since 2002, RSS has blended TMI with the Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) on the National Aeronautics and Space Administration’s Aqua satellite, using the optimal interpolation technique of Reynolds and Smith (1994).
c. GPCP rainfall
Daily 1° × 1° precipitation analyses for 1997–2008 from the Global Precipitation Climatology Project (GPCP; Adler et al. 2003) are employed to evaluate rainfall ISV in HadKPP. The GPCP analyses combine data from surface rain gauges and satellite-based infrared and microwave sounders. At this resolution, ISV in GPCP rainfall in the monsoon domain is similar to that from the Tropical Rainfall Measuring Mission analyses (not shown).
d. HadDTR diurnal SST amplitude
The mean diurnal amplitude of HadKPP SSTs is compared to the monthly-mean 5° × 5° 1990–2008 Hadley Centre diurnal temperature range (HadDTR) climatology of drifting buoy measurements. HadDTR compared well to the Stuart-Menteth et al. (2005) diurnal SST parameterization based on moored buoy data (Kennedy et al. 2007). HadKPP SSTs were interpolated onto the HadDTR grid prior to computing the diurnal amplitude (section 4b).
e. Lanczos bandpass filtering
A 30–50-day Lanczos bandpass filter with 101 weights is used to isolate rainfall and SST signals associated with the NPISO. This filter is narrower than those used for the wintertime MJO (e.g., Jones et al. 1998; Matthews et al. 1999; Gustafson and Weare 2004), to avoid including other modes of variability that occur during the summer monsoon [e.g., the 10–20-day westward-propagating mode (Annamalai and Slingo 2001)].
f. NPISO composites
Composite NPISO active events are constructed for HadKPP at the point 12.5°N, 85°E, in the southern Bay of Bengal. This point was chosen for consistency with the Lin et al. (2008) lag-correlation diagnostic used to evaluate northward propagation in section 4f. While compositing at a single point might not be representative of the entire domain, taking the area average over many points would likely result in an overly smooth composite that would not clearly show the daily and subdaily variability of interest to this study.
Active events are identified by finding, in each ensemble member, the two largest local maxima in the 30–50-day bandpass-filtered rainfall time series at 12.5°N, 85°E. The local maxima must be separated by at least 30 days to ensure that they are not from the same active event. Anomalies in atmospheric and oceanic fields for the 20 days before and after each local maximum are calculated by removing the daily ensemble mean from the full time series. The composite active event for each ensemble is the mean of all 60 active events, with each event centered on the local maximum in filtered rainfall.
4. Results
a. JJAS mean SST and rainfall
When compared to the FOAM climatological JJAS mean SST (Fig. 1a), 1M-3H shows biases of less than 0.2°C across the coupling domain (Fig. 1b). Klingaman (2008) showed that a similar HadKPP ensemble without flux corrections produced SST biases of up to ±4°C. The flux-correction technique effectively removes mean SST errors due to the lack of ocean advection and biases in HadAM3 surface forcing. SST biases for 10M-3H, 1M-24H, and 10M-24H are similarly small but are not shown for brevity.
The JJAS means of SST (°C): (a) the FOAM climatology (2002–08) and (b) 1M-3H ensemble mean minus the FOAM climatology. Contour intervals are every (a) 0.5° from 22°C. and (b) 0.1°C. In (b), positive (negative) values are shaded dark gray (light gray); the 0 contour is not shown.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
Consistent with GPCP (Fig. 2a), 10M-24H produces maxima in JJAS mean rainfall along the Western Ghats of India and in the northern Bay of Bengal (Fig. 2b). There are wet biases in the eastern Bay of Bengal, central India, and the eastern Arabian Sea (Fig. 2c). Differences against 10M-24H are small in 10M-3H (Fig. 2d), 1M-24H (Fig. 2e), and 1M-3H (Fig. 2f). The flux corrections ensure that the seasonal mean SSTs are identical among the ensembles, so changes in rainfall result from changes in atmosphere–ocean coupling strength. These differences are at most ±2 mm day−1, emphasizing that the ensembles have consistent mean states.
The JJAS means of precipitation (mm day−1): (a) the GPCP climatology (1997–2008), (b) 10M-24H ensemble mean, (c) 10M-24H minus GPCP, (d) 10M-3H minus 10M-24H, (e) 1M-24H minus 10M-24H, and (f) 1M-3H minus 10M-24H. Contour intervals are (c) every 2 until ±8, then every 4 mm day−1; (d)–(f) every 0.5 mm day−1. In (c)–(f), positive (negative) values are shaded dark gray (light gray); the 0 contour is not shown.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
b. Diurnal SST amplitude
Diurnal SST amplitudes in 1M-3H and 10M-3H are calculated from instantaneous SSTs at the 3-h coupling time steps. Sliding windows 24 h long are used at each grid point to compute eight diurnal amplitude values per day, defined as the maximum SST in the window minus the minimum. 1M-24H and 10M-24H have no diurnal amplitude in SST by definition.
HadDTR JJAS mean diurnal SST amplitudes are greatest in the Red Sea, at the head of the Bay of Bengal, near the Maritime Continent and in the South China Sea (Fig. 3a). 1M-3H agrees qualitatively with HadDTR in the locations of maxima and minima (Fig. 3b), except that 1M-3H is much smaller in the Bay of Bengal (Fig. 3c). This may be due to the absence in HadKPP of thin, highly stable freshwater barrier layers that form from river outflow during the monsoon season (e.g., Vinayachandran et al. 2002). While HadAM3 does parameterize river outflow, the outflow is not input to KPP, since KPP cannot transport the outflow away from the coast.
The ensemble means and JJAS means: (a) the amplitude of the diurnal cycle of SST from the HadDTR climatology (1990–2007) on its 5° horizontal grid; (b) the amplitude of the diurnal cycle of SST from 1M-3H, on the HadKPP grid; (c) the difference in the amplitude of the diurnal cycle of SST for 1M-3H minus HadDTR, with both on the HadDTR 5° grid; (d) as in (b), but for 10M-24H. In (c), the absolute value of the difference is shaded; negative values (i.e., where HadDTR is greater than HadKPP) are indicated with a cross.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
10M-3H (Fig. 3d) has approximately 20% of the diurnal amplitude of 1M-3H (Fig. 3b), revealing the importance of fine near-surface oceanic vertical resolution for capturing subdaily SST variability. Subdaily coupling is necessary but not sufficient to resolve diurnal SST fluctuations.
c. Rainfall–SST phase relationship
To assess the rainfall–SST phase relationship described in section 1b, the 11-day centered linear trends in longitude-averaged (75°–95°E, ocean points only) JJAS SSTs and rainfall are cross correlated at each latitude, as in Klingaman et al. (2008a). Lag correlations are computed separately for each season (ensemble member) in observations (HadKPP) and then averaged. The linear trend acts as a low-pass filter; the correlation between GPCP and TMI (Fig. 4a) is similar to that of Fu and Wang (2004), who used bandpass-filtered rainfall and SSTs (their Fig. 4).
Lead–lag correlations over JJAS between the 11-day centered linear trends in rainfall and SSTs that were first longitude averaged between 75° and 95°E for (a) GPCP rainfall and TMI SSTs for 1998–2008, (b) 1M-3H, (c) 10M-24H, and (d) a 30-member ensemble of HadAM3 integrations driven by daily observed SSTs from Klingaman et al. (2008a). Gray shading indicates correlations that are statistically significant at the 95% confidence level. On the horizontal axes, negative values indicate SST leading rainfall; positive values indicate SST lagging rainfall. Contour interval is 0.1, with negative contours dashed and the 0 contour drawn doubly thick.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
The 1M-3H ensemble produces a phase lag of 2–10 days at all latitudes (Fig. 4b), about two days shorter than the observed 4–12-day lag. The shorter lag may be the result of (a) an atmospheric convective parameterization that is overly sensitive to SST perturbations (e.g., Tompkins 2001) or (b) the inability of KPP to dissipate SST anomalies through horizontal advection, allowing gridpoint-scale anomalies to build too quickly. Spencer and Slingo (2003) concluded that HadAM3 produced precipitation too quickly in response to warm SST anomalies, which supports the first hypothesis.
The rainfall–SST correlation is weaker for the 30 1M-3H ensemble members than for the 11 yr of observations, although still statistically significant at the 5% level. A separate analysis of 11 randomly selected 1M-3H ensemble members did not increase the correlation magnitude (not shown), which suggests that the weaker magnitude in 1M-3H is due to model error, not ensemble size.
The 1M-3H phase relationship is similar to 10M-24H (Fig. 4c), and to 1M-24H and 10M-3H (not shown). By contrast, the 30-member ensemble of HadAM3 integrations forced by daily observed SSTs from Klingaman et al. (2008a) exhibited a coincident relationship (Fig. 4d). It is the presence of air–sea coupling and not its strength that controls the lagged SST–rainfall phase relationship. The relationship in HadKPP is also similar to the last 30 yr of a 100-yr HadCM3 preindustrial control simulation from Turner et al. (2005) (not shown). HadCM3 includes ocean dynamics and was configured similarly to 10M-24H. Ocean dynamics are not necessary to simulate this phase relationship in an AOGCM.
d. Intraseasonal variability in SST
To compute ISV in SSTs and rainfall, the 30–50-day Lanczos bandpass filter (section 3e) is applied to each HadKPP ensemble member, as well as to each season of TMI and GPCP. Because of the filter’s edge effects and the limited length of the HadKPP integrations, only the period 20 June–10 September (JA hereafter, as July and August dominate the period) can be analyzed. JA represents the established monsoon season, so the authors expect the results to be representative of JJAS as a whole. The monsoon onset and retreat are not included as intraseasonal events. The filtered time series from all ensemble members (seasons) are concatenated at each grid point for each ensemble (the observations), prior to taking the standard deviation to compute variability.
In TMI, ISV maxima are aligned with strong low-level monsoon winds: in the Southern Hemisphere trades, along the Somali coast, in the Arabian Sea, the Bay of Bengal, and the South China Sea (Fig. 5a). Most of the Indian coastal waters have a standard deviation exceeding 0.20°C. The high values near Somalia are associated with strong coastal upwelling and surface and subsurface currents (e.g., Izumo et al. 2008; Schott et al. 2009).
The JA: (a)–(c) the standard deviation of 30–50-day bandpass-filtered SSTs (°C) for (a) TMI; (b) 10M-24H, and (c) 1M-3H. Ratios of the standard deviations in 30–50-day bandpass-filtered SSTs for (d) 1M-3H divided by 10M-24H, (e) 10M-3H divided by 10M-24H, and (f) 1M-24H divided by 10M-24H.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
Only 1M-3H has SST ISV that approaches TMI (Figs. 5b–f). The “control” 10M-24H ensemble has less than 60% of the TMI ISV across much of the monsoon domain, including less than 40% in the Arabian Sea, near the Maritime Continent, and in the equatorial west Pacific (Fig. 5b). By contrast, 1M-3H produces comparable levels of SST ISV to TMI (Fig. 5c) and 20%–70% more than 10M-24H (Fig. 5d), with large increases in the off-equatorial west Pacific, the equatorial Indian Ocean, and the South China Sea.
10M-3H (Fig. 5e) and 1M-24H (Fig. 5f) produce only moderate increases in ISV against 10M-24H while remaining deficient relative to TMI (not shown). Shortening the coupling frequency while holding the vertical resolution constant (10M-3H) improves Indian Ocean ISV by up to 40%, but it has no effect in the Pacific. Curiously, refining the vertical resolution while holding the coupling frequency constant (1M-24H) displays a similar pattern. In the Indian Ocean, vertical resolution generally has a greater impact than coupling frequency, likely because 3-h coupling produces little subdaily SST variability with a 10-m resolution (Fig. 3d).
Increases in Pacific SST ISV appear only when both the vertical resolution and coupling frequency are improved. While 3-h coupling has no impact in the Pacific in 10M-3H (Fig. 5e), introducing 3-h coupling in addition to 1-m vertical resolution in 1M-3H greatly enhances Pacific ISV (Fig. 5d). Likewise, 1-m vertical resolution had no impact in the Pacific in 1M-24H (Fig. 5f); however, when combined with 3-h coupling, it increases Pacific ISV in 1M-3H.
Even in 1M-3H, however, HadKPP ISV remains lower than TMI, particularly near Somalia, off Sumatra, and in the equatorial west Pacific (Fig. 5c), where equatorial and coastal upwelling are strong during JJAS (Schott and McCreary 2001; Schott et al. 2009) and contribute substantially to ISV (Waliser et al. 2004). Better-resolved ocean thermodynamics in 1M-3H clearly improves ISV in regions that Waliser et al. (2004) found were driven primarily by surface heat fluxes: the northern Indian Ocean, including the Arabian Sea and the Bay of Bengal, and the northwestern tropical Pacific.
e. Intraseasonal variability in rainfall
Results for ISV in JA rainfall are similar to those for ISV in SST: finely resolved air–sea interactions in 1M-3H lead to additional ISV in rainfall, bringing HadKPP closer to observations (Fig. 6). The maxima in 30–50-day bandpass-filtered GPCP analyses (Fig. 6a) resemble the quadrupole structure of Annamalai and Slingo (2001, their Fig. 5). As for SST, 10M-24H lacks rainfall ISV in the equatorial west Pacific, with less than 40% of the GPCP ISV in a band extending 10°S–15°N (Fig. 6b). Off-equatorial rainfall variability is closer to observations, particularly around the Indian monsoon trough.
As in Fig. 5, but for 30–50-day bandpass-filtered precipitation and using GPCP precipitation instead of TMI SSTs. For clarity, areas in (d)–(f) are shaded light gray where the JJAS mean GPCP rainfall is <1 mm day−1.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
Rainfall ISV in 1M-3H approaches GPCP in the Indian Ocean, demonstrating the importance of well-represented air–sea interactions for monsoon rainfall ISV (Fig. 6c). The 1M-3H has 20%–60% more ISV than 10M-24H, with the greatest increases in the Arabian Sea and the equatorial west Pacific (Fig. 6d). Variability in the equatorial west Pacific in 1M-3H remains deficient, as for SST (Fig. 5c). Deficient ISV in either rainfall or SST can produce deficient ISV in the other quantity, leading to a positive feedback. Therefore, it is not possible to determine whether HadAM3 or KPP is at fault.
The 10M-3H (Fig. 6e) and 1M-24H (Fig. 6f) ensembles show moderately enhanced ISV against 10M-24H in the Bay of Bengal and the northwestern tropical Pacific. As in 10M-24H, 10M-3H and 1M-24H have considerably less ISV in equatorial rainfall than GPCP.
Improving only the air–sea coupling frequency (10M-3H; Fig. 6e) or ocean vertical resolution (1M-24H; Fig. 6f) leads to enhanced ISV in the Indian Ocean but little change in the west Pacific, as for SST. Vertical resolution has a slightly greater impact than coupling frequency. Equatorial and west Pacific ISV increases most when the coupling frequency and vertical resolution are improved simultaneously in 1M-3H (Fig. 6f).
f. Northward and eastward propagation
Stronger gridpoint-scale ISV in rainfall and SSTs is a necessary but not sufficient condition for a stronger propagating NPISO. Reproducing the observed propagation is essential for predicting the NPISO.
To assess the spatial and temporal evolution of the NPISO in HadKPP, lag-correlation propagation diagnostics are adapted from Lin et al. (2008), who used 24–70-day filtered rainfall. This study uses 30–50-day filtered rainfall (section 3e), because of the limited length of the HadKPP simulations. For northward propagation, the Lin et al. (2008) lag correlation of longitude-averaged (70–100°E) filtered rainfall against filtered rainfall at 12.5°N, 85°E is employed here. For eastward propagation, Lin et al. (2008) averaged over 5–25°N; this study uses 10°S–10°N, as observed NPISO events typically propagate along the equator not off it (e.g., Lawrence and Webster 2002). Lag correlations are computed of this latitude-averaged filtered rainfall against filtered rainfall at 0°, 100°E. The results for GPCP (Figs. 7a and 8a) are qualitatively similar to those for GPCP in Lin et al. (2008, their Figs. 11a and 16a).
Lead–lag correlations of 30–50-day filtered, longitude-averaged (70°–100°E) rainfall for JA, using a base point at 12.5°N, 85°E for (a) GPCP, (b) 10M-24H, (c) 10M-3H, (d) 1M-24H, and (e) 1M-3H. Shading indicates correlations that are statistically significant at the 95% confidence level.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
As in Fig. 7, but using latitude-averaged (10°S–10°N) rainfall and a base point at 0°, 95°E.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
In 10M-24H, rainfall at 12.5°N, 85°E has statistically significant correlations with only those grid points in its immediate vicinity (Fig. 7b). There is also no recognizable oscillation between active and break phases. Therefore, 10M-24H has only a very weak NPISO.
An oscillation between phases appears in 10M-3H (Fig. 7c) and 1M-24H (Fig. 7d); however, in both ensembles, the rainfall moves from the equator to 15°N too quickly compared to GPCP and then stagnates north of 15°N. This behavior indicates a standing oscillation: rainfall anomalies of opposing signs appear over the equatorial Indian Ocean and India and then swap signs without propagation. The standing oscillation is more evident in 1M-24H than in 10M-3H, but neither ensemble has a northward propagation resembling GPCP (Fig. 7a).
The 1M-3H ensemble displays consistent northward propagation from 10°N, with a phase speed similar to GPCP, in marked contrast to the other ensembles (Fig. 7e). This suggests that the improved representation of atmosphere–ocean coupling in 1M-3H enables HadKPP to produce organized tropical convection on intraseasonal temporal scales. The modeled NPISO is incoherent between the equator and 10°N in 1M-3H, however, which is consistent with reduced equatorial rainfall ISV (Fig. 6c).
No ensemble captures the observed equatorial eastward propagation of the NPISO (Fig. 8). There is an oscillation between enhanced and suppressed equatorial convection in 1M-3H (Fig. 8e), but rainfall anomalies appear at all equatorial longitudes in the Indian and Pacific Oceans simultaneously. The other ensembles do not generate a statistically significant, coherent eastward-propagating oscillation (Figs. 8b–d).
g. Composite NPISO active events
Composite NPISO active events at 12.5°N, 85°E are constructed for 1M-3H and 10M-24H using the method in section 3f (Fig. 9). No temporal filtering has been applied to the anomalies in Fig. 9; filtered rainfall was used only to identify the active events.
Composite NPISO active events at the point 12.5°N, 85°E for (a) 1M-3H and (b) 10M-24H. For each grouping: (top) daily-mean anomalies in precipitation (blue) and net surface shortwave radiation (red); (middle) daily-mean anomalies in atmospheric specific humidity; (bottom) 3-h mean ocean mixing depth (dashed line) and anomalies in ocean temperature (contours). In the top panels, dots indicate statistical significance at the 5% level, using a two-tailed Student’s t test on means. Anomalies in specific humidity and ocean temperature are shown only where statistically significant at the 5% level.
Citation: Journal of Climate 24, 10; 10.1175/2010JCLI3868.1
Active events in 1M-3H are preceded by warm ocean temperature anomalies that approach 0.3°C at the peak of the diurnal cycle (Fig. 9a). In 10M-24H, the composite ocean temperature anomalies do not exceed 0.1°C (Fig. 9b). The strong diurnal warming in 1M-3H is associated with shallow mixed layers that are regularly less than 5 m deep and occasionally less than 2 m deep, which inhibit vertical mixing of the anomalously warm upper-ocean temperatures, maintaining the warming. There is little variability in 10M-24H mixing depth, because of the lack of the diurnal cycle in surface forcing. Mixing depths are also nearly constant in 1M-24H (not shown), emphasizing that although vertical resolution permits shallow mixing depths, the diurnal warming drives the mixed layer shoaling. Ocean temperature anomalies and mixing depths in 1M-3H also show a clear diurnal cycle and stronger subseasonal variability after the active event, with cool SST anomalies of up to 0.25°C and mixing depths below 60 m. 10M-24H again shows only weak variability.
The strong oceanic ISV in 1M-3H is associated with a more intense oscillation between active and break events. Reduced rainfall, increased net surface shortwave radiation, and anomalously dry specific humidity throughout the lower and midtroposphere precede and follow the composite active event by around 15 days, consistent with break conditions. Enhanced rainfall and reduced net insolation persist for 10–15 days, accompanied by large increases in atmospheric moisture. The atmospheric anomalies are weaker in 10M-24H and less organized in time, particularly in the case of atmospheric moisture: the break events are barely distinguishable and the break–active transition is ill defined compared to 1M-3H.
While these composites are shown only at a single grid point, given the coherent northward propagation shown for 1M-3H in Fig. 7e, the temporal dimension shown in Fig. 9 can be qualitatively converted to a spatial dimension. Thus, the warm SSTs that precede the active event at this single grid point can be interpreted as occurring to the north of the enhanced convection; the cool SSTs following the event occur to the south. The authors hypothesize that the warmer (cooler) SST anomalies in 1M-3H over 10M-24H exert a much stronger “pull” on the enhanced (suppressed) convection, leading to significant northward propagation. This may occur via changes to lower-atmospheric stability. For example, considerable low-level moistening occurs 8–12 days prior to the center of the active event in 1M-3H, associated with the warmest SSTs (Fig. 9a). The much weaker SST variations surrounding the intraseasonal oscillation in 10M-24H do not induce the convection to propagate; there is only very weak low-level moistening prior to the active event in Fig. 9b. This hypothesis is consistent with studies that have concluded that high-frequency, high-amplitude SST variability led to propagating, organized, intraseasonal tropical convection in winter (Woolnough et al. 2001; Kim et al. 2008) and summer Fu et al. (2003); Klingaman et al. (2008a).
A model’s representation of the diurnal cycle is therefore critical for that model’s representation of the NPISO. The diurnal variability in surface forcing in 1M-3H enhances the upper-ocean response to NPISO events, including strong variations in mixing depth, and is associated with increased atmospheric ISV. Both the oceanic response to the atmosphere and the atmospheric response to the ocean are amplified via coupled feedbacks, leading to the maintenance and northward propagation of the intraseasonal oscillation.
5. Discussion and conclusions
A new atmosphere–ocean coupled model, HadKPP, has been assembled that uses many columns of a one-dimensional ocean boundary layer model rather than a three-dimensional dynamical ocean model, to reduce computational cost while resolving shallow mixing depths and sharp temperature gradients in the upper ocean. Four ensembles of HadKPP have been used to simulate the Northern Hemisphere summer intraseasonal oscillation, using upper-ocean vertical resolutions of 1 and 10 m and coupling frequencies of 3 and 24 h (Table 1). Depth-varying flux corrections were employed to maintain an accurate basic-state SST and to ensure that differences in intraseasonal variability among the ensembles were not due to mean-state differences.
High-frequency (3 h) atmosphere–ocean coupling and fine (1 m) upper-ocean vertical resolution led to considerable improvements in the amplitude of intraseasonal variations in SSTs (Fig. 5) and rainfall (Fig. 6), as well as in the northward propagation of organized convection from the equatorial Indian Ocean to the Bay of Bengal (Fig. 7). When HadKPP was configured similarly to most contemporary coupled GCMs—a 10-m ocean vertical resolution and 24-h coupling—the model produced less than 60% of the observed ISV across most of the domain and a very weak, statistically insignificant NPISO. Enhancing only the air–sea coupling frequency or the ocean vertical resolution led to small increases in intraseasonal variability and only a standing intraseasonal oscillation. The two must be improved simultaneously to produce realistic diurnal SST variability (Fig. 3) that rectifies onto intraseasonal periods. Well-resolved air–sea coupling did not improve the eastward propagation of the summer intraseasonal oscillation in this model (Fig. 8). Eastward propagation is therefore not a prerequisite for northward propagation in HadKPP.
Even in the highest-resolution ensemble (1M-3H), equatorial intraseasonal variability in SSTs and rainfall remained deficient, which may have been because of (i) the one-dimensional KPP ocean model does not include the dynamic ocean response to atmospheric forcing, which Waliser et al. (2004) and others have shown to be important near the equator; and (ii) the HadAM3 atmospheric model has weak variability in convection near the equator because of its poor simulation of convectively coupled equatorial waves (Yang et al. 2009). With respect to the latter, Yang et al. (2009) found that HadAM3 performed considerably better for convection coupled to off-equatorial waves. It may be, then, that errors in HadAM3 parameterizations cause variability in equatorial convection to be relatively insensitive to improved air–sea coupling in HadKPP, while off-equatorial regions show a stronger, more-organized intraseasonal oscillation. Whatever the cause, the limited equatorial variability almost certainly led to the lack of eastward propagation.
Diurnal air–sea coupling produced a considerably stronger upper-ocean temperature response to the intraseasonal oscillation than daily-mean coupling (Fig. 9). Fine vertical resolution in 1M-3H allowed KPP to capture the considerable mixed layer shoaling associated with diurnal warm events prior to the onset of enhanced convection, producing a much warmer mean SST than in 10M-24H. Similarly, stronger surface forcing and deeper mixing in 1M-3H amplified the upper-ocean cooling following the enhanced convection. The authors hypothesized that the intensified variability in upper-ocean temperatures in 1M-3H induced the intraseasonal convection to propagate north by affecting lower-atmospheric stability.
This study has shown that atmospheric and oceanic intraseasonal variability are coupled by a positive feedback mechanism: increasing the variability in upper-ocean temperatures and mixing depths led to stronger variability in convection, which in turn amplified the variability in the surface forcing, producing still further upper-ocean variability. The coupled nature of this feedback argues that predicting subseasonal variability requires continued improvements to the representation of ocean vertical mixing and to atmospheric parameterizations of convection, radiation, and boundary layer processes. The intraseasonal oscillation has been described frequently as an atmospheric mode that can be enhanced by air–sea coupled processes (Fu and Wang 2004; Rajendran and Kitoh 2006, e.g.,), but the mechanism by which SST variability feeds back onto the atmosphere remains unclear. This study supports those results by demonstrating improvements to the simulated intraseasonal oscillation from additional high-frequency upper-ocean variability. Untangling the coupled feedbacks in the intraseasonal oscillation—which are also an issue for modeling the wintertime MJO in AOGCMs—should be a focus of future work, as it would improve our understanding of the critical physical processes underlying tropical intraseasonal variability and focus efforts on improving models’ representations of those processes.
This study confirms and extends previous efforts on the effects of finely resolved atmosphere–upper ocean interactions (e.g., Bernie et al. 2007; Woolnough et al. 2007) and high-frequency SST variability (e.g., Kim et al. 2008; Klingaman et al. 2008a) on tropical subseasonal variability. This study has focused on summer, when the intraseasonal oscillation propagates north and east and influences the Indian monsoon. In HadKPP, the amplitude and coherence of the intraseasonal oscillation depended on including the diurnal cycle of SST and resolving upper-ocean vertical mixing. Most numerical weather prediction systems are atmosphere-only models driven by persisted SSTs or persisted SST anomalies, however, and so they fail to represent coupled processes even crudely. To produce socioeconomically relevant forecasts of subseasonal monsoon variability, prediction systems must include a well-resolved upper ocean, to take advantage of the predictability that arises from accurately modeling the evolution of the surface ocean in response to atmospheric forcing.
Acknowledgments
NPK was funded in part by a Marshall scholarship from the Marshall Aid Commemoration Commission. NPK, SJW, HW, and JS were funded by the National Centre for Atmospheric Science, a collaborative center of the Natural Environment Research Council, under Agreement R8/H12/83/001. HadKPP and HadAM3 simulations were performed partly on HPCx and partly on HECToR, the United Kingdom’s national high-performance computing services. GPCP combined precipitation data were provided by the NASA Goddard Space Flight Center’s Laboratory for Atmospheres (data are available from http://precip.gsfc.nasa.gov). TMI–AMSR-E microwave optimally interpolated SST data were produced by Remote Sensing Systems (www.remss.com).
REFERENCES
Adler, R. F., and Coauthors, 2003: The version-2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979–present). J. Hydrometeor., 4, 1147–1167.
Annamalai, H., and J. M. Slingo, 2001: Active/break cycles: Diagnosis of the intraseasonal variability of the Asian summer monsoon. Climate Dyn., 18, 85–102.
Annamalai, H., J. M. Slingo, K. R. Sperber, and K. Hodges, 1999: The mean evolution and variability of the Asian summer monsoon: Comparison of ECMWF and NCEP–NCAR reanalyses. Mon. Wea. Rev., 127, 1157–1186.
Bell, M. J., R. M. Forbes, and A. Hines, 2000: Assessment of the FOAM global data assimilation system for real-time operational ocean forecasting. J. Mar. Syst., 25, 1–22.
Bernie, D. J., 2006: Diurnal variability of the tropical upper ocean and its climate impacts. Ph.D. thesis, University of Reading, 163 pp.
Bernie, D. J., S. J. Woolnough, and J. M. Slingo, 2005: Modeling diurnal and intraseasonal variability of the ocean mixed layer. J. Climate, 18, 1190–1202.
Bernie, D. J., E. Guilyardi, G. Madec, J. M. Slingo, and S. J. Woolnough, 2007: Impact of resolving the diurnal cycle in an ocean-atmosphere GCM. Part 1: A diurnally forced GCM. Climate Dyn., 29, 575–590.
Bernie, D. J., E. Guilyardi, G. Madec, J. M. Slingo, S. J. Woolnough, and J. Cole, 2008: Impact of resolving the diurnal cycle in an ocean–atmosphere GCM. Part 2: A diurnally coupled CGCM. Climate Dyn., 31, 909–925.
Bhat, G. S., and Coauthors, 2001: BOBMEX: The Bay of Bengal Monsoon Experiment. Bull. Amer. Meteor. Soc., 82, 2217–2243.
Fu, X., and B. Wang, 2004: The boreal-summer intraseasonal oscillations simulated in a hybrid coupled atmosphere–ocean model. Mon. Wea. Rev., 132, 2628–2649.
Fu, X., B. Wang, T. Li, and J. P. McCreary, 2003: Coupling between northward-propagating, intraseasonal oscillations and sea surface temperature in the Indian Ocean. J. Atmos. Sci., 60, 1733–1753.
Fu, X., B. Wang, D. E. Waliser, and L. Tao, 2007: Impact of atmosphere–ocean coupling on the predictability of monsoon intraseasonal oscillations. J. Atmos. Sci., 64, 157–174.
Gadgil, S., 1990: Poleward propagation of the ITCZ: Observations and theory. Mausam, 41, 285–290.
Gustafson, W. I., Jr., and B. C. Weare, 2004: MM5 modeling of the Madden–Julian oscillation in the Indian and west Pacific Oceans: Implications of 30–70-day boundary effects on MJO development. J. Climate, 17, 1338–1351.
Ham, Y.-G., J.-S. Kug, I.-S. Kang, F.-F. Jin, and A. Timmerman, 2010: Impact of diurnal atmosphere–ocean coupling on tropical climate simulations using a coupled GCM. Climate Dyn., 34, 905–917.
Harrison, D. E., and G. A. Vecchi, 2001: January 1999 Indian Ocean cooling event. Geophys. Res. Lett., 28, 3717–3720.
Hartmann, D. L., and M. L. Michelsen, 1989: Intraseasonal periodicities in Indian rainfall. J. Atmos. Sci., 46, 2838–2862.
Inness, P. M., and J. M. Slingo, 2003: Simulation of the Madden–Julian oscillation in a coupled general circulation model. Part I: Comparison with observations and an atmosphere-only GCM. J. Climate, 16, 345–364.
Inness, P. M., J. M. Slingo, E. Guilyardi, and J. Cole, 2001: Organization of tropical convection in a GCM with varying vertical resolution: Implications for the simulation of the Madden–Julian oscillation. Climate Dyn., 17, 777–793.
Inness, P. M., J. M. Slingo, E. Guilyardi, and J. Cole, 2003: Simulation of the Madden–Julian oscillation in a coupled general circulation model. Part II: The role of the basic state. J. Climate, 16, 365–382.
Izumo, T., C. B. Montégut, J.-J. Luo, S. K. Behera, and S. Masson, 2008: The role of the western Arabian Sea upwelling in Indian monsoon rainfall variability. J. Climate, 21, 5603–5623.
Jones, C., D. E. Waliser, and C. Gautier, 1998: The influence of the Madden–Julian oscillation on ocean surface heat fluxes and sea surface temperature. J. Climate, 11, 1057–1072.
Joseph, P. V., and T. P. Sabin, 2008: An ocean–atmosphere interaction mechanism for the active break cycle of the Asian summer monsoon. Climate Dyn., 30, 553–566.
Kemball-Cook, S., and B. Wang, 2001: Equatorial waves and air–sea interaction in the boreal summer intraseasonal oscillation. J. Climate, 14, 2923–2942.
Kennedy, J. J., P. Brohan, and S. F. B. Tett, 2007: A global climatology of the diurnal variations in sea-surface temperature and implications for MSU temperature trends. Geophys. Res. Lett., 34, L05712, doi:10.1029/2006GL028920.
Kim, H.-M., C. D. Hoyos, P. J. Webster, and I.-S. Kang, 2008: Sensitivity of MJO simulation and predictability to sea surface temperature variability. J. Climate, 21, 5304–5317.
Kim, H.-M., C. D. Hoyos, P. J. Webster, and I.-S. Kang, 2010: Ocean–atmosphere coupling and the boreal winter MJO. Climate Dyn., 35, 771–784, doi:10.1007/s00382-009-0612-x.
Klingaman, N. P., 2008: The intraseasonal oscillation of the Indian summer monsoon: Air–sea interactions and the potential for predictability. Ph.D. thesis, University of Reading, 173 pp. [Available online at http://www.met.rdg.ac.uk/~ss901165/klingaman_2008_phd_thesis.pdf.]
Klingaman, N. P., P. M. Inness, H. Weller, and J. M. Slingo, 2008a: The importance of high-frequency sea surface temperature variability to the intraseasonal oscillation of Indian monsoon rainfall. J. Climate, 21, 6119–6140.
Klingaman, N. P., H. Weller, J. M. Slingo, and P. M. Inness, 2008b: The intraseasonal variability of the Indian summer monsoon using TMI sea surface temperatures and ECMWF reanalysis. J. Climate, 21, 2519–2539.
Large, W., J. McWilliams, and S. Doney, 1994: Oceanic vertical mixing: A review and a model with a nonlocal boundary layer parameterization. Rev. Geophys., 32, 363–403.
Lawrence, D. M., and P. J. Webster, 2002: The boreal intraseasonal oscillation: Relationship between northward and eastward movement of convection. J. Atmos. Sci., 59, 1593–1606.
Lin, J.-L., and Coauthors, 2006: Tropical intraseasonal variability in 14 IPCC AR4 climate models. Part I: Convective signals. J. Climate, 19, 2665–2690.
Lin, J.-L., K. M. Weickman, G. N. Kiladis, B. E. Mapes, S. D. Schubert, M. J. Suarez, J. T. Bacmeister, and M.-I. Lee, 2008: Subseasonal variability associated with Asian summer monsoon simulated by 14 IPCC AR4 coupled GCMs. J. Climate, 21, 4542–4566.
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.
Matthews, A. J., J. M. Slingo, B. J. Hoskins, and P. M. Inness, 1999: Fast and slow Kelvin waves in the Madden–Julian oscillation of a GCM. Quart. J. Roy. Meteor. Soc., 125, 1473–1498.
Pope, V. D., M. L. Gallani, P. R. Rowntree, and R. A. Stratton, 2000: The impact of the new physical parameterizations in the Hadley Centre climate model. Climate Dyn., 16, 123–146.
Rajendran, K., and A. Kitoh, 2006: Modulation of tropical intraseasonal oscillations by atmosphere–ocean coupling. J. Climate, 19, 366–391.
Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, 929–948.
Sausen, R., K. Barthel, and K. Hasselmann, 1988: Coupled ocean-atmosphere models with flux correction. Climate Dyn., 2, 145–163.
Schott, F. A., and J. P. McCreary, 2001: The monsoon circulation of the Indian Ocean. Prog. Oceanogr., 51, 1–123.
Schott, F. A., S.-P. Xie, and J. P. McCreary Jr., 2009: Indian Ocean circulation and climate variability. Rev. Geophys., 47, RG1002, doi:10.1029/2007RG000245.
Senan, R., A. D. S, and D. Sengupta, 2001: Validation of SST and windspeed from TRMM using North Indian Ocean moored buoy observations. Centre for Atmospheric and Oceanic Sciences Rep. 2001AS1, 29 pp.
Spencer, H., and J. M. Slingo, 2003: The simulation of peak and delayed ENSO teleconnections. J. Climate, 16, 1757–1774.
Stuart-Menteth, A., I. S. Robinson, R. A. Weller, and C. J. Donlon, 2005: Sensitivity of the diurnal warm layer to meteorological fluctuations. Part 2: A new parameterization for diurnal warming. J. Atmos. Ocean Sci., 10, 209–234.
Stratton, R. S., 1999: A high resolution AMIP integration using the Hadley Centre model HadAM2b. Climate Dyn., 15, 9–28.
Terray, L., E. Sevault, E. Guilyardi, and O. Thual, 1995: OASIS 2.0—Ocean Atmosphere Sea Ice Soil—user’s guide and reference manual. CERFACS Tech. Rep. TR/CMGC/95-46, 123 pp.
Tompkins, A. M., 2001: On the relationship between tropical convection and sea surface temperature. J. Climate, 14, 633–637.
Turner, A. G., P. M. Inness, and J. M. Slingo, 2005: The role of the basic state in the ENSO–monsoon relationship and implications for predictability. Quart. J. Roy. Meteor. Soc., 131, 781–804.
Vecchi, G. A., and D. E. Harrison, 2002: Monsoon breaks and subseasonal sea surface temperature variability in the Bay of Bengal. J. Climate, 15, 1485–1493.
Vinayachandran, P. N., V. S. N. Murty, and V. Ramesh Babu, 2002: Observations of barrier layer formation in the Bay of Bengal during summer monsoon. J. Geophys. Res., 107, 8018, doi:10.1029/2001JC000831.
Waliser, D. E., K. M. Lau, and J.-H. Kim, 1999: The influence of coupled sea surface temperatures on the Madden–Julian oscillation: A model perturbation experiment. J. Atmos. Sci., 56, 333–358.
Waliser, D. E., R. Murtugudde, and L. E. Lucas, 2004: Indo-Pacific Ocean response to atmospheric intraseasonal variability: 2. Boreal summer and the intraseasonal oscillation. J. Geophys. Res., 109, C03030, doi:10.1029/2003JC002002.
Wang, W., M. Chen, and A. Kumar, 2009: Impacts of ocean surface on the northward propagation of the boreal summer intraseasonal oscillation in the NCEP climate forecast system. J. Climate, 22, 6561–6576.
Webster, P. J., and C. Hoyos, 2004: Prediction of monsoon rainfall and river discharge on 15–30-day time scales. Bull. Amer. Meteor. Soc., 85, 1745–1765.
Webster, P. J., and Coauthors, 2002: The JASMINE pilot study. Bull. Amer. Meteor. Soc., 83, 1603–1630.
Wentz, F. J., 1998: Algorithm theoretical basis document: AMSR ocean algorithm. Remote Sensing Systems Tech. Rep. 110398, 65 pp.
Wentz, F. J., 2000: Satellite measurements of sea surface temperature through clouds. Science, 288, 847–850.
Woolnough, S. J., J. M. Slingo, and B. J. Hoskins, 2001: The organization of tropical convection by intraseasonal sea surface temperature anomalies. Quart. J. Roy. Meteor. Soc., 127, 887–907.
Woolnough, S. J., F. Vitart, and M. A. Balmaseda, 2007: The role of the ocean in the Madden–Julian oscillation: Implications for MJO prediction. Quart. J. Roy. Meteor. Soc., 133, 117–128.
Xavier, P. K., J.-P. Duvel, and F. J. Doblas-Reyes, 2008: Boreal summer intraseasonal variability in coupled seasonal hindcasts. J. Climate, 21, 4477–4497.
Yang, G.-Y., J. Slingo, and B. Hoskins, 2009: Convectively coupled equatorial waves in high-resolution Hadley Centre climate models. J. Climate, 22, 1897–1919.