1 NOVEMBER 2002
MALONEY AND KIEHL
2989
Intraseasonal Eastern Pacific Precipitation and SST Variations in a GCM Coupled to a Slab Ocean Model ERIC D. MALONEY
AND
JEFFREY T. KIEHL
NCAR/CGD, Boulder, Colorado (Manuscript received 15 November 2001, in final form 13 May 2002) ABSTRACT Coupling the NCAR Community Climate Model version 3.6 (CCM3.6) with relaxed Arakawa–Schubert convection to a slab ocean model (SOM) improves the simulation of eastern Pacific convection during a composite June–November intraseasonal oscillation (ISO) life cycle. Intraseasonal oscillations in the SOM simulation produce convective variability over the tropical northeastern Pacific that is similar to that produced by the observed Madden–Julian oscillation (MJO). A composite ISO life cycle in the SOM simulation exhibits stronger, more coherent, and more widespread eastern Pacific warm pool convective anomalies than in a control simulation using climatological SSTs. Competing convective forcings over land and ocean make eastern Pacific low-level circulation anomalies more complex in the SOM simulation than in the observed MJO. Off-equatorial eastern Pacific SST variations of more than 0.68C are associated with the June–November SOM simulation ISO. These variations are similar to those observed with the MJO. No significant equatorial east Pacific SST anomalies occur in the model, supporting the contention that observed MJO SST anomalies on the equator are caused by ocean dynamics. Positive off-equatorial SOM simulation SST anomalies are nearly in phase with enhanced precipitation during significant MJO events, whereas observed SST anomalies lead enhanced precipitation by just under 10 days. Latent heat flux and surface shortwave radiation anomalies are the dominant terms in controlling east Pacific intraseasonal SST in the SOM simulation, as in observations. Positive latent heat flux and shortwave radiation anomalies (positive defined as downward into the ocean) lead enhanced SST by about 10 days during significant ISO events in the SOM simulation.
1. Introduction Maloney and Kiehl (2002, hereafter MK02) found that the Madden–Julian oscillation (MJO; Madden and Julian 1994) significantly modulates eastern Pacific intraseasonal sea surface temperatures (SSTs) during June–November. SST variations during an MJO life cycle are on the order of 0.48–0.58C, as determined from Reynolds SST data (Reynolds and Smith 1994). Individual MJO events can have SST variations of magnitude 18–28C. SST variations just to the south of Mexico and Central America (the eastern Pacific hurricane region) are 1808 out of phase with those on the equator. Off-equatorial SSTs lead enhanced convection by approximately 10 days. Intraseasonal air–sea interactions, SST variations, and convection anomalies over the eastern Pacific warm pool are associated with the propagation of MJO-related wind anomalies into this region from the west. Local air–sea interactions and latent heat release in convection act to amplify intraseasonal convection and winds over Corresponding author address: Eric Maloney, College of Oceanic and Atmospheric Sciences, 104 Ocean Admin. Bldg., Oregon State University, Corvallis, OR 97331-5503. E-mail:
[email protected]
q 2002 American Meteorological Society
the eastern Pacific (Maloney and Hartmann 2000). Surface latent heat and shortwave flux variations, as obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis, largely explain the SST variations over the eastern Pacific warm pool (MK02). Observed intraseasonal surface flux anomalies, defined as positive into the ocean, typically lead the SST signal by 10–15 days. Maximum evaporation anomalies (negative latent heat flux anomalies) coincide with maximum lowlevel westerly wind anomalies and precipitation, suggesting that a wind-induced surface heat exchange mechanism similar to that proposed by Emanuel (1987) may help regulate intraseasonal east Pacific convection. Equatorial east Pacific SST anomalies cannot be explained by surface flux variations, and are likely caused by oceanic dynamics. MK02 found that SST variations lead to variations of the surface-saturation-equivalent potential temperature of 38C over the east Pacific warm pool during a June–November MJO life cycle. MK02 also showed using the diagnostic primitive equation model of Nigam and Chung (2000) that enhanced SSTs over the east Pacific MJO convective region may favor increased surface convergence. Both of these effects of SST variations may help modulate MJO convection over
2990
JOURNAL OF CLIMATE
the eastern Pacific during boreal summer. Magan˜a et al. (1999) showed that similar local processes may help modulate SSTs and monsoonal circulations in the east Pacific, particularly associated with the midsummer drought over Mexico and Central America. Previous studies have shown the MJO to modulate western Pacific and Indian Ocean SSTs during boreal winter (e.g., Kawamura 1988; Krishnamurti et al. 1988; Zhang 1996; Hendon and Glick 1997; Lau and Sui 1997). Further discussion of these studies is presented in MK02. This paper uses an atmospheric general circulation model (GCM) coupled to a slab ocean model (SOM) of specified variable mixed layer depth to determine 1) whether intraseasonal SST variations of magnitude close to those observed can be simulated over the eastern Pacific warm pool during a June–November model intraseasonal oscillation life cycle, and 2) whether these SST variations importantly modulate intraseasonal convection. We do not expect to properly simulate equatorial SST variations in this study, because the model used does not include ocean dynamics, only oceanic mixed layer thermodynamics in a slab ocean model. Processes such as Ekman transport and tropical cyclone–induced turbulent entrainment are therefore not simulated by the model (e.g., Schade and Emanuel 1999). Simulations using a dynamic ocean model will be conducted in future work to explore east Pacific intraseasonal SST variations on the equator. Previous studies have suggested the importance of oceanic mixed layer processes in producing realistic simulated intraseasonal oscillations (e.g., Wang and Xie 1998). Studies using GCMs coupled to oceanic mixed layer models have suggested the importance of SST variations to the MJO over the equatorial Indian and western Pacific Oceans. Flatau et al. (1997), Waliser et al. (1999), and Kemball-Cook et al. (2002) showed that stronger, more coherent, intraseasonal oscillations with more realistic eastward propagation speeds occur in atmospheric GCMs coupled to interactive ocean models than in uncoupled models. Waliser et al. (1999) showed that positive SST anomalies to the east of model intraseasonal oscillation convection may enhance surface moisture convergence, amplifying the model oscillation though coupled feedbacks with enhanced convection. Interactions over the eastern Pacific during boreal summer may be somewhat different than suggested by Waliser et al. (1999) since convective heating is centered to the north of the equator outside of the equatorial waveguide. Hendon (2000) did a study similar to Waliser et al. (1999) that showed no amplification of the model intraseasonal oscillation signal with inclusion of an oceanic mixed layer. Latent heat and shortwave radiation flux anomalies in the study of Hendon (2000) did not add constructively as in observations, leading to weaker SST anomalies than observed, and resulted in no amplification of the model intraseasonal oscillation. This paper will be the first study that we know of
VOLUME 15
that explores how coupling an atmospheric GCM to an oceanic mixed layer model affects intraseasonal variability over the eastern Pacific Ocean. Section 2 discusses the observational data, the GCM, and the slab ocean model of specified variable mixed layer depth used in this study, and discusses the coupled and uncoupled experiment configurations. Section 3 briefly compares coupled and uncoupled global model performance to observations during all seasons. Section 4 analyzes coupled and uncoupled model performance over the eastern Pacific during June–November, the eastern Pacific tropical cyclone season. Section 5 does an analysis of intraseasonal surface flux variations over the eastern Pacific, and section 6 will present some conclusions. 2. Data and models a. Observational data Observational data are used at several places in this study when comparisons to model fields are appropriate. MK02 should be referenced for a more detailed observational analysis. Weekly averaged Reynolds SST data (18 3 18) were available from October 1981 to December 1999 (Reynolds and Smith 1994). These data were interpolated to daily values before use in the study. Xie and Arkin (1996) merged–gridded (2.58 3 2.58) precipitation data were available in pentad format from 1979 to 1999. These data were also interpolated to daily values. Daily surface shortwave radiation and wind analyses were obtained from the NCEP–NCAR gridded reanalysis dataset (2.58 3 2.58) for the years 1979–99 (Kalnay et al. 1996). Results are subject to the assumption that the NCEP–NCAR reanalysis product produces realistic fields across the Tropics. Intraseasonal bandpass-filtered fields are constructed from a linear nonrecursive filter with half-power points at 30 and 90 days. All anomaly fields in this paper are constructed using this filter. This bandpass window allows retention of the bulk of variance at MJO timescales in the observations and model simulations. Results in this paper are not sensitive to reasonable variations in the size of this bandpass window. b. GCM and slab ocean model The GCM used here is a modified version of the NCAR Community Climate Model version 3.6 (CCM3; Kiehl et al. 1998). CCM3 simulations in this paper are conducted at T42 resolution (2.88 3 2.88) with 18 levels in the vertical, and the top of the model is at 2.9 hPa. The model time step is 20 min. The model is modified by replacing the standard CCM3 deep convection scheme of Zhang and McFarlane (1995) with the relaxed Arakawa–Schubert (RAS) scheme of Moorthi and Suarez (1992). This replacement was done to improve trop-
1 NOVEMBER 2002
MALONEY AND KIEHL
2991
modynamic effects of sea ice. A more detailed discussion of the slab ocean model can be found in Kiehl et al. (1996). c. Model simulations
FIG. 1. Annual mean mixed layer depths used in the SOM simulation. Contours are plotted every 5 m. Values less than 15 m are shaded.
ical intraseasonal variability in the model. The poor simulation of tropical intraseasonal variability produced by the CCM3 with Zhang and McFarlane (1995) convection has been previously documented by Maloney and Hartmann (2001). The Hack (1994) scheme used by the standard CCM3 to simulate shallow convection remains the shallow convection scheme in the modified version of CCM3 used here. The RAS scheme used in this study for deep convection includes a convective rain evaporation scheme, as described by Sud and Molod (1988). Maloney and Hartmann (2001) found that the parameterization of rain evaporation and convective downdrafts in a modified version of the CCM3 that uses an Arakawa–Schubert-type convection scheme (Sud and Walker 1999) greatly improves intraseasonal variability. The equation that forms the basis for the CCM3 slab ocean model (Kiehl et al. 1998) is taken from Hansen et al. (1984):
r o Co h o
]To 5 F 1 Q, ]t
where T o is the ocean mixed layer temperature, r o is the density of ocean water, C o is the heat capacity of ocean water, h o is the mixed layer depth, F is the net atmosphere to ocean heat flux, and Q is the oceanic mixed layer heat flux, which simulates deep water heat exchange and ocean transport. Here F includes surface latent heat, sensible heat, shortwave radiation, and longwave radiation fluxes. The ocean mixed layer depth h o is derived from Levitus (1982). Mixed layer depths used in the model vary spatially and employ an annual mean mixed layer depth. Figure 1 shows the east Pacific annual mean mixed depths used in the slab ocean model. The Q is specified as the heat flux required to balance the mixed layer heat budget (1) using heat fluxes F derived from a control run with climatological SSTs. Specification of Q in the heat balance equation ensures that the SST climatologies from control and slab ocean simulations are very similar. The heat balance in regions of sea ice is similar to (1), except that the heat balance equation and the specification of Q includes the ther-
A 15-yr control simulation of the modified CCM3 with fixed climatological SSTs was conducted. This simulation will hereafter be called ‘‘control.’’ Fluxes from this run were used to compute Q in (1) for use in a simulation where the modified CCM3 is coupled to the slab ocean model described above. The slab ocean run was initiated by a 20-yr spinup period to ensure that the model reaches a stabilized climate. The 15 yr of the simulation immediately following the spinup period are analyzed in this paper. This simulation will hereafter be called ‘‘SOM.’’ A global analysis of the climates of the control and SOM simulations indicates very close agreement (not shown), as would be expected given the design of the slab ocean model. Differences in tropical variability between the two simulations do exist, however, and will be described in subsequent sections. 3. All-season variability comparison This section will present a brief comparison of intraseasonal variability among observations, the control simulation, and the SOM simulation using all seasons of data. The compositing technique to be used here and in subsequent sections will also be developed. The primary purpose of this paper is not to discuss model performance over the Indian and west Pacific Oceans, but rather model performance over the east Pacific warm pool. This section will only briefly examine all-season equatorial behavior on intraseasonal timescales to ensure that the model simulations are able to capture MJOlike wind anomalies that propagate into the east Pacific. This paper’s main goal is to determine how an interactive slab ocean affects east Pacific intraseasonal convective variability during boreal summer, which will be discussed in detail in subsequent sections. Figure 2 shows wavenumber-frequency spectra of equatorial (108N–108S averaged) 850-hPa zonal winds for NCEP reanalysis and the control and SOM simulations. A Hanning window was applied in the temporal domain before spectra were computed, and the seasonal cycle was removed. Intraseasonal variance at wavenumber 1 and eastward intraseasonal periods is about 30% higher in the SOM simulation than in the control simulation, and is of comparable magnitude to observed variance. Figure 2 thus provides evidence that tropical intraseasonal variability is somewhat enhanced by coupling to an interactive ocean, supporting the findings of Flatau et al. (1997), Waliser et al. (1999), and KemballCook et al. (2002). Precipitation spectra indicate a similar enhancement of MJO-band variability in the SOM simulation over the control, although SOM precipitation
2992
JOURNAL OF CLIMATE
VOLUME 15
FIG. 3. First two EOFs of the equatorial averaged (78N–78S) 30– 90-day 850-mb zonal wind from the (a) control simulation and (b) SOM simulation. Magnitudes are normalized.
FIG. 2. Wavenumber-frequency spectrum of 108N–108S averaged 850-mb zonal wind from (a) NCEP reanalysis, (b) the control simulation, and (c) the SOM simulation. The contour interval is 2.5 m 2 s 22 , starting at 6.0 m 2 s 22 . Values greater than 8.5 m 2 s 22 are shaded.
variance in the MJO wavenumber-frequency band is lower than observed. Compositing analysis will be used throughout this paper to compare fields associated with model intraseasonal oscillations and the observed MJO. The method for compositing the model intraseasonal oscillations is
similar to that derived in Maloney and Hartmann (1998) and MK02. Intraseasonal oscillations in the model simulation will be composited using an index based on the first two empirical orthogonal functions (EOFs) of the 30–90-day equatorial averaged (78N–78S) 850-hPa zonal wind. The leading EOFs are the spatial structures that explain the greatest fraction of the variance of the intraseasonal 850-hPa zonal wind (Kutzbach 1967). Figure 3 shows the leading EOFs for the control and SOM simulations. For both simulations, EOF1 peaks in amplitude near the date line, and EOF2 peaks over the western Pacific. EOF1 (EOF2) for the control simulation explains 28% (20%) of the variance. EOF1 (EOF2) for the SOM simulation explains 30% (23%) of the variance. The leading EOFs in the control and SOM simulations are significantly different from the other EOFs
1 NOVEMBER 2002
MALONEY AND KIEHL
at the 95% confidence level based on the criterion of North et al. (1982). For the control simulation, the principal component (PC) of EOF2 (PC2) leads PC1 by 11 days at a correlation of 0.41. This correlation is statistically significant from zero at the 95% confidence level. Similarly for the SOM simulation, PC2 leads PC1 by 12 days at a correlation of 0.56. The leading EOFs for the control and SOM simulation thus form quadrature pairs that represent eastward-propagating wind anomalies along the equator. The indices for compositing the control and SOM simulation intraseasonal oscillations are derived by using linear combinations of the first two PCs, similar to the method of Maloney and Hartmann (1998). The control simulation index is constructed by adding PC2 to the value of PC1 11 days later, and the SOM simulation index is constructed by adding PC2 to the value of PC1 12 days later. Key events are chosen as periods when the indices have maxima greater than one standard deviation from zero. Seventy control simulation events and 58 SOM simulation events are selected based on this criterion during all seasons. Thirty-six control simulation events and 30 SOM simulation events occur during June–November. Phases 1–9 are then assigned for each event using the method described in Maloney and Hartmann (1998). Phase 5 is assigned to the time in each event with maximum positive amplitude, when westerly wind anomalies peak over the far west Pacific, and east Pacific wind anomalies are generally easterly. Phases 1 and 9 are assigned to the minima in the index before and after phase 5, respectively. Phases 3 and 7 correspond to zero crossing points before and after phase 5, respectively. Phases 2, 4, 6, and 8 are assigned midway between the other phases. Events are averaged to create a composite intraseasonal oscillation life cycle. Construction of the observed MJO composites uses a very similar method to the one described above for the model composites, and is described in MK02 and Maloney and Hartmann (1998). The EOFs used to composite the observed MJO are slightly different than those derived from the model, primarily because model intraseasonal variability over the Indian Ocean is weaker than observed. The strategy of using a quadrature pair of EOFs to define the observed MJO index is similar to that used for the model simulation intraseasonal oscillations, however. Maximum values of the observed MJO index (phase 5) correspond to peak 850-mb westerly anomalies over the Indian Ocean. Because Indian Ocean intraseasonal variability is weaker than observed in both model configurations, and the leading model EOFs do not have large amplitude over the Indian Ocean, the maximum amplitude of the model indices for each event (phase 5) correspond to peak westerly anomalies over the far western Pacific Ocean. A phase shift is therefore apparent between the observed and model composite life cycles, as seen below. This phase shift in no way inhibits a direct comparison of inter-
2993
FIG. 4. Composite equatorial (78N–78S averaged) 30–90-day 850mb wind (contours) and precipitation (shading) anomalies as a function of MJO phase for (a) observations (NCEP winds and Xie–Arkin precipitation), (b) the control simulation, and (c) the SOM simulation. Events during all seasons from the composites. Contour interval is 0.50 m s 21, starting at 0.25 m s 21 . Easterlies are dashed. Dark shading represents precipitation anomalies greater than 0.3 mm day 21 . Light shading represents anomalies less than 20.3 mm day 21 .
actions among east Pacific convection, winds, and SSTs among the models and observations. Figure 4 shows composite equatorial (78N–78S averaged) intraseasonal 850-hPa zonal wind and precipitation anomalies as a function of phase for the observed MJO, the control intraseasonal oscillation (ISO), and the SOM ISO. Events during all seasons are used to construct the Fig. 4 composites. June–November composite behavior will be examined below. We will use
2994
JOURNAL OF CLIMATE
the term ISO to refer to the model composite signal, because model variability does show some differences from the observed MJO. Both observations and the model simulations show eastward-propagating intraseasonal zonal wind and precipitation anomalies along the equator, with the phase relationship between zonal winds and precipitation being approximately the same in the models as in observations. Note that the definition of phases is somewhat different in the models than observations because of slight differences in the method of constructing the index. An increase in the speed of propagation occurs in all composites to the east of the date line. Coherent zonal wind and precipitation signals extend farther westward across the Indian Ocean in observations than in the models, although the SOM simulation represents a slight improvement over the control simulation in this respect. Equatorial zonal wind anomalies are also slightly stronger in the SOM simulation than in the control. The SOM simulation is not necessarily more realistic in this regard because ISO composites in both the control and SOM simulations have more vigorous wind anomalies than the observed MJO. The magnitude of SOM precipitation anomalies is larger than in the control simulation (contours not shown), although somewhat weaker than observed. The most important feature apparent in Fig. 4 for the purposes of this paper is the eastward propagation of model equatorial wind anomalies into the eastern Pacific Ocean, as in observations. The eastward propagation of zonal wind anomalies into the eastern Pacific in the model composites makes a direct comparison to the observed MJO possible. The initiation of intraseasonal boreal summer convection over the east Pacific hurricane region is associated with the extension of the observed MJO signal into the eastern Pacific (Maloney and Hartmann 2000). Intraseasonal convection and wind anomalies are amplified over the east Pacific during June–November through local feedbacks. We will compare interactions among convection, the atmospheric circulation, and SSTs during different ISO wind regimes in the two model configurations to determine whether an interactive ocean is important for producing realistic east Pacific intraseasonal variability. 4. East Pacific during June–November Eastern Pacific June–November mean surface winds and precipitation for observations, the control simulation, and the SOM simulation are shown in Figs. 5 and 6, respectively. Observed winds from NCEP reanalysis and observed precipitation from the Xie–Arkin dataset are used. Control and SOM mean winds over the east Pacific are similar to the NCEP winds (Fig. 5). The most significant difference is that cross-equatorial flow in both model simulations is stronger than in the NCEP reanalysis product. A comparison with scatterometer data indicates that the NCEP reanalysis product may underestimate cross-equatorial flow in the east Pacific
VOLUME 15
FIG. 5. Mean Jun–Nov surface winds from (a) NCEP reanalysis, (b) the control simulation, and (c) the SOM simulation.
(Milliff et al. 1999), however. Differences in mean winds between the control and SOM simulations are slight. Precipitation in both model simulations tends to be stronger near Central America than observed (Fig. 6), and also tends to be weaker in the intertropical convergence zone to the west of 1108W. It will be shown below that the strongest observed MJO convective anomalies and SOM simulation ISO anomalies are nearly collocated, so differences between the model and observed precipitation climatologies does not detract from the results discussed below on east Pacific variability. Differences in mean precipitation between the control and SOM simulations are small. Figure 7 shows an observed June–November MJO composite life cycle of intraseasonal 850-hPa winds and precipitation. This figure is identical to Fig. 3 of MK02 except that precipitation is shown instead of outgoing longwave radiation, and the size of the reference wind
1 NOVEMBER 2002
MALONEY AND KIEHL
FIG. 6. Mean Jun–Nov precipitation from (a) Xie–Arkin, (b) the control simulation, and (c) the SOM simulation. The contour interval is 4 mm day 21 , starting at 2 mm day 21 . Values greater than 2 mm day 21 are shaded.
vector has been slightly changed. Periods of westerly MJO-related wind anomalies over the eastern Pacific are associated with enhanced convection over the warm pool. MK02 further found that these westerly periods were associated with enhanced evaporation and reduced shortwave radiation at the sea surface. Weaker convection anomalies of the opposite sign tend to occur to the west of 1158W. East Pacific MJO convection anomalies to the south of Mexico and Central America tend to be quasi stationary, possibly due to the barrier provided by the mountainous terrain of the continental landmass. Easterly MJO-related wind anomalies are associated with suppressed convection, decreased surface evaporation, and positive surface shortwave radiation anomalies (MK02). Maximum warm pool SSTs in the observed composite occur during phases 3 and 4. The next two figures describe an important result of
2995
this study. Figures 8 and 9 show intraseasonal 850-hPa wind and precipitation anomalies for control and SOM simulation ISO composite life cycles. Note that observed composite phases are shifted relative to the model composite phases due to the slightly different compositing technique. Oceanic SOM simulation ISO precipitation anomalies to the south of Mexico and Central America are stronger, much better defined, and extend over a larger area than they do in the control simulation. East Pacific precipitation variations during a SOM ISO life cycle are of the same magnitude as observed MJO precipitation variations, and have similar structure. Oceanic precipitation anomalies in the SOM simulation peak just to the south of Mexico and Central America (the hurricane genesis region), as do observed MJO convective anomalies, although strong SOM precipitation anomalies extend somewhat farther east than in observations. These results clearly show that the inclusion of an interactive slab ocean improves the simulation of intraseasonal convective precipitation in the eastern Pacific during boreal summer, and suggest that SST variations are important for the modulation of eastern Pacific MJO convection. The evolution of the composite wind field in the model is more complex than in observations and is worth discussing. Observed wind anomalies peak in conjunction with the maximum in convective heating (Fig. 7). The east Pacific wind maximum in observations (to the east of 1208W) maximizes in a jetlike structure near 108N (see phases 2 and 6). The evolution of the model composite wind anomalies is much more complex. Control and SOM simulation wind anomalies show northeastward propagation, unlike observations, and tend to have maxima at more than one latitude. Peak wind anomalies in the SOM simulation also tend to lag peak convection over the ocean. Understanding the evolution of model precipitation may help explain the cause of these complex wind anomalies. Strong convective anomalies occur over land in both the control and SOM simulation composites that are not found in the observed composites. Oceanic and land precipitation anomalies are sometimes 1808 out of phase in the SOM simulation (e.g., phases 5 and 6). Further, oceanic precipitation anomalies in the SOM simulation show a tendency to propagate northeastward toward land, unlike in observations. This complexity may cause the maximum wind anomalies in SOM to lag the maximum oceanic convection, since convection over land appears very effective at forcing wind anomalies in the model. Convective anomalies over land in the control simulation appear to be effective at forcing a wind response that is as strong, or stronger, than the wind anomalies in SOM. The effect of opposite-signed convective heating over the land and ocean in the SOM simulation may be to create slightly weaker east Pacific wind anomalies than in the control simulation. A higher-resolution treatment of topography and coastlines in the model, and a more complete parame-
2996
JOURNAL OF CLIMATE
VOLUME 15
FIG. 7. Jun–Nov observed MJO composite 850-mb wind and precipitation anomalies as a function of phase. Precipitation contours are plotted every 1.2 mm day 21 , starting at 0.6 mm day 21 . Negative values are dashed. Values greater (less) than 0.6 mm day 21 (20.6 mm day 21 ) are dark (light) shaded. The reference wind vector is located at the bottom right.
terization of land surface processes are two factors that may improve the simulation of model convection over land. Future work is needed to examine whether such improvements can make the simulation of precipitation
over land, and consequently the local circulation anomalies, more realistic. The differences between model and observed composite circulation features aside, Figs. 7– 9 show that the modified CCM3 coupled to a mixed
1 NOVEMBER 2002
MALONEY AND KIEHL
2997
FIG. 8. As in Fig. 7 except for the control simulation ISO.
layer ocean produces east Pacific oceanic intraseasonal precipitation variations that are much more realistic than in a simulation with fixed SSTs. Figure 10 shows ISO composite intraseasonal SST anomalies from the SOM simulation. Figure 5 of MK02 showed a similar composite for the observed MJO. SST
variations of greater than 0.68C can be found off the equator to the south of Mexico and Central America during a SOM simulation ISO life cycle. These SST variations are of comparable magnitude to those observed with the MJO, and correspond to saturationequivalent potential temperature variations of about 48C.
2998
JOURNAL OF CLIMATE
VOLUME 15
FIG. 9. As in Fig. 7 except for the SOM simulation ISO.
Individual events can have considerably higher SST and saturation-equivalent potential temperature variations. Model SST variations are highest near 108N, 908W, where annual mean mixed layer depths are 10 m or less (see Fig. 1). Note that SST variations on the equator in the SOM simulation are very weak. The observed MJO
shows a significant SST signal on the equator that is 1808 out of phase with off-equatorial anomalies. MK02 attributed the equatorial signal to ocean dynamics. Because the SOM simulation does not include a dynamical ocean, the lack of an equatorial SST signal in the model is consistent with the idea that observed MJO equatorial
1 NOVEMBER 2002
MALONEY AND KIEHL
2999
FIG. 10. Jun–Nov ISO composite SST anomalies from the SOM simulation as a function of phase. SST contours are plotted every 0.068C, starting at 0.038C. Negative values are dashed. Values greater than 0.098C are shaded.
SST anomalies over the east Pacific are caused by ocean dynamics (e.g., Ekman transport). Whereas observed warm SST anomalies lead observed enhanced precipitation anomalies by about 10 days during significant MJO events, SOM simulation
SST anomalies tend to be closer in phase to precipitation anomalies. Figure 11 shows lag correlations between intraseasonal SSTs and precipitation at two off-equatorial locations in the east Pacific. Correlations are done at an eastern point and a western point. The observed
3000
JOURNAL OF CLIMATE
VOLUME 15
FIG. 12. Model 30–90-day specific humidity anomalies at 108N, 908W as a function of ISO phase. Contours are plotted every 0.07 g kg 21 , starting at 0.035 g kg 21 . Values greater than 0.035 g kg 21 are shaded, and negative contours are dashed.
FIG. 11. Lag correlations between intraseasonal precipitation and SST anomalies at the eastern and western points for the SOM simulation and observations. Positive lags mean precipitation lags SST. The 95% significance levels for the model data are shown.
coordinates of these points are 88N, 898W (east point) and 138N, 1048W (west point), corresponding to the averaging boxes examined in several calculations in MK02. These boxes represent the locations of highest off-equatorial modulation of SST by the MJO in the observed composites. Local results are insensitive to the exact locations of these boxes. Rather than doing an interpolation, the nearest model grid cells to the observed coordinates were used in the comparison. The model coordinates corresponding to the eastern and western points are 108N, 908W (east point) and 138N, 1048W (west point). The lag correlations in Fig. 11 are derived using data 40 days to either side of phase 5 of each significant ISO or MJO event. Conservatively assuming 30 independent samples, and using the t statistic, correlations of magnitude 0.36 or greater are signifi-
cantly different from zero at the 95% confidence level for the SOM simulation. Confidence levels for the observations are quite similar since 31 significant MJO events are used in the correlations. Confidence limits for the model correlations are shown in Fig. 11. SOM intraseasonal SST and precipitation anomalies tend to be in phase at the eastern and western points during significant ISO events. Observed intraseasonal SST anomalies tend to lead intraseasonal precipitation anomalies during significant MJO events, although by 10 days at most. The closer phase correspondence between convection and SST in the model than observations may be due to deficiencies in how model convection is parameterized. The RAS convection scheme used in the model does not account for convective inhibition near the top of the boundary layer that may delay the onset of deep convection in nature (Mapes 2000). The convection scheme may also be too insensitive to midtropospheric dryness, which must be overcome for deep convection to occur (Tompkins 2001). Waliser (1996) noted regions of warm oceanic temperatures (hot spots) where convection is suppressed due to subsidence and drying of the midand lower tropospheres. SST variations likely affect model intraseasonal convection in a couple of ways. SST variations cause surface-saturation-equivalent potential temperature variations of 48C during a SOM ISO life cycle, as noted above. These variations directly influence the buoyancy of lower-tropospheric air parcels, and therefore directly impact convective available potential energy. Figure 12 shows the composite vertical structure of 30–90-day specific humidity anomalies at 108N, 908W in the SOM simulation. Specific humidity variations of 0.5 g kg 21 occur in the lower boundary layer during an ISO life
1 NOVEMBER 2002
MALONEY AND KIEHL
cycle. Boundary layer moisture variations in the control simulation (not shown) are much weaker. The realistic spatial extent and magnitude of intraseasonal convective anomalies in the SOM simulation are likely caused by these low-level specific humidity variations. Maloney (2002) showed using the NCAR CCM3.6 with a modified relaxed Arakawa–Schubert convection scheme that model convection was particularly sensitive to variations in boundary layer specific humidity. Composite SST variations of 0.78C occur at 108N, 908W during a SOM ISO life cycle. Assuming a boundary layer relative humidity of 80% and a mean SST of 298C in this region, the SOM SST variations can account for specific humidity variations of up to 0.8 g kg 21 in the lower boundary layer. SST variations are clearly more than sufficient to account for the 0.5 g kg 21 composite humidity variation in the SOM simulation. Anomalously low evaporation leading up to phases 5 and 6 may somewhat limit the effect of SST variations on boundary layer– specific humidity, however. MK02 describe another way that SST variations can affect convection. MK02 showed using the diagnostic model of Nigam and Chung (2000) that positive SST anomalies can force enhanced surface convergence into the east Pacific MJO convective region, via the mechanism described by Lindzen and Nigam (1987). Surface convergence can produce an upward flux of water vapor in the boundary layer that feeds convection. Strengthened convection can then, in turn, further strengthen surface convergence. Composite surface convergence anomalies over the eastern Pacific are considerably stronger in the SOM simulation than in the control simulation (not shown). Because anomalous convection can also enhance surface convergence, the importance of SST anomalies in forcing SOM simulation surface convergence is difficult to determine. Since the SOM simulation only produces an off-equatorial eastern Pacific SST signal, the effect of SST anomalies in forcing surface convergence may be stronger when coupling to a dynamic ocean model that also produces equatorial SST anomalies. Equatorial intraseasonal SST anomalies, out of phase with those off the equator, can create a stronger latitudinal SST gradient. A GCM coupled to a dynamic ocean model will be examined in a future study. 5. Analysis of surface fluxes The surface flux components that dominate east Pacific SST variations in the SOM June–November ISO composite will now be determined. MK02 found that surface latent heat and shortwave radiation fluxes were responsible for observed east Pacific off-equatorial SST variations in the June–November MJO composite. Observed MJO surface latent heat and shortwave anomalies lead SST anomalies by 10–15 days. We will now determine whether similar relationships among east Pacific intraseasonal SSTs and surface latent heat and shortwave fluxes occur in the SOM simulation. The effects
3001
of observed longwave and sensible heat flux anomalies on east Pacific intraseasonal SSTs are small. Figures 13 and 14 show SOM simulation ISO composite surface latent heat flux (LH) and shortwave radiation (SW) anomalies, respectively. These flux terms clearly dominate heat variations over the east Pacific during a composite ISO life cycle, of which LH is the leading term. Sensible heat (SH) and longwave radiation flux (LW) anomalies are considerably smaller and not displayed. Positive values in Figs. 13 and 14 are defined as a net flux into the ocean. These composites can be compared to Figs. 6 and 7 of MK02, albeit with a phase shift of 908. Model LH and SW variations maximize off the equator in the east Pacific to the south of Mexico and Central America, as in observations. The magnitude of flux anomalies in the SOM simulation is slightly reduced from observations in general, although evaporation anomalies of greater than 22 W m 22 can be found over the eastern Pacific during phase 7. As mentioned above, the tendency for strong convective anomalies to occur over land, out of phase with those over ocean, may complicate the latent heat flux anomaly signal in the SOM simulation. A better indication of the magnitude of flux anomalies will be given below when examining fluxes at selected eastern Pacific grid boxes. Decreased evaporation and increased SW to the surface tend to lead enhanced SSTs by a phase or two, similar to the relationship observed, although the model flux anomalies tend to lead SSTs by about 5 fewer days than observations (see below). We will now examine composite surface fluxes in the SOM simulation at a couple of off-equatorial points in the eastern Pacific Ocean (Fig. 15). These points are the same locations as the eastern (108N, 908W) and western points (138N, 1048W) described in section 4. Remember that MK02 analyzed observed MJO surface fluxes very near these two locations. Enhanced SOM ISO precipitation maximizes at these two points during phase 6. The eastern point is near the location of largest ISO composite SST variation in the SOM simulation. At the eastern point, SST variations of magnitude greater than 0.68C occur during a model ISO life cycle. The LH and SW anomalies, which dominate the surface flux signal at the eastern point, lead SST by about two phases. Positive latent heat flux anomalies (reduced evaporation) tend to be associated with easterly wind anomalies (see Fig. 9). Negative latent heat flux anomalies are associated with westerly anomalies, suggesting that a wind evaporation heat exchange mechanism similar to that proposed by Emanuel (1987) may help regulate intraseasonal convection over the east Pacific in the SOM simulation. The climatological low-level flow is weak or slightly westerly over the eastern Pacific during June–November (Fig. 5). Maximum evaporation anomalies in the SOM simulation tend to occur slightly after the maximum east Pacific convection, whereas observed convection and evaporation anomalies are in phase. The SH and LW anomalies are comparatively small. The
3002
JOURNAL OF CLIMATE
VOLUME 15
FIG. 13. Jun–Nov ISO composite SOM simulation latent heat flux anomalies as a function of phase. Flux contours are plotted every 4 W m 22 , starting at 2 W m 22 . Positive is defined as a net flux into the ocean. Negative values are dashed. Values greater than 6 W m 22 are shaded.
dominant flux anomalies at the western point are also SW and LH, and are of comparable magnitude to those at the eastern point. Note that SST variations at the western point are on the order of 0.28C, smaller than
those at eastern point due to the deeper oceanic mixed layer depths at the western point (see Fig. 1). Now a lag correlation analysis between surface flux and SST anomalies at the eastern and western points
1 NOVEMBER 2002
MALONEY AND KIEHL
3003
FIG. 14. Same as Fig. 13 except for surface shortwave radiation anomalies.
during significant ISO events in the SOM simulation will be presented. Correlations are derived in an identical manner to the correlations presented in Fig. 11. Figure 16 shows lag correlations between flux anomalies and SST at the eastern and western points. These correlation plots can be directly compared to Fig. 9 of
MK02. The LH and SW anomalies are significantly correlated with SST at both the eastern and western points. The LH correlations peak near 0.7 at both points when LH leads SST by about 10 days. The SW is significantly correlated with SST at a lead of 5–10 days at the eastern point, and at a lead of 10–15 days at the western point.
3004
JOURNAL OF CLIMATE
VOLUME 15
FIG. 15. Jun–Nov SOM simulation composite ISO SST anomalies (dash–dot) and LH (thick solid), SH (thin dashed), SW (thick dashed), and LW (thin solid) flux anomalies for the (a) eastern point (108N, 908W) and (b) western point (138N, 1048W). Positive fluxes are defined as a net flux into the ocean.
A comparison with Fig. 9 of MK02 indicates that SOM simulation flux anomalies over the eastern Pacific tend to lead SST anomalies by fewer days than observed, although differences in phase relationship between the SOM simulation and observations are not statistically significant. Figures 11 and 16 indicate that the east Pacific lag in correlation between SOM intraseasonal precipitation and SST anomalies is smaller than the lag in correlation between negative SW and SST anomalies. A lag correlation analysis between precipitation and SW anomalies may help explain this discrepancy (Fig. 17). At the eastern point, SOM simulation SW and precipitation are correlated at about 20.45 when shortwave lags precipitation by 5 days. Although this correlation in the model is weakly significant, only about 20% of the variance of the SW anomalies can be explained by precipitation during significant model ISO events, indicating that the relationship between model intraseasonal SW and pre-
FIG. 16. Lag correlations between SOM simulation intraseasonal SST anomalies and LH (thick solid), SH (thin dashed), SW (thick dashed), and LW (thin solid) flux anomalies for the (a) eastern point (108N, 908W) and (b) western point (138N, 1048W). Negative lags mean fluxes lead SST. The 95% significance levels are shown.
cipitation anomalies is not strong. The lag between precipitation and SW may also contribute to the phase relationship among SST, SW, and precipitation mentioned above. Explained variance for the SOM simulation is just slightly higher at the western point. Much of the intraseasonal SW variance in the model is likely explained by cloudiness not associated with significant precipitation. The amount of observed SW variance explained by precipitation peaks near 50% or greater at the two east Pacific locations, indicating that the treatment of clouds and radiative forcing in the model may need improvement. To summarize, east Pacific intraseasonal SST variations within SOM ISO convective regions are primarily controlled by surface latent heat flux and shortwave ra-
1 NOVEMBER 2002
MALONEY AND KIEHL
3005
50% in the observed MJO, suggesting that the treatment of clouds and radiative forcing in the model may need improvement. 6. Conclusions
FIG. 17. Lag correlations between intraseasonal precipitation and surface shortwave radiation anomalies at the eastern and western points for the SOM simulation and observations. Positive lags mean shortwave lags precipitation. The 95% significance levels for the model data are shown.
diation anomalies, as in the observed MJO. Reduced evaporation and positive shortwave radiation anomalies maximize about 10 days before the strongest east Pacific warm pool SST and precipitation anomalies during a SOM ISO life cycle. Periods of enhanced SST are associated with intraseasonal eastern Pacific convective anomalies in the SOM simulation that are of similar spatial extent and magnitude to observed. The lag in correlation between SOM intraseasonal precipitation and SST anomalies is smaller than the lag in correlation between SOM negative surface shortwave and SST anomalies. This relationship may exist because east Pacific precipitation anomalies explain only about 20% of the shortwave radiation variance during a SOM simulation ISO life cycle. Explained variance is closer to
Experiments using the NCAR CCM3.6 with relaxed Arakawa–Schubert convection show that including an interactive slab ocean model (SOM) improves the simulation of eastern Pacific convection during a composite June–November intraseasonal oscillation (ISO) life cycle. Oceanic temperature changes in the slab ocean model are forced by surface heat fluxes and a specified transport term operating on Levitus annual mean mixed layer depths. The SOM simulation produces precipitation variations over the east Pacific warm pool (hurricane region) of similar magnitude to those associated with the observed Madden–Julian oscillation (MJO), and of similar spatial extent (Maloney and Kiehl 2002). A composite June–November ISO life cycle in a model simulation using climatological fixed SSTs (control) produces weaker, less coherent, and less widespread eastern Pacific warm pool convective anomalies than in the SOM simulation. The strength of composite low-level wind anomalies in the SOM simulation is somewhat reduced from the control simulation, likely due to the competing effects of opposite signed convective anomalies over land and ocean. Both simulations have stronger variability of convection over land than observed. Wind anomalies structures in the SOM and control simulations are therefore more complex than those in the observed MJO, where significant MJO convective anomalies are primarily confined to ocean areas to the south of Mexico and Central America. Off-equatorial SST variations of up to 0.68C occur to the south of Mexico and Central America during a June– November ISO life cycle in the SOM simulation. These SST variations are comparable to those observed with the MJO. SST anomalies in the model may foster increased convective variability through direct changes to boundary layer moist static energy, or through forcing of the surface flow by pressure perturbations as described in Lindzen and Nigam (1987). The direct thermodynamic effect of SST variations may be sufficient to account for composite boundary layer–specific humidity variations in the SOM simulation. Positive SST anomalies in the SOM simulation tend to be closer in phase to enhanced precipitation than in observations. Observed SSTs tend to lead enhanced convection by slightly less than 10 days. Inclusion of more realistic physics in the model convection parameterization may help to reduce this discrepancy between the model and observations, since mechanisms such as convective inhibition are not parameterized in the model (Mapes 2000). Although the observed MJO produces equatorial SST anomalies in the eastern Pacific (MK02), no equatorial signal is produced in the SOM simulation. This finding reinforces the assertion of MK02 that ocean dy-
3006
JOURNAL OF CLIMATE
namics are responsible for the equatorial MJO SST signal, because the slab ocean model used in this study does not include a dynamical ocean. Eastern Pacific surface latent heat flux and shortwave radiation anomalies are the dominant terms in controlling June–November intraseasonal SST variations during a SOM simulation ISO life cycle, and are also the dominant terms in controlling east Pacific intraseasonal SST during observed MJO events. Flux anomalies in the model are somewhat reduced from observations. Positive heat flux anomalies into the ocean lead enhanced SST by about 10 days in the SOM simulation during significant ISO events. Intraseasonal precipitation anomalies explain less of the intraseasonal surface shortwave radiation variance in the model than in observations, indicating room for improvement in how clouds and radiation are parameterized in the model. Acknowledgments. Two anonymous reviewers provided helpful comments that improved the manuscript. Xie–Arkin precipitation, Reynolds SST, and NCEP reanalysis data were provided by the NOAA–CIRES Climate Diagnostics Center, Boulder, Colorado. This research was supported by the NOAA Postdoctoral Program in Climate and Global Change, administered by the University Corporation for Atmospheric Research. REFERENCES Emanuel, K. A., 1987: An air–sea interaction model of intraseasonal oscillations in the Tropics. J. Atmos. Sci., 44, 2324–2340. Flatau, M., P. J. Flatau, P. Phoebus, and P. P. Niiler, 1997: The feedback between equatorial convection and local radiative and evaporative processes: The implications for intraseasonal oscillations. J. Atmos. Sci., 54, 2373–2386. Hack, J. J., 1994: Parametrization of moist convection in the National Center for Atmospheric Research community climate model (CCM2). J. Geophys. Res., 99, 5551–5568. Hansen, J., A. Lacis, D. Rind, G. Russell, P. Stone, I. Fung, R. Ruedy, and J. Lerner, 1984: Climate sensitivity: Analysis of feedback mechanisms in climate processes and sensitivity. Climate Processes and Climate Sensitivity, Geophys. Monogr., No. 29, Amer. Geophys. Union, 130–163. Hendon, H. H., 2000: Impact of air–sea coupling on the Madden– Julian oscillation in a general circulation model. J. Atmos. Sci., 57, 3939–3952. ——, 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. Kawamura, R., 1988: Intraseasonal variability of sea surface temperature over the tropical western Pacific. J. Meteor. Soc. Japan, 66, 1007–1012. Kemball-Cook, S., B. Wang, and X. Fu, 2002: Simulation of the intraseasonal oscillation in the ECHAM-4 model: The impact of coupling with an ocean model. J. Atmos. Sci., 59, 1433–1453. Kiehl, J. T., J. J. Hack, G. B. Bonan, B. A. Boville, B. P. Briegleb, D. L. Williamson, and P. J. Rasch, 1996: Description of the NCAR Community Climate Model (CCM3). NCAR Tech. Note NCAR/TN4201STR, 152 pp. [Available from NCAR, Boulder, CO 80307.] ——, ——, ——, ——, D. L. Williamson, and P. J. Rasch, 1998: The National Center for Atmospheric Research Community Climate Model: CCM3. J. Climate, 11, 1131–1150.
VOLUME 15
Krishnamurti, T. N., D. K. Oosterhof, and A. V. Metha, 1988: Air– sea interaction on the timescale of 30–50 days. J. Atmos. Sci., 45, 1304–1322. Kutzbach, J. E., 1967: Empirical eigenvectors of sea-level pressure, surface temperature, and precipitation complexes over North America. J. Appl. Meteor., 6, 791–802. Lau, K.-M., and C.-H. Sui, 1997: Mechanisms of short-term sea surface temperature regulation: Observations during TOGA COARE. J. Climate, 10, 465–472. Levitus, S., 1982: Climatological Atlas of the World Ocean. NOAA Prof. Paper 13, 173 pp. and 17 microfiche. Lindzen, R. S., and S. Nigam, 1987: On the role of sea surface temperature gradients in forcing low-level winds and convergence in the Tropics. J. Atmos. Sci., 44, 2418–2436. Madden, R. A., and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122, 814–837. Magan˜a, V., J. A. Amador, and S. Medina, 1999: The midsummer drought over Mexico and Central America. J. Climate, 12, 1577–1588. Maloney, E. D., 2002: An intraseasonal oscillation composite lifecycle in the NCAR CCM3.6 with modified convection. J. Climate, 15, 964–982. ——, and D. L. Hartmann, 1998: Frictional moisture convergence in a composite life cycle of the Madden–Julian oscillation. J. Climate, 11, 2387–2403. ——, and ——, 2000: Modulation of eastern North Pacific hurricanes by the Madden–Julian oscillation. J. Climate, 13, 1451–1460. ——, and ——, 2001: The sensitivity of intraseasonal variability in the NCAR CCM3 to changes in convective parameterization. J. Climate, 14, 2015–2034. ——, and J. T. Kiehl, 2002: MJO-related SST variations over the tropical eastern Pacific during Northern Hemisphere summer. J. Climate, 15, 675–689. Mapes, B. E., 2000: Convective inhibition, subgrid-scale triggering energy, and stratiform instability in a toy tropical wave model. J. Atmos. Sci., 57, 1515–1535. Milliff, R. F., W. G. Large, J. Morzel, G. Danabasoglu, and T. M. Chen, 1999: Ocean general circulation model sensitivity to forcing from scatterometer winds. J. Geophys. Res., 104, 11 337– 11 358. Moorthi, S., and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 978–1002. Nigam, S., and C. Chung, 2000: ENSO surface winds in CCM3 simulation: Diagnosis of errors. J. Climate, 13, 3172–3186. North, G. R., T. L. Bell, R. F. Cahalan, and F. J. Moeng, 1982: Sampling errors in the estimation of empirical orthogonal functions. Mon. Wea. Rev., 110, 699–706. Reynolds, R. W., and T. M. Smith, 1994: Improved global sea surface temperature analyses using optimum interpolation. J. Climate, 7, 929–948. Schade, L. R., and K. A. Emanuel, 1999: The oceans’s effect on the intensity of tropical cyclones: Results from a simple coupled atmosphere–ocean model. J. Atmos. Sci., 56, 642–651. Sud, Y. C., and A. Molod, 1988: The roles of dry convection, cloud– radiation feedback processes, and the influence of recent improvements in the parameterization of convection in the GLA GCM. Mon. Wea. Rev., 116, 2366–2387. ——, and G. K. Walker, 1999: Microphysics of clouds with the relaxed Arakawa–Schubert scheme (McRAS). Part I: Design and evaluation with GATE phase III data. J. Atmos. Sci., 56, 3196–3220. Tompkins, A. M., 2001: Organization of tropical convection in low vertical wind shears: The role of water vapor. J. Atmos. Sci., 58, 529–545. Waliser, D. E., 1996: Formation and limiting mechanisms for very high sea surface temperatures: Linking the dynamics and thermodynamics. J. Climate, 9, 161–188. ——, 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. Wang, B., and X. Xie, 1998: Coupled modes of the warm pool climate
1 NOVEMBER 2002
MALONEY AND KIEHL
system. Part I: The role of air–sea interaction in maintaining Madden–Julian oscillation. J. Climate, 11, 2116–2135. Xie, P., and P. A. Arkin, 1996: Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J. Climate, 9, 840–858. Zhang, C., 1996: Atmospheric intraseasonal variability at the surface
3007
in the tropical western Pacific Ocean. J. Atmos. Sci., 53, 739– 758. Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the Canadian Climate Centre General Circulation Model. Atmos.– Ocean, 33, 407–446.
