1. Introduction
Accurate simulations of tropical precipitation remain a challenge for atmospheric climate models (AGCMs). Basic dynamical issues such as the relationship between low-level convergence and precipitation remain unresolved. Recent studies suggest that intertropical convergence zones (ITCZs) identified using precipitation or outgoing longwave radiation (OLR) may not always correspond with convergence zones identified using satellite surface wind measurements (e.g., Liu and Xie 2002). Earlier Hastenrath and Lamb (1977), using ship-based wind observations, also concluded that surface convergence may exist in the absence of precipitation. Nevertheless, determining the strength of surface wind convergence in nature remains a challenge. Perhaps as a result of this observational gap, little attention has been paid to examining convergence–precipitation coupling in AGCM simulations, even though all the necessary quantities are easily accessible.
A common problem in AGCM precipitation simulations, which may be related to PBL–precipitation coupling, is the so-called double ITCZ bias (e.g., Meehl and Arblaster 1998). Many AGCMs form a spurious second ITCZ in the Southern Hemisphere (8°–10°S) under conditions in which observed precipitation is concentrated in a single ITCZ centered around 10°N. While nature does show hints of a southern ITCZ over the Pacific, particularly during March through May (Zhang 2001), this feature in AGCMs is usually too strong and persistent, lasting through the northern warm season June–September. The occurrence of double ITCZs in AGCMs leads to large rms errors in simulated precipitation since it represents a spurious rearrangement of the most intense precipitation on earth. Connections between double ITCZs and other AGCM simulation biases have not been conclusively established. However, it is clearly of concern to climate modelers if AGCMs are producing large errors in the horizontal distribution of atmospheric latent heating. Finally, the wide distribution and similar structure of this bias in a variety of AGCMs suggests the existence of a shared misunderstanding in current implementations of convection parameterizations.
In this study we will examine the connection between PBL convergence and precipitation and the double ITCZ bias in the NASA Seasonal-to-Interannual Prediction Project version 2 (NSIPP-2) AGCM. A principal motivation for performing this work is a robust sensitivity in the NSIPP AGCM’s tropical precipitation to the strength of rain reevaporation. With stronger rain reevaporation the model tends toward a realistic single ITCZ configuration. With weak reevaporation the model produces a strong double ITCZ. This sensitivity has existed in earlier versions of the NSIPP AGCM despite substantially different formulations of reevaporation. Although this sensitivity has been useful in empirical “tuning” of the NSIPP AGCM to improve precipitation simulations, the physical origin of the sensitivity has not been explained. Anecdotal evidence from other modeling groups suggests that this sensitivity may exist in some form in other AGCMs (I. M. Held 2005, personal communication) and also that other sensitivities may exist to parameters such as cumulus friction (GFDL Global Atmospheric Model Development Team 2004).
The goals of this study are to shed light on mechanisms controlling the formation of double ITCZs in the NSIPP AGCM and to suggest relevant, parameterization-independent diagnostics that can be applied to other AGCM simulations. The paper is organized as follows. Section 2 provides a description of the AGCM used in this study. Section 3 outlines the AGCM experiments performed. Section 4 presents the basic sensitivity of the model simulations to reevaporation. Seasonal mean fields are shown, as well as some analysis of vertical profiles, reevaporation tendencies, and high frequency transients. Section 5 describes three experiments with modified physics including changes to the vertical profile of reevaporation cooling and the addition of a simple diffusive cumulus momentum transport (DCMT) parameterization. Section 6 analyzes the mass and water vapor budgets in the simulations. This analysis addresses the questions of how reevaporation suppresses precipitation in the southern ITCZ and why the suppression operates preferentially on the southern ITCZ.
2. Model description
We use a development version of the NSIPP-2 AGCM (NSIPP-2.0) for this study. NSIPP-2.0 was developed from the NSIPP-1 AGCM, which was documented in Bacmeister et al. (2000) and Bacmeister and Suarez (2002). Simulated seasonal means and responses to interannual SST variation in NSIPP-1 were both in good agreement with meteorological analyses (e.g., Schubert et al. 2001, 2002). The significant modifications to NSIPP-2.0 and NSIPP-1 involve the cloud, boundary layer, and convection schemes. These include introduction of a prognostic cloud scheme in place of the Slingo (1987)-type diagnostic scheme used in NSIPP-1, as well as a simple moist boundary layer entrainment scheme, which is called in addition to the existing first-order dry turbulence parameterization of Louis et al. (1982). These modifications were aimed at improving the models simulation of subtropical marine stratus decks and, while they also impact simulated precipitation in the Tropics, they do not affect the general nature of the ITCZ sensitivities examined in this study. Cloud fields from NSIPP-2 are examined by Zhang et al. (2005).
The dynamical core of NSIPP-2.0 is the same as in NSIPP-1 and is described in Suarez and Takacs (1995). Radiative effects in NSIPP-2.0 are parameterized using the approach of Chou and Suarez (1994). Land surface effects are parameterized according to Koster and Suarez (1996), and orographic wave drag is treated according to Zhou et al. (1996).
a. Convection
Convection in the NSIPP AGCM is parameterized according to the relaxed Arakawa–Schubert (RAS) scheme of Moorthi and Suarez (1992). The implementation of RAS in NSIPP-2.0 is modified to include a convective condensate calculation with autoconversion to rain. RAS works by invoking a series of linearly entraining plumes (or “cloud types”) that detrain at selected levels in the vertical. Consistency is achieved by calculating the entrainment rate necessary to ensure zero buoyancy at the selected level. RAS is flexible as far as the number and distribution of plumes or cloud types tested. Our implementation invokes 30 cloud types per gridbox per physics time step. These are drawn at random from a uniform distribution in σ. We also emphasize that our implementation does not include an explicit downdraft parameterization.
b. Prognostic cloud condensate scheme
The NSIPP-2 prognostic condensate scheme considers only a single phase of condensate but tracks two separate species of condensate: a large-scale species qc,LS originating from gridbox condensation and an “anvil” species qc, originating from detraining convection. The rationale for this separation is that both the subgrid statistics and the microphysical properties of rapidly processed anvil condensate may be distinct from those of condensate produced by slower, large-scale dynamics (e.g., Lawson 2003). The key distinctions in our current scheme are slower autoconversion and higher number densities for qc,AN. These higher assumed number densities for qc,AN enter into the optical thickness calculation used by the model’s radiation scheme. We impose an arbitrary e-folding time of 3 h for conversion of qc,AN to qc,LS. A third species, convective condensate qc,, is calculated internally within each RAS cloud type but does not interact with the model’s radiation calculation. When ice–liquid partitioning of total condensate is needed, by the radiation scheme, for example, it is accomplished diagnostically using a linear ramp in temperature that decreases from 1 (all ice) below 263 K to 0 (all liquid) above 273 K.
c. Convective autoconversion and reevaporation
Our basic approach in parameterizing convective microphysical processes is based on a Lagrangian parcel picture. We estimate an updraft speed for each plume in RAS (Bacmeister 2005) that is combined with the model’s vertical grid spacing to give a time interval for autoconversion in a given model layer. Autoconversion rates are determined from a nonlinear temperature-dependent expression (Sundqvist 1988). The approach used is similar to that in Sud and Walker (1999), although we employ a cruder calculation for the convective updraft speed. Profiles of convective precipitating condensate qp, are accumulated over all RAS plumes and then passed to a scheme that accumulates the condensate and also calculates reevaporation, accretion, and surface precipitation fluxes. In addition to precipitating condensate produced by convection, our scheme considers autoconversion of qc,AN and qc,LS. These autoconversions are calculated separately using the Sundqvist (1988) formulation to give two additional precipitating species qp,AN and qp,LS.
Reevaporation is treated separately for each of the three streams of precipitation (“showers”) qp,CN, qp,AN, and qp,LS. This calculation also proceeds according to a Lagrangian viewpoint. First, an estimate of the local subgrid-scale precipitation rate is made using the grid mean precipitation flux and estimates of fractional shower area. A representative particle size for this precipitation is estimated from a Marshall and Palmer (1948) distribution. This particle size gives an evaporation rate, fall speed, ventilation factor, and residence time within a given model layer. These quantities are used to calculate a net loss of precipitating condensate due to evaporation during a time step. In NSIPP-2 both liquid and frozen precipitation are treated in the same way. We also allow a fraction of the convective rain shower to be “shielded” from reevaporation. This is meant to represent rain falling through a saturated environment such as a convective tower or saturated downdraft.
In the experiments discussed here bulk reevaporation of convective precipitation qp,CN was modified by changing shear-dependent parameters that control the shielded fraction, as well as the relationship between diagnosed updraft areal fraction and convective shower area. Roughly speaking more reevaporation is allowed in high shear environments in all experiments, but the strength of this shear dependence is changed to give higher or lower total reevaporation. Experience with previous versions of the NSIPP model suggests that the details of the rain reevaporation scheme are unimportant in producing the sensitivities discussed here. For clarity we will simply refer to three settings of reevaporation parameters—weak, moderate, and strong. More details on the formulation of the reevaporation calculation can be found in (Bacmeister 2005).
3. Description of experiments
We analyze results from six experiments (Table 1). The first three of these, denoted B1, B2, and B3, were performed with the same “baseline” model physics, differing only in the choices made for the rain reevaporation parameters. These experiments were initialized on 1 June 1981 from restarts derived from an existing AMIP simulation and forced with observed SSTs (Reynolds 1988). Experiments B1 and B3 ran through December 1987. Experiment B2 was an AMIP-style run conducted for the National Science Foundation (NSF) Climate Process Team (CPT) on Low-Latitude Cloud Feedbacks and ran through December 1999. For most of the analysis here we will focus on results from 1984 and 1985.
In addition to the three baseline experiments we conducted three experiments with modified physics. In the first of these, H1, the cooling produced by rain reevaporation was redistributed in the vertical; that is, at each time step, total mass-weighted reevaporation cooling below 850 hPa was found and then uniformly applied between 850 and 300 hPa. The moistening from reevaporation was not modified. Thus, moist energy conservation is violated locally but preserved in a column-integrated sense. The motivation for this experiment was to reduce the direct impact of reevaporation on boundary layer circulations, while retaining as much of the original moistening profile as possible.
The remaining two experiments M1 and M2 employed a simple diffusive cumulus momentum transport scheme devised for the Geophysical Fluid Dynamics Laboratory (GFDL) Atmospheric Model version 2 (AM2; GFDL Global Atmospheric Model Development Team 2004). The scheme simply enhances Km proportional to the total local cumulus mass flux diagnosed by RAS. The DCMT scheme has been used in the GFDL model with positive effects on both the simulated precipitation and on the simulated spectrum of ENSO variability in coupled mode. Here we apply DCMT in two experiments: M1, with weak reevaporation settings as in B1, and M2, with moderate reevaporation as in B2. Experiments H1, M1, and M2 were initialized on 1 December 1983 and run through 31 December 1985.
All experiments were conducted at a horizontal resolution of 2° × 2.5° with 40 unequally spaced σ layers. Extensive suites of diagnostic tendency outputs on σ surfaces were saved as daily averages along with standard outputs. These additional diagnostics included most of the significant water substance conversion terms such as moistening by reevaporating rain, which we denote here by
4. Basic model sensitivity to reevaporation
a. Mean seasonal precipitation
Seasonal mean precipitation fields and biases for June–August (JJA, hereafter 3-month periods are denoted by the first letter of each respective month) 1984–85 in experiments B1, B2, and B3 are shown in Figs. 1 and 2 along with observational estimates of precipitation rates from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP; Xie and Arkin 1997). The results illustrate the important climatological control exerted by the reevaporation strength in the NSIPP AGCM. Experiment B1 with weak reevaporation (Figs. 1a, 2a) tends strongly toward a “double ITCZ” configuration with precipitation rates in excess of 8 mm day−1 extending in a narrow, zonally aligned band along 10°S well into the central Pacific. As reevaporation is strengthened in B2 (Figs. 1b, 2b) and B3 (Figs. 1c, 2c) the double ITCZ in all three tropical ocean basins becomes less pronounced, although the change in the Pacific basin is most noticeable owing to its size. In connection with the weakening of the double ITCZ, a dry bias along the equator in the western Pacific in B1 is also reduced with increased reevaporation. Overall, the simulations in experiments B2 and B3 appear to be in better agreement with the CMAP climatology. Pattern correlations for the seven-season (1981–87) JJA mean are significantly lower for B1 than for B2 and B3 (Table 1). Wet biases over summertime tropical continents also appear to decrease as rain reevaporation is strengthened. Over sub-Saharan Africa as well as over the northern Amazon Basin wet biases of over 4 mm day−1 exist in experiment B1, while in B3 these regions are nearly bias free.
Unfortunately, not all precipitation biases are reduced by increasing reevaporation strength in the model. A noticeable deterioration in the simulated precipitation occurs over much of the northern tropical Pacific between Hawaii and Southeast Asia (5° to 20°N, 120°E to 150°W) as reevaporation increases. Stronger reevaporation leads to increasing wet biases in this region, culminating in the >8 mm day−1 biases evident in the “Philippine Hotspot” (15°N, 130°E) in experiment B3 (Figs. 1c, 2c). A JJA dry bias in the Indian Ocean also becomes more pronounced with increasing reevaporation. This strong wet bias is associated with excessively strong low-level monsoon westerlies over Indochina, the Philippines, and surrounding ocean. As will be shown in section 6, much of the water vapor flowing into this region does so in a strong convergent flow located above the 850-hPa surface. By contrast, in the ITCZs mass and water vapor convergence are largely restricted to the PBL.
We have focused on northern summer because the double ITCZ bias, in models which possess it, is most pronounced during the northern warm season, roughly April–November. During December–February (not shown) some double ITCZ bias remains in our weak reevaporation simulation. However, overall the DJF precipitation simulations in all three experiments are in better agreement with the CMAP climatology.
b. Fractional reevaporation
Table 2 lists domain averages of
Results for two warm season continental domains are also shown: an arid one containing the southwestern United States (“WUSA”) and a moist one containing the West African ITCZ region (“WAFR”). Despite large differences in the amounts of precipitation and in the fractions of reevaporated rain, these continental domains exhibit interesting similarities in their sensitivity to reevaporation. Both rain and surface evaporation decrease markedly with increasing rain reevaporation. This is in contrast to the situation over ocean (SITCZ and NITCZ) where surface evaporation is largely unaffected by rain reevaporation. This may reflect an additional feedback between land surface processes and precipitation. Note the large values of f (column 6, Table 2) in WUSA for all experiments. These may be related to the lack of an explicit downdraft parameterization in the model.
Observational estimates of domain-averaged precipitation from CMAP are also shown in Table 2. Comparisons of these with the simulation results tend to confirm that experiment B2 possesses the “best” precipitation, as implied by the pattern correlations and normalized variances in Table 1. Comparison of observations and simulations in WUSA show that our model has a pronounced dry bias in this region, probably related to excessive reevaporation of rain. Unfortunately, global observational estimates of ∫
c. Vertical profile of rain reevaporation
Figure 5 shows seasonal mean profiles of reevaporation tendency
Reevaporation of condensate is not a process for which we have direct observational data to validate models. However, efforts are underway to infer some of the gross features of reevaporation from Tropical Rainfall Measuring Mission (TRMM) radar precipitation rate profiles. Initial examination suggests that significant disagreements may exist between simulated reevaporation profiles and TRMM inferences (F. R. Robertson 2005, personal communication).
d. Water vapor distribution
Figure 6 shows mean water vapor profiles for experiments B1–B3 in boxes SITCZ and NITCZ along with estimates from National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR; Kalnay et al. 1996) and 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40; Simmons and Gibson 2000) reanalyses. In NITCZ (Fig. 6, top) the two reanalyses disagree by nearly 3 g kg−1 over much of the layer below 300 hPa. The three model profiles generally lie between the two reanalysis profiles. The profile for experiment B3 is up to 2 g kg−1 wetter than that for B1 with the largest differences centered around 700 hPa. The profile for B2 lies between those for B1 and B3. In Box SITCZ (Fig. 6, bottom) the situation is somewhat different. The q profiles from our three model experiments are quite similar to the corresponding profiles in NITCZ with B3 the wettest, B1 the driest, and B2 in between. The NCEP reanalysis profile in SITCZ is also similar to that in NITCZ. However, the ERA-40 reanalysis profile is significantly drier and here agrees closely with the NCEP profile. Generally speaking it is clear that increasing reevaporation in our model leads to midtropospheric moistening. However, the differences in q-profiles for different reevaporation strengths appear to be comparable to uncertainties in reanalysis q-profiles. Comparison with SSMI total precipitable water (TPW) measurements (not shown) exhibit a small but worsening global mean wet bias as reevaporation increases from B1 to B3, although the spatial distribution of simulated TPW improves.
e. Relation of low-level convergence and rainfall
The strength of the high-frequency coupling between low-level flow convergence and precipitation in nature is still not well known (e.g., Gu and Zhang 2002). However, there are indications that the connection between low-level convergence and precipitation at time scales of several days and shorter may not be as strong as commonly assumed. We examine this coupling in our simulations by looking at the correlation of daily mean vertical motion at 850-hPa ω850 with daily rainfall
The rms amplitude of ω̃850 is shown in Fig. 8. Maps of
Figure 9 shows Hovmoeller diagrams of
The simulated ITCZ disturbances in Fig. 9 appear to amplify as they move west. Figure 10a shows probability density functions (PDFs) of ω850 (unfiltered) in a box immediately west of the South American coast (10°–6°S, 110°–97.5°W) for May–October 1985. This box is situated over cool water and can be thought of as the “entrance” region for disturbances propagating westward along the southern ITCZ. This figure shows that the amplitude distribution of ω850, before strong interactions with moist heating have occurred, is remarkably similar in experiments B1 (black curve) and B3 (red curve). However, farther west (10°–6°S, 162.5°–175°E) the PDFs diverge markedly (Fig. 10b). In B1 the PDF of ω850 becomes highly skewed with an extensive tail region at large negative values of ω850, that is, upward motion, and a concentrated peak at small positive values. This means that even at the model grid scale there is a distinctly “convective” character to vertical motion, with rare but intense updrafts embedded in extensive but weak subsidence. For B3 the PDF in the western box is also somewhat more skewed than in the entrance region, but the overall shape of the PDF is much less distorted.
Taken together the results in this section imply the existence of a similar background dynamical variability in experiments B1 and B3. This “seed” variability is then amplified to different degrees by interactions with precipitation, with stronger feedback occurring when rain reevaporation is weak. The possible nature of this interaction will be examined by direct experimentation with altered model physics (section 5) and by analysis of water vapor budgets in the ITCZ (section 6).
We have not attempted a detailed comparison of our simulated x–t spectra of rainfall or vertical motion with observations (e.g., Wheeler and Kiladis 1999; Gu and Zhang 2001). However, a cursory look at our model’s background spectra of precipitation along the ITCZ suggests at least a qualitative resemblance with the background OLR spectra in Gu and Zhang (2001).
5. Experiments with altered physics
a. Vertically redistributed reevaporation cooling
From the results shown in section 4 we speculate that the formation of ITCZs in our simulations is at least partly driven by strong coupling between boundary layer convergence and precipitation-related moist heating. The disappearance of double ITCZs with increased reevaporation suggests that this coupling is interrupted by evaporative cooling in or near the PBL. To test this hypothesis, we performed an experiment, H1 (Table 1), in which we arbitrarily removed cooling driven by rain reevaporation below 850 hPa. This experiment used “strong” reevaporation settings as in B3. To conserve at least column-integrated moist static energy, we calculated a mass-weighted integral of the reevaporation cooling between 850 hPa and the surface. This mass-weighted cooling was then redistributed uniformly between 300 and 850 hPa. The corresponding moistening profile was not altered. The motivation for this procedure was to remove the low-level cooling from the strong reevaporation case while minimizing modifications to the moistening profile.
Figure 11a shows seasonal mean JJA 1984–85 precipitation from H1. A strong double ITCZ reappears in this experiment even though reevaporation parameters are as in experiment B3 (Figs. 1c, 2c). In fact, Fig. 11b shows that the fraction of reevaporated rain is generally higher than in B3. Domain averages of precipitation and reevaporation-related quantities for H1 are shown in Table 2. Figure 11c shows r(ω̃850,
b. Diffusive cumulus momentum transport
As described in the introduction, simulations of tropical precipitation in other AGCMs exhibit sensitivities to other processes. The GFDL AGCM has shown a strong sensitivity to the presence of a DCMT parameterization (GFDL Global Atmospheric Model Development Team 2004). The GFDL DCMT scheme is formulated as a momentum diffusivity proportional to the total cumulus mass flux passing through a given level. Its effects are largest at low levels where all clouds, both weakly entraining deep clouds as well as strongly entraining shallow clouds, are present. When DCMT is present in the GFDL model, the simulated tropical precipitation is improved; in particular, the model’s tendency to form double ITCZs is reduced.
We performed two experiments with DCMT (Table 1). One, M1, used reevaporation parameters as in B1. In the baseline model these parameters led to a pronounced double ITCZ bias (Figs. 1a, 2a). The second experiment with DCMT, M2, used reevaporation parameters as in B2. In the baseline model these parameters produced a reasonable simulation of precipitation overall, with a weak double ITCZ bias (Figs. 1b, 2b). The 1984–85 JJA mean precipitation for experiments M1 and M2 is shown in Figs. 12a and 12c. The small double ITCZ bias in experiment B2 is further reduced by the added DCMT in M2, although the wet bias in the northwestern tropical Pacific (0°–20°N, 120°E–180°) appears to have been exacerbated, much as in experiment B3. On the other hand, the strong double ITCZ bias in the Pacific in B1 is not qualitatively reduced by the added DCMT in M1. There are a number of subtle differences between precipitation fields in M1 and B1. The northern ITCZ in M1 is somewhat weaker and more diffuse looking than in B1, particularly in the eastern Pacific between 150° and 120°W. Nevertheless, a strong, well-developed spurious ITCZ persists in the Pacific around 10°S from 150°E to around 130°W. Interestingly, in the Atlantic, there is a more distinct reduction in the double ITCZ bias in experiment M1.
The correlation r(ω̃850,
Figures 10c and 10d show PDFs of ω850 in M1 and M2. In the eastern Pacific (Fig. 10c) the PDFs for M1 and M2 are similar to each other and to those of B1 and B3 (Fig. 10a). This suggests that the addition of DCMT to the model does not reduce the “background” level of dynamical variability in the model. However, in the core of the southern ITCZ region (Fig. 10d) it is clear that the addition of DCMT reduces the magnitude of ω fluctuations compared with the baseline experiments (Fig. 10b). This is evident in the reduced spread of the PDF for M1 (solid light blue) compared with that for B1. The PDF for M1 shows clear evidence of negative skew, which suggests that the dissipative effects of DCMT do not eliminate feedbacks between moist heating and PBL convergence. However, the negative tail in M1 is less pronounced, with more frequent weaker updrafts, while the peak associated with ubiquitous weak subsidence is also less pronounced than in B1.
6. Convective and large-scale transports in the ITCZs
Our results suggest the existence of a feedback between high-frequency, low-level convergence and precipitation, which contributes to the maintenance of the spurious ITCZ. The feedback is weakened by low-level cooling from rain reevaporation, and as a consequence our model’s double ITCZ bias is reduced when parameterized reevaporation is made stronger. The effects of DCMT are more ambiguous, yet it is clear that directly changing the character of dynamical variability in the ITCZs also has an important impact on the resulting simulation of tropical precipitation. In this section, we examine convective transports, as well as mass and water budgets in the ITCZ complex, to understand the relationship between convection, low-level convergence, and precipitation. This analysis also suggests how the southern ITCZ and the northern ITCZ differ in their response to high-frequency low-level convergence.
a. Convective fluxes
Figure 13 shows longitudinal profiles of total, cloud-base, convective mass flux ϕCB from RAS along 8°S and 8°N averaged over May–October 1985. The striking aspect of this figure is the similarity in the magnitude of the convective mass fluxes, everywhere except in the Pacific warm pool region (100° to 140°E along 8°N), where fluxes in B3 are substantially higher. Along the southern ITCZ (Fig. 13b), the total cloud-base convective mass flux in both experiments shows little sensitivity to reevaporation. In fact, somewhat higher mass fluxes exist in experiment B3 (dashed curve), which does not have a pronounced double ITCZ in precipitation. Thus, it is clear that reevaporation in our model does not suppress precipitation in the southern ITCZ by suppressing convection overall. Vertical profiles of convective mass flux (not shown) show some differences between experiments, but these are relatively small compared to the overall fluxes.
The water vapor flux entering convective updrafts is given by ϕCBqCB, where qCB is the specific humidity in the subcloud layer (here an average of the lowest two model layers). We calculate this flux from daily values of ϕCB and qCB so that transient correlations between convection and PBL humidity with time scales longer than one day are included. The mean of this quantity over May–October 1985 is plotted in Figs. 13c and 13d, along with the corresponding average of total precipitation. Here again, the differences between ϕCBqCB are opposite to those in precipitation for the southern ITCZ. Consistent with its generally higher mean values of ϕCB, experiment B3 has higher ϕCBqCB than experiment B1. The ratio of precipitation to ϕCBqCB is another measure of precipitation efficiency, like f in (1). This quantity is shown in Figs. 13e,f. As expected this efficiency is lower overall in experiment B3 with high reevaporation although, interestingly, over parts of the northern ITCZ efficiencies in B3 are higher. Nevertheless, over the southern ITCZ, efficiency in experiment B3 is systematically less than that in B1. The largest proportional decreases in precipitation efficiency caused by increased reevaporation occur around 120°W in the northern ITCZ and between 180° and 150°W in the southern ITCZ.
The mean, May–October 1985, surface evaporation
b. Horizontal transport of water vapor
Since the model evaporation field lacks structure compared with precipitation (e.g., Figs. 13c,d) and time-mean values of ∂t(TPW) are small, the patterns of
The free-tropospheric water vapor flux convergence
The major effects of reevaporation and its associated cooling appear to be 1) to reduce the strength of the time-mean PBL water vapor flux convergence in the southern ITCZ and 2) to increase water vapor removal by transport in the free troposphere above both ITCZs. Both effects contribute to eliminating the double ITCZ in experiment B3, while the lack of an appreciable reevaporation impact on
c. Time behavior of boundary layer convergence



Of greater interest is the relationship between ω̃850 and the straightforward time means of ω850, shown by the thick dashed and solid curves in Figs. 15a–d. In the southern ITCZ (dashed curves) all experiments show a remarkable correlation between −
The nearly linear relationship between
7. Summary and discussion
We examined the effect of parameterized rain reevaporation on tropical precipitation
An obvious question is why eliminating or weakening the feedbacks between low-level convergence and precipitation has an apparently much weaker effect on the northern ITCZ. We addressed this question by examining convective transports and mass and water vapor budgets in the ITCZs. Surface evaporation was found to be a relatively “featureless” field in the ITCZ complex with little variation geographically or between experiments. Therefore, the strong variations in precipitation found in our experiments must be balanced by water vapor transport. As expected, water vapor convergence below 850 hPa is the dominant term over most of the ITCZ system (Fig. 14), so precipitation largely follows this field. Nevertheless, we note that significant water vapor divergence between 850 and 200 hPa is found over the eastern portions of all ITCZs (Figs. 14c,h,m). This divergence is strongest in experiment B3 and contributes to significant reductions in the extent and magnitude of the total vertically integrated water vapor convergence in both ITCZs. The effect is more noticeable in the southern ITCZ due to relative weakness of low-level water vapor convergence there.
To understand why low-level convergence is generally weaker in the southern ITCZ we compared the variance of high-frequency ω̃850 fluctuations with the time-mean profiles of ω850 in both ITCZs (Fig. 15). In the southern ITCZ of all experiments there is a remarkable, nearly linear, relationship between
Experiments with a simple diffusive cumulus momentum transport scheme (DCMT) were encouraging in that they suggested that the Pacific ITCZs will respond directly to a reduction in dynamical variability in the Tropics. However, the effects of DCMT in our model were ambiguous. While variability in ω850 along the ITCZs was somewhat reduced, the time-mean ω850 in the two experiments with DCMT remained comparable to its values in the corresponding experiments without DCMT. The depth of the convergence layer also appeared to increase substantially with DCMT (Fig. 15e). The addition of DCMT did little to eliminate the double ITCZ in the case of weak reevaporation (Fig. 12a), although with moderate reevaporation (Fig. 12c) added DCMT largely eliminated the weak bias present in the corresponding experiment without DCMT (Fig. 1b). We conclude that our implementation of DCMT is not vigorous enough to compete with the strong convergence–precipitation feedbacks present in the model with weak reevaporation. However, in the case of stronger reevaporation, with reduced convergence–precipitation feedbacks, additional DCMT can effectively counteract the growth of disturbances moving along the southern ITCZ.
Recently, Wu et al. (2003) have shown improvement in the Community Climate Model version 3’s (CCM3’s) simulated seasonal evolution of the ITCZ, when a cumulus momentum transport (CMT) scheme (Zhang and Cho 1991) was introduced. The relationship of those results to the present study are not yet clear. However, it is of interest that some form of CMT is found to alleviate tropical precipitation biases to some degree, in at least three different AGCMs.
Gu and Zhang (2001) categorize theories of ITCZ formation into two broad categories: 1) SST forced and 2) internally forced by atmospheric dynamics. Category 2 is further divided into zonally symmetric and zonally asymmetric theories. In our model, the connection of high-frequency PBL convergence and precipitation in the southern ITCZ is suggestive of the wave-driven dynamical mechanisms proposed by Holton et al. (1971), Chang (1973), and Lindzen (1974), and later examined in aquaplanet GCM simulations by Hess et al. (1993).
The situation in our simulated northern ITCZ is not so clear. Here, PBL convergence appears to have an important mode of variability at time scales longer than 15 days. The origins of this slower variability are not clear from our analysis. However, based on its long time scale and its resilience to strong local cooling due to reevaporation, we speculate that this mode in NH ITCZ low-level convergence is part of a larger, slowly varying circulation. This circulation must also be relatively insensitive to the mean changes in the convective heating profiles resulting from changes in reevaporation. A shallow circulation forced by SST gradients (Lindzen and Nigam 1987) appears to satisfy these requirements, although we have presented no evidence here of such forcing.
We hope these results will encourage modelers to examine a number of, relatively simple, parameterization-independent diagnostics of precipitation processes in their models. Examples of such diagnostics examined here include the ratio of reevaporated rain to rain reaching the surface, profiles of domain-averaged reevaporation moistening, correlations of high-frequency time series of vertical motion and precipitation, and PDFs of PBL convergence. This list is certainly insufficient, but we believe more detailed examination of atmospheric water budgets and high-frequency precipitation and dynamical variability in climate models is called for despite the relatively poor observational basis available for validation. These analyses may reveal new dynamically significant similarities between models that suffer similar precipitation biases.
Acknowledgments
The authors thank A. Sobel and an anonymous reviewer for their comments, which led to substantial improvements in this analysis. The work described herein was funded through NASA’s Global Meeting and Assimilation Office.
REFERENCES
Bacmeister, J. T., 2005: Moist processes in the GEOS5 AGCM. [Available online at http://gmao.gsfc.nasa.gov/systems/geos5/MoistProcessesGEOSv2.pdf.].
Bacmeister, J. T., and M. J. Suarez, 2002: Wind stress simulations and the equatorial momentum budget in an AGCM. J. Atmos. Sci., 59 , 3051–3073.
Bacmeister, J. T., P. J. Pegion, S. D. Schubert, and M. J. Suarez, 2000: Atlas of seasonal means simulated by the NSIPP-1 atmospheric GCM. NASA Tech. Memo. 104606, Tech. Rep. 17, 194 pp.
Chang, C. P., 1973: A dynamical model of the intertropical convergence zone. J. Atmos. Sci., 30 , 190–212.
Chou, M-D., and M. J. Suarez, 1994: An efficient thermal infrared radiation parameterization for use in general circulation models. NASA Tech. Memo. 104606, Tech. Rep. 3, 84 pp.
Duchon, C. E., 1979: Lanczos filter in one and two dimensions. J. Appl. Meteor., 18 , 1016–1022.
GFDL Global Atmospheric Model Development Team, 2004: The New GFDL Global Atmosphere and Land Model AM2–LM2: Evaluation with prescribed SST simulations. J. Climate, 17 , 4641–4673.
Gu, G., and C. Zhang, 2001: A spectrum analysis of synoptic-scale disturbances in the ITCZ. J. Climate, 14 , 2725–2739.
Gu, G., and C. Zhang, 2002: Westward-propagating synoptic-scale disturbances and the ITCZ. J. Atmos. Sci., 59 , 1062–1075.
Hastenrath, S., and P. J. Lamb, 1977: Climatic Atlas of the Tropical Atlantic and Eastern Pacific Oceans. University of Wisconsin Press, 117 pp.
Hess, P. G., D. S. Battisti, and P. J. Rasch, 1993: Maintenance of the intertropical convergence zones and the large-scale tropical circulation on a water-covered earth. J. Atmos. Sci., 50 , 691–713.
Holton, J. R., J. M. Wallace, and J. A. Young, 1971: On boundary layer dynamics and the ITCZ. J. Atmos. Sci., 28 , 275–280.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437–471.
Koster, R. D., and M. J. Suarez, 1996: Energy and water balance calculations in the Mosaic LSM. NASA Tech. Memo. 104606, Tech. Rep. 9, 69 pp.
Lawson, P., cited. 2003: A comparison of microphysical properties of wave, cirrus and anvil clouds. CRYSTAL-FACE Science Team Meeting. [Available online at http://cloud1.arc.nasa.gov/crystalface/presentations_files/250p_Lawson.pdf.].
Lindzen, R. S., 1974: Wave-CISK in the tropics. J. Atmos. Sci., 31 , 156–179.
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.
Liu, W. T., and X. Xie, 2002: Double intertropical convergence zones—A new look using scatterometer. Geophys. Res. Lett., 29 .2072, doi:10.1029/2002GL015431.
Louis, J., M. Tiedtke, and J. Geleyn, 1982: A short history of the PBL parameterization at ECMWF. Proc. ECMWF Workshop on Planetary Boundary Layer Parameterization, Reading, United Kingdom, ECMWF, 59–80.
Maloney, E. D., and D. L. Hartmann, 2001: The Madden–Julian oscillation, barotropic dynamics, and North Pacific tropical cyclone formation. Part I: Observations. J. Atmos. Sci., 58 , 2545–2558.
Marshall, J. S., and W. M. Palmer, 1948: The distribution of raindrops with size. J. Meteor., 5 , 165–166.
Meehl, G. A., and J. M. Arblaster, 1998: The Asian–Australian Monsoon and El Niño–Southern Oscillation in the NCAR Climate System Model. J. Climate, 11 , 1356–1385.
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.
Raymond, D. J., and Coauthors, 2004: EPIC2001 and the Coupled Ocean–Atmosphere System of the tropical east Pacific. Bull. Amer. Meteor. Soc., 85 , 1341–1354.
Reynolds, R. W., 1988: A real-time global sea surface temperature analysis. J. Climate, 1 , 75–86.
Schubert, S. D., M. J. Suarez, Y. H. Chang, and G. Branstator, 2001: The impact of ENSO on extratropical low-frequency noise in seasonal forecasts. J. Climate, 14 , 2351–2365.
Schubert, S. D., M. J. Suarez, P. J. Pegion, M. A. Kistler, and A. Kumar, 2002: Predictability of zonal means during boreal summer. J. Climate, 15 , 420–434.
Simmons, A. J., and J. K. Gibson, 2000: The ERA-40 project plan. ERA-40 Project Series No. 1, ECMWF, Reading, United Kingdom, 62 pp.
Slingo, J., 1987: The development and verification of a cloud prediction scheme for the ECMWF model. Quart. J. Roy. Meteor. Soc., 113 , 899–927.
Suarez, M. J., and L. L. Takacs, 1995: Documentation of the Aries/GEOS dynamical core Version 2. NASA Tech. Memo. 104606, Tech. Rep. 5, 56 pp.
Sud, Y., 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.
Sundqvist, H., 1988: Parameterization of condensation and associated clouds in models for weather prediction and general circulation simulation. Physically Based Modelling and Simulation of Climate and Climatic Change, M. E. Schlesinger, Ed., Reidel, 433–461.
Wheeler, M., and G. N. Kiladis, 1999: Convectively coupled equatorial waves: Analysis of clouds and temperature in wavenumber–frequency domain. J. Atmos. Sci., 56 , 374–399.
Wu, X., X-Z. Liang, and G. J. Zhang, 2003: Seasonal migration of ITCZ precipitation across the equator: Why can’t GCMs simulate it? Geophys. Res. Lett., 30 .1824, doi:10.1029/2003GL017198.
Xie, P., and P. Arkin, 1997: Global precipitation, a 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 2539–2558.
Zhang, C., 2001: Double ITCZs. J. Geophys. Res., 106 , 11785–11792.
Zhang, G. J., and H. R. Cho, 1991: Parameterization of the vertical transport of momentum by cumulus clouds. Part I: Theory. J. Atmos. Sci., 48 , 1483–1492.
Zhang, M. H., and Coauthors, 2005: Comparing clouds and their seasonal variations in 10 atmospheric general circulation models with satellite measurements. J. Geophys. Res., 110 .D15S02, doi10.1029/2004JD005021.
Zhou, J., Y. C. Sud, and K-M. Lau, 1996: Impact of orographically induced gravity-wave drag in the GLA GCM. Quart. J. Roy. Meteor. Soc., 122 , 903–927.
JJA averages of precipitation for 1984–85 (a) from expt B1 with weak reevaporation, contours are drawn for 1.0, 2.0, 4.0, 8.0, and 16.0 mm day−1; (b) for expt B2 (moderate reevaporation); (c) for expt B3 (strong reevaporation); and (d) for Xie–Arkin precipitation data.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
JJA average precipitation biases with respect to Xie–Arkin (CMAP) climatology for 1984–85: (a) expt B1 with weak reevaporation, contours drawn for −16, −8, −4, −2, 2, 4, 8, and 16 mm day−1; (b) for expt B2 (moderate reevaporation); and (c) for expt B3 (strong reevaporation).
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Ratio of reevaporated precipitation to precipitation reaching the surface, f defined in (1). Seasonal means for JJA 1984–85 are shown for (a) B1, weak reevaporation; (b) B2, moderate reevaporation; and (c) B3, strong reevaporation.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Boxes used for regional precipitation and evaporation calculations presented in Table 2. Four regions are defined: northern and southern Pacific ITCZ (NITCZ and SITCZ), tropical West Africa (WAFR), and the southwestern United States (WUSA). Dashed contours show mean 4 mm day−1 contour for JJA 1984–85 precipitation in expt B1.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Mean profile of reevaporation moistening
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
(top) Specific humidity q in g kg−1 as a function of pressure in boxes (left) NITCZ and (right) SITCZ. Profiles are averages for JJA 1984–85. Solid lines show result for expt B3, dashed lines for B2, and dotted lines for B1. Filled triangles show the ERA-40 q profiles, and filled diamonds show q for the NCEP–NCAR reanalysis. (bottom) Average TPW in mm for same period, as function of latitude and longitude in (left) expts B1 and (right) B3.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Correlation of daily 15-day, high-pass filtered vertical motion ω̃ and precipitation
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Rms value of 15-day high-pass filtered vertical motion field
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Hovmoeller diagrams of (right) vertical motion and (left) precipitation along 10°S for JAS 1985 in (top) expt B1 and (bottom) expt B3. The 15-day, high-pass filtered vertical motion ω̃ and unfiltered daily precipitation are shown.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
(a)–(d) PDFs of unfiltered, daily ω850 for JJA 1985 for two regions along the southern ITCZ. (a) and (b) show results for expts B1 (solid black curve) and B3 (dashed red). (c) and (d) show results for perturbed physics expts H1 (solid black), M1 (solid light blue), and M2 (dashed dark blue). (a) and (c) show PDFs for a region in the eastern Pacific between 12° and 6°S bounded by 110° and 97.5°W. This region immediately off the coast of South America is meant to represent atmospheric variability that is as yet relatively unaffected by interactions between moist heating and dynamics. Sharp peaks at small positive values indicate ubiquitous weak subsidence. (b) and (d) show PDFs for a region in the central Pacific between 12° and 6°S bounded by 162.5° and 175°E. (e) PDFs for B1 and B3 for a box in the eastern portion of the northern ITCZ (6°–12°N, 110°–97.5°W) for comparison.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
(a) Seasonal mean JJA 1984–85 precipitation for expt H1, scale as in Fig. 1; (b) f, fraction of reevaporated rain to surface rain, scale as in Fig. 3; and (c) correlation of ω̃ and
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
(a) JJA 1984–85 precipitation for expt M1. (b) Correlation of daily ω̃ and
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Convective mass and moisture fluxes and precipitation efficiency for JJA 1985 in B1–B3, along (left) northern (8°–14°N) and (right) southern (12°–6°S) ITCZs. Key for lines in each row is shown in right panels. (a) and (b) show cloud-base mass flux summed over all cloud types invoked by RAS. (c) and (d) show diagnosed cloud-base convective moisture fluxes (thick solid and dashed lines). Thin solid and dashed lines in (c) and (d) show precipitation. Symbols in (c) and (d) show surface evaporation—large circles correspond to expt B3 and small diamonds to B1. (e) and (f) show the ratio of precipitation to convective moisture flux at cloud base.
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Mean, vertically integrated, horizontal transport tendencies for water vapor, defined in (2)–(4) of the text, as functions of longitude and latitude. Results are shown for JJA 1985; in (a)–(e) expt B1, (f)–(j) expt B3, and (k)–(o) expt H1. All quantities are displayed in units of mm day−1. (a), (f), and (k) show total, column-integrated water vapor flux convergence
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
(a)–(d) Mean vertical motion
Citation: Journal of the Atmospheric Sciences 63, 12; 10.1175/JAS3791.1
Summary of experiments performed. First column gives short-hand designation. Second column indicates strength of reevaporation. Third column indicates additional modifications to physical parameterizations. Fourth column gives pattern correlation of simulated precipitation with the CMAP climatology for a seven-season JJA 1981–87 mean. Fifth column is the rms error with respect to CMAP (mm day−1) for the same period.
Domain-averaged precipitation, reevaporation, and related quantities in the four boxes illustrated in Fig. 4 for JJA 1984–85. First column gives experiment designation from Table 1. Second column indicates box for averages. Third column gives average precipitation