1. Introduction
Shallow cumuli in the subtropics and midlatitudes have a considerable influence on the earth's climate because of their impact on the vertical transport of water vapor and entropy in the lower troposphere, and the surface energy and moisture budgets. Over land, shallow moist convection and its associated cloudiness are intimately tied to the diurnal cycle of the atmospheric boundary layer, which is also an important issue for climate simulation (Wilson and Mitchell 1986).
Despite many efforts over the past decades, the representation of shallow cumuli (i.e., cumuli in which precipitation plays a secondary role) in models remains a difficult parametric problem. Current regional forecast or global climate models use simple parameterizations for shallow cumuli. Some schemes treat shallow moist convection similar to deep convection, for example, the Tiedtke (1989) scheme used at the European Centre for Medium-Range Weather Forecasts (ECMWF) model, the Gregory and Rowntree (1990) scheme used by the Met Office, the Bechtold et al. (2001) scheme, and some spectral ensemble schemes such as the relaxed Arakawa–Schubert (RAS; Moorthi and Suarez 1992) scheme. Other models have a separate module for shallow convection, for example, the Hack scheme used in the National Center for Atmospheric Research (NCAR) Community Climate Model (CCM) version 3 (Hack 1994). Some models, such as the fifth-generation (Pennsylvania State University) PSU–NCAR Mesoscale Model (MM5; Grell et al. 1994), are commonly run without a parameterization that can effectively represent shallow cumulus convection.
To improve the representation of shallow moist convection in regional or global models, Bretherton et al. (2004) and McCaa and Bretherton (2004) developed a new parameterization for shallow moist convection (hereafter the BM scheme) based on a buoyancy-sorting plume model similar to that of Kain and Fritsch (1990) with cloud-base mass flux controlled by a convective inhibition (CIN; Mapes 2000) criterion and the turbulent kinetic energy (TKE) of the boundary layer. Incorporating the new scheme together with a new boundary layer parameterization (Grenier and Bretherton 2001, hereafter GB) into a regional model, they successfully simulated the transition from stratocumulus to shallow cumulus convection in the northeast and southeast Pacific Ocean. Sensitivity tests indicate that the seasonal-mean cloud cover and albedo in the model depend heavily on the parameterization of shallow convection.
The BM scheme needs to be further evaluated against extensive in situ observations both over ocean and particularly over land, since in frequency of occurrence shallow cumuli outranks any other type of low-latitude clouds (Warren et al. 1988), and their characteristics can be quite different under various external forcings and ambient meteorological conditions. Observations and large eddy simulations (LESs; Zhu and Albrecht 2002, 2003) indicate that continental shallow cumuli forced by strong surface sensible heat fluxes are often negatively buoyant throughout the cloud layer, a situation contrasting with most oceanic trade wind cumuli, which have weak positively buoyant cores extending down nearly to cloud base. Since the forced shallow cumuli are intermediate between a dry convective boundary layer and buoyant shallow cumuli, they are a potential challenge to parameterizations for most models.
The major objectives of this paper are: First, to examine the performance of the BM shallow cumulus scheme in a regional model under different meteorological conditions by comparing model output with the comprehensive conventional and state-of-art remote sensing measurements of clouds, radiation, and thermodynamic profiles at the sites in the tropical western Pacific (TWP) and southern Great Plains (SGP) operated by the Atmospheric Radiation Measurements program (ARM; Stokes and Schwartz 1994). We focus on continental shallow cumulus convection since it has received less attention than its marine counterpart. Second, we investigate the impact of the parameterized shallow cumulus convection on the boundary layer structure and surface energy budgets under different conditions. Such a study is helpful to understand and characterize the basic role of shallow convection in the climate system. The organization of the paper is as follows. In section 2, we present our basic research strategy and briefly describe the observational data and the model used in this study. A short discussion of the BM scheme is also provided in this section. Sections 3 and 4 present a series of comparisons between simulations and observations for over-land and over-sea conditions. The effects of shallow cumulus convection on the atmosphere are discussed in section 5. Finally, a summary is presented in section 6.
2. Approach
a. Model and experiment design
The ultimate test of a parameterization is its performance in a global climate model (GCM). However, running a full GCM is very time consuming, and interpreting GCM results can be complex. Thus, it is not always an optimal method for testing a new scheme. Recently, the single-column model (SCM) has emerged as an alternate test bed for evaluating parameterization schemes (Betts and Miller 1986; Iacobellis and Somerville 1991; Randall et al. 1996). The usefulness of SCMs for developing and diagnosing representations of diabatic processes in weather and climate models is discussed by Bergman and Sardeshmukh (2003). Although an SCM is attractive since it is a more controllable modeling framework and runs much faster than a GCM, providing the necessary forcings and boundary conditions for SCMs from measurements has proven to be a challenging task due to sampling and measurement errors in winds and thermodynamic forcings (Zhang and Lin 1997; Mace and Ackerman 1996; Randall et al. 1996) and uncertainties in horizontal condensate advection. For highly advective conditions, an SCM may not be a suitable test bed for cloud parameterizations (Ghan et al. 2000).
In this study we choose a regional model forced at its lateral boundaries by the analyses from the ECMWF forecast model as a test bed for shallow cumulus parameterization due to several considerations. First, running a regional model is much cheaper than running a full GCM. Nevertheless, the regional model can internally generate appropriate dynamical forcings including condensate advection for all grid columns set well away from the lateral domain boundaries. Second, the space– time variability in a regional model simulation is directly comparable to daily or even hourly time-scale observations, since on a shorter time period the verification data can be more complete and comprehensive, for example, data obtained in intensive observation periods (IOPs). Finally, a regional model can be executed at different locations, which allows one to test the performance of a parameterization scheme under typical climate conditions anywhere in the world, without the computational burden of a full GCM.
The regional model chosen as a parameterization test bed for the BM scheme is the MM5 regional model version 3.5. Other major parameterizations used in MM5 include turbulence, radiation, deep convection, stratus clouds, and cloud microphysics. Following Bretherton et al. (2004), we use the GB boundary layer parameterization scheme, which combines a 1.5-order turbulence closure model with an entrainment parameterization at the top of the boundary layer. The NCAR CCM version 2 radiation scheme is used, but is revised to calculate cloud radiative fluxes by using the cloud cover fraction and liquid water content diagnosed by stratiform and convective cloud parameterizations instead of using relative humidity (McCaa 2001). Deep convection and microphysics are parameterized in the model using the Kain–Fritsch scheme (1993) and Reisner mixed-phase scheme (Reisner et al. 1998), respectively. The atmospheric model is coupled with a simple multilayer soil temperature model (Dudhia 1996).
The simulations are performed on a single-grid mesh (51 × 51) at 40-km horizontal resolution centered at the central facility of the SGP site (36.605°N, 97.485°W) and the TWP Nauru site (0.521°S, 166.916°E), respectively. Thirty-eight full sigma levels (1.00, 0.995, 0.99, 0.985, 0.98, 0.97, 0.96, 0.95, 0.94, 0.93, 0.92, 0.91, 0.90, 0.88, 0.86, 0.83, 0.80, 0.77, 0.74, 0.71, 0.68, 0.64, 0.60, 0.56, 0.52, 0.48, 0.44, 0.40, 0.36, 0.32, 0.28, 0.24, 0.20, 0.16, 0.12, 0.08, 0.04, 0.0) are specified in the vertical. The pressure at the model top is set to 100 hPa, and there are 15 levels below 850 hPa. Both initial fields and lateral boundary conditions are supplied by ECMWF Tropical Ocean Global Atmosphere (TOGA) 2.5° global surface and upper-air analysis data obtained from NCAR's Data Support Section, available at 6-h intervals.
b. Verification data
The verification data used in this study are from the ARM SGP site and TWP Nauru site. By comparing the data collected at these two sites with the simulations, we attempt to characterize the behavior of continental and maritime shallow cumulus convection in models. Three periods of time (a month long each) are chosen. These are 13 June–13 July 1997 at the SGP site (representing the continental summer season), 1 January– 31 January 1998 at the SGP site (representing the continental winter season), and 17 January–17 February 2000 at the TWP Nauru site (representing maritime tropical shallow cumuli). These particular periods were chosen based on the following considerations. The summer month is approximately an ARM IOP running from 2330 UTC 17 June to 2330 UTC 17 July, and includes two continental shallow cumulus cases simulated by LES models (Brown et al. 2002; Zhu and Albrecht 2003) that provide attractive comparisons for the BM shallow cumulus scheme. The reason to choose 13 June instead of 17 June as the starting point for the summer month is that from 13 June, shallow cumuli were observed at the SGP site in three consecutive days, which provide good cases to examine how well the revised BM scheme can represent the typical continental shallow cumuli induced by the boundary layer processes. The over-land winter month and the time period chosen for tropical ocean conditions are not IOPs, but do have relatively complete observational data. In addition, in January 1998, the SGP site was often covered by stratus and stratocumulus clouds, which provides a good opportunity to test the performance of the model for a variety of continental shallow cloud types when the boundary layer temperature is lower. In the chosen period at the TWP Nauru site, shallow cumuli were consistently observed throughout the month (Fig. 2, to be discussed in detail later).
Four ARM datasets are used in this study. First, we use surface measurements from the Surface Meteorological Observation System (SMOS) and the Energy Balance Bowen Ratio System (EBBR), averaged to 30-min intervals. The SMOS provides conventional in situ meteorological data, while the EBBR is a ground-based system for estimating the surface fluxes of sensible, latent, and soil heat. Second, we use radiative measurements by the Solar Infrared Radiation Station (SIRS), which provides downward solar and infrared radiative fluxes at 60-s time resolution. Third, we use atmospheric profiling measured by the Balloon-Borne Sounding System (BBSS), which provides vertical profiles of both thermodynamic (temperature and relative humidity) and dynamic (wind speed and direction) state of the atmosphere. Soundings are available every 3-h during IOPs. Under standard operations, five soundings (0530, 1130, 1430, 1730, and 2030 UTC) are launched per day at the central facility of the SGP site and two soundings (1130 and 2330 UTC) at the TWP Nauru site. Fourth, we use cloud observations from the Active Remote Sensing of Clouds (ARSCL) suite, which combines the measurements from the Millimeter Wave Cloud Radar (MMCR), Belfort Laser Ceilometer (BLC), and Micropulse Lidar (MPL). From the ARSCL, we can obtain cloud boundary information. For details of the instruments and the data, refer to Zhu and Albrecht (2001, 2002, and www.arm.gov).
c. Shallow convective triggering criterion
The BM scheme only considers “active” buoyant shallow cumuli with cloud top higher than the LFC, but neglects the “forced” shallow cumuli which never reach the LFC. Such an assumption is understandable based on observations of marine trade wind cumuli from the Barbados Oceanographic and Meteorological Experiment (BOMEX; Holland and Rasmusson 1973) and the Atlantic Trade-Wind Experiment (ATEX; Augstein et al. 1973), for which the majority of cumuli were active. Figure 2 shows the radar-observed cloud boundaries for clouds with base heights below 3 km and the nighttime LFC estimated from the near-surface properties recorded by the BBSS soundings at the TWP Nauru site, a location of oceanic trade-cumulus convection fairly similar to BOMEX (note that the Nauru daytime BBSS-estimated LFCs seem to be contaminated by the island heating as discussed later in section 4). The top of most cumuli is above the estimated LFC. The few clouds that appear to lie below the LFC would be neglected by the BM scheme.
LES analyses of the BOMEX and ATEX cumulus cases indicate that the buoyancy production in the subcloud layer is weak compared with that in the cloud layer (Stevens et al. 2001). Buoyancy production in the cloud layer is the key for maintaining a shallow cumulus layer, since over the subtropical oceans the surface sensible heat flux is typically small, for example, about 10 W m−2 in the BOMEX and ATEX experiments. In contrast, over land the buoyancy production in the subcloud layer can be much larger than that in the cloud layer due to the strong surface sensible heat fluxes (Zhu and Albrecht 2003). In this case, the buoyancy production in the subcloud layer alone may be sufficient for maintaining a forced cumulus layer. Figure 3 shows a similar plot to Fig. 2 but at the SGP site. As the figure indicates, the majority of shallow cumuli are below their LFC including the two shallow cumulus cases (21 June 1997 and 6 July 1997) simulated by Brown et al. (2002) and Zhu and Albrecht (2003) using LES models. Zhu and Albrecht found that although these clouds are not active, they substantially affect turbulence transport, and their effects should be parameterized in models.
Based on these considerations, we revise the BM scheme slightly so that it is able to represent forced shallow cumuli. To do so, we redefine CIN as the shaded area in Fig. 1b. The critical velocity wc is redefined such that the cumulus air released at the inversion with wc reaches the LCL with zero velocity. The BM bulk updraft scheme is then applied as before but with purely penetrative mixing in the forced case if there is no LFC. This is a necessary adjustment if the scheme is used to parameterize continental forced shallow moist convection.
3. Simulations of continental shallow moist convection
In this section, we take a close look at the regional simulations of the two continental cumulus cases observed 21 June and 6 July 1997 at the SGP site, which have been analyzed by Zhu and Albrecht (2002) and simulated by LES models (Brown et al. 2002; Zhu and Albrecht 2003). Then, we present the simulated diurnal variation of the boundary layer during monthlong regional simulations for summer and winter.
A straightforward way to evaluate the performance of a parameterization scheme is to see whether the scheme is able to appropriately generate the processes that the model cannot resolve. In the case of shallow moist convection over land, the focus is on whether the scheme can capture the basic diurnal feature of subgrid cloud fields based on the model-resolved external meteorological conditions. The top of Fig. 4 shows the radar-detected cloud boundaries, the LCL estimated from the surface observations, and the boundary layer height estimated from the soundings for these two cases. The coincidence between the development of cumulus fields and the deepening of the boundary layer indicates that the observed shallow cumuli are controlled by the subcloud layer processes. The nearly perfect match between the LCL and the cloud-base height also confirms that the clouds originate from source air in the surface layer. These clouds are basically maintained by the buoyancy production in the subcloud layer. Without sufficient buoyancy support from underneath, the clouds dissipate quickly after sunset as indicated by Figs. 4a1 and 4a2, since the buoyancy production of the forced cumuli themselves is minimal. This radar-observed evolution of cloud fields is consistent with satellite images and SGP WSI (Whole Sky Imager) measurements.
The middle of Fig. 4 show the corresponding MM5 simulations with the modified BM parameterization combined with the GB turbulence scheme. Henceforth, we refer to this combination as the UW (University of Washington) scheme. To allow an optimal comparison with observations on these 2 days, the simulations are initialized at 1200 UTC 21 June and 6 July, respectively, from ECMWF operational analyses rather than simply using the monthlong simulations shown later, which build up small but significant regional-scale errors at these times.
Since there are no direct observations of cumulus-induced mass flux available, we compare the parameterized mass fluxes with those obtained from LESs (Brown et al. 2002; Zhu 2002). The bottom of Fig. 4 show the vertical profiles of mass fluxes at 2030 UTC. In Fig. 4c1, the LES-simulated mass flux is the hourly and domain-averaged profile from eight different models executed on a domain with 6.4 km on a side. Thus, the profile at 2030 UTC represents the average between 2000 and 2100 UTC. The mass flux profile from MM5 shown in the figure, on the other hand, is the 10-min averaged profile at the SGP column. The LES-determined mass flux profile of cumulus case 6 July (Fig. 4c2) is from a RAMS (Regional Atmospheric Modeling System) simulation (Zhu 2002). The model domain is 5.9 km × 3.1 km. Statistics are averaged over the entire domain and every 10 min. It appears that the revised BM scheme successfully captures the diurnal variation of the observed cumuli. The relative strength and thickness of the cloud layer of these two cases are also reasonably simulated by MM5. The magnitude of the parameterized mass fluxes are close to those from LESs as indicated by Figs. 4c. This is also true for the time periods other than in the vicinity of 2030 UTC (not shown here). These evidences indicate that overall the evolution of the cloud fields on these 2 days is qualitatively reproduced by MM5.
There are some differences between the simulated and observed clouds. The most significant is that simulated cumuli generally get stronger and thicker in the afternoon, but the radar-observed cloud layer appears to be the thickest right before local noon (1830 UTC). In addition, in the case of 21 June, the simulated clouds form and dissipate earlier than the observed ones; while for the other case, the simulated clouds are lower and thicker than the observations. These biases may reflect errors in the entire regional modeling system, not just the BM parameterization. For instance, in this simulation both the initial fields and boundary conditions are specified solely based on ECMWF analyses (2.5° resolution), which means that the model profiles at the domain center located at the SGP site may not exactly represent the real meteorological conditions of the site. As an example, Fig. 5 compares the initial profiles (1200 UTC) of case 21 June at the SGP cell with the sounding available at 1130 UTC. The biases in the initial fields, especially the bias in moisture below 2 km, are significant enough to generate the differences between simulations and observations shown in Fig. 4.
We next turn to statistical comparisons of the entire month of simulations with the SGP observations to provide a more climatological evaluation of the overall performance of the model. Our comparisons focus on the summer month since there is little shallow moist convection in winter at the SGP site.
Figures 6a,b compare the observed and simulated 2-m air temperature at the central facility of the SGP site for the summer month. For clarity, both the time series and the monthly averaged diurnal cycle are plotted in the figure. Overall, the observed diurnal cycle of 2-m air temperature is well simulated by the model, although the model tends to underestimate the temperature slightly in the later afternoon. A detailed analysis of model output suggests that this bias may not be caused by the boundary layer scheme and low-level cloud scheme, but is related to excessive high-level ice clouds. In the model, ice clouds are considered to cover the whole grid box whenever the ice mixing ratio is greater than 0.1 g kg−1. On 19 June, and 6 and 7 July, when only shallow cumuli were detected at the site, the simulated temperature tracks the observed one nicely. The warm bias caused by high-level cirrus is supported by Fig. 7, which shows that the simulated temperature at 300 hPa, where cirrus clouds usually occur, does have a warm bias.
The comparisons of 2-m humidity are shown in Figs. 6c,d. The day-to-day variability of water vapor mixing ratio is reasonably simulated. However, the weak diurnal variation of 2-m moisture, which is more obvious when averaged over 10 facilities at the entire SGP site (thick line in the figure), is not well simulated by MM5. The simulated diurnal variation has a small nighttime high bias and daytime low bias. For comparison, the 2-m mixing ratio determined by parallel simulations using the Medium-Range Forecast Model (MRF; Troen and Mahrt 1986; Hong and Pan 1996) and Gayno–Seaman (GS; Gayno 1994; Shafran et al. 2000) boundary layer schemes available in the MM5 standard distribution are also plotted in Fig. 6c. Unlike the UW scheme, both the MRF and GS schemes tend to systematically overestimate the surface layer moisture, perhaps due in part to inadequate subcloud ventilation associated with the lack of a shallow cumulus scheme.
In addition to the surface-layer thermodynamic properties, the model also reasonably reproduces the observed vertical structure of the boundary layer. Figure 8 compares the observed and simulated vertical profiles of potential temperature and water vapor mixing ratio in the summer month. The bias pattern in both fields is rather complicated, which may be partly attributed to the uncertainties coming from the comparison between a point observation and simulation with a 40-km grid cell. Neither of these fields show large systematic biases, and both follow the observed day-to-day variability. This is more clear in the time series of 850 and 500 hPa potential temperature and mixing ratio shown in Fig. 9. The synoptic features are well reproduced by MM5. However, precipitation events are not as well simulated. Note that most large biases in temperature and moisture correspond approximately to these events.
Figure 10 compares the monthly averaged diurnal cycle of potential temperature for both observations and simulations. The diurnal cycle of the entire boundary layer is reasonably simulated. There is a nighttime near-surface warm bias and a daytime near-surface cold bias. Early in this period, the bias is localized near the surface, then it spreads through the lowest 1 km as the boundary layer deepens. This bias may be more related to the GB boundary layer scheme than the shallow convection scheme since shallow convection is usually weak in the early morning. The monthly averaged moisture fields show little diurnal signal, thus they are not shown here.
The good simulations of surface thermodynamic properties and the boundary layer structure reflect a realistic surface energy budget generated by the model. Figure 11a shows the monthly averaged observed and simulated diurnal variation of surface heat fluxes. The largest error is an underestimate of the daytime surface latent heat fluxes. This is a likely cause of the model's daytime dry bias shown in Fig. 6d. The underestimate of surface latent heat fluxes is partly compensated by a slight underestimate in daytime net radiative fluxes. The underestimate of surface downward shortwave radiative fluxes (Fig. 11b) suggests that the model overestimates cloudiness. However, since the model actually underpredicts downwelling longwave radiative fluxes during the day as well, the excessive cloudiness is likely high-level cirrus as suggested by Fig. 7. These high-level ice clouds have a negligible surface greenhouse effect.
Figure 12 shows the monthly composite diurnal cycle of the parameterized shallow cumuli at the SGP cell for both summer and winter months. In summer, the simulated shallow moist convection occurs almost every day in the model domain although its intensity and location may vary depending on the external conditions. It has a distinct diurnal cycle, forming during the day when the buoyancy production in the mixed layer gets strong enough in the morning and dissipating after sunset when the subcloud layer is no longer positively buoyant. Compared with the summer month, the simulated shallow moist convection in the winter month is more intermittent and much weaker, reflecting the weaker surface sensible and latent heat fluxes. Therefore, in addition to the diurnal cycle, the seasonal variation of continental shallow convection is also well represented by the revised BM scheme.
4. Simulations of maritime shallow cumulus convection
In this section, we discuss simulations of maritime shallow cumuli based on a monthlong (17 January–17 February 2000) regional numerical experiment with the model domain centered at the ARM TWP Nauru site.
Figure 13a shows the radar-detected cloud boundaries and the LCL estimated from the surface-layer properties recorded by the BBSS at 1130 UTC (∼2240 LST) indicated by circles and 2330 UTC (∼1040 LST) indicated by stars. The observed cloud bases match the LCLs at 1130 UTC very well, but the LCLs at 2330 UTC are systematically higher than cloud bases by 100– 800 m. This bias indicates that the daytime soundings do not represent the lower boundary layer thermodynamic structure of the TWP regime as well as nighttime soundings because of an apparent local island effect due to solar heating at the island surface during daytime, when the measured surface temperature and humidity may not be representative of those over the ocean surface. The island influence on boundary layer structure can be clearly seen from the composite profiles at 1130 and 2330 UTC shown in Fig. 14 (to be discussed shortly).
Figure 13b shows the simulated cumulus-induced mass fluxes at the TWP Nauru site. The observed mean cloud height and thickness are well reproduced by the BM shallow cumulus scheme, as is the suppression of shallow cumuli on 20 January and 12 February. The simulated cloud base tends to be too high by about 200 m. This bias may be caused by the model's overestimate of the LCL because of an underestimate of the mixed layer moisture as indicated by Fig. 15c2. In contrast with the continental shallow cumuli, no distinct diurnal variation can be seen in either observations or simulations. The simulated mass fluxes are concentrated in a relative thin layer in the lower part of clouds, which suggests that the cloud-layer buoyancy is relatively weak, a situation slightly different from BOMEX. Figure 14 compares the monthlong composite virtual potential temperature profile at the Nauru site with that from BOMEX. To estimate the buoyancy of shallow cumuli rising through the lower troposphere, the adiabatic profile of their assumed source air (specified to be the monthly mean of the surface-layer air at 1130 UTC, which is believed to be less influenced by the island because of a close match between the LCL and the observed cloud base) is plotted along with the composite sounding profile. The undiluted updraft buoyancy is far larger for BOMEX than that for the ARM TWP Nauru case. Note that for both TWP Nauru and BOMEX, undilute lifted near-surface air is positively buoyant in and above the trade inversion. Nevertheless, only shallow convection is observed, emphasizing the importance of entrainment dilution in controlling the depth of moist convection.
Figure 15 compares the observed and simulated vertical profiles of potential temperature and water vapor mixing ratio for the TWP Nauru. The model overestimates the depth of the subcloud mixed layer slightly and underestimates both potential temperature and mixing ratio in the surface layer. It is not clear what are the main reasons for these biases. Numerical experiments with other boundary layer schemes (available in MM5) other than the UW scheme produce even larger biases (not shown here). This bias may be associated with the broad errors in the MM5 modeling framework or caused by the inaccuracy of initial and lateral boundary conditions. Since the bias is mainly in the lower boundary layer, it may also partly reflect the daytime local island effect in soundings. Despite these biases, the main features of the observed vertical thermodynamic structure are well reproduced.
These comparisons indicate that the boundary layer scheme and the shallow cumulus scheme used in the study can reproduce the observed boundary layer structure under different meteorological conditions, although not without biases. We now use these simulations to further investigate the effects of parameterized shallow cumuli on the model boundary layer structure and surface energy budget.
5. Impact of shallow cumuli on the atmosphere
Shallow cumulus convection affects the thermodynamic structure of the lower troposphere and the near-surface temperature and humidity, feeding back on such things as cloudiness, surface fluxes, and deep convective triggering. To study the effects of shallow cumuli, we compared the earlier simulations with parallel monthlong simulations with the shallow cumulus scheme inactivated.
Figure 16 shows the difference of level-averaged potential temperature and water vapor mixing ratio profiles with and without the shallow cumulus scheme for the continental summertime and the maritime simulations. For both cases, shallow convection moistens and cools the cloud layer and dries the subcloud layer. This occurs since shallow cumuli strengthen turbulent mixing near the top of the subcloud layer and enhance the entrainment of drier air down into the subcloud layer. The maritime shallow cumuli are more persistent and have a somewhat larger impact on the thermodynamic structure of the boundary layer than their continental counterparts. The impact of the continental shallow cumuli is strongest during daytime as one would expect, when it frequently exceeds 0.5 K in temperature and 0.5 g kg−1 in moisture. Thus, neglect or misrepresentation of the effects of continental shallow cumuli may cause noticeable mean biases in a climate model.
The simulated maritime shallow cumuli also warm and dry the overlying free troposphere up to a pressure of 50 kPa. The drying effect of shallow cumuli on the subcloud layer may suppress sporadic deeper convection by reducing conditional instability. This effect is less apparent in the continental case, perhaps because it is overwhelmed by the strong diurnal surface forcing.
Figure 17 shows the domain-averaged surface sensible and latent heat fluxes with and without the shallow cumulus scheme for the continental summertime and the maritime conditions. The maritime shallow cumuli increase the surface latent heat fluxes by about 30 W m−2 because they dry the subcloud layer. The surface sensible heat fluxes almost remain unchanged because the warming in the subcloud layer by shallow cumuli is quite small. Such an effect was also reported in previous studies. For example, Tiedtke et al. (1988) found that the inclusion of a shallow cumulus scheme in the model increases surface evaporation in the subtropical oceans by as much as 50 W m−2, leading to an increase of precipitation in the ITCZ up to 10 mm day−1. However, in our continental simulations, the surface latent and sensible heat fluxes are barely changed by the shallow cumulus scheme. One possible reason is that over land during the daytime the radiative forcing is paramount in setting the surface turbulent fluxes. Since the cloud fraction and liquid water path of continental shallow cumuli are normally small, shallow cumuli have minor effects on the shortwave and longwave radiative fluxes. As a result, the surface energy budget over land is insensitive to shallow cumuli.
It is important to appropriately couple separate deep and shallow convection schemes in models. In reality, there is no clean boundary that separates deep convection and shallow convection. In our simulations, the BM shallow convection scheme is activated before the Kain– Fritsch deep convection scheme is called. What the shallow convection scheme does not do is left over for the deep convection scheme. The simulations for the continental summer month indicate that the domain-averaged accumulated convective precipitation produced by the deep convection scheme at the end of the month is reduced by about 5% with the shallow convection scheme compared with that without the shallow convection scheme (not shown here). The reason for this precipitation reduction is that the shallow convection has reduced the conditional instability within the lower troposphere by transporting moisture upward and regulating the stratification of the atmosphere.
Another interesting question about shallow moist convection is how it interacts with stratiform clouds when they coexist, for example, in the transition from trade wind regime to stratocumulus regime over oceans and across vast areas over land in midlatitudes. The role of cumuli rising into stratocumulus layers has been studied observationally (e.g., Miller and Albrecht 1995; Albrecht et al. 1995; Miller et al. 1998). The shallow cumulus scheme and the stratiform cloud scheme activated in the numerical experiments in this study provide a modeling perspective on this issue. Figure 18 compares the domain-averaged profiles of cloud fraction and liquid water content determined by the stratiform cloud scheme with and without the shallow cumulus scheme. Shallow convection usually tends to reduce cloud fraction and level-averaged liquid water content in the lower 10 kPa of the cloud layer, while the reverse occurs in the overlying 20 kPa. We attribute this to the upward moisture transport in the cumuli.
6. Summary
Shallow cumulus convection can influence weather and climate by altering the cloud cover, and energy and moisture budgets of the boundary layer and the lower troposphere. In this paper, we have evaluated the overall performance of a recently developed shallow cumulus parameterization under various meteorological conditions. We have also used this parameterization to investigate the impact of continental and maritime shallow cumuli on the boundary layer structure and the surface energy and moisture budget. Our approach is to incorporate the shallow cumulus scheme into a regional model, and then compare the results of 1-month boundary-forced simulations with observations collected at the ARM SGP and TWP sites. By using a regional model as the parameterization test bed, we can test a scheme under different climate conditions without much computational burden by centering the model domain at different locations. In this study we chose two model domains centered at the ARM SGP and TWP Nauru site, respectively, so that we can simulate and study the characteristics of continental and maritime shallow cumuli by fully utilizing the conventional and state-of-art remote sensing observations measured at these two sites.
Observations indicate that although maritime shallow cumuli are buoyancy-driven clouds, most continental shallow cumuli are negatively buoyant, overshooting updrafts supported mainly by the buoyancy production in the subcloud layer due to a strong surface forcing. For this reason, we modified the original BM shallow cumulus scheme slightly so that it can allow for both active and forced shallow cumuli. Our simulations and the comparisons with observations indicate that
The shallow cumulus scheme used in this simulation realistically represents maritime and continental shallow cumuli. In particular, it reproduces the observed vertical thermodynamic structures, the surface energy budget, and the diurnal cycle over land.
Shallow cumulus convection significantly alters the vertical thermodynamic structure of the atmosphere by moistening and cooling the cloud layer and drying and slightly heating the subcloud layer. Over the oceans, such effects considerably enhance surface latent heat fluxes as noted by previous studies, and appear to dry the midtroposphere by inhibiting deep convection. Over land, in summer these effects are weaker, but are still significant during the day. In our monthlong wintertime SGP simulation, shallow convective mass fluxes were sporadic and averaged only a tenth as strong as in summer.
Our simulations suggest that shallow convection usually but not always reduces low-level stratiform clouds at pressures exceeding 80 kPa and enhances them in 80–60 kPa.
In this study, our focus has been using a regional model to explore the parameterization and systematic effects of shallow moist convection. The large-scale feedbacks of shallow cumuli are more fully realized in a GCM. This study suggests that the GB boundary layer scheme and the BM shallow convection scheme perform sufficiently realistically for both continental and maritime conditions to warrant implementation and testing these schemes in a climate model, and this will be our next focus.
Acknowledgments
The authors want to thank Dr. James McCaa for providing the boundary layer scheme and the shallow cumulus scheme. We are grateful to two anonymous reviewers for their constructive comments. Their helpful suggestions lead to improvements of this paper. This research was supported by NASA Grant NAG5S-10624. The data used in this paper are provided by the ARM Data Archive.
REFERENCES
Albrecht, B. A., M. P. Jensen, and W. J. Syrett, 1995: Marine boundary layer structure and fractional cloudiness. J. Geophys. Res, 100 , 14209–14221.
Augstein, E., H. Riehl, F. Ostapoff, and V. Wagner, 1973: Mass and energy transports in an undisturbed Atlantic trade-wind flow. Mon. Wea. Rev, 101 , 101–111.
Bechtold, P., E. Bazile, F. Guichard, P. Mascart, and E. Richard, 2001: A mass flux convection scheme for regional and global models. Quart. J. Roy. Meteor. Soc, 127 , 869–886.
Bergman, J. W., and P. D. Sardeshmukh, 2003: Usefulness of single column model diagnosis through short-term predictions. J. Climate, 16 , 3803–3819.
Betts, A. K., and M. J. Miller, 1986: A new convective adjustment scheme. Part II: Single column tests using GATE wave, BOMEX, ATEX, and Arctic air-mass data sets. Quart. J. Roy. Meteor. Soc, 112 , 693–709.
Bretherton, C. S., J. R. McCaa, and H. Grenier, 2004: A new parameterization for shallow cumulus convection and its application to marine subtropical cloud-topped boundary layers. Part I: Description and 1D results. Mon. Wea. Rev, 132 , 864–882.
Brown, A. R., and Coauthors, 2002: Large-eddy simulation of the diurnal cycle of shallow cumulus convection over land. Quart. J. Roy. Meteor. Soc, 128 , 1059–1073.
Dudhia, J., 1996: A multi-layer soil temperature model for MM5. Preprints, Sixth PSU/NCAR Mesoscale Model User's Workshop, Boulder, CO, NCAR, 49–50.
Gayno, G. A., 1994: Development of a higher-order, fog-producing boundary layer model suitable for use in numerical weather prediction. M.S. thesis, Dept. of Meteorology, The Pennsylvania State University, 104 pp.
Ghan, S., and Coauthors, 2000: A comparison of single column model simulations of summertime midlatitude continental convection. J. Geophys. Res, 105 , 2091–2124.
Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability dependent closure. Mon. Wea. Rev, 118 , 1483–1506.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1994: A description of the fifth generation Penn State/NCAR mesoscale model (MM5). NCAR Tech. Note NCAR/TN-398, 80 pp.
Grenier, H., and C. S. Bretherton, 2001: A moist PBL parameterization for large-scale models and its application to subtropical cloud-topped marine boundary layers. Mon. Wea. Rev, 129 , 357–377.
Hack, J. J., 1994: Parameterization of moist convection in the National Center for Atmospheric Research community climate model (CCM2). J. Geophys. Res, 99 , 5541–5568.
Holland, J. Z., and E. M. Rasmusson, 1973: Measurement of the atmospheric mass, energy, and momentum budgets over a 500-kilometer square of tropical ocean. Mon. Wea. Rev, 101 , 44–55.
Hong, S-Y., and H-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. Mon. Wea. Rev, 124 , 2322–2339.
Iacobellis II, S. F., and R. C. J. Somerville, 1991: Diagnostic modeling of the Indian monsoon onset. Part I: Model description and validation. J. Atmos. Sci, 48 , 1948–1959.
Kain, J. S., and J. M. Fritsch, 1990: Nonlocal boundary layer vertical diffusion in a medium-range forecast model. J. Atmos. Sci, 47 , 2784–2802.
Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 165–170.
Mace, G. G., and T. P. Ackerman, 1996: Assessment of error in synoptic-scale diagnostics derived from wind profiler and radiosonde network data. Mon. Wea. Rev, 124 , 1521–1534.
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.
McCaa, J. R., 2001: A new parameterization of marine stratocumulus and shallow cumulus clouds for climate models. Ph.D. dissertation, Dept. of Atmospheric Sciences, University of Washington, 161 pp.
McCaa, J. R., and C. S. Bretherton, 2004: A new parameterization for shallow cumulus convection and its application to marine subtropical cloud-topped boundary layers. Part II: Regional simulations of marine boundary layer clouds. Mon. Wea. Rev, 132 , 883–896.
Miller, M. A., and B. A. Albrecht, 1995: Surface-based observations of mesoscale cumulus–stratocumulus interaction during ASTEX. J. Atmos. Sci, 52 , 2809–2826.
Miller, M. A., M. P. Jensen, and E. E. Clothiaux, 1998: Diurnal cloud and thermodynamic variations in the stratocumulus transition regime: A case study using in situ and remote sensors. J. Atmos. Sci, 55 , 2294–2310.
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.
Randall, D. A., K-M. Xu, R. J. S. Somerville, and S. Iacobellis, 1996: Single-column models as links between observations and climate models. J. Climate, 9 , 1683–1697.
Reisner, J., R. J. Rasmussen, and R. T. Bruintjes, 1998: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc, 124 , 1071–1107.
Shafran, P. C., N. L. Seaman, and G. S. Gayno, 2000: Evaluation of numerical predictions of boundary layer structure during the Lake Michigan ozone study. J. Appl. Meteor, 39 , 412–426.
Stevens, B., and Coauthors, 2001: Simulations of trade wind cumuli under a strong inversion. J. Atmos. Sci, 58 , 1870–1891.
Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the cloud and radiation test bed. Bull. Amer. Meteor. Soc, 75 , 1201–1221.
Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parametrization in large-scale models. Mon. Wea. Rev, 117 , 1779–1800.
Tiedtke, M., W. A. Heckley, and J. Slingo, 1988: Tropical forecasting at ECMWF: The influence of physical parameterization on the mean structure of forecasts and analyses. Quart. J. Roy. Meteor. Soc, 114 , 639–664.
Troen, I. B., and L. Mahrt, 1986: A simple model of the atmospheric boundary layer: Sensitivity to surface evaporation. Bound.-Layer Meteor, 37 , 129–148.
Warren, S. G., J. Hahn, J. London, R. M. Chervin, and R. L. Jenne, 1988: Global distribution of total cloud cover and cloud type amounts over the ocean. NCAR Tech. Note NCAR/TN-317+STR, 42 pp. [Available from NCAR, P.O. BOX 3000, Boulder, CO 80307.].
Wilson, C. A., and J. F. B. Mitchell, 1986: Diurnal variation and cloud in a general circulation model. Quart. J. Roy. Meteor. Soc, 112 , 347–369.
Zhang, M. H., and J. L. Lin, 1997: Constrained variational analysis of sounding data based on column-integrated budgets of mass, heat, moisture, and momentum: Approach and application to ARM measurements. J. Atmos. Sci, 54 , 1503–1524.
Zhu, P., 2002: Evolution of shallow cumulus convection and its parameterization. Ph.D. dissertation, University of Miami, 174 pp.
Zhu, P., and B. A. Albrecht, 2001: Formation and development of nocturnal boundary layer clouds over the southern Great Plains. J. Atmos. Sci, 58 , 1409–1426.
Zhu, P., and B. A. Albrecht, 2002: A theoretical and observational analysis on the formation of fair-weather cumuli. J. Atmos. Sci, 59 , 1983–2005.
Zhu, P., and B. A. Albrecht, 2003: Large eddy simulations of continental shallow cumulus convection. J. Geophys. Res.,108, 4453, doi:10.1029/ 2002JD003119.
Schematic illustration of CIN defined by the shaded area. (a) Original definition of CIN according to Bretherton et al. (2004). (b) Modified CIN that allows for forced cumuli
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Cloud boundaries with cloud-base height less than 3 km from the ARSCL dataset at the TWP Nauru site. Shading: cloud boundaries; star: the LFC estimated from the nighttime surface thermodynamic properties recorded by the BBSS. Solid lines represent precipitation rate referenced to the right coordinate. The time period is from 17 Jan to 16 Feb 2000
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
The same as Fig. 2 but for the observations at the SGP site
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a1) Cloud boundaries from the ARSCL dataset 21 Jun 1997 at the SGP site. Dark dot: cloud top; light dot: cloud base; circle: boundary layer height estimated from radiosondes; line: the LCL estimated from the surface meteorological observations. (a2) The same as (a1) but for 6 Jul 1997. (b1) The simulated cumulus-induced mass fluxes with unit kg m−2 s−1 on 21 Jun 1997. (b2) The same as (b1) but for 6 Jul 1997. (c.1), (c.2) The parameterized vertical profiles of mass fluxes at 2030 UTC along with those obtained from LESs (Brown et al. 2002; Zhu 2002)
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Comparison between initial profiles and observations at the SGP cell for cumulus case 21 Jun 1997
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Observed and simulated 2-m air temperature and moisture at the SGP site. (a) Time series of temperature 13 Jun to 13 Jul 1997. Dashed line: obs temperature at the central facility. Light line: simulated temperature. (b) Monthly averaged daily variation of obs temperature at the central facility (dashed line) and simulated temperature (line). Shading: std dev of the obs temperature. Vertical bar: standard deviations of the simulated temperature. (c) The same as (a) but for the water vapor mixing ratio. Light and dot–dashed lines represent the mixing ratio determined by the MRF and GS boundary layer schemes, respectively. (d) The same as (b) but for the mixing ratio. Thick solid line is the value averaged over various facilities at the entire SGP site
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
The temperature difference (MM5 − observation) at 300 hPa for the summer month
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a1) Observed vertical structure of potential temperature by radiosondes at the SGP site from 13 Jun to 13 Jul 1997; (a2) simulated vertical structure of potential temperature in this period; (a3) difference between simulations and observations. (b1)–(b3) The same as (a1)–(a3) but for water vapor mixing ratio
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a) Comparison of 850- and 500-hPa potential temperature from MM5 simulation, ECMWF analyses, and observations at the SGP cell. (b) The same as (a) but for the mixing ratio. (c) Observed and MM5 simulated precipitation
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Monthly averaged diurnal cycle of potential temperature at the SGP site from 13 Jun to 13 Jul 1997: (a) observations, (b) simulations, and (c) difference between simulations and observations
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a) Monthly averaged surface heat fluxes at the SGP site. Dashed line: observed values; solid line: simulated values. (b) Monthly averaged obs (dashed line) and simulated (solid line) downward shortwave radiative fluxes. Shading: std devs of the obs values. Vertical bar: std devs of the simulated values. (c) The same as (b) but for downward longwave radiative fluxes
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Monthly averaged diurnal variation of the parameterized cloud properties of shallow cumuli at the SGP cell. (left) 13 Jun–13 Jul 1997. (right) 1–31 Jan 1998
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a) Cloud boundaries from the ARSCL dataset 17 Jan–16 Feb 2000. Circles and stars are the LCL estimated from the surface thermodynamic properties recorded by the BBSS at 1130 UTC (∼2240 LST) and 2330 UTC (∼1040 LST), respectively. (b) Simulated cumulus-induced mass fluxes with unit kg m−2 s−1
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a) The composite virtual potential temperature profiles at 1130 UTC (solid) and 2330 UTC (light dot– dashed) at the TWP Nauru site from 17 Jan to 16 Feb 2000. Dashed lines indicate the adiabatic profiles estimated from the surface layer air at 1130 UTC. (b) The same as (a) but for the BOMEX case
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Observed and simulated vertical thermodynamic structure at the Nauru site from 17 Jan to 16 Feb 2000. (a1) Potential temperature obs by radiosondes. (a2) Water vapor mixing ratio obs by radiosondes. (b1), (b2) Potential temperature and mixing ratio differences between simulations and observations. (c1), (c2) Monthly averaged obs and simulated profiles of potential temperature and water vapor mixing ratio
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
Domain-averaged differences of potential temperature and water vapor mixing ratio (simulated values with the BM shallow cumulus scheme minus those without the shallow cumulus scheme). (a1) Δθ from 13 Jun to 13 Jul 1997 at the SGP site. (a2) Δq from 13 Jun to 13 Jul 1997 at the SGP site. (b1), (b2) The same as (a1) and (a2) but for the TWP Nauru site
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a) Domain-averaged surface sensible and latent heat fluxes with and without the shallow cumulus scheme for simulations of continental shallow cumuli. (b) The same as (a) but for maritime shallow cumuli
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2
(a1) Domain-averaged cloud fraction determined by the stratiform cloud scheme with the BM shallow cumulus scheme for the summer month at the SGP site. (a2) The same as (a1) but for liquid water content. (b1) Domain-averaged cloud fraction determined by the stratiform cloud scheme without the BM shallow cumulus scheme. (b2) The same as (b1) but for liquid water content. (c1) The diff between (a1) and (b1). (c2) The diff between (b1) and (b2)
Citation: Monthly Weather Review 132, 10; 10.1175/1520-0493(2004)132<2391:ASSOSM>2.0.CO;2