1. Introduction
Significant precipitation with surface rain rates exceeding 1 mm h−1 is frequently observed from shallow cumuli in the Caribbean trade wind region (Short and Nakamura 2000; Nuijens et al. 2009; Snodgrass et al. 2009). Rain evaporation cools the surrounding air and generates convective downdrafts that form cold pools at the surface (Zuidema et al. 2012). The suppression of convection within the stable surface cold pool region and the invigoration of convection at the cold pool downwind boundary are responsible for the arc-shaped mesoscale cloud organizations (Snodgrass et al. 2009; Zuidema et al. 2012) observed during the Rain in Cumulus over the Ocean (RICO) experiment held in the eastern Caribbean from December 2004 to January 2005 (Rauber et al. 2007). The influence of the shallow cumulus cold pools on cloud fraction and rain generation is important for the radiative budget and hydrological cycle of the expansive trade wind region. Although shallow precipitation is now considered in atmosphere general circulation models (e.g., Park and Bretherton 2009), the physics of shallow cumulus cold pools are typically not included. Recent studies attempt to better understand the processes related to shallow cumulus cold pools through fine-resolution (25–100-m horizontal grid spacing) model simulations. The studies have been focusing on the convective invigoration at the cold pool downwind boundary (Li et al. 2014), the mesoscale cloud organization due to cold pools (Seifert and Heus 2013), and the aerosol effects on the organization of shallow cumulus convection through cold pools (Xue et al. 2008; Jiang et al. 2009). Li et al. (2014) discuss the ability of cold pool propagation to increase the humidity, coverage, and strength of subcloud updrafts near the cold pool downwind boundary, with a similar finding for tropical cold pools reported in Schlemmer and Hohenegger (2014). Xue et al. (2008) and Seifert and Heus (2013) report a sensitivity of the shallow cumulus cold pool mesoscale organization to imposed cloud drop number or aerosol number concentration, as well as domain size and grid spacing.
The parameterization of rain microphysics may influence the trade wind cumulus convection independent of aerosol effects. Stevens and Seifert (2008) examine the sensitivity of simulations of shallow cumulus convection to bulk microphysical schemes but do not examine rain evaporation and cold pools. Abel and Shipway (2007) improve the Met Office model’s representation of trade wind cumulus convection, primarily by reducing the specified cloud droplet number concentration
The goal of this study is to understand the shallow cumulus precipitation and cold pool properties in response to cloud and rain microphysics by comparing model simulations that are configured in the exact same setup, except for the microphysical scheme. The analyzed day of 19 January 2005 contains cold pools embedded within the more synoptically defined cloud lines of a dissipated cold front (Caesar 2005). Although this case does not fit the archetypical trade wind regime description, with cloud tops that remained below 4 km, it does represent one of the typical regimes of the wintertime Caribbean. This day has also been selected by other studies for examining trade wind regime cloud properties; for instance, the well-sampled cloud microphysics from this day has been studied and implemented to improve the Met Office’s Large Eddy Model depictions of precipitating shallow cumuli (Abel and Shipway 2007).
This study builds on the previous study (Li et al. 2014) of a nested Weather Research and Forecasting (WRF) Model simulation, with the exact same simulation configuration, except for the microphysical scheme. In addition to the double-moment rain microphysical scheme developed by Greg Thompson [an extension of Thompson et al. (2008)] used in Li et al. (2014), this study also employ the Morrison double-moment scheme (Morrison et al. 2005). With these two simulations, we explore the systematic differences in precipitating convection and their associated cold pools. These two schemes are chosen because both are commonly used available options within WRF, and Li et al. (2014) have already demonstrated that nested WRF simulations invoking the Thompson microphysical scheme produce cold pools comparable to those observed, if weaker and smaller. This study augments that of Van Zanten et al. (2011), who assessed the behavior of 12 LES models against a composite of the less-disturbed 16 December 2004–8 January 2005 RICO time period.
The clouds and precipitation produced by the Thompson and Morrison schemes are also evaluated within the one-dimensional Kinematic Driver model described by Shipway and Hill (2012a), with prescribed updraft velocities. This serves to assess the rain production process rates independently from interactions with other dynamical processes in WRF and allows for a more thorough investigation of the sensitivity of evaporation to the fall speed parameterization.
2. WRF simulations with Thompson and Morrison schemes
a. Description of the microphysical schemes
Aerosol effects are not examined, and the cloud droplet number concentration is specified at 100 cm−3 in both simulations. This is the default value in the Thompson scheme and is slightly higher than the prescription of Nd = 70 cm−3 in Van Zanten et al. (2011), but it falls within the
Autoconversion (the conversion of water from cloud to rain through collision and coalescence among cloud drops) and accretion (the further growth of the raindrops through collecting cloud drops) are parameterized collectively in the Thompson scheme following a modified version of Berry and Reinhardt (1974, hereafter BR). The Thompson double-moment rain scheme evaluated here is an undocumented evolution of a previous single-moment rain scheme (Thompson et al. 2008). BR schemes are explicitly reviewed by Gilmore and Straka (2008), who note that BR schemes are effective because they include information on both the mean cloud droplet size and cloud dispersion, but large rain mixing ratio growth-rate differences are possible between different implementations of the same scheme. The Morrison scheme is based on the separate autoconversion and accretion schemes of Khairoutdinov and Kogan (2000, hereafter KK). The KK scheme is parameterized from explicit microphysics produced by LESs of marine stratocumulus. The KK autoconversion and accretion parameterizations have straightforward dependencies upon the cloud
Rain self-collection and breakup can have a significant impact on surface precipitation rates for shallow cumulus (Stevens and Seifert 2008). Both the Morrison and Thompson schemes include rain self-collection, following Beheng (1994). This is an extension of the original KK parameterization within the Morrison scheme. Rain breakup is implemented within both schemes using slightly different variations of Verlinde and Cotton (1993) [see also Morrison and Milbrandt (2011)].
The KK scheme is employed within two different LES models participating in Van Zanten et al. (2011) and is also used within the Met Office Large Eddy Model simulation for this day (Abel and Shipway 2007). Shipway and Hill (2012a) compare the Thompson (as Thompson09) and Morrison schemes within a one-dimensional kinematic framework that we also employ and report more accumulated water due to more autoconversion, less accretion, and more evaporation for the Thompson scheme. A newer version of the KK scheme, which is developed from explicit microphysics simulations driven by the Van Zanten et al. (2011) composite case, has also been developed by (Kogan 2013) to optimize the scheme for trade wind cumulus. This has an autoconversion rate of
For the same environmental conditions, more evaporation should result in colder cold pools. Small drops evaporate more readily, and a parameterized rain size distribution that has more small drops with a higher total drop number concentration for a given rain rate will evaporate more efficiently. In both the Thompson and Morrison schemes, the raindrop size distribution (RSD) follows an inverse exponential distribution:
b. Simulation setup
The simulations use the updated version 3.2 of the Weather Research and Forecasting Model with multiple nested domains (Li et al. 2014). The parent domain and four two-way nested inner domains are centered at 18°N, 61.7°W within the operational domain of the Research Vessel (R/V) Seward Johnson during the RICO experiment (Zuidema et al. 2012). The simulations span from 0000 UTC 19 January to 0600 UTC 20 January, with only the 1-min output from the last 24 h analyzed unless otherwise indicated. The R/V Seward Johnson radiosondes are assimilated to improve the depiction of the boundary layer vertical structure, and the NCEP Final Analyses (NCEP FNL; NOAA/National Centers for Environmental Prediction 2000) supply the initial and lateral boundary conditions. Only the simulation results from the innermost domain are analyzed. This has a domain size of 24 km × 24 km from the surface to 10 hPa at a constant horizontal grid spacing of 100 m and vertical grid spacing varying from 6 to 200 m between the surface and the 4-km level. The use of realistic large-scale forcing and the nudging to radiosondes enables the cold pools to be produced in a more realistic environment than that of LESs with doubly periodic lateral boundary conditions. In addition, it may overcome some of the deficiencies of doubly periodic LESs; for example, Van Zanten et al. (2011) find that their doubly periodic LESs tend to produce a uniform rain field and a lack of strong rain events. Other specifications of the simulations are described in Li et al. (2014).
In contrast to the one-way input mode in the more traditional LES, the two-way interactions in our simulations allow feedbacks from small to large scales. This means that the differences in microphysics are given more freedom to interact with mesoscale dynamics as well as the convective organization.
The simulated potential temperature θ, water vapor mixing ratio qυ, and wind profiles averaged over the entire day and over three 4-h segments through the day are shown in Fig. 1. The averages over the three 4-h segments make clear that the boundary layer winds change significantly over the course of the day, with the near-surface winds becoming stronger and more northerly with time and northwesterly winds at 3 km transitioning to calmer southwesterlies. In addition, the atmosphere becomes more stably stratified at about 2 km as the day evolves. As would be expected from simulations nudged to the prevailing environmental conditions, the simulations produce rain similar to (if slightly exceeding) that observed. The area-averaged rainfall rate derived from a scanning precipitation radar for this day is 1.87 mm day−1 (Snodgrass et al. 2009), equivalent to 53 W m−2. The Thompson and Morrison simulations produce 1-day domain-averaged surface rain rates of 2.1 and 2.4 mm day−1, respectively, equivalent to 60–70 W m−2 of latent heating. For comparison, the Van Zanten et al. (2011) simulations corresponded to a mean radar-derived latent heating estimate of 21 W m−2.
The domain-averaged (a) potential temperature, (b) water vapor mixing ratio, and (c) zonal and (d) meridional wind speed profiles averaged over the 24 h for the WRF Thompson (black) and Morrison (red) simulations, compared with the averaged soundings from 0700 UTC 19 Jan to 0300 UTC 20 Jan (green). Domain-averaged profiles averaged over 0800–1200 UTC 19 Jan (black dotted), 1400–1800 UTC 19 Jan (black dashed), and 0200–0559 UTC 20 Jan (black dashed–dotted) for the WRF Thompson simulation. Blue profiles in (a) and (b) indicate the initial profiles used in KiD experiments of section 3.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
c. Cloud and rain in WRF simulations
The evolution of the domain maximum surface rain rate (
For the WRF Thompson (black) and Morrison (solid red) simulations with Nd = 100 cm−3, the 1-h smoothed (a) domain maximum surface rain rate, (b) domain-averaged surface rain rate, (c) accumulated domain-averaged surface rain, (d) domain-mean 3–1-km wind (speed) shear, (e) domain maximum updraft vertical velocity at the cloud-base level, (f) cloud-base-level updraft mass flux, (g) cloud cover, and (h) domain-averaged liquid water path. Dashed red lines in (a),(b), and (e) correspond to the Morrison scheme with Nd = 250 cm−3. Only this figure uses the first 6-h simulation data.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
For the WRF Thompson (black) and Morrison (red) simulations, (a),(e),(i) the cloud (solid) and cloud core (dotted) fraction; (b),(f),(j) the cloud mixing ratio within cloud; (c),(g),(k) the rain fraction; and (d),(h),(l) the rain mixing ratio within rain shaft. Domain-averaged profiles averaged over (a)–(d) 0800–1200 UTC 19 Jan, (e)–(h) 1400–1800 UTC 19 Jan, and (i)–(l) 0200–0559 UTC 20 Jan. Dashed red profiles indicate the Morrison scheme with Nd = 250 cm−3.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
The cloud cover of the Thompson simulation is lower than that of the Morrison simulation for most of the integration period (Fig. 2g), with mean cloud cover over the last 24 h of 12.7% for Thompson versus 19.8% for Morrison and mean cloud cover over the last 16 h (after 1400 UTC) of 14.8% for Thompson versus 22.4% for Morrison. The difference in mean cloud cover between the two simulations is significant above the 99% confidence level. The microphysics-caused difference is also evident in LWP as well as the accumulated surface rain, with a mean LWP of 60 g m−2 for Morrison and 48 g m−2 for Thompson and a mean accumulated surface rain of 1.29 mm for Morrison and 0.95 mm for Thompson.
The domain maximum cloud-base vertical velocities slightly exceed those from aircraft observations on this day (Abel and Shipway 2007) as well as from a shipboard Doppler lidar (Zuidema et al. 2012), but this may represent sampling differences. In the observations, the shipboard rain measurements and scanning radar images show precipitation occurring most frequently between 1200 and 1800 UTC (Zuidema et al. 2012). This is better captured by the Morrison simulation with a later rain initiation and shorter quiescent period. In the latter half of the simulation, when the vertical wind shear becomes larger, the Morrison simulation using the higher
Further insight is provided through examining the vertical structure of the cloud and rain fractions and mixing ratios shown in Fig. 3, averaged over the same 4-h segments as in Fig. 1. Cloudy regions are defined by a cloud water mixing ratio qc > 0.1 g kg−1, the convective core through cloudy regions with positive vertical velocity (
From 1400 to 1800 UTC, wind shear at 3 km (Fig. 2d) is evident as enhanced cloud detrainment, more pronounced for the higher-extending clouds in the Morrison simulation. As the simulation progresses, the cloud tops lower in both simulations (Figs. 3a,e,i). Given the strengthened surface winds, one might expect a deeper boundary layer (Nuijens and Stevens 2012). The lowered cloud tops instead appear to reflect a new inversion advected in at 2 km of about 4 g kg−1 for specific humidity and 5 K for potential temperature within 300-m vertical depth, apparent in the individual radiosondes (Zuidema et al. 2012) and to some extent in Fig. 1. The Morrison simulation with Nd = 250 cm−3 shows a higher cloud and rain mixing ratio reaching higher altitudes earlier in time (Figs. 3b,d) but less rain produced at all vertical levels as the day progresses. This may reflect the influence of the increasing wind shear.
The relationship of the cloud cover to the domain-averaged surface rain rate is shown statistically in Fig. 4. Cloud cover is clearly related to the average surface rain rate within both simulations, but more clearly within the Thompson simulation. The cloud cover within the Morrison simulation is generally greater, reaching 40% at times. At low rain rates, the Morrison cloud cover is almost independent of the surface rain rate. Van Zanten et al. (2011) did not find a strong correspondence between the cloud cover and precipitation for their overall weaker precipitation rates. This could be consistent with the Morrison results at low rain rates shown here. The clear increase in cloud cover with surface rain at higher rain rates may reflect the shearing off of upper cloud through wind shear, as seen in observations (Zuidema et al. 2012).
Contoured frequency distribution of the top-down cloud cover as a function of the domain-averaged rain rates for the WRF Thompson (black) and Morrison (red) simulations. Contour levels indicate the percentile of output minutes included.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
A more difficult aspect of the two simulations to assess is their fidelity to the observed mesoscale organization. Two-dimensional fast-Fourier transforms (2D FFT) provide one objective approach to comparing the two simulations and, in addition, allow for an assessment of which spatial scales provide more of the overall variance (e.g., Seifert and Heus 2013). Time series of power spectra of the vertical velocity and total water at 2-km height reveal more variance for the Morrison simulation than for the Thompson simulation in total water, but not in the vertical velocity (except between 0800 and 1200 UTC; figures not shown), consistent with Figs. 2 and 4. The greater variance in total water is not linked to a particular eddy size, consistent with an overall slower autoconversion rate.
We include two snapshots from when RRmax is 55 mm h−1 for Thompson and 45 mm h−1 for Morrison at 0300 UTC 20 January (Fig. 5). The mean power spectra of vertical velocity within the cloud layer do not differ significantly between the two simulations, yet the snapshots do reveal more subtle differences. Horizontal snapshots of the second innermost domain at 0300 UTC show that both simulations capture the synoptic cloud line and are clearly producing cold pools (Fig. 6). Examination of cloud field and surface temperature movies indicates these features are commonly present in both simulations through the model integration time. The Thompson simulation contains two smaller cloud lines within the innermost domain, as compared to the longer cloud line produced by the Morrison simulation. The clouds within the Morrison simulation are more congregated than those in the Thompson simulation (Figs. 5a,d), and the surface cold pool corresponding to the domain maximum surface rain rate is colder and larger, with updrafts located at the cold pool boundaries (Figs. 5b,e). A cross section is taken along the 3-m-level domain-mean wind direction across the point of domain maximum surface rain rate for each simulation (Figs. 5c,f). The cloud associated with the heaviest rain in the Thompson simulation is tilted by the vertical wind shear, with the heaviest rain occurring in the middle of the cloud layer and the downdraft dislocated from the cloud-base-level updraft (Fig. 5c). This geometry provides a positive feedback for furthering secondary convection, explored more deeply in Li et al. (2014). The Morrison simulation is much cloudier, consistent with Fig. 3i. Cloud overlies much of the cold pool, in contrast to observations. Different from the Thompson simulation, the rain in the Morrison simulation occurs near the cloud top, which is sheared off by the change of wind direction above 2 km (Figs. 1c,d). This explains the reason that the vertical cross section across the maximum surface rainrate grid point for the Morrison simulation does not capture the cloud mass at cloud-base level (Fig. 5f).
For the WRF (a)–(c) Thompson and (d)–(f) Morrison simulations at 0300 UTC 20 Jan, (left) horizontal snapshot of cloud water path (g m−2); (middle) horizontal snapshot of 3-m-level
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
The 3-m-level air temperature (shaded) and nonzero cloud water path (green contours) for the (left) Thompson and (right) Morrison simulations at 0300 UTC 20 Jan in the outer simulation domain of the innermost domain. The black square indicates the boundaries of the innermost domain.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
d. Cold pool properties and rain evaporation in WRF simulations
The surface air buoyancy is modified by the cold pools present within the domain. More larger cold pools would result in greater coverage of negative buoyancy anomaly. The surface air buoyancy anomaly
For the WRF Thompson (black) and Morrison (red) simulations, the total fractional area containing surface
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
The 3-m-level cold pool properties are examined according to their associated domain maximum surface rain rate (Fig. 8). The domain maximum surface rain rate is used as a reference for the convective intensity, as it correlates better with the maximum surface changes within cold pools than the domain-averaged surface rain rate, except for the surface change in
For the WRF (top) Thompson and (bottom) Morrison simulations, contoured frequency distribution of the 3-m-level maximum horizontal anomalies of (left)–(right) θ,
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
The properties of cold pools are first and most dominantly determined by the convective intensity, as previously shown in observations (Zuidema et al. 2012) and captured by both the simulations (Fig. 8). The differences in cold pool production by the two schemes are magnified under close examinations that focus on moderate and intense convection, respectively. The consistencies between Figs. 7 and 8 point to the conclusion that the Thompson scheme is more likely to produce colder cold pools under the more frequently occurring moderate rain condition, resulting in larger area with weak negatively buoyant air. However, the few intense rain cases should not be ignored, because they are responsible for a large portion of surface precipitation and often further convection (Zuidema et al. 2012). For these intense rain cases, the Morrison scheme produces colder and stronger cold pools.
The evaporation process should explain the different cold pool temperature depressions produced by the two simulations as a function of rain rate. The column-integrated evaporation rate
For the WRF Thompson (black) and Morrison (red) simulations, contoured frequency distribution of the column-integrated evaporation rate at the grid points containing the domain maximum surface rain rate of each output minute (top) integrated over the entire model column and (bottom) integrated over the column from the surface to 1-km altitude.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
In Fig. 10, the vertical profiles of the rain mixing ratio
For the WRF Thompson (black) and Morrison (red) simulations, composite profiles of (a) rain mixing ratio, (b) evaporation rate, and (c) evaporation efficiency, averaged over grid columns containing the domain maximum surface rain rate: 5 < RRmax < 10 mm h−1 for moderate (solid) and 40 < RRmax < 80 mm h−1 for intense (dotted) rain cases.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
3. 1D column model experiments
Experiments with the one-dimensional Kinematic Driver model (KiD; Shipway and Hill 2012a) using the Thompson and Morrison microphysical schemes confirm that the dynamical feedbacks are not masking the differences in rain and evaporation that can be attributed to the schemes themselves. The analysis also allows for a cleaner comparison on the parameterizations of the processes contributing to the rain mass water budget: autoconversion, accretion, sedimentation, and rain evaporation.
As rain falls, the RSD is altered by the processes of rain self-collection and raindrop breakup, and the raindrop fall speed associated with different drop sizes. Size sorting is handled somewhat differently by the two schemes, in part through different parameterizations of the fall speeds, and in part by different treatments of sedimentation. The Morrison scheme determines the sedimentation of the number and mass contents by applying analytic formulations to the number- and mass-weighted fall speeds. The Thompson scheme introduces an ad hoc increase in the number-weighted fall speed to reduce the difference from the mass-weighted fall speed and thereby reduce the size sorting. In this study, we experiment with modifying the fall speed parameterization only to examine how differences in parameterized fall speed affect precipitation and associated evaporation. Condensation onto the raindrops is negligible (Shipway and Hill 2012a) and is not considered in either scheme.
The microphysical processes in KiD are driven by prescribed vertical velocity profiles. All experiments are run for 5400 s, with a 1-s time step, and outputting prognostic variables every 10 s. The column is set to be 7 km high, with 120 full-z levels and 119 half-z levels at a vertical grid spacing of approximately 58 m.
The initial thermodynamic profiles resemble the average profiles from WRF simulations (Fig. 1). In the default setting of this model, as suggested in Shipway and Hill (2012a), the θ profile is held constant throughout the simulation to minimize the feedbacks from the two different schemes, and only the water vapor mixing ratio is updated. The imposed vertical velocity profiles resemble an updraft core modeled on data from Abel and Shipway (2007) [test case 5 in Shipway and Hill (2012b)], varying exponentially with height and time. The profiles reach their peak value
a. Cloud and rain comparisons in KiD
Two KiD simulations are performed using the same updraft profiles and configured with the two microphysical schemes, respectively. The averaged updraft profile reaches its maximum
Profiles from the KiD Thompson (black) and Morrison (red) simulations driven by the same vertical velocities at all altitudes and all times, averaged over time for (a) cloud water mixing ratio when cloud is present (qc > 0.1 g kg−1) at each vertical level, (b) rain mixing ratio when rain is present (qr > 0.1 g kg−1) at each vertical level, and (c) autoconversion, (d) accretion, and (e) evaporation rates when these rates are nonzero at each vertical level.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
Consistent with the WRF simulations, the Morrison scheme clearly produces more cloud than the Thompson scheme (Fig. 11a). A breakdown by the rain budget processes reveals, for a similar rain mixing ratio (Fig. 11b), the autoconversion rate is higher for the Thompson scheme (Fig. 11c), and the accretion rate is higher for the Morrison scheme (Fig. 11d), along with a higher evaporation rate (Fig. 11e) corresponding to higher surface rain rate.
The significantly smaller autoconversion rate of the Morrison scheme (Fig. 11c) is capable of explaining the higher Morrison cloud fractions in the WRF simulations. This is especially true for the lower cloud layer (below 1.5 km in Fig. 11c). The autoconversion process of these two schemes is discussed explicitly in Shipway and Hill (2012a).
The accretion contributes to much more rain production than the autoconversion for both schemes (Fig. 12), as expected for cumulus clouds. The Morrison scheme’s accretion rate is more active than that of the Thompson scheme (Fig. 12), especially at higher altitude (Fig. 11d). This may explain why rain is present at higher altitudes in the Morrison WRF simulation than in the Thompson simulation (Fig. 3).
Time series of the column-integrated autoconversion rate (black dotted), accretion rate (black dashed), the sum of autoconversion and accretion rates (black solid), and evaporation rate (green), along with the surface rain rate (light blue) for the same KiD (a) Thompson and (b) Morrison simulations shown in Fig. 11. Both evaporation and surface rain rates are expressed as sink terms within the rainwater mass budget.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
The Thompson evaporation rate is significantly greater than that of the Morrison scheme for a similar averaged rain mixing ratio (Fig. 11), which is also seen in the WRF simulation results (Fig. 10). The evaporation all occurs below 1.1 km in the KiD simulations, maximized near cloud base, in contrast to the near-surface maximum evaporation rate in the WRF simulations (Fig. 10). The rain and evaporation in WRF simulations occur in downdrafts that are not represented in the KiD simulations, likely causing the difference in vertical structure of the evaporation.
The time series comparison of the column-integrated rain production rate and evaporation rate for the two schemes are shown in Fig. 12. For the same dynamical forcing, the Morrison scheme produces larger surface rain rates, and less rain evaporation, than the Thompson scheme, consistent with Shipway and Hill (2012a).
b. KiD rain evaporation comparison and fall speed dependence
The KiD simulations driven by updraft profiles reaching
Composite profiles of (a) rain mixing ratio, (b) rain evaporation rate, and (c) rain evaporation efficiency for the KiD Thompson (black) and Morrison (red) simulations averaged over 1000 s containing surface rain rates of approximately 5 mm h−1 for moderate (solid) and 60 mm h−1 for intense (dotted) rain cases. Green profiles represent the Thompson simulation with fall speed parameters replaced by the ones used in the Morrison scheme for moderate (green solid) and intense (green dotted) rain cases.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
One factor that can influence the modeled rain evaporation is the parameterized number- and mass-weighted fall speed. The Thompson scheme parameterizes a higher fall speed than the Morrison scheme for raindrop sizes up to about 3 mm, with the maximum difference occurring around 2 mm (Fig. 14; note that the fall speed is truncated at 9.1 m s−1). The effects of the fall speed difference are investigated by replacing the fall speed parameters in the Thompson scheme with the ones used in the Morrison scheme (thus, only the results from Thompson simulations may be affected). Figure 13c shows that the evaporation efficiency of the modified moderate-rainrate Thompson simulation is significantly reduced, matching that of the Morrison simulation at altitudes below 800 m. For the intense rain rates, the replacement of the fall speed parameterization only slightly reduces the rain mixing ratio and rain evaporation of the Thompson simulation, with no discernible impact on the evaporation efficiency (Fig. 13).
The parameterized terminal fall speed of raindrops over the size spectrum 100 μm–3 mm for the Thompson (black) and the Morrison (red) scheme.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
These changes are likely explained through the treatment of size sorting. The number-weighted raindrop fall speeds are applied to the sedimentation of
The cloud-base-level (460 m) RSDs corresponding to (a) moderate KiD simulations by the Thompson scheme (black solid), Morrison scheme (red solid), and Thompson scheme with fall speed parameter replaced (green solid) and (b) intense KiD simulations by the Thompson scheme (black dotted), Morrison scheme (red dotted), and Thompson scheme using the Morrison fall speed parameterization (green dotted). The RSDs are averaged over the 1000 s corresponding to Fig. 13. (c) The cloud-base-level (450 m) RSDs in the WRF Thompson (black dotted) and Morrison (red dotted) simulations, averaged over grid columns with RRmax between 50 and 70 mm h−1.
Citation: Journal of the Atmospheric Sciences 72, 9; 10.1175/JAS-D-14-0099.1
4. Conclusions and discussion
In this study, the impact of microphysics on precipitating shallow convection and cold pools is assessed by examining the characteristics of rain production and evaporation within two popular microphysical schemes available in WRF—namely, the Thompson and Morrison schemes—using nested WRF simulations of one particular day. The Thompson scheme is applied within Li et al. (2014) to examine secondary convection at cold pool boundaries, with this study providing a measure of the sensitivity of the Li et al. (2014) findings to the microphysical scheme employed.
The two schemes have different influence on cold pool characteristics for moderate and intense rain cases. For moderate rain rates, the Thompson simulation contains more colder and stronger cold pools, consistent with more rain evaporation and smaller mean drop sizes. The situation is reversed for the intense rain rates that occur much less often but contribute importantly to the overall rain rates. The Morrison scheme generates more rain evaporation for the very intense rain cases. One-dimensional kinematic model experiments show the same relationships among the Thompson and Morrison simulations for moderate and intense rain cases. The consistency between the 1D model and WRF simulations indicates that the dynamical feedbacks do not mask the impact of microphysics on precipitation and evaporation.
The sensitivity of the evaporation to the fall speed parameterization is assessed in the 1D model by replacing the raindrop terminal fall speed parameters in the Thompson scheme with the ones used in the Morrison scheme. The Thompson scheme parameterizes faster raindrop fall speeds than does the Morrison scheme. With a lowered fall speed, the modified moderate Thompson rain mixing ratio increases, and the number of small raindrops decreases, decreasing evaporation and the evaporation efficiency to match that of the Morrison simulation. For the very intense rain cases that occur much less frequently in the 3D simulations, however, fall speed differences are not responsible for the differences in rain mixing ratio, evaporation rate, and evaporation efficiency between the two schemes.
Cold pools are related to observed cloud mesoscale organization; however, the relative ability of the two microphysical schemes to produce a realistic cloud and rain mesoscale organization is more difficult to assess. Although the detailed appearance and structure of mesoscale cloud organizations are different in the two simulations, the power spectra of both dynamic (winds) and thermodynamic (θ,
The most robust difference in cloud production between the two simulations is the higher average cloud fraction within the Morrison simulation (19.8% vs 12.7% for the Thompson scheme). Clouds in the Morrison simulation also reach higher altitudes more often, where they are subsequently sheared. The cloud cover is correlated to the surface rain rates in both simulations but more clearly in the Thompson simulation. At low rain rates, the Morrison cloud cover is almost independent of the surface rain rate. The differences in cloud cover and LWP are significantly larger than the difference caused by random initial conditions discussed in Grabowski (2014); thus, it should not be simply considered as a sort of uncertainty, but, rather, the outcome resulted from differently parameterized microphysical processes. The higher cloud fraction in the Morrison simulation is consistent with its slower autoconversion rate. Recently, Kogan (2013) has modified the KK scheme to better represent shallow cumulus clouds. The newer autoconversion parameterization depends more strongly on
A further consideration is the breadth of the cloud and raindrop size distributions. The shape of the cloud drop spectrum is important for establishing the autoconversion rate within the Thompson scheme (the Morrison autoconversion depends on the number and mass of the cloud droplets but not the spectral width). Both schemes prescribe cloud dispersions according to the observational analysis of Martin et al. (1994) for marine stratocumulus (dispersion is 0.278 for a gamma shape parameter μ = 12). This adequately captures the dispersion of the RICO shallow cumulus drop size distributions at cloud base, but the dispersion increases significantly at higher altitudes within the cloud (Arabas et al. 2009). As such, the Thompson scheme should underestimate the autoconversion rates above cloud base, all else being equal. Neither scheme considers a raindrop size distribution that incorporates a nonzero μ, but Geoffroy et al. (2014) confirm that an inverse exponential distribution fits in situ RSDs from RICO well. Nevertheless, exponential raindrop size distributions encourage excessive size sorting that can produce peak surface precipitation rates exceeding those from other schemes when employed in double-moment schemes (Wacker and Seifert 2001; Shipway and Hill 2012a). Simulations in which μ is allowed to differ from zero produce vertical rain mass mixing ratio profiles that are more weighted to the cloud base rather than to the surface (Stevens and Seifert 2008; Kogan 2013).
The assessment provided here adds to the growing literature evaluating the ability of high-resolution models to emulate the mesoscale organization of precipitating shallow cumulus. Neither scheme evaluated here is ideally suited for precipitating shallow marine cumulus, and the substantial difference in cloud cover and LWP has profound implications for the model radiative balance of the trade wind regime, suggesting more work remains to be done. A useful initial further step would be to evaluate microphysical parameterizations developed specifically for shallow marine cumulus, such as Kogan (2013), with the in situ microphysical datasets gathered during RICO and other shallow marine cumulus deployments, similar to the marine stratocumulus assessments of Wood (2005). A BR-based scheme would likely benefit from a revisitation of dedicated bin-model experiments (Gilmore and Straka 2008), such as those of Kogan (2013). Further dedicated simulations based on such assessed microphysics would contribute to a more confident understanding of the microphysical–dynamical feedbacks for the strongly precipitating shallow marine cumulus that influence a large portion of our planet. In particular, the influence of wind shear on cold pool mesoscale organization in the climatological trade wind region would benefit from further assessment through ensembles of simulations, building on the simulations evaluated here.
Acknowledgments
We acknowledge support from the National Science Foundation through Grant AGS-1114521 and the National Oceanic and Atmospheric Administration Climate Variability Program through Grant NA13OAR4310157 for this work. Input from Greg Thompson and the anonymous comments of three reviewers are greatly appreciated. Movies of the horizontal cloud field, surface temperature, and precipitation from the WRF simulations are available upon request by personal contact with the author.
REFERENCES
Abel, S. J., and B. J. Shipway, 2007: A comparison of cloud-resolving model simulations of trade wind cumulus with aircraft observations taken during RICO. Quart. J. Roy. Meteor. Soc., 133, 781–794, doi:10.1002/qj.55.
Arabas, S., H. Pawlowska, and W. W. Grabowski, 2009: Effective radius and droplet spectral width from in-situ aircraft observations in trade-wind cumuli during RICO. Geophys. Res. Lett., 36, L11803, doi:10.1029/2009GL038257.
Beheng, K. D., 1994: A parameterization of warm cloud microphysical conversion processes. Atmos. Res., 33, 193–206, doi:10.1016/0169-8095(94)90020-5.
Berry, E. X., and R. L. Reinhardt, 1974: An analysis of cloud drop growth by collection: Part II. Single initial distributions. J. Atmos. Sci., 31, 1825–1831, doi:10.1175/1520-0469(1974)031<1825:AAOCDG>2.0.CO;2.
Caesar, K.-A. L., 2005: Summary of the weather during the RICO project, 30 pp. [Available online at http://www.eol.ucar.edu/projects/rico/kathy_caesar.pdf.]
Geoffroy, O., A. P. Siebesma, and F. Burnet, 2014: Characteristics of the raindrop distributions in RICO shallow cumulus. Atmos. Chem. Phys. Discuss., 14, 677–705, doi:10.5194/acpd-14-677-2014.
Gilmore, M. S., and J. M. Straka, 2008: The Berry and Reinhardt autoconversion parameterization: A digest. J. Appl. Meteor. Climatol., 47, 375–396, doi:10.1175/2007JAMC1573.1.
Grabowski, W., 2014: Extracting microphysical impacts in large-eddy simulations of shallow convection. J. Atmos. Sci., 71, 4493–4499, doi:10.1175/JAS-D-14-0231.1.
Hudson, J. G., and S. Noble, 2014: Low-altitude summer/winter microphysics, dynamics, and CCN spectra of northeastern Caribbean small cumuli, and comparisons with stratus. J. Geophys. Res. Atmos.,119, 5445–5463, doi:10.1002/2013JD021442.
Jiang, H., G. Feingold, and I. Koren, 2009: Effect of aerosol on trade cumulus cloud morphology. J. Geophys. Res., 114, D11209, doi:10.1029/2009JD011750.
Khairoutdinov, M., and Y. Kogan, 2000: A new cloud physics parameterization in a large-eddy simulation model of marine stratocumulus. Mon. Wea. Rev., 128, 229–243, doi:10.1175/1520-0493(2000)128<0229:ANCPPI>2.0.CO;2.
Kogan, Y., 2013: A cumulus cloud microphysics parameterization for cloud-resolving models. J. Atmos. Sci., 70, 1423–1436, doi:10.1175/JAS-D-12-0183.1.
Li, Z., P. Zuidema, and P. Zhu, 2014: Simulated convective invigoration processes at trade wind cumulus cold pool boundaries. J. Atmos. Sci., 71, 2823–2841, doi:10.1175/JAS-D-13-0184.1.
Martin, G., D. W. Johnson, and A. Spice, 1994: The measurement and parameterization of effective radius of droplets in warm stratocumulus clouds. J. Atmos. Sci., 51, 1823–1842, doi:10.1175/1520-0469(1994)051<1823:TMAPOE>2.0.CO;2.
Morrison, H., and J. Milbrandt, 2011: Comparison of two-moment bulk microphysics schemes in idealized supercell thunderstorm simulations. Mon. Wea. Rev., 139, 1103–1130, doi:10.1175/2010MWR3433.1.
Morrison, H., J. A. Curry, and V. I. Khvorostyanov, 2005: A new double-moment microphysics parameterization for application in cloud and climate models. Part I: Description. J. Atmos. Sci., 62, 1665–1677, doi:10.1175/JAS3446.1.
Morrison, H., S. Tessendorf, K. Ikeda, and G. Thompson, 2012: Sensitivity of a simulated midlatitude squall line to parameterization of raindrop breakup. Mon. Wea. Rev., 140, 2437–2460, doi:10.1175/MWR-D-11-00283.1.
NOAA/National Centers for Environmental Prediction, 2000: NCEP FNL Operational Model Global Tropospheric Analyses, continuing from July 1999 (updated daily). National Center for Atmospheric Research Computational and Information Systems Laboratory Research Data Archive, doi:10.5065/D6M043C6.
Nuijens, L., and B. Stevens, 2012: The influence of wind speed on shallow marine cumulus convection. J. Atmos. Sci., 69, 168–184, doi:10.1175/JAS-D-11-02.1.
Nuijens, L., B. Stevens, and A. P. Siebesma, 2009: The environment of precipitating shallow cumulus convection. J. Atmos. Sci., 66, 1962–1979, doi:10.1175/2008JAS2841.1.
Park, S., and C. S. Bretherton, 2009: The University of Washington shallow convection and moist turbulence schemes and their impact on climate simulations with the Community Atmosphere Model. J. Climate, 22, 3449–3469, doi:10.1175/2008JCLI2557.1.
Rauber, R. M., and Coauthors, 2007: Rain in shallow cumulus over the ocean: The RICO campaign. Bull. Amer. Meteor. Soc., 88, 1912–1928, doi:10.1175/BAMS-88-12-1912.
Rutledge, S. A., and P. V. Hobbs, 1983: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. VIII: A model for the “seeder-feeder” process in warm-frontal rainbands. J. Atmos. Sci., 40, 1185–1206, doi:10.1175/1520-0469(1983)040<1185:TMAMSA>2.0.CO;2.
Schlemmer, L., and C. Hohenegger, 2014: The formation of wider and deeper clouds as a result of cold-pool dynamics. J. Atmos. Sci., 71, 2842–2858, doi:10.1175/JAS-D-13-0170.1.
Seifert, A., and T. Heus, 2013: Large-eddy simulation of organized precipitating trade wind cumulus clouds. Atmos. Chem. Phys., 13, 5631–5645, doi:10.5194/acp-13-5631-2013.
Shipway, B. J., and A. A. Hill, 2012a: Diagnosis of systematic differences between multiple parametrizations of warm rain microphysics using a kinematic framework. Quart. J. Roy. Meteor. Soc., 138, 2196–2211, doi:10.1002/qj.1913.
Shipway, B. J., and A. A. Hill, 2012b: Kinematic Driver model (KiD) user guide, version 2.3.2625. Met Office, 25 pp. [Available online at http://appconv.metoffice.com/microphysics/doc.shtml.]
Short, D. A., and K. Nakamura, 2000: TRMM radar observations of shallow precipitation over the tropical oceans. J. Climate, 13, 4107–4124, doi:10.1175/1520-0442(2000)013<4107:TROOSP>2.0.CO;2.
Snodgrass, E. R., L. D. Girolamo, and R. M. Rauber, 2009: Precipitation characteristics of trade wind clouds during RICO derived from radar, satellite, and aircraft measurements. J. Appl. Meteor. Climatol., 48, 464–483, doi:10.1175/2008JAMC1946.1.
Srivastava, R. C., and J. L. Coen, 1992: New explicit equations for the accurate calculation of the growth and evaporation of hydrometeors by the diffusion of water vapor. J. Atmos. Sci., 49, 1643–1651, doi:10.1175/1520-0469(1992)049<1643:NEEFTA>2.0.CO;2.
Stevens, B., and A. Seifert, 2008: Understanding macrophysical outcomes of microphysical choices in simulations of shallow cumulus convection. J. Meteor. Soc. Japan, 86A, 143–162, doi:10.2151/jmsj.86A.143.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 5095–5115, doi:10.1175/2008MWR2387.1.
van Weverberg, K., N. van Lipzig, L. Delobbe, and A. Vogelmann, 2012: The role of precipitation size distributions in km-scale NWP simulations of intense precipitation: Evaluation of cloud properties and surface precipitation. Quart. J. Roy. Meteor. Soc., 138, 2163–2181, doi:10.1002/qj.1933.
Van Zanten, M. C., and Coauthors, 2011: Controls on precipitation and cloudiness in simulations of trade-wind cumulus as observed during RICO. J. Adv. Model. Earth Syst.,3, M06001, doi:10.1029/2011MS000056.
Verlinde, J., and W. R. Cotton, 1993: Fitting microphysical observations of nonsteady convective clouds to a numerical model: An application of the adjoint technique of data assimilation to a kinematic model. Mon. Wea. Rev., 121, 2776–2793, doi:10.1175/1520-0493(1993)121<2776:FMOONC>2.0.CO;2.
Wacker, U., and A. Seifert, 2001: Evolution of rain water profiles resulting from pure sedimentation: Spectral vs. parameterized description. Atmos. Res., 58, 19–39, doi:10.1016/S0169-8095(01)00081-3.
Wood, R., 2005: Drizzle in stratiform boundary layer clouds. Part II: Microphysical aspects. J. Atmos. Sci., 62, 3034–3050, doi:10.1175/JAS3530.1.
Xue, H., G. Feingold, and B. Stevens, 2008: Aerosol effects on clouds, precipitation, and the organization of shallow cumulus convection. J. Atmos. Sci., 65, 392–406, doi:10.1175/2007JAS2428.1.
Zuidema, P., and Coauthors, 2012: On trade wind cumulus cold pools. J. Atmos. Sci., 69, 258–280, doi:10.1175/JAS-D-11-0143.1.