Results from control simulations:
STEPS—cold-based clouds: control simulation
Hydrometeors profiles associated with warm-rain and ice crystal processes: cold-based clouds
Figure 2 displays the five microphysical species and the warm (red line) and cold (blue line) components of rain and graupel. In the simulated storm, most graupel is from the ice crystal process with riming of snow (Fig. 2e, f). In the STEPS case of cold-based clouds, the abundance of continental aerosol particles (~2500 cm-3 at supersaturation of 1% at ~1 km mean sea level (MSL) 27) makes droplets too small for coalescence (< 20 mean diameter). Both the high CCN concentration and cold cloud base suppress the warm-rain process (red curves in Figs. 2a-d).
The mass and number mixing ratios of cold graupel (blue curves in highlighted regions of Figs. 2a, b) are higher than the corresponding warm components by almost one order of magnitude above freezing level. The simulated ice concentration is highest (~105 m-3) at about -40oC (Fig. 2f) in the upper troposphere, due to homogenous freezing. The conditionally averaged number concentration of cloud droplets is just over half an order of magnitude higher near cloud base (~0oC) than at about -30oC (Fig. 2g) due to riming during ascent.
Contribution of warm-rain and ice-crystal processes to surface precipitation: cold-based clouds
Figure 3a shows that by the end of the simulation about 90% of surface precipitation is from the ice crystal process. The warm-rain process is suppressed by the small mean size of cloud droplets from the cold cloud base (~0oC) and continental aerosol conditions (~2500 cm-3), as noted above. At temperatures colder than -10oC, any trace amounts of supercooled "warm-rain" from coalescence can freeze to form "warm graupel", followed by melting as it falls out (Fig. 3a). The cold surface precipitation is from melting of snow and cold graupel. Accumulated stratiform precipitation (Fig. 3c) in this simulated "downburst event"46 on 19 June is smaller by ~95% compared to convective precipitation (Fig. 3b), because its particles are too small to survive evaporation in the dry deep sub-cloud layer. Overall, cold precipitation is predicted to dominate both the convective and stratiform components of surface precipitation (upper panel of Fig. S20 (schematic) highlights this result).
MC3E—slightly warm-based clouds: control simulation
Validation of AC
In the second case study, we simulated slightly warm-based deep convective clouds observed during the MC3E campaign on 10-13 May 201147. Among these three days, the MC3E campaign data show that the major convective event was observed on 11 May 201147. AC is evaluated by comparing the observed and modeled cloud properties and aerosol activity. The observed cloud and precipitation features were realistically reproduced, such as the evolution, storm propagation, precipitation rate, radar reflectivity, and vertical velocity statistics of the case. In our case, the predicted cloud parameters show a reasonable agreement with the observations to within the uncertainty limit (Fig. 4).
The simulated and observed CCN spectra (~2000 CCN cm-3 at supersaturation of 1% near ground) agree adequately (Fig. 4a). AC also predicts ice nuclei (IN) activity from these aerosol conditions, and this matches nearby IN observations from a similar campaign over the same region and month of another year.
The predicted vertical profiles of concentration (Fig. 4b), diameter (Fig. 4c) of cloud droplets and LWC (Fig. 4d) closely match the observations to within 20-25% at most levels. The predicted ice particle concentrations with their sizes larger than 200(at vertical velocity > 2 ms-1) agrees well with the aircraft observations. We have also validated the predicted ice particle concentrations in the downdraft and stratiform regions. The predicted ice concentration differs from the observations by less than 40% (Fig. 4e). Similarly, predicted and observed vertical profiles of radar reflectivity (Ka-Band ARM Zenith Radars (KAZR)) agree (Fig 4f). Differences between observed and predicted accumulated surface precipitation on 11 May 2011 during 0000-1800 UTC is less than < 10% (Fig. 4g).
Hydrometeors profiles associated with warm-rain and ice crystal processes: slightly warm-based clouds
Above the freezing level, the mass and number mixing ratios of cold graupel are higher and lower by an order of magnitude, respectively, than the corresponding warm components (Fig. 5a, b). Numerous small supercooled drops of rain/drizzle freeze aloft. This contrasts with the cold-based clouds noted above (Fig. 2a, b). Below the freezing level, cold components of mass and number of rain and graupel dominate, but the warm components cannot be ignored (Fig. 5b, d).
Contribution of warm-rain and ice-crystal processes to surface precipitation: slightly warm-based clouds
Figure 6 shows that the warm and cold components of total precipitation accumulated at the ground are comparable throughout this case. The same is true for the convective and stratiform regions. Nearly 60% of the total surface precipitation is from cold precipitation, and the remainder is from the warm-rain process. Overall surface precipitation over the entire domain is mostly from stratiform clouds (>~80%). As compared to STEPS (Fig. 3), the warm-rain process is more influential due to a warmer cloud base and larger cloud droplets aloft in MC3E (Fig. 6).
In summary, the predicted precipitations at the ground for both convective storms are dominated by the ice-crystal process.
Results from sensitivity simulations:
Sensitivity of warm-rain and ice-crystal processes to cloud base, aerosol conditions, and SIP
To assess the impact from cloud base temperature and CCN loadings on the warm-rain and ice-crystal processes, a sequence of sensitivity tests is done for both STEPS and MC3E. For each of the two cases (STEPS, MC3E), the following perturbation simulations are compared with the corresponding control run:
- cloud base (CB) is lowered to a warmer level ('Low CB simulation'),
- only CCN concentrations are reduced ('Low CCN simulation'),
- both CB height and CCN concentrations are reduced ('Low CB/CCN simulation'),
- IN concentrations are reduced ('Low IN simulation'), and
- all SIP is prohibited ('no-SIP simulation')
The low CB simulation involved lowering cloud base without altering the in-cloud ascent statistics (see the Material and Methods section). This isolates the microphysical effects from the warmer temperature of cloud base (e.g., higher adiabatic LWC aloft) in the analysis. Note that in nature, cloud-bases can be lower due to the lower troposphere being either cooler (higher relative humidity) or moister (higher absolute humidity), causing slower or faster in-cloud ascent aloft respectively. There is no simple correspondence between cloud-base height and in-cloud ascent aloft in reality.
In the low CCN and low IN simulations, the loadings of soluble and insoluble solid aerosol species respectively were altered by another height-dependent factor of 0.1 at the ground. This factor was linearly interpolated over height to unity at an altitude of 12 km MSL. This reflects the land-ocean contrast in aerosol loadings48.
All four types of SIPs were prohibited in the no-SIP simulation. More details of the sensitivity experiments are provided in Material and Method section.
Sensitivity simulations for STEPS
In the low CB sensitivity simulation, the mass and number mixing ratios of "warm graupel" are higher by almost one order of magnitude above freezing level relative to the control simulation (Fig. 7a, b). The other hydrometeor species, such as snow and ice mass, slightly increased in the low CB sensitivity simulation relative to the control simulation (Fig. 7e, f). Above enhancement in the mass mixing ratios of different hydrometeor profiles might affect both the warm and cold components of surface precipitation.
Moreover, in the low CB simulation of STEPS, warming the cloud base, from ~1C in the control run to 18C, augmented the adiabatic LWC and cloud droplet size aloft. This boosted the warm and cold components of surface precipitation by 1000% and 200%, respectively (Fig. 8a) over the entire simulation. Similar changes are found over convective and stratiform regions (Fig. 8b, c). We also use pie charts (Figure 9) to describe fractional contributions of the warm and cold components to the total surface precipitation at the end of the simulation. Figure 9a, b shows that the fractional contribution of the warm-rain process to the total surface precipitation is increased to 54% due to a strengthening of coalescence in low CB run as compared to 20% in the control simulation. This is a remarkable change in the balance between both processes of precipitation, reflecting the paramount importance of cloud base temperature for the microphysical regime of condensate generation. Such a high enhancement of the warm component of surface precipitation is due to the faster rates of collision and coalescence of larger cloud droplets aloft when the adiabatic LWC is boosted by more moisture in the PBL. The moistened PBL intensifies raindrop freezing aloft, producing more warm graupel and enhancing the warm precipitation at the surface. The lower cloud base also slightly promotes cold surface precipitation—yet less so than its strong boost of warm surface precipitation—from the ice crystal process because more supercooled cloud liquid (Fig. S6) intensifies riming of snow and graupel.
In low CCN simulation, Fig. 9b, c shows that the fractional contribution of the warm components to the total surface precipitation is slightly increased to 23% at most in-cloud levels compared to 20% in control. The absolute amounts of the warm and cold components of total surface precipitation in the low CCN simulation are moderately changed by the end of the simulation relative to the control run (Fig. 8d-f). These changes in the warm and cold components of total surface precipitation are minimal because the STEPS control simulation involves little activity of the CCN-sensitive warm-rain process as cloud droplets were too small to coalesce27. Fewer CCN can reduce the aerosol-induced invigoration of convection, reducing the supply of moisture for conversion to precipitation49. This reduction of CCN loadings can also lower the level of warm-rain during ascent with less evaporation during shorter fallout to the ground (solid red line in Fig. 8d), causing a slight increase in the warm-rain process led to total surface precipitation.
In the low CB/CCN simulation, Fig. 8g shows that the absolute contribution from the warm (cold) component to the total surface precipitation is more than 1300% (130%) higher at the end of the simulation than the corresponding control run. The total surface precipitation increased by more than a factor of 4 beyond the control run. Fig. 9b, d shows that the fractional contribution of the warm-rain process to the total surface precipitation is increased to 65% compared to 20% in the control. The fractional contribution from the ice-crystal process is reduced to 35% from 80%. The maximum of the average number concentrations of cloud droplets is reduced by one order of magnitude when compared to that of the control run (Fig. S12g). Also, the average diameter of cloud droplets was doubled (an increase by ~8-10) (Fig. S6a). These larger cloud droplets coalesce more efficiently to form rain22, which may freeze if supercooled, strongly augmenting the warm components of total surface precipitation. The strengthening of raindrop freezing in the mixed-phase region (0 to -36C) and the melting and freezing of warm graupel could also increase the warm component of surface precipitation. This is consistent with the alteration of ice multiplication by warm graupel (Fig. S12f). Above results explain why the low CB/CCN simulation produces more warm components of the total surface precipitation.
In the low IN simulation, Figure 8j shows that the warm and cold components of total surface precipitation are relatively less sensitive to the reduction of active IN concentrations than the control run. The fractional contribution of the warm components to total surface precipitation is increased to 19.9% from 19.5% than the control run (Fig. 9b, e). The slight increase in the warm components in the low IN simulation could be because the fewer active IN enhance the concentrations of supercooled cloud droplets, thereby increasing the riming and freezing of raindrops, increasing the warm graupel and warm components of surface precipitation. Ice multiplication causes the average number concentration of crystals to be almost unchanged in this sensitivity test (Fig. S13f), as also explained by Phillips et al27.
In the no-SIP simulation, the fractional contribution of the warm component to total surface precipitation is increased to 25% from 20% in the control run (Fig. 9b, f). The absolute increase in warm and cold components is, respectively, 300% and 350% (Fig. 8m). The total surface precipitation in the no-SIP simulation is increased by a factor of 2 beyond the control run (Fig. 8m). In the no-SIP simulation, the increased average size of cloud ice crystals (and decrease in ice-crystal number concentrations) intensify the overall cold precipitation and thereby total surface precipitation compared to the control run (Fig. S6b). Larger crystals grow by aggregation and vapor deposition to form snow more rapidly. Overall, the no-SIP simulation showed little effect on the average size and number concentrations of cloud droplets (Fig. S6a, c). Thus, the increase in the warm component of total surface precipitation in the no-SIP simulation is primarily associated with the strengthening of the collisional raindrop freezing from the larger ice crystals and more formation of warm graupel.
Sensitivity simulations for MC3E
In the low CB simulation, lowering cloud base by warming it by ~11K (from ~17 in the control) affects the warm components of total surface precipitation more strongly than the cold components. Figure 10a shows the absolute contribution from the warm (cold) component to the total surface precipitation is increased in low CB simulation by about 170% (35%) relative to the corresponding control run at the end of the simulation. Moreover, the fractional contribution of the warm-rain process to the total surface precipitation is increased to 42% from 24% in the control simulation (Fig. 11a, b). This is because of the lowering of cloud base involves moistening of the lower troposphere which yields a greater mass of condensate from condensation for coalescence to form precipitation. In this low CB simulation, the mass mixing ratio of rain (particularly warm component) intensifies below freezing level, yielding more warm precipitation at the surface (Fig. S15c). As stated above, the moistened PBL intensifies raindrop freezing aloft and increases the warm graupel mass. This in turn also contributes to the enhancement of the warm precipitation at the surface. Qualitatively, the manner of this response to lowering cloud base is similar to that noted above for STEPS case.
In the low CCN simulation, Figure 11b, c shows that the fractional contribution of the warm component to the total surface precipitation is increased to 29% at most in-cloud levels compared to 24% in control run. By the end of the simulation, the absolute contributions from the warm and cold components show an increase by ~10% and a decrease by ~10% relative to the control run respectively, causing little change in the total surface precipitation due to this compensation (Fig. 10d). In this low CCN simulation, the increase in cloud droplet size by more than 30% throughout the vertical column relative to the control run causes an enhancement in the warm component of total surface precipitation (Fig. S8a). However, the mean size of ice crystals is slightly reduced in the mixed-phase region (Fig. S8b), causing the reduction of the cold contributions to the total surface precipitation.
In the low CB/CCN simulation, there is almost the strongest response out of all the sensitivity tests presented here because now the lowering of both cloud base and CCN concentrations act in concert in the same direction, enhancing coalescence. The fractional contribution of the warm-rain process to the total surface precipitation is increased to 49%, due to a strengthening of coalescence at most in-cloud levels, compared to 24% in control (Fig. 11b, d). The fractional contribution from the ice-crystal process is reduced to 51% from 76%. The absolute increase in warm (cold) components of total surface precipitation is more than 250% (58%) at the end of the simulation relative to the control run. The total surface precipitation increased by nearly a factor of 2 beyond the control run. Interestingly, over convective regions, the fractional contribution of warm components to the surface precipitation is increased to 82% from 41% (Fig. 12b, d). This increase in the low CB/CCN simulation is noted to be greatest relative to the control run. The average diameter of cloud droplets was increased by ~33% in the lower to middle troposphere relative to the control run (Fig. S8a). As compared to the control run, the higher growth of average diameter of cloud droplets and reduced adiabatic LWC (by almost 66%) can enhance the coalescence efficiency and hence the warm components of total surface precipitation. The partial increase in the cold components of total surface precipitation relative to control run is due to enhanced growth of ice crystal diameter and reduced ice water content (IWC) below -30 (Fig. S8d).
In the low IN simulation, Figure 10j-l shows that the warm and cold components of total surface precipitation are relatively less sensitive to the lowering of ice nuclei than the control run. This lack of response is similar to that found with the STEPS case noted above.
In the no-SIP simulation, at the end of the simulation, the absolute increase in warm components is 130% more than the control run (Fig. 10m), whereas the absolute decrease in cold components is 40%. The fractional contribution of the warm components to total surface precipitation arises to 52% from 24% than the control run (Fig. 11b, f). The no-SIP simulation showed a relatively stronger effect on the cloud droplet size and cloud droplets number concentrations than the control run, which contrasts with the STEPS case. The increase in the warm components of total surface precipitation in the no-SIP simulation is associated with the strengthening of both the ice crystal size and cloud droplet sizes above -25 (Fig. S8a, b). The increase ice cloud fraction by 100% relative to the control show that the increase in ice-cloud lifetime and the lesser cold precipitation (Fig. S10b).
In summary, among all the MC3E sensitivity tests, the greater increase in the fractional contribution from warm components to the total precipitation is mainly associated with the low CB and no-SIP simulations relative to the control run. This indicates the strong sensitivity of warm components of liquid and ice precipitation with respect to the cloud base temperature and multiple mechanisms of ice multiplication.