Global Evidence of Aerosol-induced Invigoration in Marine Cumulus Cloud

Aerosol-cloud-precipitation interactions can lead to a myriad of responses within shallow cumulus clouds including an invigoration response, whereby aerosol loading results in a higher rain rate, more turbulence, and deepening of the cloud layer. However few global studies have found direct evidence that invigoration occurs. The few satellite based studies that report evidence for such effects generally focus on only the deepening response. Here, we show evidence of invigoration beyond a deepening response. Using latent heating and vertical motion profiles derived from CloudSat radar observations, we show 5 precipitating cumulus clouds in unstable, polluted environments exhibit a marked increase in precipitation formation rates and cloud top entrainment rates. However, invigoration is only discernible when the stability of the boundary layer is explicitly accounted for in the analysis. Without this environmental constraint, the mean polluted and pristine cloud responses are indiscernible from each other due to offsetting cloud responses in stable and unstable environments. Invigoration, or suppression depending on the environment, may induce possible feedbacks in both stable and unstable conditions that could subdue or 10 enhance these effects, respectively. The strength of the invigoration response is found to additionally depend on cloud organization defined here by the size of the warm rain system. These results suggest that warm cloud parameterizations must account for not only the possibility of aerosol-induced cloud invigoration, but also the dependence of this invigorated state on the environment and the organization of the rain system.

Environmental information is provided by MERRA-2 reanalysis. We define the stability of the atmosphere using the estimated inversion strength (EIS) (Wood and Bretherton, 2006). Stability of the boundary layer controls the depth of the cloud making it imperative that this relationship is constrained in order to separate aerosol effects from environmental forcings (Zuidema et al., 2009). Unstable environments are defined as having an EIS below 1 K while stable environments are defined 95 as having an EIS above 3 K. This partitions environments into two main regimes: trade cumuli (unstable) and cumuli from stratocumulus to cumulus transitions (stable). A dry free atmosphere alters the distribution of liquid throughout the cloud layer, thereby directly impacting precipitation formation processes as well. In order to control for these interactions, clouds are further subset into a dry regime whereby the RH 700 is below 30% to analyze how dry air entrainment may impact invigoration processes.

Latent heating profiles
The Wisconsin Algorithm for Latent heating and Rainfall Using Satellites (WALRUS) provides information on the latent heating and vertical motion profiles in the atmosphere. The algorithm combines CloudSat's CPR observations with a database of warm rain states derived from the Regional Atmospheric Modeling System (RAMS) simulations to emulate realistic latent heating rates and related vertical motion (Nelson et al., 2016). WALRUS limits our analysis to maritime clouds with heights 105 less than the freezing level and only those that exhibit reflectivity greater than 0 dBZ somewhere in the column, consistent with the Rain Certain flag in CloudSat's 2C-PRECIP-COLUMN product ). Our results do not include the effects of drizzle on possible invigoration processes. This should also focus our results on only the growing and mature stages of shallow convection. Signals of invigoration are derived based on changes in the latent heating within the cloud. Precipitation formation rates correspond to the latent heat release within the cloud, while evaporation due to entrainment at the cloud top 110 or vigra below the cloud are indicated by cooling from WALRUS. Enhanced turbulence, or the change in vertical velocity, is determined by the difference in vertical velocity between polluted and pristine environments.
WALRUS employs a Bayesian Monte Carlo method in order to derive probabilistic latent heating profile. While precipitation amounts alone can be used to infer total latent heating in the column, vertically-resolved reflectivity profiles allow the inference of the distribution of latent heating throughout the profile, below, within, and above the cloud. The Bayesian Monte Carlo 115 method relies on an a priori distribution of possible characteristics to connect to the CloudSat observations. The a priori database is created using the RAMS model with simulations based on the Atlantic Trade Wind Experiment field campaign.
The model is run at a 250 m horizontal and 100 m vertical resolution for a set of sea surface temperatures (293 K, 298 K, and 303 K). Quick Beam produces radar reflectivity profiles and attenuation signals from the RAMS simulation, which are sampled every 40 minutes for the database. Overall, WALRUS had 1.4 million possible a priori warm rain structures against 120 which observed CloudSat reflectivities are compared to retrieved the most physically realistic latent heating and associated vertical motion rates. For more information please refer to Nelson et al. (2016).

Partitioning clouds
Cloud profiles are partitioned according to the individual cloud base and cloud top heights determined for each profile using CloudSat's 2B-CLDCLASS-LIDAR CloudLayerBase and CloudLayerTop products. These heights are used to distinguish the 125 in-cloud region from the environment below or above it. The maximum above cloud cooling due to evaporation is found by taking the maximum of all the cooling rates (which include cooling by evaporation) starting at the cloud top to the top of the profile. The cloud top is obvious in the latent heating profiles (Figure 1); the abrupt shift from heating to cooling indicates the entrainment zone of the cloud near the cloud top. The mean below cloud cooling rate is similarly found using the cloud base height from 2B-CLDLCASS-LIDAR and taking the sum of all evaporative cooling rates at the cloud base to the bottom 130 of the profiles (approximately ground level). The geometrical center of the cloud is the midpoint of the cloud (e.g. for a 7 km cloud as seen in Figure 1, the midpoint is 3.5 km) therefore the profiles on either side are used to determine the behavior of the geometrical center of the cloud.

Results and Discussion
3.1 Aerosol effects on warm rain formation rates 135 Theoretical arguments for warm rain invigoration predict that in a more polluted environment, the rate of collision coalescence and therefore precipitation production increases. Our analysis suggests that, on average, clouds in polluted environments do not show an increased rates of precipitation formation relative to those in pristine environments (black solid line, Figure 2). The  difference between polluted (solid) and pristine (dashed) conditions is minimal when clouds in environments are considered together. However, when separated according to the environmental stability, it is evident that the reason for this is not that the 140 warm rate intensity is unaffected by aerosol loading, it is that clouds react differently under stable and unstable conditions.
In unstable environments, polluted conditions lead to a marked increase in precipitation rate relative to unstable, pristine conditions (blue, dotted line) for all rain systems smaller than ∼6 km. Conversely, stable, polluted conditions (red, solid line) lead to a decrease in precipitation rate relative to stable, pristine conditions (red, solid line). The opposite reactions in stable vs.
unstable conditions offset each other, giving the impression that warm rain is unaffected by aerosol loading when in actuality 145 its sensitivity is environmentally dependent. Invigoration is only identifiable when stability is accounted for and this suggests that aerosol-induced invigoration of shallow convection may exhibit marked spatial patterns globally.
Polluted clouds exhibit a non-linear relationship between the size of the rain system and the maximum rain formation rate, pristine clouds show a steady, linear increase in the rain formation rate as the size of the rain system increases (Figure 2). Rain formation in polluted clouds, on the other hand, exhibits an inflection point first increasing gradually with size up to 5 km and 150 then after that size they show much greater rates of rain formation. This inflection point depends on both the stability of the boundary layer and the humidity of the free atmosphere. Pristine conditions do not show this acceleration.
The core of a warm convective system should theoretically exhibit the greatest invigoration of precipitation. Our results indicate this conceptual model is correct: as invigoration of the warm rain formation rate due to aerosol is most pronounced in the geometrical center of the rain system ( Figure 3). Mean precipitation rates increases in the center of unstable, polluted 155 clouds relative to both cleaner and more stable conditions. This effect is exacerbated in dry conditions (Figure 3, bottom) until the rain system seems to hit a size inflection point around 7 km. While instability in polluted clouds leads to greater formation rates in the center, clouds in stable but equally polluted environments show a decrease in rain production relative to pristine conditions. This supressive behavior is observed regardless of the overlying free atmosphere, as clouds in dry environments

Aerosol effects on evaporative processes
However, that is not to say that the free atmosphere does not play a role in altering the thermodynamics or possible invigorate state of warm rain systems. Evaporative processes link entrainment, below cloud evaporation, precipitation formation, and the energy budget of a cloud. When focusing on how aerosol may affect entrainment, the moisture content of the free atmosphere becomes a controlling factor. A drier atmosphere fosters greater evaporation rates above the cloud in more polluted environ-165 ments ( Figure 4). While increased mixing with the free atmosphere may lead to cloud deepening, it may also lead to an early onset of cloud breakup processes through evaporation-entrainment (Small et al., 2009). In some cases, increased entrainment and evaporation at the cloud top could lead to reduced cloud top heights, opposite of an invigoration effect . Whether the growth of a particular cloud is enhanced or inhibited may depend on the distribution of liquid water near the cloud top and the ability of the cloud to penetrate the free atmosphere.

170
A drier atmosphere enhances cloud top evaporation in only unstable conditions; clouds in stable conditions are unaffected by a drier free atmosphere. This is likely due to the stronger capping inversion in stable conditions which limits mixing with