Spatial Organisation Affects the Pathway to Precipitation in Simulated Trade‐Wind Convection

We investigate whether and how spatial organization affects the pathway to precipitation in large‐domain hectometer simulations of the North Atlantic trades. We decompose the development of surface precipitation (P) in warm shallow trade cumulus into a formation phase, where cloud condensate is converted into rain, and a sedimentation phase, where rain falls toward the ground while some of it evaporates. With strengthened organization, rain forms in weaker updrafts from smaller cloud droplets so that cloud condensate is less efficiently converted into rain. At the same time, organization creates a locally moister environment and modulates the microphysical conversion processes that determine the raindrops' size. This reduces evaporation and more of the formed rain reaches the ground. Organization thus affects how the two phases contribute to P, but only weakly affects the total precipitation efficiency. We conclude that the pathway to precipitation differs with organization and suggest that organization buffers rain development.

• The development of surface precipitation in simulated trade-wind convection is decomposed into a formation and sedimentation phase • As organization strengthens, less cloud condensate is converted into rain, but more rain reaches the ground as evaporation is suppressed • Organization affects rain formation by modulating the local moisture environment, cloud vertical motion and microphysical properties Supporting Information: Supporting Information may be found in the online version of this article.
differences.Organization may affect how efficiently rain forms and how much evaporates through modulating mesoscale circulations or the local moisture environment (Narenpitak et al., 2021;Seifert & Heus, 2013).Moreover, understanding the relationship between spatial organization and precipitation may also be key to disentangle the mechanisms of organization and explain its influence on the total cloud cover in the trades, a prerequisite to further constrain the climate feedback of the trades (Bony et al., 2020;Nuijens & Siebesma, 2019).
Analyzing rain radar measurements upstream of Barbados taken during the EUREC 4 A field campaign (Hagen et al., 2021;Stevens et al., 2021), Radtke et al. (2022) show that while the occurrence of trade-wind precipitation is related to organized cells, the mean rain rate varies largely independently of the cells' degree of organization.However, scenes with similar precipitation but different degrees of spatial organization also differed in the moisture environment.Similarly, Yamaguchi et al. (2019) find that in idealized LES, shallow cumulus precipitation varies little, but the sizes and spatial distribution of clouds differ in response to large changes in the aerosol environment.Could spatial organization be a mechanism to maintain precipitation in different environments, enabling or creating a different pathway to precipitation?
To answer this question on a process-level, we make use of large-domain hm-scale simulations of the North Atlantic trades that were run for the period January-February 2020 during the EUREC 4 A campaign (Bony et al., 2017;Schulz & Stevens, 2023;Stevens et al., 2021) and were designed to explore spatial organization on the mesoscale (20-200 km).We follow the method of Langhans et al. (2015) and Lutsko and Cronin (2018) and decompose the development of surface precipitation into two phases, (a) a formation phase, in which cloud condensate is converted into rain water, and (b) a sedimentation phase, in which the formed rain water falls toward the ground while part of it evaporates.Section 2 describes the setup and microphysical scheme of the simulations and our analysis method.Section 3.1 shows that spatial organization in scenes of  (100 km) influences how these two phases contribute to the development of trade-wind precipitation.Section 3.2 explains such behavior and interprets it as a form of buffering, before we conclude in Section 4.

EUREC 4 A Large-Domain ICON Hm-Scale Simulations
The simulations are conducted with the LES configuration of the ICOsahedral Non-hydrostatic (ICON) model (Dipankar et al., 2015).ICON solves the compressible Navier-Stokes equations on an unstructured grid as detailed in Zängl et al. (2015) and Dipankar et al. (2015).To explore mesoscale convective variability, the simulations are performed with a relatively fine hectometer-scale mesh over a large domain of   (1,000 km) for an extended EUREC 4 A campaign period from 9 January to 19 February 2022.They are realistically forced with initial and boundary data from a storm resolving simulation at 1.25-km grid spacing, which is in turn initialized and nudged at its lateral boundaries to the atmospheric analysis of the European Centre for Medium-Range Weather Forecasts (similar to Klocke et al., 2017).Here, we analyze a simulation with 625 m grid spacing that covers the western tropical Atlantic from 60.25 to 45.0°W and 7.5-17.0°N,spanning about 1,650 km in the eastwest direction, and 1,050 km in the north-south direction.In the vertical, 150 levels are used, resulting in 70 and 85 m vertical resolution at 1,000 and 2,000 m height, respectively.Schulz and Stevens (2023) show that this simulation reproduces differences in the mesoscale structure underlying the canonical forms of trade cumulus organization of Stevens et al. (2020), as well as variability in precipitation, which makes them a good starting point to investigate how the process of precipitation may vary with spatial organization.A nested 312 m simulation does not show a substantially greater skill in representing different cloud organizations or rain rates (Schulz & Stevens, 2023) and shows the same qualitative behavior in our analysis (not shown).We refer the reader to Schulz and Stevens (2023) for an in-depth observational evaluation and comparison of the simulations.In Supporting Information S1 we summarize and discuss the evaluation of the simulation properties relevant for our work.
In the simulations, turbulence is parameterized with the Smagorinsky scheme, microphysics with the two-moment mixed-phase bulk microphysics scheme of Seifert and Beheng (2006) and a cloud condensation nuclei concentration of 130 cm −3 is prescribed.Warm rain is produced by autoconversion and accretion, defined following Seifert and Beheng (2001) as of rain, we recalculate the autoconversion and accretion rates from the instantaneous 3D model output of cloud water, rain water and cloud effective radius r eff , from which the volume radius r v is derived by  (v∕eff) 3 = 0.8 (Freud & Rosenfeld, 2012) to calculate the cloud droplets' mean mass.The 3D output is available every 3 hr.

Investigating Spatial Organisation and the Pathway to Precipitation
We investigate spatial organization in scenes of 4 × 4° (about 450 × 450 km), an area extent similar to previous studies (George et al., 2023;Radtke et al., 2022).Figure 1a shows three scenes with different degrees of organization as an example.To mask high ice clouds, scenes with outgoing longwave radiation <275 W m −2 are excluded (changes in the threshold do not affect our qualitative results), as well as scenes with little precipitation P < 0.01 mm hr −1 .Here and if not indicated otherwise, we refer to domain mean values.In total about 2,000 scenes are used in the analyses (about 7 scenes across the domain every 3 hr).
Following Bretherton and Blossey (2017) and Narenpitak et al. (2021), we assess the degree of spatial organization as mesoscale variability in the moisture field, which is closely connected to the cloud structure.First, the total water path, W T , defined as the sum of vertically integrated water vapor, cloud condensate, and rain, is coarse-grained into tile sizes of 20 × 20 km 2 , representing variability associated with the mesoscale (Orlanski, 1975).Subsequently, the coarse-grained W T is binned into quartiles and the difference between the fourth and first quartile is calculated as organization metric  Δ T .This metric classifies the three example scenes from weakly organized (low  Δ T ) on the left, to more strongly organized (high  Δ T ) on the right.This is , with W R as rain water path, shown for the three precipitation regimes, separated into a clustered (  Δ T > 70 th percentile, filled bars) and scattered sample (  Δ T < 30 th percentile, empty bars).The green line denotes the median, the dot the mean, the box the interquartile range and the whiskers denote the 5th and 95th percentile.consistent with a visual subjective classification of the cloud field, the nearest neighbor clustering technique I ORG (Tompkins & Semie, 2017), and the cloud pattern classification of Stevens et al. (2020).According to this classification, the left scene depicts a Gravel pattern, characterized by scattered convection (low I ORG ), and the right scene a Fish pattern, characterized by more clustered convection (higher I ORG ).An overview and more in-depth discussion of different organization metrics has recently been given by, for example, Janssens et al. (2021).
To investigate the pathway to trade-wind precipitation, we decompose the development of surface rain following Langhans et al. (2015) and Lutsko and Cronin (2018) into (a) a formation phase and (b) a sedimentation phase.In phase (a), warm rain initially forms by the merging of small cloud droplets, parameterized by the autoconversion rate.Additionally, rain is produced as falling raindrops collect cloud droplets, parameterized by the accretion rate.Autoconversion dominates the formation of rain especially for young or short-lived clouds, while accretion contributes more to the formation of rain as clouds live longer and there is more time available for the collision-coalesence process to take place (Feingold et al., 2013).To quantify how efficient the formation of rain water is, we define a conversion efficiency where C R = C Auto + C Acc with C Auto and C Acc denoting the vertically integrated autoconversion and accretion rates and W L the cloud liquid water path.In phase (ii), the rain formed by autoconversion and accretion sediments toward the ground.During this process, some rain evaporates.The rain that does not evaporate reaches the ground as surface precipitation, P, so that we call the sedimentation efficiency with ϵ evap the evaporation efficiency.Please note that we do not refer to in-cloud sedimentation here, but, following Langhans et al. (2015) and Lutsko and Cronin (2018), we use sedimentation efficiency to refer to how much rain reaches the ground instead of evaporating.
The product of the conversion and sedimentation efficiencies describes how much cloud water in a given time interval is returned to the surface as precipitation, representing an overall precipitation efficiency ϵ P , for example, as used in Lau and Wu (2003): Said differently, the inverse of ϵ P is the time it takes to remove all cloud water at the given precipitation rate, thus describing a typical residence time.It is to note that precipitation efficiency itself has no unique definition (e.g., Sui et al., 2020).Different results may emerge for different definitions and also depend on local versus domainmean views.However, using an approximation of the condensation rate following Muller and Takayabu (2020) instead of liquid water path in Equation 3 results in the same qualitative behavior.Here, we mainly exploit precipitation efficiency and its decomposition into conversion and sedimentation efficiencies as a proxy for the pathway that precipitation development takes.

The Pathway to Precipitation Varies With Organisation
The hm-scale simulations reproduce EUREC 4 A observations in that scene precipitation in the trades varies mainly independently of organization (Radtke et al., 2022).This is depicted in the example scenes in Figure 1a, which display a similar rain rate but vastly different degrees of organization, and is more quantitatively shown in Figure 1b.In the simulations, scene rain rates vary up to 0.2 mm hr −1 as shown in Figure 1b, which compares well to rain rates observed in the RICO (Nuijens et al., 2009) and EUREC 4 A campaign (Radtke et al., 2022).In the following, we will show whether also the pathway to these rain rates is similar or in how far organization affects how these rain rates are generated, and could thus be a process to maintain precipitation in different environments.
To investigate this, we group our sample of scenes into three precipitation regimes, a weak, a moderate, and a strong precipitation regime, as visualized in Figure 1b.In each regime, we define more organized scenes as 10.1029/2023GL103579 5 of 11  Δ T > 70th percentile and less organized scenes as  Δ T < 30th percentile, which we refer to as clustered and scattered sample, respectively.The conclusions are insensitive to the exact choice of threshold.This relative way of differentiating between more and less organized scenes is based on the simulated distribution of organization, which will vary depending on how variable the large-scale conditions are and possibly which specific model is used.
Figure 1c shows that, instead of the mean rain rate, organization tends to increase the rain intensity, which is again in line with observational studies of trade-wind (Radtke et al., 2022) and deep convection (Louf et al., 2019).That is, clustered convection produces the same amount of scene precipitation as scattered convection with more intense rain covering a smaller area.Detecting cold pools based on the calculation and criterion of a mixed layer height smaller than 400 m following Touzé-Peiffer et al. ( 2022), clustered scenes are also populated by more cold pools as shown in Figure 1d, possibly associated with this increase in rain intensity.In clustered scenes, the cold-pool fraction is about four times greater than the rain fraction, whereas in scattered scenes it is about three times greater.These findings may already hint to an altered precipitation process in more organized compared to less organized scenes.
We investigate the relationship between organization and the conversion, sedimentation and total precipitation efficiency (Equation 3), shown in Figure 2a.Organization maximizes toward the lower right of the phase space, at low conversion and high sedimentation efficiencies.An increase in the degree of organization is thus related to a decrease in how efficiently cloud water is converted into rain and an increase in how efficiently rain sediments as a greater fraction of rain reaches the ground instead of evaporating.The sedimentation efficiency varies between 0.1 and 0.3, emphasizing that much of the rain evaporates, as reported by Naumann and Seifert (2016) or Sarkar et al. (2022).Figure 2b shows that precipitation maximizes toward the upper right of the same phase space, that is, at high sedimentation and conversion efficiencies.Within a precipitation regime, as shown in Figures 1e and 1f, rain thus sediments more efficiently but forms less efficiently in clustered compared to scattered scenes.This behavior is slightly enhanced in regimes with stronger precipitation.
The product of the conversion and sedimentation efficiencies gives the overall precipitation efficiency, denoted in the contour lines in Figure 2. Precipitation efficiency varies closely with precipitation and lies mostly between 1 hr −1 and 3 hr −1 .That one to three times the cloud liquid water path precipitates per hour emphasizes the rapid turnover and rain formation in trade-wind clouds, which with tops greater than 2,500 m "usually rain within half an hour" (Squires, 1958).Because conversion efficiency decreases but sedimentation efficiency increases with organization, contours of precipitation efficiency and organization lie perpendicular to each other in Figure 2a.This means that organization and precipitation efficiency, like precipitation, vary mainly independently of each other.Composited on the three different precipitation regimes, Figure 1g shows that precipitation efficiency compared to the conversion and sedimentation efficiency varies only weakly with organization with a slight tendency to increase with organization.Analyzing the ratio of rain water path to cloud water path instead of the ratio between precipitation and cloud liquid water path gives the same result (Figure 1h).
Our analysis thus suggests that organization weakly affects precipitation efficiency in terms of how much cloud water precipitates on average, but changes the pathway to precipitation in terms of how the formation versus sedimentation phases contribute to the development of surface precipitation.Next, we investigate physical mechanisms behind this behavior.

Sedimentation Efficiency
The sedimentation efficiency describes how much rain reaches the ground instead of evaporating.Following Lutsko and Cronin (2018), we suggest that ϵ sed should scale to a first approximation with the moisture environment through which the rain falls, or more explicitly with the saturation deficit, and the time it takes the rain to fall: where  rain is the averaged relative humidity the falling rain experiences, that is, conditioned on pixels with rain water q r > 0.001 g kg −1 (van Zanten et al., 2011), and t fall the average fall time, which depends on the average fall height h fall and fall velocity v fall .The higher the saturation deficit or the longer the rain falls and thus has time to evaporate, the higher evaporation and the lower the amount of rain reaching the ground.
We hypothesize that organization influences the moisture environment through which rain falls, since it manifests itself in an uneven (horizontal) distribution of moisture, as also used in our metric of organization.Figure 2c shows that in the simulations, rain in clustered scenes indeed typically falls through a more humid environment with a lower saturation deficit than in scattered scenes.This is true for all precipitation regimes, with little variations in  rain between precipitation regimes.We find that rain falls through a moister environment because the environment outside of or beneath clouds is closer to saturation (about 3%, Figure 2d), not just because more rain may fall within than outside of clouds, for example, due to different wind shears and cloud tilts.This is in line with the idea that clouds in more organized scenes develop preferentially in the parts of the domain with moister, more favorable thermodynamic conditions, for example, preconditioned by former clouds (sometimes called mutual-protection hypothesis, Seifert & Heus, 2013) or established through enhanced moisture transport into anomalously moist patches by mesoscale circulations (George et al., 2023;Narenpitak et al., 2021).That way, clusters may form, clouds may be better protected from updraft buoyancy reduction through entrainment (Becker et al., 2018;Mapes & Neale, 2011), and less rain evaporates.
Besides the moisture environment, organization could also influence the fall time of the rain drops by modulating the fall height or fall velocity (Equation 4).We define the fall height h fall as average height where rain is produced by autoconversion and accretion.Analyzing h fall shows that rain in clustered convection falls on average from slightly higher heights than in scattered convection (Figure 2e), related to a tendency of clouds to grow deeper and inversion heights to increase with organization (not shown).If the fall velocity stays unchanged, this would suggest that organization slightly increases the time it takes for rain to fall to the ground, which would act to enhance, not to reduce evaporation in more strongly organized scenes.
For the mean fall velocity, multiple factors, for example, the strength of up-and downdrafts and the raindrops' size are decisive.The raindrop size was not included in the model output but the way rain is produced, that is, in how far autoconversion versus accretion dominates the formation of rain, may serve as a proxy for the raindrops' size.Figure 3a shows that in how far autoconversion versus accretion contributes to rain formation explains 79% of the variations in sedimentation efficiency.Because autoconversion produces initial "embryo" raindrops when cloud droplets merge, whereas accretion is responsible for the growth of raindrops through further collection of cloud droplets, an increased contribution of accretion to rain formation indicates that raindrops have grown larger.Figure 2f shows that in more organized scenes the contribution of accretion to rain formation is increased.Raindrops in more organized scenes are thus likely larger.Larger rain drops fall faster, reducing the fall time and hence evaporation.
When including, in addition to the relative importance of autoconversion and accretion,  rain as additional predictor, 85% of the variations in sedimentation efficiency can be explained.Additionally including h fall does not explain further variations.Our analysis thus suggests, as illustrated in Figure 4, that organization reduces evaporation and increases the sedimentation efficiency because rain in more organized scenes is increasingly produced by accretion so that raindrops are larger and fall faster, through an environment that is moister.

Rain Formation Efficiency
Rain starts to form when sufficient cloud water has been produced and cloud droplets have grown to raindrop size (e.g., Seifert & Stevens, 2010).To initiate and grow cloud particles the air's saturation is important and influenced by thermodynamic conditions as well as vertical motions (Rogers & Yau, 1996).
In the simulations, organization influences the clouds' vertical motion.
Figure 2g shows that in clustered scenes the mean in-cloud vertical motion near cloud base,  cld 900 (cloud-conditioned, i.e., where cloud water q c > 0.01 g kg −1 , and at 900 hPa), is weaker than in more scattered scenes.This initially appears surprising.It can be attributed, in part, to the presence of stronger downdrafts, for example, the 25th percentile of  cld 900 is lower, and in part to weaker updrafts as the mean and median cloud upward motion is reduced (not shown).Bao and Windmiller (2021) found a similar decrease in vertical motions with organization in deep convection.Because 4. Conceptual illustration of the simulated pathway to surface precipitation in weakly (left) versus strongly (right) organized convection: For similar surface precipitation, as organization strengthens, rain forms in a locally moister environment (shading) in weaker updrafts (arrow) and increasingly from accretion indicating larger raindrops (dots).As a consequence, evaporation is reduced, so that more rain reaches the ground compared to scattered convection (ϵ sed is larger), but rain forms less efficiently (ϵ conv is smaller), with both changes having a compensating, that is, buffering effect on surface rain development.
organization creates more favorable thermodynamic conditions for cloud and rain formation with a local increase in humidity (Figures 2c and 2d), this may allow clouds and rain to develop in less favorable dynamic conditions, that is, at weaker mean upward motions.Additionally, the cloud population in more organized scenes could also consist of more long-lived, older clouds as indicated by the increased contribution of accretion to rain formation.In these, updrafts may have already started to weaken.Future analysis of the cloud lifecycle with respect to organization may therefore potentially reconcile the apparent contradiction between the expected larger raindrops resulting from increased accretion and the presence of weaker updrafts.
More organized scenes also differ from less organized scenes in the mean cloud droplet radius.Figure 2h shows that in clustered scenes, the mean cloud droplet radius is smaller by about 1.3 μm than in scattered scenes.From moderate to high precipitation, this difference increases, which is in line with the strong decrease in conversion efficiency at high precipitation.The smaller cloud droplet size in more organized scenes agrees with the weaker vertical motions.Besides, Cooper et al. (2013) showed that mixing and entrainment affect cloud droplet growth and the onset of precipitation.By changing the mixing characteristics of clouds, organization might also influence the cloud droplets' size.
Figure 3b shows that 70% of the variations in conversion efficiency are explained by the mean vertical motion at cloud base, to which the mean cloud droplet size is correlated.To conclude and as illustrated in Figure 4, our analyses suggest that organization reduces the efficiency with which cloud water is converted into rain water because rain in clustered scenes forms in weaker updrafts that correlate with smaller mean cloud droplets.We hypothesize that this is because favorable thermodynamic conditions may compensate for weaker dynamic conditions and organization may affect the lifetime as well as the mixing characteristics of the cloud population.

Buffering
Organization is associated with an increase in the sedimentation efficiency, but a decrease in the formation efficiency and thus influences rain development in an opposing or stabilizing way.This may be interpreted as a form of buffering.Buffering as introduced in Stevens and Feingold (2009) denotes that if there are different pathways to reach the same final state, these buffer or stabilize the system against disruptions to any particular pathway.For example, cloud deepening as a dynamical response to increased droplet number concentration has been shown to buffer the microphysical suppression of precipitation (Seifert et al., 2015).Our analyses suggest that organization can have a stabilizing, that is, buffering, relationship to rain development, due to opposing effects on rain formation and sedimentation efficiency, as illustrated in Figure 4.While in more scattered convection, rain development is characterized by efficient conversion of cloud water into rain water but also subsequent evaporation is strong, in more clustered convection, increased sedimentation efficiency compensates for a decreased conversion efficiency.These variations on the pathway to precipitation by organization may be an additional explanation for why rain development is so common in the trades.They also offer an explanation for the observed (Radtke et al., 2022) and simulated large independence between rain amount and organization rather than a reinforcement.
Based on our results, we hypothesize that such buffering effect of organization on rain development is related to (a) an interplay of thermodynamic and dynamic conditions and (b) (life)time effects.Regarding (a), more favorable thermodynamic conditions may allow clouds to develop under less favorable dynamic conditions.Regarding (b), the increased contribution of accretion to rain formation with organization indicates longer-lived, older clouds, in agreement with the expectation of more sustained convection with organization.In older, more mature clouds, the rain formation process had time to evolve (Feingold et al., 2013), so that raindrops may grow larger and fall out more efficiently, while updrafts may have already weakened so that further rain forms less efficiently.

Summary and Conclusions
We exploit realistic large-domain hm-scale simulations of the North Atlantic trades to investigate whether and how organization affects the pathway to trade-cumulus precipitation.We decompose the development of surface precipitation following Langhans et al. (2015) into a formation phase, where cloud condensate is converted to rain, and a sedimentation phase, where the formed rain falls to the ground while some of it evaporates.In the simulations, organization affects how these two phases contribute to rain development, summarized schematically in Figure 4.
With strengthened organization, rain in the hm-scale simulations forms in and falls through a locally more humid environment.Additionally, rain is increasingly produced by accretion rather than autoconversion, which indicates that clouds live longer and raindrops grow larger.Larger raindrops, that fall through a more humid environment experience less evaporation, leading to an increase in the sedimentation efficiency.The relative importance of accretion and autoconversion explains 79% of the variations in sedimentation efficiency, increasing to 85% when including the rain-conditioned relative humidity as an additional predictor.A locally more humid environment is in line with the idea that an increase in organization is related to more humid patches in which clouds develop and which protect clouds from dilution and raindrops from evaporation.It may suggest that organization also increases the efficiency with which cloud condensate is converted to rain.However, in more organized scenes rain forms in weaker updrafts (as in Bao and Windmiller, 2021), and from smaller cloud droplets.This leads to cloud water being less efficiently converted to rain, in agreement with radiative-convective equilibrium simulations by Lutsko and Cronin (2018).71% of the variations in conversion efficiency are explained by the in-cloud vertical motion at cloud base, to which the cloud droplet size is correlated.Possibly because the thermodynamic environment is more favorable with organization, less favorable dynamic conditions already allow for rain formation, or lifetime effects may play a role here.Both effects, the increase in sedimentation efficiency and the decrease in conversion efficiency, largely compensate, so that organization does not substantially affect the total precipitation efficiency.
Our analyses suggest that organization can buffer rain development via opposing effects on the rain formation and sedimentation efficiencies.They offer an explanation for the observed and simulated large independence between rain amount and organization.While in less organized scenes rain development is characterized by efficient conversion of cloud condensate into rain, in more organized scenes more efficient sedimentation, as evaporation is suppressed, increasingly contributes to surface rain development.It remains to be shown in how far these results carry over to other models and observations.In our simulations, we conclude that the pathway to precipitation differs with spatial organization.
| ∼ cr , where L r is rain water content, L c cloud water content and  c =  c  c mean mass of cloud droplets with cloud droplet number concentration N c .To quantify the production 10

Figure 1 .
Figure 1.(a) Three example scenes with similar scene-averaged precipitation P (i.e., rain amount, blue) but different degrees of organization characterized by  Δ T (orange) and the I ORG (green).Color shading denotes rain rate R. Gray shading denotes cloud albedo calculated from simulated cloud liquid water path.(b) P as a function of  Δ T .Three different rain regimes with weak P = (0.024, 0.037), mod P = (0.042, 0.064) and high P = (0.07, 0.12) are distinguished.(c) Rain intensity I, (d) cold pool fraction F C per rain fraction F R , (e) sedimentation efficiency ϵ sed , (f) conversion efficiency ϵ conv , (g) precipitation efficiency ϵ P , and (h) rain water loading efficiency  W R =  R  L

Figure 2 .
Figure 2. (a) Degree of mesoscale organization  Δ T and (b) precipitation P (shading) as a function of conversion efficiency ϵ conv and sedimentation efficiency ϵ sed .Contour lines denote precipitation efficiency from Equation3.Relative frequency of (c) rain-conditioned relative humidity  rain , (d) rain-and-no-cloud-conditioned relative humidity  rnc , (e) fall height h fall , (f) ratio of autoconversion C Auto to accretion C Acc , (g) cloud-conditioned vertical velocity at 900 hPa  cld 900 and (h) mean cloud droplet radius r qc for the scenes separated into scattered (dashed line, empty bars) and clustered (solid line, filled bars) convection and divided into three precipitation regimes (as in Figure1, hP denoting the high P regime, mP the mod P regime and lP the low P regime).Horizontal boxes denote the interquartile range, vertical lines the median.

Figure 3 .
Figure 3. (a) Sedimentation efficiency ϵ sed as a function of the relative importance of autoconversion C Auto and accretion C Acc .Shading denotes the rain-conditioned relative humidity  rain (b) Conversion efficiency ϵ conv as a function of cloud-conditioned vertical velocity at 900 hPa  cld 900 .Shading denotes the mean cloud droplet radius r qc .