1. Introduction
It is widely understood that the largest uncertainty in estimates of climate sensitivity stems from the uncertain response of clouds to global warming. In particular, the feedback between low clouds and warming is a major source of uncertainty, and it is representing the full physical processes that determine the properties of these clouds, which continues to represent a challenge in global change research (Flato et al. 2013). The accurate representation of warm cloud microphysics, in particular, is considered essential if we are to reduce uncertainties in climate change projections.
A process central in defining the life cycle of low clouds, in establishing microphysical properties that affect radiation (Wood 2012) and in influencing their mesoscale structure (Jensen et al. 2000), is warm rain formation. Previous studies (e.g., Golaz et al. 2013; Suzuki et al. 2013) showed how the historical record of global-mean surface temperature change simulated by models do so with unrealistic representation of warm rain and that more realistic representations of the rain processes, compared to observations, produce inferior simulations of the historical temperature record. A common problem is that models tend to produce rain more frequently than observed that is too light (Stephens et al. 2010; Kay et al. 2018). Since drizzle and rain fundamentally controls the life cycle of clouds (e.g., Miller et al. 1998; Albrecht 1989; Albrecht 1993; Wood 2000) and their mesoscale structure (e.g., Jensen et al. 2000), these model drizzle biases are a source of additional problems in representing feedbacks associated with them (e.g., Mülmenstädt et al. 2020).
The motivation to compare the warm rain formation processes in the real world to model simulations is high. However, this is an observational challenge since the warm rain formation processes in clouds would have to be first monitored as droplets rise (the condensation process), undergo transformation via coalescence and fall from clouds as precipitation. In situ microphysical measurements from aircraft observations tend to be limited in space and time, offering an incomplete glimpse at particle growth processes. Conversely, satellite-based observations offer a large enough sample but typically cannot differentiate the rain process from other cloud physical processes and are typically in the form of column-integrated quantities or limited to near the cloud top (Nakajima et al. 1991; Nakajima and Nakajima 1995; Rosenfeld and Lensky 1998; Rosenfeld 2000). Understanding how microphysical processes evolve in real clouds requires, in part, ways of observing vertical profiles of cloud microphysical properties.
To overcome such challenges, Stephens and Haynes (2007) first suggested that a combination of radar information from active sensors and cloud properties from passive sensors provides such a means, and gives direct insight into warm cloud microphysical processes. Nakajima et al. (2010) and Suzuki et al. (2010) expanded these ideas and devised a novel methodology named the contoured frequency by optical depth diagram (CFODD), which reveals both condensation and coalescence processes in warm clouds by using the radar information from CloudSat, together with optical depth and cloud-top effective radius (Re) from the Moderate Resolution Imaging Spectroradiometer (MODIS). Previous results based on CFODDs (e.g., Suzuki et al. 2010; Takahashi et al. 2017) show that condensation processes are dominant when Re < 10 μm, and cloud-to-drizzle and drizzle-to-rain transitions under coalescence processes typically occur when Re > 10 μm.
Since a CFODD provides insights into the connections between cloud microphysics as characterized by Re and warm rain processes described by the radar reflectivity, it can be a process-level model evaluation tool when it is applied to climate model results and compared to observations. Therefore, many previous studies conducted multimodel comparisons based on CFODDs (e.g., Suzuki et al. 2013, 2015; Jing et al. 2017) and revealed that the coalescence process occurs even when Re < 10 μm in most GCMs.
Although multimodel intercomparisons of these diagnostics are useful for documenting similarities and differences between models and observations, or among different versions of models, they are difficult to interpret due to the large disparity of differences in multiple models’ formulation of many different physical processes. Therefore, they usually cannot help us find sources of model errors or clues to improve models. To do that, it can be more useful to compare the results among different experiments from a single model, similar to the recent CFODD analyses based on a global climate model. Jing et al. (2019) used five different auto conversion schemes in Model for Interdisciplinary Research on Climate (MIROC) to reveal how warm rain formation and aerosol indirect effect (AIE) are largely controlled by the tunable parameter β (the autoconversion rate on the cloud droplet number concentration). Michibata and Suzuki (2020) used different precipitation schemes in MIROC to highlight that both cloud-to-rain transition and aerosol–cloud interactions are better represented in prognostic precipitation schemes than those in diagnostic precipitation schemes. While these recent studies focused on the differences in auto conversion and precipitation schemes, we explore the impact of a wider range of schemes, taking advantage of some of the experiments of the multiphysics ensemble of Bodas-Salcedo et al. (2019), which used two configurations of the Hadley Center Global Environment Model, version 3 (HadGEM3; Hewitt et al. 2011), to track down the main processes that control the changes in feedbacks between these two model configurations. We also use different configurations of HadGEM3 to delve into what configuration changes impact the warm rain formation processes and what the diagnostics are telling us about the process. In this way, we develop an understanding of the limitations of the diagnostics and are able to propose future avenues for model development.
In this study, we use a controlled experimental setup in which we make physical changes in a selected number of physical processes to pursue three main objectives: 1) evaluate the representation of physical processes related with the simulation of warm rain, 2) understand the limitations of the diagnostics used in the analysis, and 3) explore aerosol impacts on model physics. The present paper demonstrates and documents the utility and limitation of CFODDs by introducing a new process-level model evaluation tool and analyzing how the warm rain formation processes in HadGEM3 can be affected by the changes to the aerosol scheme and to warm rain microphysics. The rest of the paper is organized as follows. Section 2 describes the data and diagnostic methods used to analyze them. Results and interpretations are presented in section 3. Section 4 summaries the study and discusses future avenues for model development.
2. Data and diagnostics
In this paper we seek to determine what changes affect the warm rain formation processes in HadGEM3. We focus on single-layered oceanic warm clouds whose cloud-top temperatures exceed 273 K and compared A-Train satellite observations to a series of different experiment runs performed using modified versions of the latest climate configurations of the Hadley Centre Global Environmental Model, version 3.1 (HadGEM3-GC3.1; Williams et al. 2018).
a. A-Train satellite observations
CloudSat and Aqua both fly as a part of the A-Train constellation. CloudSat carries a 94 GHz Cloud Profiling Radar (CPR; Stephens et al. 2008), which has a horizontal resolution of 1.7 km along-track × 1.4 km across track, with a vertical resolution of 480 m. CloudSat CPR is sensitive to both cloud- and precipitation-sized particles and provides us with information of detailed vertical structures of clouds. On the other hand, the MODIS instrument on Aqua, which is also a member of the A-Train satellite constellation, provides horizontal information not only on cloud-top properties (pressure, temperature, and height), but also on cloud optical properties (optical thickness, effective particle radius, and water path) by measuring the amount of sunlight reflected from clouds. Following previous studies (e.g., Suzuki et al. 2010; Takahashi et al. 2017), we use the CloudSat CPR profiles from the geometric profile product (2B-GEOPROF) product (e.g., Mace 2007; Marchand et al. 2008) and the MODIS collection 5.1 level 2 MYD06 cloud product (e.g., Platnick et al. 2003), whose retrievals collocated to the daytime (1330 local time) CloudSat CPR footprint. The 2B-GEOPROF includes not only reflectivity, but also a cloud mask to identify clouds. In this study, we use a cloud mask value ≥ 30, which is a high-confidence detection of clouds. For effective particle radius (Re), total optical thickness (τc), and cloud-top temperature, we use the MYD06 cloud product (e.g., Platnick et al. 2003). These data are combined with the European Centre for Medium-Range Weather Forecasts–auxiliary (ECMWF-AUX) temperature profiles matched to the CloudSat CPR footprint. In this study, we only select the profiles of warm clouds when both MODIS cloud-top temperature and ECMWF-AUX temperature at cloud tops exceed 273 K. Cloud-top height is determined by CloudSat CPR echo top of ~−30 dBZ. Also, profiles containing multilayered clouds are excluded from analysis.
b. Model description and experimental setup
The latest climate configuration of the Hadley Centre Global Environmental model is HadGEM3-GC3.1. The Global Coupled (GC) configurations of the HadGEM3 contain the following subcomponents: Global Atmosphere (GA), Global Land (GL), Global Ocean (GO), and Global Sea Ice (GSI). The atmospheric model configuration of GC3.1 is GA7.1 coupled to GL7.0.
Here, we follow the experimental setup used in Bodas-Salcedo et al. (2019), by disabling some of the individual model changes that were included in the latest GA configuration. Our control experiment is GA7.1, which is an incremental evolution from GA7.0. (Walters et al. 2019; Williams et al. 2018). GA7.1 was developed to target a better representation of aerosol radiative forcing (Mulcahy et al. 2018). We target model changes that we believe significantly affect the warm rain processes in HadGEM3. Those changes are described in the following discussion.
GA7.1 includes the Global Model of Aerosol Processes (GLOMAP-mode) aerosol scheme described in Mann et al. (2010), which is part of the U.K. Chemistry and Aerosols (UKCA) code. GLOMAP-mode is a two-moment scheme that simulates speciated aerosol mass and number in four soluble modes and an insoluble Aitken mode. GLOMAP-mode replaces the Coupled Large-Scale Aerosol Simulator for Studies in Climate (CLASSIC; Bellouin et al. 2011), which is a single-moment scheme used in previous configurations of the Hadley Centre model since HadGEM2 (Collins et al. 2011; Martin et al. 2011). Bodas-Salcedo et al. (2019) found that the replacement of the CLASSIC aerosol scheme by the GLOMAP-mode scheme was one of the main drivers of the changes in feedbacks between GA6.0 and GA7.1. GLOMAP-mode also had a significant impact on the cloud climatology, reducing the climatological effective radius. The reduction in effective radius is mainly due to a change in cloud droplet number concentration. GLOMAP-mode produces realistic distributions of cloud droplet number concentration when compared to MODIS retrievals (Mulcahy et al. 2018). However, it should be noted that potential biases exist in GLOMAP-mode since the uncertainties in satellite retrievals of cloud droplet number concentration are known to be large.
A new warm rain microphysics scheme was implemented as part of the GA7.0 developments. This scheme includes: an improved representation of the autoconversion and accretion rates (Boutle and Abel 2012), using the formulation by Khairoutdinov and Kogan (2000), which replaces the Tripoli and Cotton (1980) scheme; a representation of subgrid variability of cloud and precipitation, and its effects on autoconversion and accretion rates (Boutle et al. 2014); and a better representation of the second aerosol indirect effect (Hill et al. 2011). Although the old warm rain microphysics scheme (OWR) includes a second indirect effect, the rain production scheme is relatively insensitive to cloud droplet number concentration (CDNC).
The calculation of Re is different between CLASSIC and GLOMAP-mode configurations. CLASSIC provides the aerosol number concentration, and CDNC is calculated using the parameterization by Jones et al. (1994). Then, Re is calculated as in Martin et al. (1994), which takes cloud liquid water content and CDNC as inputs. GLOMAP-mode directly provides CDNC, and Re is calculated from CDNC and liquid water content, but accounting for the effect of spectral dispersion (Liu et al. 2008).
Given that GLOMAP-mode has a large impact on the Re (Bodas-Salcedo et al. 2019), we want to understand how this impact translates into an influence on warm rain formation processes, as expected from the relationships identified in the CFODDs based on the satellite observations. Also, knowing that the new warm rain scheme was specifically developed to address some of the known deficiencies of warm rain processes in the old scheme, we investigate how a representation of the second aerosol indirect effect and how the switching from the old to new warm rain microphysics scheme impact warm rain formation processes. Therefore, in addition to our GA7.1 control experiment, we run three additional experiments, which are modifications of the control GA7.1 experiment:
CLASSIC: replace GLOMAP-mode by CLASSIC aerosol scheme.
AIE: second aerosol indirect effect switched off.
OWR: use the GA6.0 old warm rain microphysics scheme.
List of key differences between Control and other experiments are summarized in Table 1. In the AIE experiment, switching off the second indirect means that autoconversion uses a land–sea split with constant CDNC of 300 cm−3 over land and 100 cm−3 over ocean. The Re used by the radiation scheme is still calculated using the CDNC provided by GLOMAP-mode. This experiment is not meant to represent a realistic perturbation, but a sensitivity experiment that allows us to explore the impact of the second indirect effect of clouds, radiation, and precipitation.
List of key differences between Control and other HadGEM3 experiments.
All the experiments are run at N96 horizontal resolution, with an equal-angle grid of 1.875° in longitude and 1.25° in latitude. The atmosphere is divided into 85 layers in the vertical, with the model top at 85 km high. We use boundary conditions and forcings requested by the amip experiment of phase 6 of the Coupled Model Intercomparison Project (CMIP6; Eyring et al. 2016). All model runs were started in January 1979, but the lengths differ. The control and AIE experiments are run for 36 years (January 1979–December 2014), whereas the CLASSIC and OWR experiments are run for 12 years (January 1979–December 1990). The CLASSIC and OWR experiments belong to the ensemble of experiments used by Bodas-Salcedo et al. (2019). They were only run for 12 years because the main aim of these experiments was to apply an initial screening of those model changes that contributed most to the changes in radiative feedbacks between two major model configurations. After this initial screening, a small subset of longer experiments was run. More details about the experimental setup of these runs are given in Bodas-Salcedo et al. (2019). We produce instantaneous, 3-hourly diagnostics for one April month of each model run. Model diagnostics of radar reflectivity are produced with the Cloud Feedback Model Intercomparison Project (CFMIP) Observation Simulator Package, version 1.4 (COSP; Bodas-Salcedo et al. 2011).
c. CFODD
3. Results
a. Assessments of model physics
1) Conventional CFODD
The top panels of Fig. 1 demonstrate how the warm rain formation processes in the A-Train observations are visualized by CFODDs, classified into four categories of Re: Re = 5–10, 10–15, 15–20, and 20–25 μm. For the CFODD with Re = 5–10 μm, cloud mode (Zmax < −15 dBZ) dominates throughout the entire cloud layer from the cloud top (τd = 0–10) to bottom (τd = 40–50), and the behavior of the profile is associated with condensation processes and little coalescence (Suzuki et al. 2010). For the CFODDs shown for the other three categories with Re = 10–25 μm, reflectivity increases with τd in the manner consistent with coalescence and the slope of the CFODDs increases as Re increases indicative of enhanced coalescence in the larger Re category. These CFODDs composited on equal intervals of Re (which we call conventional CFODDs) clearly capture the features of condensation and coalescence processes as a function of Re, and exhibit transitions from cloud to drizzle and drizzle to rain. Many previous studies have used this conventional CFODD approach to evaluate model performance (e.g., Michibata et al. 2014; Suzuki et al. 2015; Jing et al. 2019).
Conventional CFODDs based on the observations and different HadGEM3 runs over global ocean, classified according to Re into four groups: Re = 5–10, 10–15, 15–20, and 20–25 μm.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Conventional CFODDs based on the observations and different HadGEM3 runs over global ocean, classified according to Re into four groups: Re = 5–10, 10–15, 15–20, and 20–25 μm.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Conventional CFODDs based on the observations and different HadGEM3 runs over global ocean, classified according to Re into four groups: Re = 5–10, 10–15, 15–20, and 20–25 μm.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Compared to the observations, the four HadGEM3 experiments each exhibit coalescence associated with smaller droplet size Re = 5–10 μm than is observed and the consequence is that these model warm clouds all produce more drizzle than observed. This is a common problem in many GCMs (e.g., Kay et al. 2018). Unlike the observations, the transition from drizzle to rain does not frequently occur as radar reflectivities in HadGEM3 tend to not exceed 0 dBZ regardless of the value of Re. This result gives the impression that HadGEM3 substantially underestimates the occurrence of heavier rain but all HadGEM3 experiments suggest attenuation of radar reflectivity due to heavy rain indicated by radar reflectivities that decrease with increasing τd when Re = 15–20 μm, and the attenuation effect becomes even stronger when Re = 20–25 μm, especially for OWR experiment, which is not found in the observations. The reflectivity is calculated from large-scale condensate (cloud and precipitation) and convective precipitation is not included, and it is possible that the calculation of the reflectivity is affected by errors in the particle size distribution. A recent study by Michibata and Suzuki (2020) reveals that the assumption of raindrop size can largely impact the CFODD statistics. To determine the dominant driver causing the reflectivity less than 0 dBZ tendency in HadGEM3, further investigation is needed, including the examination of the nonattenuated reflectivity values in COSP as well as the assumption of raindrop size. This will be a topic of a future study.
We know that the new aerosol scheme (GLOMAP-mode) has a large impact in the Re. Due to the nature of the coupling between Re and warm rain formation processes, one may be expected that the CLASSIC experiment, since it replaces GLOMAP-mode with the CLASSIC aerosol scheme, produces different warm rain characteristics that appear in the conventional CFODDs compared to the Control experiment. However, since conventional CFODDs represent the relationship of rain formation characteristics relative to the effective particle radius that is fundamentally determined by cloud microphysics itself, they tend to not be affected by the aerosol–cloud interrelationships. Therefore, we do not expect the CLASSIC as well as AIE produce different conventional CFODDs compared to Control. OWR is the one to be expected to produce different conventional CFODDs compared to Control since the new warm rain scheme was specifically developed to address some of the known deficiencies of warm rain processes in the old scheme. As we expected, CLASSIC and AIE both have little impact on the warm rain characteristics as their conventional CFODDs are comparable to Control, while OWR produces a significant impact on the warm rain characteristics as expected. Therefore, the most intriguing feature in Fig. 1 lies in the difference between OWR and the results from other HadGEM3 configurations, especially for the range Re = 5–15 μm. Unlike other HadGEM3 runs that illustrate that most of warm clouds experience the transition from cloud to light drizzle mode, the warm clouds of the OWR simulations contain only a drizzle mode. Switching to the new warm rain microphysics scheme successfully reduces overly produced drizzle mode when Re = 5–15 μm. This result is also consistent with research by Boutle and Abel (2012), which shows that the new autoconversion parameterization and a new representation of the raindrop size distribution produce more realistic surface drizzle rates. The results here extend these conclusions to the global oceans, beyond the more limited case studies of Boutle and Abel (2012).
2) Cloud-top effective radius and the column maximum radar reflectivity
As discussed above, Re is a useful cloud diagnostic property for understanding warm rain formation processes. It is important to identify how the distribution of Re varies between different HadGEM3 experiments, and how the changes in Re distributions link to the changes in Zmax, and in this context to an understanding of the rain intensity. Figure 2 are histograms and cumulative histograms of Re and Zmax derived from observations and from the different HadGEM3 experiments. The differences in Re are small among Control, AIE, and OWR, whereas the difference between Control and CLASSIC are large. This confirms that disabling second aerosol indirect effect or switching to a new warm rain microphysics scheme do not grossly affect the Re distribution. However, as suggested by Bodas-Salcedo et al. (2019), switching to the new aerosol scheme has a large impact on reducing Re, degrading the comparisons to the observations. We expect that the reduction of Re would impact the distribution of Zmax, and thus CLASSIC would have a different Zmax distribution compared to the different HadGEM3 experiments. However, surprisingly such difference in Zmax is observed in OWR, not in CLASSIC despite the fact that CLASSIC has a different Re distribution compared to the different HadGEM3 experiments. These results merely imply that a weak relation between Re and Zmax distributions exist in HadGEM3 and not monotonic in character like that of the observations.
The histograms of (a) cloud-top effective radius (Re) and (b) the column maximum radar reflectivity (Zmax), together with normalized cumulative histograms of (c) Re and (d) Zmax based on the observations and different experiments from HadGEM3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
The histograms of (a) cloud-top effective radius (Re) and (b) the column maximum radar reflectivity (Zmax), together with normalized cumulative histograms of (c) Re and (d) Zmax based on the observations and different experiments from HadGEM3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
The histograms of (a) cloud-top effective radius (Re) and (b) the column maximum radar reflectivity (Zmax), together with normalized cumulative histograms of (c) Re and (d) Zmax based on the observations and different experiments from HadGEM3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Figure 2 also reveals that all HadGEM3 experiments underestimate Re with respect to observations. The fact that the distributions of Re show little overlap between the observations and the HadGEM3, especially the simulations with the new aerosol scheme (Control, AIE, and OWR), and little connection to the existence of rain and drizzle, raises the question about the approach of compositing the conventional CFODDs in predefined intervals of Re and whether the results of Fig. 1 are an artifact of the analysis approach. For the HadGEM3 simulations, the fraction of the population of warm rain clouds with Re > 15 μm is small for simulations with the new aerosol scheme. For those simulations, the last two columns of plots in the conventional CFODDs (Fig. 1) exaggerates this population difference between simulations giving too much weight to a small fraction of the population.
3) Pros and cons of conventional CFODD
Conventional CFODDs anchor the real-world dependence of warm rain formation processes to particle radius size, as condensation and coalescence processes are dominated when Re <10 μm and Re > 10 μm, respectively. Therefore, by using four groups of Re = 5–25 μm with 5 μm interval, conventional CFODDs are able to make comparison between the observations and models under the same category of Re, which helps us evaluate how realistically the timing and intensity of simulated warm rain formation processes are performed as a function of Re. However, these fixed references of Re make conventional CFODDs very sensitive to the underlying respective observed and modeled distributions of Re. This means that a conventional CFODD can only be a robust evaluation tool when it applies to models whose Re distributions are comparable to the observations. If models and observations have different Re distributions, a conventional CFODD could misguide the diagnosis of model errors because it ignores the sample biases in models over each range of Re and model results in this format might be significantly controlled by outliers. Because of these reasons, we now develop a modified representation of the data to reflect a representative and robust way of comparing observations to models.
4) Quartile-based CFODD
We propose the bin data into four quartiles (0%–25%, 25%–50%, 50%–75%, and 75%–100%) of Re as a way of presenting the CFODDs (herein quartile-based CFODDs) as shown in Fig. 3. With this approach, each quartile-based CFODD contains the same fraction of warm rain cloud samples, in both models and in the observations.
Quartile-based CFODDs over global ocean, classified according to four quartiles (0%–25%, 25%–50%, 50%–75%, and 75%–100%) of Re distributions based on the observations and different HadGEM3 runs.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Quartile-based CFODDs over global ocean, classified according to four quartiles (0%–25%, 25%–50%, 50%–75%, and 75%–100%) of Re distributions based on the observations and different HadGEM3 runs.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Quartile-based CFODDs over global ocean, classified according to four quartiles (0%–25%, 25%–50%, 50%–75%, and 75%–100%) of Re distributions based on the observations and different HadGEM3 runs.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Although all the HadGEM3 simulations still produce more drizzle than the observations, and the largest differences among the HadGEM3 experiments remains between Control and OWR, some conclusions about the model results change with this new approach. First, warm clouds in the observations and HadGEM3 with the new warm rain scheme (Control, CLASSIC, and AIE) both experience comparable early stages of the coalescence process that have similar transition from cloud to drizzle (more frequently) and drizzle to rain (less frequently) in the first quartile. Second, there are some clear differences between Control and CLASSIC between the second to fourth quartiles, as CLASSIC tends to produce slightly larger reflectivities in the drizzle and cloud modes in contrast to the results in the conventional CFODDs. This is consistent with the expected impact of larger Re. CLASSIC also seems to transition to drizzle earlier than the control experiment, suggesting that the upgrade to GLOMAP-mode has contributed to reducing the problem of persistent drizzle, although this is a much smaller effect compared to the impact of the new warm rain scheme. Third, OWR has virtually no transition from cloud to drizzle or precipitation (e.g., OWR has too-frequent drizzle or precipitation), consistent with the persistent drizzle problem noted in other studies (e.g., Bodas-Salcedo et al. 2008; Abel and Boutle 2012). The quartile-based CFODDs clearly demonstrate that this problem has been improved when using the new warm rain scheme (Control, CLASSIC, and AIE). Finally, all HadGEM3 runs exhibit a few percentages of rain samples over most quartiles, with the attenuation of radar reflectivity being only vaguely evident in the fourth quartile. Nevertheless, all versions of HadGEM3 still substantially underestimate the occurrence of heavier rain, which is also evident in Fig. 2b. The new warm rain scheme substantially reduces the autoconversion rate and increases accretion rate. Although this improves the simulation of rainfall, it is probably not producing enough warm rain (Williams and Bodas-Salcedo 2017). Again, this may also be affected by uncertainties in reflectivity simulation due to biases in the representation of the size distribution either from precipitable shallow convection, or deeper convection that is still warm phase and single-layer cloud that is picked up in the observations but not in the model. This will be a topic of a future study.
Compared to conventional CFODDs, quartile-based CFODDs provide fairer comparisons between the observations and models because they overcome the sample biases and also focus more on the models’ individualities in the connection between Re and warm rain formation processes. For a model whose timing of starting coalescence differs from observations over a certain range of Re, it is still possible to produce a realistic occurrence and intensity of warm rain formation processes over the whole range of Re. Quartile-based CFODDs shed light on such a model and reveal how occurrence, timing, and intensity of condensation and coalescence processes are realistically distributed in each quartile of Re. However, since observations and models have different quartiles of Re, the model results are being evaluated without the real-world references to Re that guide us to knowing when the condensation and coalescence processes are supposed to be happening. Conventional and quartile-based CFODDs both have pros and cons and reveal informative model biases from different angles. To complement each other, our results recommend using both methods as a set to evaluate simulated warm rain formation processes, unless the range of simulated Re distributions are very different from observations. In such a case, quartiles-based CFODDs are recommended over the conventional CFODDs as evaluation tool.
In addition, HadGEM3 has large biases in dBZ (Fig. 2). This makes it hard to simply applied the observation-based definitions of cloud, drizzle, and rain modes (Zmax < −15 dBZ, −15 < Zmax < 0 dBZ, and Zmax > 0 dBZ, respectively) to model outputs to capture the cloud-to-rain transition in the sequence of conventional CFODDs. When models have large biases in dBZ, conventional CFODDs may give the impression that HadGEM3 does not progress from cloud to rain, which is misleading. In order for these models to properly monitor the transition, statistical-quartile-based approach is needed to look at the progression. This is again, not entirely anchored to the real-world dependence of warm rain formation processes, but we can observe the progression through the processes going from cloud to rain to statistically evaluate how the progression is realistic. As a result, quartile-based CFODDs, show some shifts of CFODD from lower to higher value of dBZ more continuous ways than conventional CFODDs. Quartile-based CFODDs tell that HadGEM3 is still progressing from cloud to rain without consider the thresholds of dBZ. Some biases are physical-based but some are caused by the wrong estimation of dBZ. Understanding the sources of those biases are also topic of ongoing research.
b. Second indirect effect and regional precipitation constraints
The suppression of the second indirect effect is a drastic perturbation in the rainfall–aerosol interaction. Control generates its own internal aerosol mass and number concentration from surface emissions, and the activation scheme calculates the CDNC interactively (West et al. 2014). When the second indirect effect is switched off, a constant land–sea split CDNC is used: 100 cm−3 over ocean and 300 cm−3 over land. When compared to satellite retrievals of CDNC (e.g., Mulcahy et al. 2018), Fig. 4a shows that Control produces a realistic geographical pattern of CDNC, with low concentrations in remote clean areas like the Southern Ocean, and concentrations in excess of 300 cm−3 over some of the highly polluted continental regions. Control implies large biases in amount of CDNC concentrations although the satellite retrievals of CDNC themselves remain questionable. Compared to Control, CLASSIC tends to underestimate CDNC over most of the globe, and only some land regions, except the Asian continent, tend to overestimate CDNC (Figs. 4b,c). Overall, their differences are not as large as those between Control and AIE.
Warm cloud droplet number concentration annual climatology (2009–13) for (a) Control, (b) CLASSIC, (c) difference between CLASSIC and Control, and (d) difference between AIE and Control. Units are cm−3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Warm cloud droplet number concentration annual climatology (2009–13) for (a) Control, (b) CLASSIC, (c) difference between CLASSIC and Control, and (d) difference between AIE and Control. Units are cm−3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Warm cloud droplet number concentration annual climatology (2009–13) for (a) Control, (b) CLASSIC, (c) difference between CLASSIC and Control, and (d) difference between AIE and Control. Units are cm−3.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
In the AIE experiment, the switch to a constant land–sea split imposes a large perturbation in CDNC, with an increase in concentration over most of the globe (Fig. 4d). Only some oceanic regions surrounding emission hotspots show a substantial decrease in concentration, which show very little change in large-scale precipitation (Fig. 5c) in contrast with the expected increase in the autoconversion rate due to the reduction in CDNC. In the tropics, AIE shows a decrease (increase) in large-scale rainfall in regions where CDNC increases (decreases), and the fractional changes in large-scale precipitation (Fig. 5e) in those areas are relatively large. It is somewhat surprising that the CFODDs discussed above do not show any appreciable impact from theses precipitation changes. Moreover, AIE shows a significant decrease in CDNC over Arabian Sea and Bay of Bengal, but these regions are dominated by convective precipitation, and the model does not represent any interaction between convection and aerosols. We can only conclude that the changes in the amount of precipitation induced by the AIE perturbation do not modify the characteristics of the cloud-to-precipitation transition in the model.
Control annual climatologies (2009–13) of (a) rainfall and (b) cloud LWP. Impact of second aerosol indirect effect on (c) climatological rainfall difference, (d) cloud LWP difference, (e) fractional change in rainfall, and (f) TOA net radiation difference (W m−2). Rainfall units are mm day−1 and LWP units are g m−2.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Control annual climatologies (2009–13) of (a) rainfall and (b) cloud LWP. Impact of second aerosol indirect effect on (c) climatological rainfall difference, (d) cloud LWP difference, (e) fractional change in rainfall, and (f) TOA net radiation difference (W m−2). Rainfall units are mm day−1 and LWP units are g m−2.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Control annual climatologies (2009–13) of (a) rainfall and (b) cloud LWP. Impact of second aerosol indirect effect on (c) climatological rainfall difference, (d) cloud LWP difference, (e) fractional change in rainfall, and (f) TOA net radiation difference (W m−2). Rainfall units are mm day−1 and LWP units are g m−2.
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
The most interesting aspect of the changes in AIE is the muted rainfall response (Fig. 5e) in the midlatitude, especially in the Southern Ocean region (e.g., 30°–60°S). In these regions, AIE shows a significant increase in CDNC, which is not translated into a reduction in precipitation. Observational studies have shown that rainfall in the midlatitudes is mainly originated from snow aloft (Mülmenstädt et al. 2015; Field and Heymsfield 2015). Since the model’s ice microphysics is not linked to the aerosol concentration, one could argue that the lack of rainfall response in AIE is expected. However, Figs. 5d,f do not support this because AIE shows a large increase in liquid water path (LWP) and a large decrease in TOA net radiation in the midlatitudes. It suggests that large-scale constraints act to reduce the impact of such a big change in CDNC.
It is well established that global-mean precipitation is controlled by the atmospheric energy budget (e.g., Mitchell et al. 1987; Allen and Ingram 2002). Regional controls are less well understood, but Muller and O’Gorman (2011) show that precipitation and diabatic cooling become tightly linked for spatial scales larger than 7000 km. To fulfil these energetic large-scale constraints on precipitation, changes in other variables must occur to compensate for the imposed changes in CDNC. Given that the autoconversion rate depends on CDNC and liquid water content (LWC), a natural pathway for this compensation is to look at changes in LWP (Fig. 5d). Indeed, the large changes in LWP over the Southern Ocean region act to reduce the direct impact of the CDNC changes on rainfall. The large increase in LWP acts to increase the autoconversion rate, compensating for the decrease in autoconversion due to the increase in CDNC. The end result is a small change in large-scale rainfall in this region. The LWP adjustment induces large changes in top-of-the-atmosphere (TOA) net radiation (Fig. 5f). The radiative changes occur mainly through shortwave changes, producing a large regional negative forcing. It is worth noting that the large-scale precipitation constraints in the midlatitudes could be of dynamical nature, like the balance between precipitation and moisture ingested into midlatitude cyclones along the warm conveyor belt (Field and Wood 2007; McCoy et al. 2019, 2020).
Figure 6 shows 5-yr time averages of vertical profiles of autoconversion and accretion rates, and cloud liquid mass fraction for the Southern Ocean region (30°–60°S). These vertical profiles seem to support our interpretation of the results. Initially, a big drop in the autoconversion rate is forced by the increase in CDNC. The large-scale constraint makes the cloud liquid to increase, leading to increasing accretion and autoconversion rates. The resulting balance of processes leads to changes in accretion dominated by the cloud liquid increase, and to reduced autoconversion dominated by the CDNC increase. Surface precipitation is probably influenced by other processes that can complicate the picture. Only considering warm rain, evaporation between cloud base and the surface is significant (Boutle and Abel 2012), which could play a significant role in the new equilibrium in the perturbed climate. In the midlatitudes, ice processes are probably the dominant source of mean surface precipitation. Given that the perturbation introduced in the AIE experiment does not impact ice processes, it is likely that the new balance is achieved by adjustments in autoconversion and accretion rates and subcloud evaporation. However, we cannot rule out that more riming due to the increase in cloud liquid water also plays a role in balancing an initial decrease in precipitation. We cannot provide a full description of all the terms in this balance with the available data, but Fig. 5 shows a large potential radiative impact through aerosol–precipitation interactions.
Vertical profiles of 5-yr time averages of (a) autoconversion (Qaut) and accretion rates (Qacc) and (b) cloud liquid mass fraction for the Southern Ocean region (30°–60°S).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Vertical profiles of 5-yr time averages of (a) autoconversion (Qaut) and accretion rates (Qacc) and (b) cloud liquid mass fraction for the Southern Ocean region (30°–60°S).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
Vertical profiles of 5-yr time averages of (a) autoconversion (Qaut) and accretion rates (Qacc) and (b) cloud liquid mass fraction for the Southern Ocean region (30°–60°S).
Citation: Journal of the Atmospheric Sciences 78, 7; 10.1175/JAS-D-20-0321.1
One could argue that a large-scale constraint may be anticipated for a perturbation like AIE, since the changes in CDNC are not directly perturbing the diabatic cooling (unlike a CO2 perturbation, for instance), and the fastest pathway for the system to adjust is through a rearrangement in the microphysical rates. Given the large radiative impact, this is a topic that deserves more attention and we plan to explore this in more detail in future work with dedicated experiments.
It must be noted that we are using the AIE experiment as a tool to better understand the information that can be extracted from complex diagnostics like CFODDs, and it is not intended to be interpreted as a realistic experiment. It is possible that no significant difference in CFODD may be a result from activation of the accretion process that does not directly depends on aerosols. The impact of accretion on the CFODD statistics obtained from the model is our future research.
4. Conclusions and summary
To track down the impacts of changes in warm rain formation processes between the latest two configurations of HadGEM3, the performance of the model with different configurations of cloud and aerosol parameterizations (Control, AIE, CLASSIC, and OWR) are evaluated against A-Train observations and it is found out that all experiments underestimate Re compared to observations. The fact that the distributions of observed Re show little correspondence to that from the HadGEM3 experiments, especially the simulations that use the GLOMAP-mode aerosol scheme (Control, AIE, and OWR), raises the question about the approach of compositing the CFODDs in predefined intervals of Re (i.e., conventional CFODDs). We introduce a new analysis that splits the population into four quartiles of Re, deriving CFODDs for each quartile (i.e., quartile-based CFODDs). This way, each CFODD contains the same fraction of samples, which gives a more direct way of comparing model and observation. Nevertheless, we wish to stress that since conventional and quartile-based CFODDs expose model biases from different perspective, it is important to use both in order to gain a better understanding of model performance. The key findings of this study are as follows:
Although the largest Re differences are found between the Control and CLASSIC experiments, the largest warm rain formation differences are found between Control and OWR, which suggests that Re is only weakly linked in all HadGEM3 configurations to warm rain formation processes in contrast to observations.
Switching to the new aerosol scheme (Control, AIE, and OWR) also contributes to reducing the drizzle mode, yet switching to the new warm rain microphysics scheme (Control, CLASSIC, and AIE) causes the largest difference in warm rain formation processes reducing overly produced drizzle mode in HadGEM3.
A large impact in warm rain is expected in the AIE experiment due to the large changes in CDNC. However, regions with large fractional changes in CDNC show only a small change in precipitation because large-scale constraints act to reduce the impact of microphysical changes involving CDNC, which leads us to speculate that the large-scale precipitation constraints regulate the radiative forcing sensitivity to autoconversion parameterization.
The changes in the amount of precipitation induced by the AIE perturbation do not modify the characteristics of the cloud to precipitation transition in the model.
Within the context of the radiative impact of the second indirect effect, it is important to evaluate the simulation of the relative contributions of autoconversion and accretion to warm rain processes, since only autoconversion is modeled as directly sensitive to changes in CDNC. However, the separation of these two processes is a huge challenge in the observations. Future research will focus on the design of diagnostics that could help relate this separation with observations. The AIE experiment used here, with a constant land–sea split CDNC has been used as a sensitivity study for convenience, but it is not the best choice for as a realistic experiment. In future studies, a more realistic experimental setup will be used to study the impact of perturbations to the second indirect effect.
Acknowledgments
The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration (NASA). CloudSat observations can be found from the CloudSat Data Processing Center (www.cloudsat.cira.colostate.edu). This study was supported by the CloudSat funding (103428/8.A.1.6.). A. B.-S. was supported by the Met Office Hadley Centre Climate Programme funded by BEIS and Defra. A. B.-S. was also supported by the CloudSat funding to visit the Jet Propulsion Laboratory. The authors thank Dr. Ian Boutle for insightful comments and discussions.
REFERENCES
Abel, S. J., and I. A. Boutle, 2012: An improved representation of the raindrop size distribution for single-moment microphysics schemes. Quart. J. Roy. Meteor. Soc., 138, 2151–2162, https://doi.org/10.1002/qj.1949.
Albrecht, B. A., 1989: Aerosols, cloud microphysics, and fractional cloudiness. Science, 245, 1227–1230, https://doi.org/10.1126/science.245.4923.1227.
Albrecht, B. A., 1993: Effects of precipitation on the thermodynamic structure of the trade wind boundary layer. J. Geophys. Res., 98, 7327–7337, https://doi.org/10.1029/93JD00027.
Allen, M. R., and W. J. Ingram, 2002: Constraints on future changes in climate and the hydrologic cycle. Nature, 419, 228–232, https://doi.org/10.1038/nature01092.
Bellouin, N., J. Rae, A. Jones, C. Johnson, J. Haywood, and O. Boucher, 2011: Aerosol forcing in the Climate Model Intercomparison Project (CMIP5) simulations by HadGEM2-ES and the role of ammonium nitrate. J. Geophys. Res., 116, D20206, https://doi.org/10.1029/2011JD016074.
Bodas-Salcedo, A., M. J. Webb, M. E. Brooks, M. A. Ringer, K. D. Williams, S. F. Milton, and D. R. Wilson, 2008: Evaluating cloud systems in the Met Office Global Forecast Model using simulated CloudSat radar reflectivities. J. Geophys. Res., 113, D00A13, https://doi.org/10.1029/2007JD009620.
Bodas-Salcedo, A., and Coauthors, 2011: COSP: Satellite simulation software for model assessment. Bull. Amer. Meteor. Soc., 92, 1023–1043, https://doi.org/10.1175/2011BAMS2856.1.
Bodas-Salcedo, A., J. P. Mulcahy, T. Andrews, K. D. Williams, M. A. Ringer, P. R. Field, and G. S. Elsaesser, 2019: Strong dependence of atmospheric feedbacks on mixed-phase microphysics and aerosol-cloud interactions in HadGEM3. J. Adv. Model. Earth Syst., 11, 1735–1758, https://doi.org/10.1029/2019MS001688.
Boutle, I. A., and S. J. Abel, 2012: Microphysical controls on the stratocumulus topped boundary-layer structure during VOCALS-REx. Atmos. Chem. Phys., 12, 2849–2863, https://doi.org/10.5194/acp-12-2849-2012.
Boutle, I. A., S. J. Abel, P. G. Hill, and C. J. Morcrette, 2014: Spatial variability of liquid cloud and rain: Observations and microphysical effects. Quart. J. Roy. Meteor. Soc., 140, 583–594, https://doi.org/10.1002/qj.2140.
Brenguier, J. L., H. Pawlowska, L. Schüller, R. Preusker, J. Fischer, and Y. Fouquart, 2000: Radiative properties of boundary layer clouds: Droplet effective radius versus number concentration. J. Atmos. Sci., 57, 803–821, https://doi.org/10.1175/1520-0469(2000)057<0803:RPOBLC>2.0.CO;2.
Collins, W. J., and Coauthors, 2011: Development and evaluation of an Earth-System model–HadGEM2. Geosci. Model Dev., 4, 1051–1075, https://doi.org/10.5194/gmd-4-1051-2011.
Comstock, K. K., R. Wood, S. E. Yuter, and C. S. Bretherton, 2004: Reflectivity and rain rate in and below drizzling stratocumulus. Quart. J. Roy. Meteor. Soc., 130, 2891–2918, https://doi.org/10.1256/qj.03.187.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Field, P. R., and R. Wood, 2007: Precipitation and cloud structure in midlatitude cyclones. J. Climate, 20, 233–254, https://doi.org/10.1175/JCLI3998.1.
Field, P. R., and A. J. Heymsfield, 2015: Importance of snow to global precipitation. Geophys. Res. Lett., 42, 9512–9520, https://doi.org/10.1002/2015GL065497.
Flato, G., and Coauthors, 2013: Evaluation of climate models. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 741–866.
Golaz, J. C., L. W. Horowitz, and H. Levy, 2013: Cloud tuning in a coupled climate model: Impact on 20th century warming. Geophys. Res. Lett., 40, 2246–2251, https://doi.org/10.1002/grl.50232.
Hewitt, H. T., D. Copsey, I. D. Culverwell, C. M. Harris, R. S. R. Hill, A. B. Keen, A. J. McLaren, and E. C. Hunke, 2011: Design and implementation of the infrastructure of HadGEM3: The next-generation Met Office climate modelling system. Geosci. Model Dev., 4, 223–253, https://doi.org/10.5194/gmd-4-223-2011.
Hill, P. G., J. Manners, and J. C. Petch, 2011: Reducing noise associated with the Monte Carlo independent column approximation for weather forecasting models. Quart. J. Roy. Meteor. Soc., 137, 219–228, https://doi.org/10.1002/qj.732.
Jensen, J. B., S. Lee, P. B. Krummel, J. Katzfey, and D. Gogoasa, 2000: Precipitation in marine cumulus and stratocumulus.: Part I: Thermodynamic and dynamic observations of closed cell circulations and cumulus bands. Atmos. Res., 54, 117–155, https://doi.org/10.1016/S0169-8095(00)00040-5.
Jing, X., K. Suzuki, H. Guo, D. Goto, T. Ogura, T. Koshiro, and J. Mülmenstädt, 2017: A multimodel study on warm precipitation biases in global models compared to satellite observations. J. Geophys. Res. Atmos., 122, 11 806–11 824, https://doi.org/10.1002/2017JD027310.
Jing, X., K. Suzuki, and T. Michibata, 2019: The key role of warm rain parameterization in determining the aerosol indirect effect in a global climate model. J. Climate, 32, 4409–4430, https://doi.org/10.1175/JCLI-D-18-0789.1.
Jones, A., D. L. Roberts, and A. Slingo, 1994: A climate model study of indirect radiative forcing by anthropogenic sulphate aerosols. Nature, 370, 450–453, https://doi.org/10.1038/370450a0.
Kay, J. E., T. L’Ecuyer, A. Pendergrass, H. Chepfer, R. Guzman, and V. Yettella, 2018: Scale-aware and definition-aware evaluation of modeled near-surface precipitation frequency using CloudSat observations. J. Geophys. Res., 123, 4294–4309, https://doi.org/10.1002/2017JD028213.
Khairoutdinov, M. F., and Y. L. Kogan, 2000: A new cloud physics parameterization in a large-eddy simulation model of marine stratocumulus. Mon. Wea. Rev., 128, 229–243, https://doi.org/10.1175/1520-0493(2000)128<0229:ANCPPI>2.0.CO;2.
Liu, Y., P. H. Daum, H. Guo, and Y. Peng, 2008: Dispersion bias, dispersion effect, and the aerosol–cloud conundrum. Environ. Res. Lett., 3, 045021, https://doi.org/10.1088/1748-9326/3/4/045021.
Mace, G., 2007: Level 2 GEOPROF product process description and interface control document algorithm version 5.3. NASA Jet Propulsion Laboratory, 44 pp., http://www.cloudsat.cira.colostate.edu/sites/default/files/products/files/2B-GEOPROF_PDICD.P_R04.20070628.pdf.
Mann, G. W., and Coauthors, 2010: Description and evaluation of GLOMAP-mode: A modal global aerosol microphysics model for the UKCA composition-climate model. Geosci. Model Dev., 3, 519–551, https://doi.org/10.5194/gmd-3-519-2010.
Marchand, R., G. G. Mace, T. Ackerman, and G. Stephens, 2008: Hydrometeor detection using CloudSat—An Earth-orbiting 94-GHz cloud radar. J. Atmos. Oceanic Technol., 25, 519–533, https://doi.org/10.1175/2007JTECHA1006.1.
Martin, G. M., D. W. Johnson, and A. Spice, 1994: The measurement and parametrization of effective radius in warm stratocumulus clouds. J. Atmos. Sci., 51, 1823–1842, https://doi.org/10.1175/1520-0469(1994)051<1823:TMAPOE>2.0.CO;2.
Martin, G. M., and Coauthors, 2011: The HadGEM2 family of Met Office Unified Model climate configurations. Geosci. Model Dev., 4, 723–757, https://doi.org/10.5194/gmd-4-723-2011.
McCoy, D. T., and Coauthors, 2019: Cloud feedbacks in extratropical cyclones: Insight from long-term satellite data and high-resolution global simulations. Atmos. Chem. Phys., 19, 1147–1172, https://doi.org/10.5194/acp-19-1147-2019.
McCoy, D. T., P. Field, A. Bodas-Salcedo, G. S. Elsaesser, and M. D. Zelinka, 2020: A regime-oriented approach to observationally constraining extratropical shortwave cloud feedbacks. J. Climate, 33, 9967–9983, https://doi.org/10.1175/JCLI-D-19-0987.1.
Michibata, T., and K. Suzuki, 2020: Reconciling compensating errors between precipitation constraints and the energy budget in a climate model. Geophys. Res. Lett., 47, e2020GL088340, https://doi.org/10.1029/2020GL088340.
Michibata, T., K. Kawamoto, and T. Takemura, 2014: The effects of aerosols on water cloud microphysics and macrophysics based on satellite-retrieved data over East Asia and the North Pacific. Atmos. Chem. Phys., 14, 11 935–11 948, https://doi.org/10.5194/acp-14-11935-2014.
Miller, M. A., M. P. Jensen, and E. E. Clothiaux, 1998: Diurnal cloud and thermodynamic variations in the stratocumulus transition regime: A case study using in situ and remote sensors. J. Atmos. Sci., 55, 2294–2310, https://doi.org/10.1175/1520-0469(1998)055<2294:DCATVI>2.0.CO;2.
Mitchell, J. F., C. A. Wilson, and W. M. Cunnington, 1987: On CO2 climate sensitivity and model dependence of results. Quart. J. Roy. Meteor. Soc., 113, 293–322, https://doi.org/10.1256/smsqj.47516.
Muhlbauer, A., I. L. McCoy, and R. Wood, 2014: Climatology of stratocumulus cloud morphologies: Microphysical properties and radiative effects. Atmos. Chem. Phys., 14, 6695–6716, https://doi.org/10.5194/acp-14-6695-2014.
Mulcahy, J. P., and Coauthors, 2018: Improved aerosol processes and effective radiative forcing in HadGEM3 and UKESM1. J. Adv. Model. Earth Syst., 10, 2786–2805, https://doi.org/10.1029/2018MS001464.
Muller, C. J., and P. A. O’Gorman, 2011: An energetic perspective on the regional response of precipitation to climate change. Nat. Climate Change, 1, 266–271, https://doi.org/10.1038/nclimate1169.
Mülmenstädt, J., O. Sourdeval, J. Delanoë, and J. Quaas, 2015: Frequency of occurrence of rain from liquid-, mixed-, and ice-phase clouds derived from A-Train satellite retrievals. Geophys. Res. Lett., 42, 6502–6509, https://doi.org/10.1002/2015GL064604.
Mülmenstädt, J., and Coauthors, 2020: Reducing the aerosol forcing uncertainty using observational constraints on warm rain processes. Sci. Adv., 6, eaaz6433, https://doi.org/10.1126/sciadv.aaz6433.
Nakajima, T., M. D. King, J. D. Spinhirne, and L. F. Radke, 1991: Determination of the optical thickness and effective particle radius of clouds from reflected solar radiation measurements. Part II: Marine stratocumulus observations. J. Atmos. Sci., 48, 728–750, https://doi.org/10.1175/1520-0469(1991)048<0728:DOTOTA>2.0.CO;2.
Nakajima, T. Y., and T. Nakajima, 1995: Wide-area determination of cloud microphysical properties from NOAA AVHRR measurements for FIRE and ASTEX regions. J. Atmos. Sci., 52, 4043–4059, https://doi.org/10.1175/1520-0469(1995)052<4043:WADOCM>2.0.CO;2.
Nakajima, T. Y., K. Suzuki, and G. L. Stephens, 2010: Droplet growth in warm water clouds observed by the A-Train. Part II: A multisensor view. J. Atmos. Sci., 67, 1897–1907, https://doi.org/10.1175/2010JAS3276.1.
Platnick, S., M. D. King, S. A. Ackerman, W. P. Menzel, B. A. Baum, J. C. Riedi, and R. A. Frey, 2003: The MODIS cloud products: Algorithms and examples from Terra. IEEE Trans. Geosci. Remote Sens., 41, 459–473, https://doi.org/10.1109/TGRS.2002.808301.
Rosenfeld, D., 2000: Suppression of rain and snow by urban and industrial air pollution. Science, 287, 1793–1796, https://doi.org/10.1126/science.287.5459.1793.
Rosenfeld, D., and I. M. Lensky, 1998: Satellite-based insight into precipitation formation processes in continental and maritime convective clouds. Bull. Amer. Meteor. Soc., 79, 2457–2476, https://doi.org/10.1175/1520-0477(1998)079<2457:SBIIPF>2.0.CO;2.
Sato, Y., T. Y. Nakajima, and T. Nakajima, 2012: Investigation of the vertical structure of warm-cloud microphysical properties using the cloud evolution diagram, CFODD, simulated by a three-dimensional spectral bin microphysical model. J. Atmos. Sci., 69, 2012–2030, https://doi.org/10.1175/JAS-D-11-0244.1.
Stephens, G. L., and J. M. Haynes, 2007: Near global observations of the warm rain coalescence process. Geophys. Res. Lett., 34, L20805, https://doi.org/10.1029/2007GL030259.
Stephens, G. L., and Coauthors, 2008: CloudSat mission: Performance and early science after the first year of operation. J. Geophys. Res., 113, D00A18, https://doi.org/10.1029/2008JD009982.
Stephens, G. L., and Coauthors, 2010: Dreary state of precipitation in global models. J. Geophys. Res., 115, D24211, https://doi.org/10.1029/2010JD014532.
Suzuki, K., T. Y. Nakajima, and G. L. Stephens, 2010: Particle growth and drop collection efficiency of warm clouds as inferred from joint CloudSat and MODIS observations. J. Atmos. Sci., 67, 3019–3032, https://doi.org/10.1175/2010JAS3463.1.
Suzuki, K., G. L. Stephens, and M. D. Lebsock, 2013: Aerosol effect on the warm rain formation process: Satellite observations and modeling. J. Geophys. Res., 118, 170–184, https://doi.org/10.1002/jgrd.50043.
Suzuki, K., G. L. Stephens, A. Bodas-Salcedo, M. Wang, J.-C. Golaz, T. Yokohata, and T. Koshiro, 2015: Evaluation of the warm rain formation process in global models with satellite observations. J. Atmos. Sci., 72, 3996–4014, https://doi.org/10.1175/JAS-D-14-0265.1.
Szczodrak, M., P. H. Austin, and P. B. Krummel, 2001: Variability of optical depth and effective radius in marine stratocumulus clouds. J. Atmos. Sci., 58, 2912–2926, https://doi.org/10.1175/1520-0469(2001)058<2912:VOODAE>2.0.CO;2.
Takahashi, H., K. Suzuki, and G. Stephens, 2017: Land–ocean differences in the warm-rain formation process in satellite and ground-based observations and model simulations. Quart. J. Roy. Meteor. Soc., 143, 1804–1815, https://doi.org/10.1002/qj.3042.
Tripoli, G. J., and W. R. Cotton, 1980: A numerical investigation of several factors contributing to the observed variable density of deep convection over south Florida. J. Appl. Meteor., 19, 1037–1063, https://doi.org/10.1175/1520-0450(1980)019<1037:ANIOSF>2.0.CO;2.
Walters, D., and Coauthors, 2019: The Met Office Unified Model Global Atmosphere 7.0/7.1 and JULES Global Land 7.0 configurations. Geosci. Model Dev., 12, 1909–1963, https://doi.org/10.5194/gmd-12-1909-2019.
West, R. E. L., P. Stier, A. Jones, C. E. Johnson, G. W. Mann, N. Bellouin, D. G. Partidge, and Z. Kipling, 2014: The importance of vertical velocity variability for estimates of the indirect aerosol effects. Atmos. Chem. Phys., 14, 6369–6393, https://doi.org/10.5194/acp-14-6369-2014.
Williams, K. D., and A. Bodas-Salcedo, 2017: A multi-diagnostic approach to cloud evaluation. Geosci. Model Dev., 10, 2547–2566, https://doi.org/10.5194/gmd-10-2547-2017.
Williams, K. D., and Coauthors, 2018: The Met Office global coupled model 3.0 and 3.1 (GC3.0 and GC3.1) configurations. J. Adv. Model. Earth Syst., 10, 357–380, https://doi.org/10.1002/2017MS001115.
Wood, R., 2000: Parametrization of the effect of drizzle upon the droplet effective radius in stratocumulus clouds. Quart. J. Roy. Meteor. Soc., 126, 3309–3324, https://doi.org/10.1002/qj.49712657015.
Wood, R., 2012: Stratocumulus clouds. Mon. Wea. Rev., 140, 2373–2423, https://doi.org/10.1175/MWR-D-11-00121.1.