1. Introduction
A deep convective cloud cluster meets the criterion for a mesoscale convective system (MCS) when its contiguous precipitation area reaches a scale of 100 km or more in one direction (e.g., Houze 1993, 2004). An MCS often contains both convective towers and stratiform regions (Houze 1993) and can develop mesoscale circulations during its mature stage (e.g., Houze 1993, 2004; Chen and Frank 1993) that differentiate it from isolated deep convection.
MCSs significantly affect the hydrologic cycle and radiative fluxes and reshape large-scale circulations while redistributing moisture, heat, momentum (Houze 2004), and aerosols (Crumeyrolle et al. 2008). TRMM satellite estimates demonstrate that MCSs are generally responsible for more than 50% of the annual precipitation across the tropic and subtropics (Nesbitt et al. 2006; Schumacher and Rasmussen 2020). A total of 30%–70% of warm-season precipitation is contributed by MCSs in the central United States (Fritsch et al. 1986; Feng et al. 2016). MCSs have a peak in latent heating that is shifted higher in the troposphere than discrete deep convection as a result of stratiform precipitation (e.g., Schumacher et al. 2004; Liu et al. 2015; Feng et al. 2018; Schumacher and Rasmussen 2020). This upper-level heating shift coupled with vertical vorticity generated by MCSs can prolong the MCS lifetime and amplify synoptic ridges and troughs downstream of the MCS (e.g., Yang et al. 2017; Clarke et al. 2019; Schumacher and Rasmussen 2020). In the tropics, MCS heating and mesoscale circulation anomalies alter the structure of equatorial waves and contribute to teleconnections between the tropics and higher latitudes. For example, the increase of stratiform to convective rain partitioning in the tropics associated with La Niña conditions can enhance large-scale wave propagation to the extratropics as compared with El Niño periods (Schumacher et al. 2004). MCS stratiform rain regions also produce stronger net top-of-the-atmosphere (TOA) radiative cooling than convective regions, whereas anvil regions have a net warming effect (e.g., Feng et al. 2011). Thus, the relative areal coverages of convective, stratiform rainfall, and anvil regions impact both the latent heating and radiative budgets in the tropics (e.g., Hartmann et al. 2001; Lin et al. 2002) and extratropics (e.g., Lindzen et al. 2001).
Although MCSs significantly impact weather and climate, there are still significant errors in their representation within models. Relatively coarse resolution climate models largely fail to capture precipitation associated with MCSs because cumulus parameterization schemes do not consider mesoscale processes within MCSs (e.g., Bukovsky and Karoly 2009; Kooperman et al. 2015; Lin et al. 2017; Feng et al. 2019). Regional convection permitting models (CPMs; grid spacing < 4 km) like the High Resolution Rapid Refresh (HRRR; Benjamin et al. 2016) can better predict MCS cloud cover (e.g., Griffin et al. 2017) and extreme rainfall than coarser resolution models (e.g., Chan et al. 2014; Prein et al. 2013; Kendon et al. 2014). In addition, some CPM simulations can approximate climatological MCS number, size, lifetime, and orientation characteristics fairly well in some regions (e.g., Cai and Dumais 2015; Pinto et al. 2015; Prein et al. 2020; Feng et al. 2018). However, some model–observation differences remain. Simulated convective updrafts in case studies have been shown to be stronger than those in observational retrievals (Varble et al. 2011, 2014a; Fan et al. 2016). These overly strong updrafts coupled with biased parameterized microphysical processes cause an overproduction of heavily rimed ice and underproduction of stratiform precipitation (Caine et al. 2013; Varble et al. 2014a,b; Stanford et al. 2017). These biases can also produce errors in downdraft properties, development of cold pools, and mesoscale circulations that impact the MCS life cycle (e.g., Varble et al. 2014b, 2020) through interactions with ambient environmental thermodynamic and kinematic conditions that impact convective updraft initiation, tilt, strength, and width (e.g., Rotunno et al. 1988; Tompkins 2001; Schlemmer and Hohenegger 2014; Feng et al. 2015; Torri et al. 2015).
Accurate prediction of MCSs depends on accurate representation of deep convective upscale growth (e.g., Coniglio et al. 2010; Mulholland et al. 2019), which is the collection of processes that produce mature MCSs from initially isolated deep convection. The upscale growth process also matters for hazardous weather forecasts. For example, during the transition from discrete deep convection to MCSs, the risk of large hail and severe tornadoes decreases and the likelihood for damaging winds and widespread flash flooding increases (Johns and Doswell 1992; Gallus et al. 2008), although this is not true of all MCSs (e.g., Smith et al. 2012).
CPM case study and coarse resolution long-term simulations have been used to identify factors associated with MCS development. Idealized case study simulations have shown that large-scale lifting promotes cloud organization (Tao and Simpson 1984), and large-scale convergence prolongs MCS lifetimes (Crook and Moncrieff 1988). Coarse resolution analyses and reanalyses (100–250-km grid spacing) show that synoptic conditions associated with low-level jets (LLJs) bringing moisture south from the Amazon capped with subsiding, steep lapse rate dry air aloft correspond to MCSs with large horizontal extents and wide convective regions over subtropical South America (e.g., Salio et al. 2007; Rasmussen and Houze 2011, 2016). Relatively higher resolution analyses (20-km grid spacing) over the contiguous United States have been used to show that the rapid development of initial deep convection into MCSs may be promoted by relatively stronger LLJs, more environmental instability, weaker vertical wind shear above 3 km, and lesser geostrophic potential vorticity in the upper troposphere (Coniglio et al. 2010). However, these studies match observed MCSs with environments derived from analyses or reanalyses that are too coarse to capture mesoscale orographic impacts on MCS growth. A recent 18-h CPM simulation indicates that complex terrain can impact the upscale growth process through its impact on near-storm vertical wind shear, instability, and cold pools (Mulholland et al. 2019).
Underestimation of MCS stratiform precipitation by CPMs relative to observations for case studies (e.g., Luo et al. 2010; Tao et al. 2016; Varble et al. 2011, 2014b, 2020) indicates that CPMs may not represent some aspects of upscale growth well. Similar stratiform underproduction is also demonstrated in long-term (2–4 months) CPM simulations with hourly model output (e.g., Hagos et al. 2014; Feng et al. 2018). However, hourly MCS life-cycle assessments (e.g., Prein et al. 2020; Feng et al. 2018) cannot capture evolution of relatively small and short-lived MCSs or discrete life-cycle stages well. Accordingly, MCS growth rate and duration during the growth stage are not quantified in past MCS assessments (e.g., Pinto et al. 2015; Prein et al. 2020; Feng et al. 2018), and environmental conditions associated with the isolated growth-stage properties remain unclear. In addition, environmental conditions associated with MCSs have typically been derived from global reanalyses/analyses (Salio et al. 2007; Coniglio et al. 2010; Rasmussen and Houze 2011, 2016; Feng et al. 2019), which are too coarse to resolve interactions with complex terrain that finer resolution output can overcome.
Some of these issues may stem from past model evaluation tending to focus on either high-resolution case studies that are not necessarily representative of all MCSs or regional climate simulations with insufficient output to examine ordinary MCSs that are not very large or very long-lived. In this study, we seek to overcome these issues by merging these two approaches with a 6.5-month regional convection-permitting simulation with 3-km grid spacing over an MCS initiation hotspot in Argentina (e.g., Zipser et al. 2006; Mulholland et al. 2018) with subhourly output that is comparable to geostationary satellite retrievals and surface precipitation measurements collected during the Remote Sensing of Electrification, Lightning, And Mesoscale/Microscale Processes with Adaptive Ground Observations (RELAMPAGO) and Clouds, Aerosols, and Complex Terrain Interactions (CACTI) field campaigns. Evaluation of simulations with 3-km grid spacing is extremely relevant given the wide usage of this grid spacing in current regional operational weather prediction models (e.g., High Resolution Rapid Refresh; Benjamin et al. 2016). Global climate research models and Earth system models (ESMs) are also in development at or are approaching this scale (e.g., DYAMOND; Stevens et al. 2019).
Our tracking and comparisons of observed and simulated MCSs differ from some previous studies. The subhourly model output better depicts the MCS life cycle than hourly evaluations in previous studies and supports inclusion of more numerous and relatively smaller, shorter-lived MCSs. The MCS life-cycle stages are identified using subhourly horizontal and vertical cloud growth rates of individual MCS, which differs from previous studies using rainfall area and major axis length thresholds (e.g., Coniglio et al. 2010; Feng et al. 2018). We diagnose similarities and differences between observed and simulated life-cycle stages, and test sensitivities of model–observation differences to MCS size, depth, and growth duration and rate. The convection-permitting simulation as compared with reanalysis allows for better resolved environmental conditions and their interaction with complex terrain in exploring how atmospheric conditions differentiate MCS growth rate and sustenance. Last, our focus on subtropical South America can elucidate similarities and differences with many previous studies focused on the central and eastern United States (e.g., Pinto et al. 2015; Prein et al. 2020; Feng et al. 2018) to provide a more holistic view of MCS growth processes.
The remainder of the study is organized as follows. Section 2 describes the satellite retrievals, field campaign measurements, the simulation used in this study, the MCS tracking algorithm, MCS life-cycle-stage identification, and matching of MCSs with ambient atmospheric conditions. Section 3 presents a general evaluation of the simulation performance. A comparison of simulated and observed MCSs and causes for their differences are explored in section 4. Meteorological factors impacting upscale growth are analyzed in section 5. Conclusions and future work are discussed in section 6.
2. Data and method
a. Model setup
A 6.5-month-long, 1800 km × 1500 km regional climate simulation was conducted with the Weather Research and Forecasting (WRF; Skamarock and Klemp 2019) Model, version 4.1.1, using 3-km horizontal grid spacing and 80 vertical levels. The simulation covers the primary experimental period of the field campaign from 15 October 2018 to 30 April 2019, which is the warm season for this region and an active period for MCSs. The model’s integration time step is 15 s. The model domain covers central and northern Argentina extending from west of the Andes to the Atlantic Ocean including the Sierras de Córdoba (SDC) range near the center of the domain where the RELAMPAGO-CACTI experiment was performed (Fig. 1). Key two-dimensional (2D) kinematic, thermodynamic, and microphysical fields relevant to observational comparisons are recorded every 15 min (although only every 30 min is used to match observed fields), and full three-dimensional output is saved every hour. Vertical grid spacing below 5-km altitude is kept to 250 m or less to better resolve the frequently sharp temperature inversion layers east of the Andes. Initial and boundary conditions are obtained from the fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF ERA-5; Copernicus Climate Change Service 2017) that assimilates global observations every 6 h. Microphysical processes are parameterized by the Thompson aerosol aware scheme that allows for variable cloud condensation nuclei and ice nucleating particle concentrations based on climatology (Thompson and Eidhammer 2014). Boundary layer processes are parameterized using the Mellor–Yamada–Nakanishi–Niino (MYNN) level-2.5 (Nakanishi and Niino 2006, 2009) eddy diffusivity mass flux scheme. The simulation is restarted every 4 simulation days from WRF-outputted restart files, thus providing output that is the same as a continuous simulation. Surface/spectral nudging is not used.
The regional WRF run domain (color filled) with the d1 region considered for MCS tracks outlined in black. Red plus signs represent the locations of 15 rain gauge sites, and the black plus sign represents the CACTI AMF site where routine soundings were launched.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
b. Observations
During the 6.5-month analysis period, 30-min Integrated Multisatellite Retrievals for GPM V06B (IMERG; Huffman et al. 2018) at 0.1° grid spacing are used to evaluate simulated MCS precipitation because of the high spatiotemporal resolution, and the 3-h Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation Analysis (TRMM 3B42; Huffman et al. 2007) at 0.25° grid spacing is used to evaluate simulated seasonal total precipitation because of its advantageous estimation at longer time scales. The 30-min TOA infrared brightness temperature (IR Tb) from the National Aeronautics and Space Administration (NASA) merged infrared (MERG-IR) 4-km product coupled with precipitation from IMERG is used to identify observed MCSs. The IR Tb dataset is regridded to match the 10-km IMERG dataset to facilitate the analysis. The WRF outgoing longwave flux is converted to IR Tb using the empirical formula in Yang and Slingo (2001). The derived IR Tb and precipitation in the WRF simulation are regridded (averaged) to coarser observational IMERG grid (~10-km spacing) to identify simulated MCSs. Regridding slightly alters MCS area, but only accounts for a small fraction of MCS total area. MCS identification is limited to domain d1 (black rectangle in Fig. 1), which excludes regions near the model domain boundaries that are highly impacted by MCSs initiating outside the domain that cannot be reproduced in the WRF simulation. Fifteen surface tipping-bucket rain gauge sites (red pluses in Fig. 1) deployed by the National Center for Atmospheric Research Research Applications Laboratory (NCAR RAL) during RELAMPAGO-CACTI (Gochis et al. 2019) are used to evaluate satellite retrieved and simulated rainfall. Simulated thermodynamic conditions relevant to deep convection are evaluated using radiosondes launched every 3–4 h between 0900 and 2200 UTC at the CACTI Atmospheric Radiation Measurement Mobile Facility (AMF) site (Holdridge et al. 2018).
c. MCS tracking
Observed and simulated MCSs are tracked consistently by applying an updated version of the FLEXTRKR algorithm (Feng et al. 2021) to 30-min TOA IR Tb (e.g., Figs. 2a,b) and GPM IMERG surface precipitation (e.g., Figs. 2c,d). An MCS (e.g., purple contours in Figs. 2a–d) in this study needs to meet three requirements: 1) a contiguous cold cloud-shield (Tb < 241 K) area is larger than 1 × 104 km2 and contains a precipitation feature (PF: contiguous area with rain rate > 1 mm h−1) with a major axis length > 100 km [following Houze (2004)], 2) PF area, mean rain rate, rain-rate skewness, and heavy rain (rain rate > 10 mm h−1) ratio exceed the lowest lifetime-dependent thresholds defined in Feng et al. (2021), and 3) both conditions 1 and 2 last continuously for 2 h or more. The PF criteria are designed to exclude weakly precipitating or nonprecipitating cold clouds that are not MCSs (Feng et al. 2021). Smaller and/or short-lived cold cloud clusters that do not meet the MCS definition thresholds but merge with or split from an identified MCS are considered as part of that tracked MCS.
Example of observed and simulated MCSs at 2000 UTC 18 Oct 2018 for (a),(b) observed and simulated TOA IR Tb, with purple contours identifying the MCS, and (c),(d) IMERG-retrieved and WRF-simulated surface precipitation associated with the outlined MCSs.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
d. Life-cycle-stage identification
MCS life-cycle stages are identified by tracking changes in cloud growth (Futyan and Del Genio 2007; Takahashi and Luo 2014). Cloud growth in the vertical direction is associated with a decrease of TOA minimum IR Tb and growth in the horizontal direction is associated with an increase of the cold cloud-shield area. However, transitions between these stages are often not smooth, and some MCSs enter and exit these life-cycle stages multiple times. Thus, a zero-phase second-order low-pass Butterworth filter with a 55.6-μHz cutoff frequency is used to smooth the temporal evolution of the cloud-shield area and TOA minimum IR Tb (e.g., red line in Fig. 3a). As shown in Fig. 3b, after smoothing, the growth stage is defined as deepening and expanding clouds, the mature stage as expanding clouds following peak depth, and the decay stage as shrinking and warming clouds. The unsustained (<1 h) or undefined transitional stages that connect these life-cycle stages are excluded. The smoothed and sustained evolution is then used to separate stages for each MCS event. Growth rate is quantified on the basis of the smoothed cloud-shield area during the isolated growth stage. Growth duration is defined as the time length of the growth stage. Other MCS properties are calculated from original cloud-shield and precipitation parameters extracted during each life-cycle stage.
Example of (a) the low-pass-filtering process applied to cloud properties for the purposes of identifying life-cycle stages and (b) identification of stages after smoothing for the same case. Each point in (b) is separated by 30 min.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
e. Statistical quantification
The Kolmogorov–Smirnov (KS) test is conducted for testing the significance of differences between observed and simulated cumulative distribution functions (CDFs). The smaller the p value calculated from the test is, the more likely it is that observed and simulated distributions do not come from the same distribution. A p value lesser than 5% (0.05) is chosen to signify a significant difference of CDFs. When the observed and simulated CDFs are similar, the relative difference of simulated to observed mean values (RMD) can still be large because mean values can be impacted by extremes. Thus, KS test and RMD are jointly utilized to quantify the model–observation differences during isolated MCS life-cycle stages.
3. Evaluation of the regional climate simulation
a. Total precipitation
Throughout the 6.5-month period, the simulated domain-average precipitation rate (red line in Fig. 4a) in d1 maintains a greater correlation (r = 0.83) with 3B42 (black line in Fig. 4a) than ERA-5 (r = 0.79). The spatial correlation coefficient between WRF accumulated precipitation (Fig. 4d) and 3B42 (Fig. 4b) is 0.87 (Fig. 4e), which is the same as that between ERA-5 (Fig. 4c) and 3B42 (Fig. 4f). However, the 2D precipitation bias maps show that WRF better reproduces the 3B42 spatial variability than ERA-5 apart from the northeastern portion of the domain. This is especially true over the SDC range in the west-central portion of the domain. This may be related to the SDC being poorly resolved by the coarser resolution model generating ERA-5 datasets and inadequate handling of MCSs by the ECMWF cumulus parameterization. Despite improvements, WRF simulated rainfall exhibits a southeastern biased spatial shift as compared with observations (Figs. 4e,f). This may be related to a bias in MCS placement as found in U.S. studies (e.g., Peters et al. 2017; Vertz et al. 2021) and bias related to MCSs advecting across domain boundaries that cannot be reproduced by the simulation.
(a) Domain-average precipitation rate (every 3 h) over the entire 6.5-month period, and (b)–(g) 6.5-month accumulated precipitation within WRF and ERA-5 as compared with TRMM 3B42. Correlation coefficient r values are also shown. The 1- and 3-km levels of terrain height are shown in gray, highlighting the Andes and SDC range centered near 65°W and 32°S. Simulated rainfall maps are regridded to observational grids for comparison.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
b. Environmental conditions
To further evaluate the quality of the WRF simulation near the SDC in the middle of the domain far from boundary condition influence, simulated lifted parcel parameters (Fig. 5) at the AMF site on the eastern slope of the SDC are compared with the same parameters calculated from AMF soundings at the radiosonde launch times. As shown in Figs. 5a and 5b, throughout the 6.5-month period, the model reasonably reproduces the observed evolution of most unstable convective available potential energy (MUCAPE; r = 0.77) and most unstable convective inhibition energy (MUCIN; r = 0.75) that are computed for an undilute air parcel rising from the most unstable lifted parcel layer (LPL) to the equilibrium level (EL) where its buoyancy is neutralized. The LPL (r = 0.67) and EL (r = 0.79) heights are also reasonably correlated (Fig. 5c), which indicates that the model captures the general evolution of environmental instability over this long time period. The level of free convection (LFC) height (r = 0.43) is less well represented than the other parameters (Fig. 5d), which is likely a result of inversion layers being insufficiently resolved with limited model vertical levels despite stacking of vertical levels below 5-km altitude. This is an unavoidable problem for a simulation of this length due to computing constraints.
Simulated lifted parcel parameters at the CACTI AMF site as compared with radiosonde observations over the entire CACTI field campaign.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
4. Evaluation of simulated MCS characteristics
a. MCS bulk statistics
There are 363 observed and 343 simulated tracked MCSs initiating in domain d1, whose initial cloud-shield areas are required to be less than 10 000 km2. MCSs initiating to the north of domain d1 that propagate southward into d1 are excluded in observations and the simulation. Figure 6 shows that there are slight differences between simulated and observed MCS numbers, lifetimes, and accumulated hours. Simulated MCS numbers and accumulated hours peak in January 2019 like observed (Figs. 6a,c), but the WRF simulation (343) produces slightly fewer MCSs than observed (363). Simulated MCS lifetimes (Fig. 6b) generally agree well with those observed with average MCS lifetimes that are very comparable (11.2 vs 10.8 h). Total identified growth-stage hours in percentage of total MCS hours (Fig. 6d) are a bit less in WRF than observations but the proportioning of time between stage is similar in WRF and observations. One caveat is that most MCSs propagate outside of domain d1 before they decay, which will mostly limit mature and decay stage hours. Focusing on the better characterized growth stage, its diurnal cycle is quite similar between the model and observations (Fig. 6e), peaking strongly in the late afternoon with a minimum in the morning (local time is UTC − 3 h). Upscale growth occurs into the nighttime hours supporting previous studies showing a nighttime maximum of large MCSs attributed to the LLJ transporting moisture into the region, much like the nocturnal LLJ–MCS relationship over the U.S. Great Plains (Maddox 1983; Cotton et al. 1989; Laing and Fritsch 1997; Houze 2004; Salio et al. 2007).
Comparison of observed and simulated MCS (a) numbers by month, (b) lifetimes, (c) accumulated hours by month, (d) hours by life-cycle stage, and (e) growth-stage diurnal cycles.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
WRF generally captures the spatial variability of MCS cloud shields and their associated rainfall with a significant increase in coverage from between the Andes and SDC to ~60°W (Fig. 7). However, accumulated MCS cloud-shield hours and rainfall in WRF are less than observed over most of the domain. Observed MCS cloud shields and MCS rainfall are particularly more likely to occur in the northeast portion of the domain, likely associated with aforementioned boundary condition limitations in WRF. We estimate MCS propagation direction and speed by using the 2D cross correlation of precipitation coverage between consecutive time steps (Feng et al. 2018). WRF generally captures the observed eastward propagation, but systematically exaggerates the propagation speed (Figs. 7a,b), especially over La Plata Basin (eastern part of domain d1). To better interpret these cumulative differences, we comprehensively compare distributions of a number of MCS cloud and rainfall properties as a function of the MCS life cycle in section 4b.
Observed and simulated MCS accumulated (a),(b) cold cloud-shield time (color shading), system propagation (vectors), and (c),(d) surface precipitation. The dashed contours show the coastal boundary of South America.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
b. MCS life cycles
WRF slightly underestimates the MCS cloud-shield equivalent diameter [D = 2(area/π)1/2; Fig. 8a] but significantly underestimates the equivalent diameter of precipitation under the MCS cloud shield (Fig. 8b) by ~20% (RMD). Rainfall value CDFs are not significantly different, but the simulated mean values are less than those observed by more than 20% (Fig. 8d). This may be caused by WRF struggling to reproduce observed MCSs that produce extremely large rainfall volume that can significantly impact mean values. Simulated cloud tops have a minimum IR Tb that is slightly colder than observed with a slightly larger cloud aspect ratio, but differences are not significant (Figs. 8c,g). WRF also reproduces the areal ratio of cold cloud area (TOA IR Tb < 225 K) to total cloud-shield area (i.e., the cold core ratio, Fig. 8e), but the contribution of simulated heavy rain (>10 mm h−1) to the total volumetric rain (i.e., the heavy rain ratio) is significantly overestimated (Fig. 8f) by 44%–51% (RMD). Last, simulated MCS propagation is significantly faster than those observed (Fig. 8h) by 15%–33% (RMD). Nearly all of these differences are consistent across all MCS life-cycle stages. In general, cloud-shield property differences are less than precipitation property differences. The most significant and largest difference is between the observed and simulated heavy rain ratio, in which WRF produces more heavy and less light rain rates than retrieved by IMERG.
Violin plots of observed (light red) and simulated (light blue) MCS characteristics for different life-cycle stages for (left) cloud-shield and (right) precipitation properties. Each side of the violin shows the probability density distribution smoothed by a kernel density estimator. Horizontal lines and circles represent mean and median values, respectively. Vertical black bars in the center of the violins represent the interquartile range (IQR), and SD signifies statistical significance at 95% using a KS test. Black values are the RMD (%) between observed and simulated mean values.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
Figure 9 shows the evolution of MCS properties as a function of normalized MCS lifetime. The simulation (light blue) slightly underestimates the cloud-shield size for most of the life cycle (Fig. 9a). The simulated minimum IR Tb is slightly colder than observed during the first 10% of the MCS life cycle (Fig. 9b), indicating faster deepening of the simulated deep convective cores than observed. Heavy rain ratio remains constant through much of the life cycle and then gradually decreases in the later life cycle, and WRF produces a more than 15% greater heavy rain ratio than observations (Fig. 9c). There are either no trends or slight changes in terms of propagation speed, and significant WRF overestimates of propagation speeds are present throughout most of the life cycle (Fig. 9d). The cause of faster propagation in WRF remain unclear and requires more investigation. Possible culprits are not properly simulating back-building processes, cold pool outflow, and/or MCS mergers with one another.
Evolution of observed (light red) and simulated (light blue) MCS properties selected from Fig. 8 as a function of normalized MCS lifetime. The solid red and blue lines represent mean values at each time. Shaded areas represent the IQR.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
MCS rainfall area equivalent diameter and volume (Figs. 9e,f) follow the similar unimodal curve as cloud-shield diameter in Fig. 9a. WRF significantly underestimates the rainfall area equivalent diameter (Fig. 9e) but cloud-shield diameter (Fig. 9a) is only slightly underestimated. Rainfall volumes are better simulated than areas, but mean values are similar to 75th percentile values indicating a very skewed distribution that is more skewed for observations than the simulation (Fig. 9f).
Unlike past studies focusing on large, long-lived MCSs, the subhourly model output allows us to use less strict MCS lifetime thresholds to investigate potential bias dependencies on MCS lifetime. Figure 10 shows that the model underestimation of rainfall area equivalent diameter is not significant for short-lived (≤8.5 h) MCSs but is quite large for long-lived (>8.5 h) MCSs. In addition, WRF overestimates the rainfall volume for short-lived MCSs but underestimates it for long-lived MCSs. The reasons for these differences based on MCS lifetime are not clear and could result from a combination of WRF and satellite rainfall retrieval biases associated with more widespread stratiform rain and thick anvil regions in larger, longer-lived MCSs.
Evolution of observed (light red) and simulated (light blue) MCS rainfall (a),(b) area equivalent diameter and (c),(d) volume as a function of normalized lifetime of (left) short-lived and (right) long-lived MCSs divided by median lifetime (8.5 h).
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
c. MCS upscale growth
Differences in WRF-simulated and IMERG retrieved cloud-shield diameter, rainfall area equivalent diameter, and volumetric rainfall growth rate and growth hours are shown in Fig. 11 for the MCS upscale growth stages alone. Cloud-shield and volumetric rainfall growth rates (Figs. 11a,e) are similar in WRF (blue line) and satellite retrievals (red line), but the mean difference of volumetric rainfall growth (RMD = −16%) is greater than that for the cloud shield (RMD = −4%), possibly because WRF underestimates the probability of extremely rapid volumetric rainfall growth rates. The largest (RMD = −64%), most significant underestimation by WRF is rainfall area equivalent diameter growth rate (Fig. 11c) consistent with the underestimation of rainfall area already shown.
PDFs of mean growth rate and growth time of (a),(b) cloud-shield equivalent diameter; (c),(d) precipitating area equivalent diameter; and (e),(f) volumetric rainfall in observations (red line) and the simulation (blue line). Vertical dashed lines show mean values. Magenta and black characters show the KS significant difference and RMD, respectively.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
Simulated growth times of the cloud shield, rainfall area, and rainfall volume are shorter than those observed (Figs. 11b,d,f) with statistically significant RMDs from −7% to −11%. The simulated shorter growth time might be related to differences in simulated precipitation driven cold pools. For example, CPM simulations of a squall line case (Varble et al. 2020) suggest that under-resolved, overly large downdrafts too efficiently transport cool, dry midlevel air downward, biasing cold pool outflows. This could impact MCS upscale growth through altered cold pool interactions with environmental vertical wind shear that impact updraft strength and squall line maintenance (Rotunno et al. 1988; Weisman and Rotunno 2004). Microphysics parameterization may also bias MCS growth. Too efficient precipitation can result in less moisture detraining to stratiform regions and an underproduction of rainfall areas (Gilmore et al. 2004; Varble et al. 2014b). Particle size also plays a role by modulating cold pool intensity. For example, slower falling smaller hail increases melting and evaporation relative to larger hail, leading to stronger downdrafts and more intense cold pools (van den Heever and Cotton 2004). In addition, the number of predicted hydrometeor size distribution moments impacts cold pool size and intensity by altering evaporation and size sorting (e.g., Dawson et al. 2010).
Because MCS precipitation properties depend on upscale growth, we investigate whether the magnitude of differences between WRF and IMERG depends on several cloud-shield growth metrics. We separate MCSs into two groups using median values shown in parentheses as (observed median, simulated median) as a divider for the following metrics: maximum equivalent diameter extent (253, 254 km), maximum depth (208, 207 K), upscale growth duration (4.5, 3.5 h), and upscale growth rate (mean growth rate during the MCS growth stage; 41, 40 km h−1). Figure 12 shows that rainfall area and volume are sensitive to MCS maximum areal extent, maximum vertical depth, growth duration, and growth rate of the MCS cloud shield.
Following definitions in Fig. 8, violin plots of observed (light red) and simulated (light blue) maximum MCS rainfall (a) area and (b) volumetric rainfall for MCSs categorized by size, depth, upscale growth sustenance, and growth speed.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
Differences in WRF and IMERG rainfall area and volume are greatest for large, deep, sustained-growth, and rapid-growth MCSs (Fig. 12). Simulated rainfall equivalent diameter is underestimated in all MCS categories. Differences vary from 23% to 29% for these systems, whereas differences remain between 6% and 16% for relatively small, shallow, slowly growing, or unsustained-growth systems. MCS maximum depth has the most significant impact on differences between WRF and IMERG rainfall areas (27% difference in equivalent diameter for deep systems as compared with only a 6% difference for shallow systems). Simulated volumetric rainfall is underestimated by 26%–28% for large and deep MCSs but slightly overestimated by 10%–13% for small and shallow MCSs. WRF-IMERG volumetric rainfall differences are less sensitive to MCS growth speed and sustenance than maximum size and depth because unsustained- and slow-growth MCS rainfall is underestimated in WRF by 11%–13%, although not as severely as sustained- and rapid-growth MCS rainfall (26%–29%). Whereas many differences in rainfall area are statistically significant, differences in volumetric rainfall are not because WRF exaggerates the heavy rainfall contribution, counteracting the negative effect of underestimated rainfall area. This is highlighted in the similar simulated and retrieved probability density functions (PDFs; Fig. 13c). The worsening performance of WRF as MCSs grow in area and depth, and as MCS growth rate and sustenance increase, indicates that there could be processes disproportionately affecting these MCSs that WRF particularly struggles to simulate. We speculate that large, long lived MCS may be more dominated by stratiform rainfall that accumulates over time. Because WRF struggles to produce stratiform rainfall, the bias is thus greatest for these MCSs. However, it is also possible that WRF has additional problems representing cold pools or mesoscale circulations that could be important to rainfall maintenance in large, long-lived MCSs.
Probabilities of (a) rain rates from all precipitating events and (b) rain rates only from MCS tracks, (c) contribution of MCS rain rates values to total MCS rainfall, (d) MCS total raining time from surface rain gauges (black), GPM IMERG (red), and the WRF simulation (blue).
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
d. Investigation of rain-rate differences
The underestimation of MCS rainfall area coupled with the overestimation of heavy rain rates and underestimation of light rain rates (Figs. 8 and 9) are likely a combination of model bias and an IMERG retrieval bias toward excessive coverage of relatively low rain rates. To evaluate this hypothesis, we compare WRF-simulated and IMERG-retrieved MCS rain rates with 30-min rain rates from 15 NCAR-RAL surface rain gauges (locations signified by red pluses in Fig. 1) during the 6.5-month RELAMPAGO-CACTI experimental period. Rain rates less than 1 mm h−1 are excluded because of rain gauge and IMERG retrieval uncertainties (Cui et al. 2020). Rain gauge rain rates are classified as MCS rain rates when the rain gauge location is covered by identified and tracked MCS cloud shields.
Figures 13a and 13b shows the probability distributions of rain rate contributed by all precipitating events and tracked MCSs, respectively. The probabilities of simulated rain rates from all precipitating events agree well with those measured by rain gauges and retrieved from IMERG (Fig. 13a). However, as shown in Fig. 13b, the 1–3 mm h−1 MCS rain rate probability is underestimated by WRF (blue line) and overestimated by IMERG (red line) as compared with rain gauge estimates (black line). IMERG and WRF both slightly underestimate 3–5 mm h−1 rain-rate probabilities. For rain rates > 5 mm h−1, IMERG and rain gauge retrievals agree but WRF overestimates probabilities. Figure 13c confirms hypothesized WRF and IMERG biases. Relative to rain gauge estimated rain rates, contribution of rain rates between 1 and 3 mm h−1 to total MCS rainfall is overestimated in IMERG. Contribution of rain rates between 3 and 9 mm h−1 to MCS rainfall are underestimated in both WRF and IMERG but more severely in WRF. WRF significantly overestimates the contribution of rain rates exceeding 9 mm h−1 to MCS rainfall while IMERG agrees well with rain gauges over these relatively heavier rain rate ranges. Despite these probability differences, the MCS total raining time between 1 and 9 mm h−1 is much better estimated in WRF and significantly overestimated in IMERG (Fig. 13d).
Overall, the WRF rain rate probability distribution is similar to rain gauges and IMERG. WRF agrees better with rain gauges than IMERG in the amount of time that it is raining from MCSs, but IMERG better reproduces the probability distribution of MCS rain rates and their contribution to total MCS rainfall from rain gauges than WRF. The underproduction of light rainfall and overproduction of heavy rainfall in the simulation may be related to overly strong simulated updrafts (e.g., as in Varble et al. 2011, 2014a; Fan et al. 2016) coupled with biased microphysical parameterization that causes the overproduction of heavily rimed ice (Stanford et al. 2017; Varble et al. 2014a,b, 2020). Heavily rimed ice precipitates more efficiently than vapor grown ice, potentially leaving less moisture and ice to be detrained into stratiform anvil regions. In addition, overly strong or large updrafts may detrain higher in the troposphere such that much of the ice detrained sublimates before reaching lower altitudes (e.g., as in Varble et al. 2014b). These factors may cause insufficient light-moderate stratiform rainfall and areal coverage in the simulation and are the focus of future research that is beyond the scope of this study.
5. Meteorological impacts on MCS upscale growth
Although some precipitation biases exist in WRF-simulated MCSs, previous sections show that WRF reproduces MCS cloud-shield evolution well and reasonably shifts distributions of precipitation as a function of maximum MCS size and growth characteristics. Hence, it is reasonable to suspect that the 6.5-month, 3-km horizontal grid spacing WRF simulation can be used to examine environmental controls on MCS growth speed and sustenance. 52 simulated MCSs (trajectories shown in Fig. 14a) that initiate over and downstream of the SDC (500 km × 500 km magenta rectangle in Fig. 14, i.e., domain d2) are matched with ambient environmental conditions that are derived from hourly model output at the nearest time preceding the MCS initiation. Similar to section 4, these MCSs are divided into two groups by medians of cloud-shield growth rate (46 km h−1), growth duration (4.5 h), and maximum cloud-shield extent (271 km). Mean meteorological conditions are then computed for each group.
WRF-simulated meteorological conditions averaged at the initiation time of (left) slow-growth and (right) rapid-growth MCSs (genesis locations and trajectories are shown in the magenta filled circles and lines, respectively). Fields shown are (a),(b) MUCIN (color shading) and MUCAPE (contours at 1000, 1300, and 1600 J kg−1); (c),(d) 850-hPa water vapor mass flux (color shading) and speed of horizontal wind vector (contours at 7, 9, and 10 m s−1); (e),(f) sea level pressure (color shading), 850-hPa temperature (contours every 1 K between 290 and 294 K), and horizontal wind vectors; (g),(h) 500-hPa vertical winds (color shading) and 200-hPa horizontal wind vectors and speed (contours at 40, 42, and 44 m s−1); and (i),(j) 0–3-km vertical wind shear (color shading) and 850-hPa horizontal moisture convergence (contours at 0.2, 3, and 8 × 10−7 s−1). Thicker contours indicate larger absolute values. The magenta-outlined rectangle shows the region selected for MCS initiation locations.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
As compared with slow-growth MCSs, rapid-growth MCSs are associated with thermodynamically more favorable conditions for deep convection. On average, their environments have greater MUCAPEs over the northern SDC range and northward (Figs. 14a,b). MUCINs for both groups of MCSs are minimized in the locations where MCSs form. Rapid-growth MCSs are also associated with warmer temperatures, stronger northerly winds, and greater moisture transport at 850 hPa (Figs. 14c–f) in and to the north of the initiation region. This greater large-scale moisture and heat transport in rapid-growth cases results from greater absolute humidity and temperature in addition to stronger LLJ winds along and east of the Andes.
Several variables highlight strong synoptic controls on the varying thermodynamic conditions between slow- and rapid-growth MCSs. The low pressure center to the northwest of the SDC range in the lee of the Andes (i.e., the northwestern Argentinean low) is on average ~5 hPa lower in rapid-growth than slow-growth situations (Figs. 14e,f), producing a greater pressure gradient that is consistent with the stronger northerly low-level winds and transport of heat and moisture that produce greater convective instability. Rapid-growth MCSs have warmer 850-hPa temperatures over and to the north of the MCS genesis region than slow-growth MCSs with similar relatively colder temperatures to the south, which produces greater baroclinicity in rapid-growth situations (Figs. 14e,f). The 850-hPa isotherms are clearly distorted by the SDC range with flow that is deflected around the high terrain, indicating a potentially important role for the complex terrain in modulating MCS initiation locations. Many of the initiating MCSs concentrate over or south of the SDC range. Greater low-level baroclinicity is also correlated with a stronger upper-level jet (ULJ) and greater 500-hPa upward vertical motion (Figs. 14g,h) for rapid-growth cases at the time of MCS initiation consistent with Salio et al. (2007), although regions of enhanced ascent are also associated with regions of enhanced low-level baroclinicity that is partly controlled by SDC–LLJ–cold front interactions.
When the northerly LLJ encounters high terrain impeding its progress, it splits into western and eastern components on each side of the SDC range (Figs. 14c–f). The northeasterly winds with an upslope component result in moisture convergence along the eastern side of the SDC (Figs. 14j,i). At the southern end of the SDC range, low-level winds cyclonically rotate and converge with the northerly flow to the west of the SDC range, potentially supporting some MCSs that form in this area. Because of the enhanced LLJ, rapid-growth MCSs are associated with slightly enhanced moisture convergence to the west and east of the SDC (Figs. 14j,i). However, the peak moisture convergence locations are not the primary locations for generating MCSs, which instead are associated with the lowest-MUCIN- value regions, which indicates that CIN removal processes may be critical for dictating where an MCS will form. However, once an MCS forms, there is potential for greater low-level moisture convergence in the rapid-growth group as MCS cold pools hinder the southward progression of the stronger northerly winds. Rapid-growth cases also experience greater 0–3 km vertical wind shear to the south and east of SDC mountains (Figs. 14j,i). This wind shear increase from slow- to rapid-growth cases is more significant than that found in Coniglio et al. (2010) analyses over U.S. Great Plains.
Similar to the difference between rapid- and slow-growth cases, sustained-growth MCSs are associated with greater MUCAPE, moisture and heat transport via the LLJ, and baroclinicity with a deeper northwestern Argentinean low than unsustained-growth MCSs (Figs. 15a–f). MCS initiation locations are correlated with relatively low MUCIN locations, where they are mostly limited to the SDC range for unsustained-growth cases (Fig. 15a), indicating the importance of orographic forcing in these seemingly meteorologically less favorable situations. However, sustained growth corresponds to weaker 200-hPa ULJ winds, lesser 0–3-km vertical wind shear south and east of the SDC range, and similar magnitudes of 500-hPa vertical motion as compared with unsustained growth (Figs. 15g–j). This suggests that growth sustenance may be more controlled by processes at low levels than upper levels. More research is needed to ascertain whether these differences in vertical wind shear and ULJ winds between rapid- and sustained-growth cases are robust and whether cold pools differ such that interactions with vertical wind shear may impact growth sustenance (e.g., as in Rotunno et al. 1988; Weisman and Rotunno 2004).
Meteorological conditions as in Fig. 14, averaged at the initiation time of (left) unsustained-growth and (right) sustained-growth MCSs.
Citation: Monthly Weather Review 149, 10; 10.1175/MWR-D-20-0411.1
The combined effects of growth rate (Fig. 14) and sustenance (Fig. 15) jointly dictate the MCS maximum extent, where large and small MCSs are separated by maximum cloud-shield extent. Interestingly, 92% of large MCSs experience sustained growth while only 50% of them have rapid growth. This indicates that growth sustenance rather than the growth rate may predominantly control the maximum size of the MCSs that initiate over and near the SDC range.
6. Conclusions
Accurate prediction of MCSs depends on accurate representation of upscale growth, which is a collection of processes that produce an MCS from isolated deep convection. Our study objectives are to evaluate how well a regional model using grid spacing typical of regional convection-permitting weather and climate forecast models could reproduce MCS characteristics during the upscale growth life-cycle stage and investigate critical environmental factors that impact deep convective growth rate.
A 6.5-month simulation over subtropical South America was conducted using the WRF model at 3-km grid spacing covering the warm season from October 2018 to April 2019 during the RELAMPAGO-CACTI field campaign. The FLEXTRKR storm tracking algorithm (Feng et al. 2021) was applied consistently to observations and the simulation to obtain observed and simulated MCS tracks with identified growth, mature, and decay stages for which a range of MCS characteristics were analyzed.
Results show that observed and simulated MCS numbers and lifetimes are similar to one another, showing that convection-permitting WRF simulations can reproduce some aspects of MCSs well in subtropical South America like they can in other geographical regions (e.g., Cai and Dumais 2015; Griffin et al. 2017; Prein et al. 2020; Feng et al. 2018). We also show that the successes of WRF even extend to relatively small and short-lived MCSs, and different MCS life-cycle stages, that have not been the focus of previous studies. Simulated cloud-shield area, minimum cloud top temperature, growth rate, cold core areal ratio, and aspect ratio agree well with those observed regardless of life-cycle stage. The MCS growth-stage diurnal cycle is also well reproduced by the simulation. However, WRF produces shorter cloud-shield growth durations and faster cooling of the cloud top minimum IR Tb, indicating potentially faster vertical deepening and evolution of the simulated deep convective region than observed early in the MCS life cycle. These evolutionary differences may be related to overly intense simulated updrafts at kilometer-scale grid spacing due to insufficient buoyancy dilution by entrainment, as suggested in some earlier studies (e.g., Varble et al. 2014a). The simulated MCS propagation speed is also faster than that observed, possibly driven by overly strong, under-resolved convective downdrafts and cold pools as highlighted in Varble et al. (2020).
While WRF with 3-km horizontal grid spacing generally simulates MCS cloud properties well, it significantly underestimates the MCS rainfall area, to a lesser degree underestimates rainfall volume, and significantly overestimates 30-min rain rates. These results highlight similar stratiform underproduction over subtropical South America to that indicated in past case studies (Luo et al. 2010; Tao et al. 2016; Varble et al. 2011, 2014b, 2020) and seasonal assessments (Hagos et al. 2014; Feng et al. 2018) over other regions. We show that rainfall area becomes significantly underestimated during the growth stage, and thus, the early MCS life cycle should be a focus for further study into causes for this bias. In addition, these rainfall differences during growth stages are sensitive to MCS horizontal extent, depth, and lifetime. Small, shallow, and short-lived MCSs do not exhibit large IMERG-WRF differences in rainfall area, and produce more rainfall volume in WRF, while large, deep, and long-lived MCSs produce much smaller rainfall areas and slightly lesser rainfall volume in WRF. This may be related to large, deep, and long-lived MCSs potentially containing a greater fraction of rainfall coming from stratiform precipitation that WRF may particularly struggle to simulate. However, as demonstrated through comparison with 15 distributed rain gauges, these differences are a combination of model bias and a bias in the IMERG dataset used for satellite-retrieved rainfall. IMERG reasonably reproduces the rain gauge MCS 30-min rain rate probability distribution but rains much too frequently, which agrees with a previous study focusing on Brazil (Freitas et al. 2020). WRF, on the other hand, produces too few light and too many heavy rain rates as compared with rain gauge estimates. Simulated rainfall area and volume growth time lengths are also shorter than that observed. Similar to early life-cycle TOA IR Tb differences, these rainfall differences may be related to overly intense convective drafts, but also may be coupled with biased microphysical parameterizations that precipitate too efficiently because of excessive riming and/or detrainment too high in the troposphere that inhibits efficient formation of stratiform regions (e.g., Varble et al. 2014a,b; Stanford et al. 2017).
Simulated MCS cloud-shield growth and ambient environmental condition statistics that are unbiased and reasonably correlated with observations lend confidence to examination of critical environmental factors impacting the MCS growth rate and sustenance using the WRF simulation. The growth rate of simulated MCSs near the SDC range is strongly controlled by synoptic features such as synoptic-scale lift, baroclinicity, the LLJ, and the surface northwestern Argentinean low in the lee of the Andes to the northwest of the SDC. The growth rate accelerates as the surface low pressure to the northwest of the MCS initiation locations deepens. The northwestern Argentinean low increases LLJ northerly winds that amplify moisture and heat transport into the region. The enhanced low-level moisture and heat flux coupled with westerly free tropospheric flow descending in the lee of the Andes significantly increases CAPE while CIN inhibits initiation of deep convection over the entire region outside of the SDC vicinity. These synoptic conditions supporting rapid growth agree with those that support large, intense MCSs in subtropical South America studied previously (Salio et al. 2007; Rasmussen and Houze 2011, 2016), but here we show that they also exist at more muted amplitudes for more ordinary, smaller MCSs.
Unlike analyses and reanalyses used in past studies, this CPM simulation better resolves the influence of complex terrain on key fields including CIN, wind shear, and moisture convergence. The relatively lower CIN values over the SDC range and southeastward potentially relates to the combined positioning of low-level baroclinicity, synoptic ascent, orographic ascent, and LLJ-SDC interactions. MCSs initiate in this region of relatively low MUCIN near and south of the SDC rather than in regions of maximized MUCAPE, low-level moisture transport, or low-level moisture convergence, although these fields are important for distinguishing growth rate. Sustained-growth MCSs have similar environments as rapid-growth MCSs, including surface low, baroclinicity, LLJ, and instability. Unlike rapid growth, sustained growth is less sensitive to ULJ, synoptic-scale ascent or low-level vertical shear at initiation time. Although the maximum size of MCSs is a product of both growth rate and its sustenance, the latter is more controlling in this region.
MCS biases and environmental analyses in 3-km grid spacing simulations are extremely relevant given the push of weather and climate models toward this scale. Our evaluation of this 3-km-grid-spacing simulation indicates that convection-permitting weather prediction and ESMs will likely be capable of predicting subtropical and midlatitude continental MCS numbers, lifetimes, cloud-shield extents, aspect ratios, and temperatures, as well as properties of growth, mature, and decay life-cycle stages given sufficient physics parameterizations. This will be critical for predicting high impact weather as well as the hydrologic cycle and radiative budget in these regions. While better simulated than in current ESMs with cumulus parameterizations, MCS rainfall area and volume will potentially be underestimated with exaggerated heavy rainfall contributions that impact regional hydrologic cycles and latent heating distributions. As highlighted in previous studies, this is likely a result of under-resolved convective circulations (e.g., Bryan and Morrison 2012; Lebo and Morrisson 2015; Varble et al. 2020) and biased microphysics parameterization (e.g., Stanford et al. 2017; Feng et al. 2018).
Nevertheless, how model resolution coupled with imperfect parameterizations affect deep convective upscale growth processes as a function of environmental conditions remains unclear. Updraft core properties may nonmonotonically vary with a decrease of model grid spacing. Bryan and Morrison (2012) show that convective updrafts increase in strength moving from 4- to 1-km grid spacing but decrease in strength moving from 1-km to 250-m grid spacing, possibly because of decreasing opposing pressure gradients and increasing buoyancy dilution via entrainment as grid spacing decreases (e.g., Lebo and Morrison 2015). Under-resolved, overly wide updrafts at 1–4-km grid spacing may too efficiently precipitate and not sufficiently detrain condensate to stratiform regions with feedbacks to mesoscale circulations (e.g., Varble et al. 2020), which may contribute to MCS rain rate biases shown in this study. This will be a focus of future research.
Acknowledgments
The study was funded by NSF Project 1661662. Additional support was provided by the U.S. Department of Energy Office of Science Biological and Environmental Research as part of the Atmospheric System Research program. Pacific Northwest National Laboratory is operated by Battelle for the U.S. Department of Energy under Contract DE-AC05-76RLO1830. Computing and disk storage resources were provided by the NCAR Computational and Information Systems Laboratory (CISL) and the University of Utah Center for High Performance Computing (CHPC). Some data analysis was also performed using computational resources provided by the National Energy Research Scientific Computing Center (NERSC), a DOE Office of Science User Facility supported by the Office of Science of the U.S. Department of Energy under Contract DE-AC02-05CH11231. We thank Bastian Bechtold for sharing code to generate the violin plots. We also thank three anonymous reviewers for providing insightful suggestions to improve the paper.
Data availability statement
Geostationary satellite MERG-IR data from NOAA/NCEP were obtained through a NASA repository (https://doi.org/10.5067/P4HZB9N27EKU). Satellite-retrieved surface rainfall data were obtained from the NASA GPM precipitation data repository (https://doi.org/10.5067/GPM/IMERG/3B-HH/06). Surface rain gauge measurements from the NCAR Research Applications Laboratory are available through the RELAMPAGO data repository hosted by the NCAR Earth Observing Laboratory (https://data.eol.ucar.edu/master_lists/generated/relampago). Relevant radiosondes are available through the Atmospheric Radiation Measurement (ARM) archive (https://adc.arm.gov/discovery/). Raw model output is available by contacting the authors.
REFERENCES
Benjamin, S. G., and Coauthors, 2016: A North American hourly assimilation and model forecast cycle: The Rapid Refresh. Mon. Wea. Rev., 144, 1669–1694, https://doi.org/10.1175/MWR-D-15-0242.1.
Bryan, G. H., and H. Morrison, 2012: Sensitivity of a simulated squall line to horizontal resolution and parameterization of microphysics. Mon. Wea. Rev., 140, 202–225, https://doi.org/10.1175/MWR-D-11-00046.1.
Bukovsky, M. S., and D. J. Karoly, 2009: Precipitation simulations using WRF as a nested regional climate model. J. Appl. Meteor. Climatol., 48, 2152–2159, https://doi.org/10.1175/2009JAMC2186.1.
Cai, H., and R. E. Dumais, 2015: Object-based evaluation of a numerical weather prediction model’s performance through forecast storm characteristic analysis. Wea. Forecasting, 30, 1451–1468, https://doi.org/10.1175/WAF-D-15-0008.1.
Caine, S., T. P. Lane, P. T. May, C. Jakob, S. T. Siems, M. J. Manton, and J. Pinto, 2013: Statistical assessment of tropical convection-permitting model simulations using a cell-tracking algorithm. Mon. Wea. Rev., 141, 557–581, https://doi.org/10.1175/MWR-D-11-00274.1.
Chan, S. C., E. J. Kendon, H. J. Fowler, S. Blenkinsop, N. M. Roberts, and C. A. T. Ferro, 2014: The value of high-resolution Met Office regional climate models in the simulation of multihourly precipitation extremes. J. Climate, 27, 6155–6174, https://doi.org/10.1175/JCLI-D-13-00723.1.
Chen, S. S., and W. M. Frank, 1993: A numerical study of the genesis of extratropical convective mesovortices. Part I: Evolution and dynamics. J. Atmos. Sci., 50, 2401–2426, https://doi.org/10.1175/1520-0469(1993)050<2401:ANSOTG>2.0.CO;2.
Clarke, S. J., S. L. Gray, and N. M. Roberts, 2019: Downstream influence of mesoscale convective systems. Part 1: Influence on forecast evolution. Quart. J. Roy. Meteor. Soc., 145, 2933–2952, https://doi.org/10.1002/qj.3593.
Coniglio, M. C., J. Y. Hwang, and D. J. Stensrud, 2010: Environmental factors in the upscale growth and longevity of MCSs derived from Rapid Update Cycle analyses. Mon. Wea. Rev., 138, 3514–3539, https://doi.org/10.1175/2010MWR3233.1.
Copernicus Climate Change Service, 2017: ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. ECMWF, Reading, United Kingdom, Copernicus Climate Change Serv. Climate Data Store (CDS), accessed 23 May 2019, https://cds.climate.copernicus.eu/cdsapp#!/home.
Cotton, W. R., M. Lin, R. L. McAnelly, and C. J. Tremback, 1989: A composite model of mesoscale convective complexes. Mon. Wea. Rev., 117, 765–783, https://doi.org/10.1175/1520-0493(1989)117<0765:ACMOMC>2.0.CO;2.
Crook, N. A., and M. W. Moncrieff, 1988: The effect of large-scale convergence on the generation and maintenance of deep moist convection. J. Atmos. Sci., 45, 3606–3624, https://doi.org/10.1175/1520-0469(1988)045<3606:TEOLSC>2.0.CO;2.
Crumeyrolle, S., L. Gomes, P. Tulet, A. Matsuki, A. Schwarzenboeck, and K. Crahan, 2008: Increase of the aerosol hygroscopicity by cloud processing in a mesoscale convective system: A case study from the AMMA campaign. Atmos. Chem. Phys., 8, 6907–6924, https://doi.org/10.5194/acp-8-6907-2008.
Cui, W., X. Dong, B. Xi, Z. H. E. Feng, and J. Fan, 2020: Can the GPM IMERG final product accurately represent MCSs’ precipitation characteristics over the central and eastern United States? J. Hydrometeor., 21, 39–57, https://doi.org/10.1175/JHM-D-19-0123.1.
Dawson, D. T., M. Xue, J. A. Milbrandt, and M. K. Yau, 2010: Comparison of evaporation and cold pool development between single-moment and multimoment bulk microphysics schemes in idealized simulations of tornadic thunderstorms. Mon. Wea. Rev., 138, 1152–1171, https://doi.org/10.1175/2009MWR2956.1.
Fan, J., Y. Wang, D. Rosenfeld, and X. Liu, 2016: Review of aerosol–cloud interactions: Mechanisms, significance, and challenges. J. Atmos. Sci., 73, 4221–4252, https://doi.org/10.1175/JAS-D-16-0037.1.
Feng, Z., X. Dong, B. Xi, C. Schumacher, P. Minnis, and M. Khaiyer, 2011: Top-of-atmosphere radiation budget of convective core/stratiform rain and anvil clouds from deep convective systems. J. Geophys. Res., 116, D23202, https://doi.org/10.1029/2011JD016451.
Feng, Z., S. Hagos, A. K. Rowe, C. D. Burleyson, M. N. Martini, and S. P. De Szoeke, 2015: Mechanisms of convective cloud organization by cold pools over tropical warm ocean during the AMIE/DYNAMO field campaign. J. Adv. Model. Earth Syst., 7, 357–381, https://doi.org/10.1002/2014MS000384.
Feng, Z., L. R. Leung, S. Hagos, R. A. Houze, C. D. Burleyson, and K. Balaguru, 2016: More frequent intense and long-lived storms dominate the springtime trend in central U.S. rainfall. Nat. Commun., 7, 13429, https://doi.org/10.1038/ncomms13429.
Feng, Z., L. R. Leung, R. A. Houze, S. Hagos, J. Hardin, Q. Yang, B. Han, and J. Fan, 2018: Structure and evolution of mesoscale convective systems: Sensitivity to cloud microphysics in convection-permitting simulations over the United States. J. Adv. Model. Earth Syst., 10, 1470–1494, https://doi.org/10.1029/2018MS001305.
Feng, Z., R. A. Houze, L. R. Leung, F. Song, J. C. Hardin, J. Wang, W. I. Gustafson, and C. R. Homeyer, 2019: Spatiotemporal characteristics and large-scale environments of mesoscale convective systems east of the Rocky Mountains. J. Climate, 32, 7303–7328, https://doi.org/10.1175/JCLI-D-19-0137.1.
Feng, Z., L. R. Leung, N. Liu, J. Wang, R. A. Houze, J. Li, J. C. Hardin, and J. Guo, 2021: A global high-resolution mesoscale convective system database using satellite-derived cloud tops, surface precipitation, and tracking. J. Geophys. Res. Atmos., 126, e2020JD034202, https://doi.org/10.1029/2020JD034202.
Freitas, E. da S., and Coauthors, 2020: The performance of the IMERG satellite-based product in identifying sub-daily rainfall events and their properties. J. Hydrol., 589, 125128, https://doi.org/10.1016/j.jhydrol.2020.125128.
Fritsch, J. M., R. J. Kane, and C. R. Chelius, 1986: The contribution of mesoscale convective weather systems to the warm-season precipitation in the United States. J. Climate Appl. Meteor., 25, 1333–1345, https://doi.org/10.1175/1520-0450(1986)025<1333:TCOMCW>2.0.CO;2.
Futyan, J. M., and A. D. Del Genio, 2007: Deep convective system evolution over Africa and the tropical Atlantic. J. Climate, 20, 5041–5060, https://doi.org/10.1175/JCLI4297.1.
Gallus, W. A., N. A. Snook, and E. V. Johnson, 2008: Spring and summer severe weather reports over the midwest as a function of convective mode: A preliminary study. Wea. Forecasting, 23, 101–113, https://doi.org/10.1175/2007WAF2006120.1.
Gilmore, M. S., J. M. Straka, and E. N. Rasmussen, 2004: Precipitation uncertainty due to variations in precipitation particle parameters within a simple microphysics scheme. Mon. Wea. Rev., 132, 2610–2627, https://doi.org/10.1175/MWR2810.1.
Gochis, D., and Coauthors, 2019: NCAR-RAL surface hydrometeorological observation network data, version 1.0. UCAR/NCAR–Earth Observing Laboratory, accessed 4 August 2020, https://doi.org/10.26023/KW8Z-F2WX-H0Y.
Griffin, S. M., J. A. Otkin, C. M. Rozoff, J. M. Sieglaff, L. M. Cronce, C. R. Alexander, T. L. Jensen, and J. K. Wolff, 2017: Seasonal analysis of cloud objects in the High-Resolution Rapid Refresh (HRRR) model using object-based verification. J. Appl. Meteor. Climatol., 56, 2317–2334, https://doi.org/10.1175/JAMC-D-17-0004.1.
Hagos, S., Z. Feng, C. D. Burleyson, K. S. S. Lim, C. N. Long, D. Wu, and G. Thompson, 2014: Evaluation of convection-permitting model simulations of cloud populations associated with the Madden–Julian oscillation using data collected during the AMIE/DYNAMO field campaign. J. Geophys. Res. Atmos., 119, 12–52, https://doi.org/10.1002/2014JD022143.
Hartmann, D. L., L. A. Moy, and Q. Fu, 2001: Tropical convection and the energy balance at the top of the atmosphere. J. Climate, 14, 4495–4511, https://doi.org/10.1175/1520-0442(2001)014<4495:TCATEB>2.0.CO;2.
Holdridge, D., E. Keeler, and J. Kyrouac, 2018: Balloon-Borne Sounding System (SONDEWNPN). Atmospheric Radiation Measurement (ARM) user facility, accessed 14 October 2019, https://doi.org/10.5439/1021460.
Houze, R. A., 1993: Cloud Dynamics. Academic Press, 573 pp.
Houze, R. A., 2004: Mesoscale convective systems. Rev. Geophys., 42, RG4003, https://doi.org/10.1029/2004RG000150.
Huffman, G. J., and Coauthors, 2007: The TRMM Multisatellite Precipitation Analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeor., 8, 38–55, https://doi.org/10.1175/JHM560.1.
Huffman, G. J., andCoauthors, 2018: NASA Global Precipitation Measurement (GPM) Integrated Multi-satellitE Retrievals for GPM (IMERG). NASA Algorithm Theoretical Basis Doc., version 5.2, 35 pp., https://docserver.gesdisc.eosdis.nasa.gov/public/project/GPM/IMERG_ATBD_V5.pdf.
Johns, R. H., and C. A. Doswell, 1992: Severe local storms forecasting. Wea. Forecasting, 7, 588–612, https://doi.org/10.1175/1520-0434(1992)007<0588:SLSF>2.0.CO;2.
Kendon, E. J., N. M. Roberts, H. J. Fowler, M. J. Roberts, S. C. Chan, and C. A. Senior, 2014: Heavier summer downpours with climate change revealed by weather forecast resolution model. Nat. Climate Change, 4, 570–576, https://doi.org/10.1038/nclimate2258.
Kooperman, G. J., M. S. Pritchard, and R. C. J. Somerville, 2015: The response of U.S. summer rainfall to quadrupled CO2 climate change in conventional and superparameterized versions of the NCAR community atmosphere model. J. Adv. Model. Earth Syst., 6, 859–882, https://doi.org/10.1002/2014MS000306.
Laing, A. G., and J. M. Fritsch, 1997: The global population of mesoscale convective complexes. Quart. J. Roy. Meteor. Soc., 123, 389–405, https://doi.org/10.1002/qj.49712353807.
Lebo, Z. J., and H. Morrison, 2015: Effects of horizontal and vertical grid spacing on mixing in simulated squall lines and implications for convective strength and structure. Mon. Wea. Rev., 143, 4355–4375, https://doi.org/10.1175/MWR-D-15-0154.1.
Lin, B., B. A. Wielicki, L. H. Chambers, Y. Hu, and K. M. Xu, 2002: The Iris hypothesis: A negative or positive cloud feedback? J. Climate, 15, 3–7, https://doi.org/10.1175/1520-0442(2002)015<0003:TIHANO>2.0.CO;2.
Lin, Y., W. Dong, M. Zhang, Y. Xie, W. Xue, J. Huang, and Y. Luo, 2017: Causes of model dry and warm bias over central U.S. and impact on climate projections. Nat. Commun., 8, 881, https://doi.org/10.1038/s41467-017-01040-2.
Lindzen, R. S., M. D. Chou, and A. Y. Hou, 2001: Does the Earth have an adaptive infrared iris? Bull. Amer. Meteor. Soc., 82, 417–432, https://doi.org/10.1175/1520-0477(2001)082<0417:DTEHAA>2.3.CO;2.
Liu, C., S. Shige, Y. N. Takayabu, and E. Zipser, 2015: Latent heating contribution from precipitation systems with different sizes, depths, and intensities in the tropics. J. Climate, 28, 186–203, https://doi.org/10.1175/JCLI-D-14-00370.1.
Luo, Y., Y. Wang, H. Wang, Y. Zheng, and H. Morrison, 2010: Modeling convective-stratiform precipitation processes on a mei-yu front with the Weather Research and Forecasting model: Comparison with observations and sensitivity to cloud microphysics parameterizations. J. Geophys. Res., 115, D18117, https://doi.org/10.1029/2010JD013873.
Maddox, R. A., 1983: Large-scale meteorological conditions associated with midlatitude, mesoscale convective complexes. Mon. Wea. Rev., 111, 1475–1493, https://doi.org/10.1175/1520-0493(1983)111<1475:LSMCAW>2.0.CO;2.
Mulholland, J. P., S. W. Nesbitt, R. J. Trapp, K. L. Rasmussen, and P. V. Salio, 2018: Convective storm life cycle and environments near the Sierras de Córdoba, Argentina. Mon. Wea. Rev., 146, 2541–2557, https://doi.org/10.1175/MWR-D-18-0081.1.
Mulholland, J. P., S. W. Nesbitt, and R. J. Trapp, 2019: A case study of terrain influences on upscale convective growth of a supercell. Mon. Wea. Rev., 147, 4305–4324, https://doi.org/10.1175/MWR-D-19-0099.1.
Nakanishi, M., and H. Niino, 2006: An improved Mellor-Yamada Level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor., 119, 397–407, https://doi.org/10.1007/s10546-005-9030-8.
Nakanishi, M., and H. Niino, 2009: Development of an improved turbulence closure model for the atmospheric boundary layer. J. Meteor. Soc. Japan, 87, 895–912, https://doi.org/10.2151/jmsj.87.895.
Nesbitt, S. W., R. Cifelli, and S. A. Rutledge, 2006: Storm morphology and rainfall characteristics of TRMM precipitation features. Mon. Wea. Rev., 134, 2702–2721, https://doi.org/10.1175/MWR3200.1.
Peters, J. M., E. R. Nielsen, M. D. Parker, S. M. Hitchcock, and R. S. Schumacher, 2017: The impact of low-level moisture errors on model forecasts of an MCS observed during PECAN. Mon. Wea. Rev., 145, 3599–3624, https://doi.org/10.1175/MWR-D-16-0296.1.
Pinto, J. O., J. A. Grim, and M. Steiner, 2015: Assessment of the high-resolution Rapid Refresh model’s ability to predict mesoscale convective systems using object-based evaluation. Wea. Forecasting, 30, 892–913, https://doi.org/10.1175/WAF-D-14-00118.1.
Prein, A. F., A. Gobiet, M. Suklitsch, H. Truhetz, N. K. Awan, K. Keuler, and G. Georgievski, 2013: Added value of convection permitting seasonal simulations. Climate Dyn., 41, 2655–2677, https://doi.org/10.1007/s00382-013-1744-6.
Prein, A. F., C. Liu, K. Ikeda, R. Bullock, R. M. Rasmussen, G. J. Holland, and M. Clark, 2020: Simulating North American mesoscale convective systems with a convection-permitting climate model. Climate Dyn., 55, 95–110, https://doi.org/10.1007/s00382-017-3993-2.
Rasmussen, K. L., and R. A. Houze, 2011: Orogenic convection in subtropical South America as seen by the TRMM satellite. Mon. Wea. Rev., 139, 2399–2420, https://doi.org/10.1175/MWR-D-10-05006.1.
Rasmussen, K. L., and R. A. Houze, 2016: Convective initiation near the Andes in subtropical South America. Mon. Wea. Rev., 144, 2351–2374, https://doi.org/10.1175/MWR-D-15-0058.1.
Rotunno, R., J. B. Klemp, and M. L. Weisman, 1988: A theory for strong, long-lived squall lines. J. Atmos. Sci., 45, 463–485, https://doi.org/10.1175/1520-0469(1988)045<0463:ATFSLL>2.0.CO;2.
Salio, P., M. Nicolini, and E. J. Zipser, 2007: Mesoscale convective systems over southeastern South America and their relationship with the South American low-level jet. Mon. Wea. Rev., 135, 1290–1309, https://doi.org/10.1175/MWR3305.1.
Schlemmer, L., and C. Hohenegger, 2014: The formation of wider and deeper clouds as a result of cold-pool dynamics. J. Atmos. Sci., 71, 2842–2858, https://doi.org/10.1175/JAS-D-13-0170.1.
Schumacher, C., R. A. Houze, and I. Kraucunas, 2004: The tropical dynamical response to latent heating estimates derived from the TRMM precipitation radar. J. Atmos. Sci., 61, 1341–1358, https://doi.org/10.1175/1520-0469(2004)061<1341:TTDRTL>2.0.CO;2.
Schumacher, R. S., and K. L. Rasmussen, 2020: The formation, character and changing nature of mesoscale convective systems. Nat. Rev. Earth Environ., 1, 300–314, https://doi.org/10.1038/s43017-020-0057-7.
Skamarock, W. C., and Coauthors, 2019: A description of the Advanced Research WRF Model version 4. NCAR Tech. Note NCAR/TN-556+STR, 145 pp., https://doi.org/10.5065/1dfh-6p97.
Smith, B. T., R. L. Thompson, J. S. Grams, C. Broyles, and H. E. Brooks, 2012: Convective modes for significant severe thunderstorms in the contiguous United States. Part I: Storm classification and climatology. Wea. Forecasting, 27, 1114–1135, https://doi.org/10.1175/WAF-D-11-00115.1.
Stanford, M. W., A. Varble, E. Zipser, J. W. Strapp, D. Leroy, A. Schwarzenboeck, R. Potts, and A. Protat, 2017: A ubiquitous ice size bias in simulations of tropical deep convection. Atmos. Chem. Phys., 17, 9599–9621, https://doi.org/10.5194/acp-17-9599-2017.
Stevens, B., and Coauthors, 2019: DYAMOND: The DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains. Prog. Earth Planet. Sci., 6, 61, https://doi.org/10.1186/s40645-019-0304-z.
Takahashi, H., and Z. J. Luo, 2014: Characterizing tropical overshooting deep convection from joint analysis of CloudSat and geostationary satellite observations. J. Geophys. Res. Atmos., 119, 112–121, https://doi.org/10.1002/2013JD020972.
Tao, W. K., and J. Simpson, 1984: Cloud interactions and merging: Numerical simulations. J. Atmos. Sci., 41, 2901–2917, https://doi.org/10.1175/1520-0469(1984)041<2901:CIAMNS>2.0.CO;2.
Tao, W. K., D. Wu, S. Lang, J. D. Chern, C. Peters-Lidard, A. Fridlind, and T. Matsui, 2016: High-resolution NU-WRF simulations of a deep convective-precipitation system during MC3E: Further improvements and comparisons between Goddard microphysics schemes and observations. J. Geophys. Res. Atmos., 121, 1278–1305, https://doi.org/10.1002/2015JD023986.
Thompson, G., and T. Eidhammer, 2014: A study of aerosol impacts on clouds and precipitation development in a large winter cyclone. J. Atmos. Sci., 71, 3636–3658, https://doi.org/10.1175/JAS-D-13-0305.1.
Tompkins, A. M., 2001: Organization of tropical convection in low vertical wind shears: The role of cold pools. J. Atmos. Sci., 58, 529–545, https://doi.org/10.1175/1520-0469(2001)058<0529:OOTCIL>2.0.CO;2.
Torri, G., Z. Kuang, and Y. Tian, 2015: Mechanisms for convection triggering by cold pools. Geophys. Res. Lett., 42, 1943–1950, https://doi.org/10.1002/2015GL063227.
van den Heever, S. C., and W. R. Cotton, 2004: The impact of hail size on simulated supercell storms. J. Atmos. Sci., 61, 1596–1609, https://doi.org/10.1175/1520-0469(2004)061<1596:TIOHSO>2.0.CO;2.
Varble, A., and Coauthors, 2011: Evaluation of cloud-resolving model intercomparison simulations using TWP-ICE observations: Precipitation and cloud structure. J. Geophys. Res., 116, D12206, https://doi.org/10.1029/2010JD015180.
Varble, A., and Coauthors, 2014a: Evaluation of cloud-resolving and limited area model intercomparison simulations using TWP-ICE observations: 1. Deep convective updraft properties. J. Geophys. Res. Atmos., 119, 13 891–13 918, https://doi.org/10.1002/2013JD021371.
Varble, A., and Coauthors, 2014b: Evaluation of cloud-resolving and limited area model intercomparison simulations using TWP-ICE observations: 2. Precipitation microphysics. J. Geophys. Res. Atmos., 119, 13 919–13 945, https://doi.org/10.1002/2013JD021372.
Varble, A., H. Morrison, and E. Zipser, 2020: Effects of under-resolved convective dynamics on the evolution of a squall line. Mon. Wea. Rev., 148, 289–311, https://doi.org/10.1175/MWR-D-19-0187.1.
Vertz, N. J., W. A. Gallus, and B. J. Squitieri, 2021: Can pre-storm errors in the low-level inflow help predict spatial displacement errors in MCS initiation? Atmosphere, 12, 7, https://doi.org/10.3390/atmos12010007.
Weisman, M. L., and R. Rotunno, 2004: “A theory for strong long-lived squall lines” revisited. J. Atmos. Sci., 61, 361–382, https://doi.org/10.1175/1520-0469(2004)061<0361:ATFSLS>2.0.CO;2.
Yang, G.-Y., and J. Slingo, 2001: The diurnal cycle in the tropics. Mon. Wea. Rev., 129, 784–801, https://doi.org/10.1175/1520-0493(2001)129<0784:TDCITT>2.0.CO;2.
Yang, Q., R. A. Houze, L. R. Leung, and Z. Feng, 2017: Environments of long-lived mesoscale convective systems over the central United States in convection permitting climate simulations. J. Geophys. Res. Atmos., 122, 13–288, https://doi.org/10.1002/2017JD027033.
Zipser, E., D. Cecil, C. Liu, S. Nesbitt, and D. Yorty, 2006: Where are the most intense thunderstorms on Earth? Bull. Amer. Meteor. Soc., 87, 1057–1072, https://doi.org/10.1175/BAMS-87-8-1057.
