Near‐Storm Environmental Relationships With Tropical Oceanic Convective Structure Observed During NASA CPEX and CPEX‐AW

Deep tropical oceanic convection (TOC) is a prevailing component of the tropical atmosphere and plays a significant role in modulating global weather and climate. Despite its importance, prediction challenges remain, partly attributed to a lack of understanding of how TOC relates to its near‐storm environments. Prior studies suggest location‐dependent relationships between TOC structure and associated environments, necessitating targeted regional studies. The NASA 2017 Convective Processes Experiment (CPEX) and 2021 CPEX—Aerosols & Winds (CPEX‐AW) field campaigns collected high‐resolution measurements of convective storms and their environments in the Gulf of Mexico, Caribbean, and western Atlantic basins, providing a rare opportunity to investigate near‐storm environmental relationships with 3‐D TOC structure where in situ non‐tropical cyclone‐related deep TOC research is comparatively lacking. Collocated CPEX and CPEX‐AW airborne observations from the multi‐wavelength Airborne Precipitation Radar, Doppler Aerosol Wind Lidar, and dropsondes revealed large near‐storm environmental variability across TOC of similar convective type (i.e., isolated, organized) and within individual convective systems. However, trends still emerged amongst the large environmental variability. Horizontal TOC structure was most consistently linked to planetary boundary layer and mid‐tropospheric near‐storm environments, with organized TOC being associated with generally greater relative humidity (RH) and vertical speed shear than isolated TOC. TOC intensity was linked to upper tropospheric (i.e., above melting level) near‐storm environments, with isolated TOC intensity most consistently associated with upper tropospheric CAPE and organized TOC intensity associated with upper tropospheric RH. Mesoscale low‐level convergence was also linked to greater organized TOC intensity, motivating further research using these unique data sets.

x 15 Same data as Figure

Introduction
Oceanic deep convection is a prevalent feature of the tropical atmosphere and plays a key role in driving both regional and global weather and climate. It highly influences the large-scale tropical atmospheric circulation through its redistribution of mass, momentum, heat, and moisture, while the precipitation, through buoyancy fluxes, instigates upper ocean responses (Alexander and Young, 1992;Brown and Zhang, 1997;LeMone et al., 1998;Saxen and Rutledge, 2000;Tompkins, 2001;Cetrone and Houze, 2006;Liu and Lian, 2010). Frequent, widespread threedimensional (3-D) convective cloud structures significantly impact radiative fluxes, and thus enhanced warming/cooling, particularly near cloud-top levels (LeMone et al., 1998;Igel and van den Heever, 2015). Deep tropical oceanic convection (TOC) also frequently produces thermodynamically driven cold pools, which can trigger development of non-precipitating cumulus congestus clouds (a further radiative influencer), initiate new deep convection, alter planetary boundary layer (PBL) characteristics, and modify air-sea exchange (Chandra et al., 2018;Houze, 2018;Touzé-Peiffer et al., 2021).
With TOC modulating tropical weather and climate, accurate TOC representation and parameterization is critical for the success of weather and climate modeling efforts. However, despite decades of TOC research, challenges persist in modeling TOC. This challenge is partly attributed to a lack of understanding of TOC structure, initiation, and evolution, including how TOC relates to its near-storm environments throughout storm system lifecycle (Cetrone and Houze, 2006;Igel and van den Heever, 2015;Minamide and Posselt, 2022). Studying near-storm environmental relationships is challenging, as it requires frequent, high-resolution measurements to capture fundamental small-scale convective processes, features, and environments. In situ data collection is the ideal means of garnering such measurements, and it is historically too few and far in between.
Observational analysis from field campaigns in the western Pacific basin (e.g., TOGA-COARE, KWAJEX, EMEX, PISTON) has sought to investigate near-storm environmental wind shear relationships with TOC structure using in situ dropsonde and radiosonde deployments. The west Pacific analyses have linked greater low-tropospheric speed shear to a greater degree of spatial organization of western Pacific mesoscale convective systems (MCSs) (Figure 1a), with quasilinear convective systems oriented perpendicular to the low-level shear vector (Alexander and Young, 1992;LeMone et al., 1998;Chudler and Rutledge, 2021). Guy and Jorgensen (2014) found differing results in the Indian Ocean during the DYNAMO field campaign, with quasi-linear MCSs oriented more parallel to the low-tropospheric shear.
Strong deep-layer vertical speed shear in the western Pacific has been linked to larger, more intense linear precipitating systems (i.e., squall lines), owing to more expansive stratiform precipitation and anvil advection (Saxen and Rutledge, 2000). Research efforts using cloud-resolving model (CRM), spaceborne remote sensing, and European Centre for Medium-Range Weather Forecasts Reanalysis-Interim (ERA-Interim) reanalysis datasets have extended the regional, western Pacific observational analysis to the broad oceanic tropics and found similar deep-layer speed shear relationships with tropical squall lines (Tompkins, 2001) and tropical MCSs (Igel and van den Heever, 2015) in general ( Figure 2). However, Tompkins (2001) acknowledges CRM limitations (e.g., limited vertical dimension and unrealistic cyclic boundary conditions), while Igel and van den Heever (2015) acknowledges biases stemming from inherently inconsistent storm-relative CloudSat measurements.
While the aforementioned studies have made great strides towards understanding near-storm wind shear relationships with TOC structure, the research is predominately centered on (quasi-) linear convection. Non-linear TOC is abundant in the tropics (e.g., Houze et al., 2015) and warrants more targeted investigations. Additionally, the aforementioned studies are largely based in the western Pacific and Indian Ocean. With TOC structure varying across the tropics (Houze et al., 2015), an evaluation of TOC relationships with near-storm wind shear necessitates detailed in situ studies in other tropical oceanic regions.
Field campaign data in the western Pacific and Indian Ocean basins, along with broader tropical oceanic CRM and ERA-Interim reanalysis data, have also been used to investigate near-storm environmental moisture relationships with TOC structure. Analysis of these datasets have shown mid-tropospheric relative humidity (RH) to positively correlate with TOC precipitation area and intensity due to decreased dry air entrainment (Brown and Zhang, 1997;LeMone et al., 1998;Tompkins, 2001;Cetrone and Houze, 2006;Chen et al., 2017). Airborne data collected during DYNAMO over the Indian Ocean additionally showed lesser near-storm, mid-tropospheric moisture associated with isolated TOC to correspond to deeper convectively generated cold pools and, consequently, longer PBL recovery times (Savarin et al., 2014;Chen et al., 2016). However, the relationships between lower-tropospheric moisture and TOC precipitation area and intensity vary, in both strength and sign, across tropical oceanic studies, even within similar regions (Tompkins, 2001;Cetrone and Houze, 2006;Chen et al., 2017). The inconsistencies may relate to data collection during differing TOC lifecycle stages, with low-level inflow shown to potentially be more important during early lifecycle stages compared to mid-level inflow for later lifecycle stages (Mechem et al., 2002).
Similarly, inconsistent relationships between environmental CAPE and TOC structure exist across prior studies of the western Pacific basin. Cetrone and Houze (2006) and Chudler and Rutledge (2021) found KWAJEX and PISTON MCSs, respectively, to be associated with lesser CAPE compared to smaller, less organized convective systems (Figure 1b), while Kingsmill and Houze (1999) found opposing results using TOGA-COARE data. The conflicting results could stem from thermodynamic instability being asymmetrically concentrated in the lower troposphere in the KWAJEX soundings and near-surface modification by MCSs in the PISTON soundings, thereby negatively biasing CAPE measurements near the MCSs (Cetrone and Houze, 2006;Chudler and Rutledge, 2021).
Collectively, both the strength and sign of near-storm mean-layer environmental moisture, wind shear, and CAPE relationships with TOC structure are inconsistent across studies. These inconsistent findings could be attributed to a multitude of factors, like differing analysis methods and data sources. However, a lack of regional context may be a major culprit. Most modern studies of environmental relationships with non-tropical cyclone related TOC structure are focused in the western Pacific or broadly across the oceanic tropics. This lack of modern, regionally diverse, and regionally distinct research is an issue because TOC structure and its relationships with near-storm environments have been shown through TRMM observations and ERA-Interim reanalysis to exhibit regional dependencies ( Figure 3) (Houze et al., 2015;Chen et al., 2017). Synoptic-scale climatological differences, such as sea surface temperature distribution, jet presence, and MJO influences, likely play a role in these regional dependencies by influencing TOC initiation, organization, and microphysics. Therefore, evaluations of TOC structure relationships with nearstorm environments necessitate targeted regional studies, particularly in understudied areas, so that the impacts of region-relative enhanced/diminished mean-layer moisture and wind shear can be assessed (Cetrone and Houze, 2006). Satellite and reanalysis datasets provide the ability to examine each tropical oceanic region separately, but their limited spatial and temporal resolutions cannot sufficiently capture essential small-scale near-storm environmental variability and convective processes. Consequently, targeted regional in situ studies equipped with collocated high-resolution hydrometeor, moisture, and wind measurement capabilities are imperative to adequately analyze near-storm environmental relationships with TOC structure.
Two such targeted in situ regional studies were the 2017 NASA Convective Processes Experiment Caribbean Sea, and western Atlantic-regions that were notably lacking recent in situ, non-tropical cyclone related deep convective research (Cui et al., 2020). The DC-8 aircraft was equipped with, amongst other instrumentation, a multi-wavelength airborne precipitation radar, a Doppler wind lidar, and dropsondes ( Figure 5). Together, these instruments provided rare, coincident, high-resolution profiling of 3-D convective structure, near-cloud winds, and near-storm moisture for convective systems of different spatial scales and intensities (Turk et al., 2020;Hristova-Veleva et al., 2021). Given the uniqueness of this suite of observations, the CPEX and CPEX-AW field campaigns present an exceptional opportunity to analyze their region's near-storm environmental relationships with tropical oceanic 3-D convective structure, which will be the focus of this study.

In particular, how does 3-D TOC structure relate to near-storm environmental RH, vertical speed shear, and CAPE in different tropospheric vertical layers in the CPEX(-AW) observational domain?
Prior studies suggest the strength and sign of these environmental relationships may be layerdependent. The three aforementioned DC-8 instruments' high vertical resolutions provide sufficiently numerous observations in any given tropospheric layer to adequately assess multiple tropospheric layers' environmental characteristics. We hypothesize that larger, more intense (i.e., more organized) TOC will be associated with greater near-storm mean RH in each defined tropospheric layer, due to less dry air entrainment into the convective system (i.e., less evaporative cooling, limiting cold pool development, and less negative buoyancy introduction) (Brown and Zhang, 1997;Savarin et al., 2014;Chen et al., 2016). Similar to western Pacific field campaign analysis, more organized, multi-core convection in the CPEX(-AW) domain is also hypothesized to be connected to greater near-storm mean vertical speed shear in each defined tropospheric layer, due to more effective separation of convective updrafts and downdrafts and continued initiation along cold pool boundaries (Yuter and Houze, 1995;Houze, 2018). Conflicting arguments in the literature make it difficult to predict how CAPE will relate to 3-D TOC structure. Some studies link lesser CAPE to more organized convection (Cetrone and Houze, 2006;Chudler and Rutledge, 2021), while other studies link greater CAPE to more organized convection (e.g., Kingsmill and Houze, 1999). Meanwhile, an argument exists that CAPE does not correlate with TOC structure at all, due to typically small CAPE variations within a given tropical oceanic region for a given time of year (Lucas et al., 1994). Therefore, it is unclear what we expect with regards to CAPE relationships with 3-D TOC structure in the CPEX(-AW) domain, and it will be interesting to evaluate the results.
To address the research question and evaluate the associated hypotheses, the organization of the paper is as follows. Section 2 offers a description of the CPEX(-AW) instrumentation and data used for the analysis. Section 3 outlines the methods used to filter and contextualize the observational data, along with the approaches and techniques used to analyze 3-D TOC structure and its near-storm environments. Section 4 introduces and discusses the results of the analysis of near-storm, mean-layer environmental relationships with 3-D TOC structure in the CPEX(-AW) region, first by comparing convective systems with notably different 3-D structure (e.g., singlecore vs. multi-core) and then by comparing convective systems with similar 3-D structure (e.g., multi-core vs. multi-core). Section 5 provides a discussion of the results in the context of prior field campaign studies from different tropical oceanic regions. Section 6 concludes the paper with main takeaways from the analysis and next steps for future research.     . This paper will focus on analysis of the higher spatial resolution airborne datasets from the following instrumentation: dropsondes, the Doppler Aerosol WiNd Lidar (DAWN), and the Third-Generation Airborne Precipitation Radar (APR-3). Together, these three instruments provided coincident, detailed measurements of near-storm moisture, winds, and 3-D convective structure (e.g., Figure 6) at sufficient resolutions to analyze characteristics of distinct vertical layers.
Seventeen of the 23 CPEX(-AW) science flights sampled 20 separate convective systems (hereafter referred to as convective cases; Figure 4) with this instrument payload. As such, only observational data from these 17 science flights were used for analysis for this paper.

b) Dropsondes
Vertical profiles of pressure, temperature, horizontal wind velocity, and humidity were collected dropsondes sampled pressure at a rate of 2-Hz, accuracy of 1.5 mb (at 25°C), and resolution of 2.5 mb (Greco et al., 2018;Black et al., 2017). Temperature was collected at a 2-Hz rate with an accuracy of 0.148°C and resolution of 0.0168°C (Greco et al., 2018;Black et al., 2017). Horizontal wind velocity was sampled at a rate of 4-Hz, accuracy of 0.5 m s −1 , and precision of 0.2 m s −1 (Greco et al., 2018;Black et al., 2017). RH was collected at a 2-Hz rate, with an accuracy of 1.8% (at 25°C) and resolution of 0.1% (Greco et al., 2018;Black et al., 2017). RH was collected with an accuracy of 3% and resolution of 0.01% (AVAPS Dropsondes, 2023).
Given the scales that are being explored for comparing mean-layer wind shear and RH (differences hypothesized to be several m s −1 and several percent, respectively), the CPEX(-AW) dropsonde accuracies and precisions for each metric are sufficient for the purposes of this study. RH accuracies were notably worse in actively precipitating environments as the dropsondes exhibited moisture biases when encountering precipitation, likely due to water ingress (Greco et al., 2018).
As such, moisture data from dropsondes deployed in actively precipitating environments was excluded from analysis in this paper. The dropsonde data was processed using the NCAR Atmospheric Sounding Processing ENvironment (ASPEN) software, which, amongst other quality control practices, removes outlier data and smooths profiles accordingly (Greco et al., 2018;Vömel et al., 2021;Martin and Suhr, 2021). Post-mission GPS correction was also employed (CPEX Dropsonde, 2019).
Data for which the dropsonde fall speed did not exceed 15 m s −1 was filtered out, and usage of hydrostatic altitude, when available, was prioritized over GPS altitude due to greater height accuracy (CPEX Dropsonde, 2019). Temperature, relative humidity, dewpoint, and both component wind values all had to be present for the dropsonde data to be included at a given height level. Dropsonde profiles with frequent, graphically visible anomalous spikes in equivalent potential temperature were excluded from further analysis. In total, this filtering amounted to 195 usable dropsondes across the 20 convective cases (Figure 4).

c) Doppler Aerosol WiNd Lidar (DAWN)
High-resolution vertical profiles of wind near convection were collected by the DAWN instrument aboard the NASA DC-8 aircraft during CPEX(-AW) ( Figure 5; Kavaya et al., 2014;Greco et al., 2020). DAWN is equipped with a 2-m, 10-Hz laser that utilizes atmospheric aerosols to measure horizontal wind components (Kavaya et al., 2014;Turk et al., 2020;Greco et al., 2020). The laser scanned conically 30° off nadir, collecting two 66 m resolution line-of-sight (LOS) wind profiles at -45° and 45° azimuth angles relative to the flight direction (CPEX DAWN, 2019; Turk et al., 2020). DAWN vertical wind profiles were obtained at horizontal resolutions as fine as 3-7 km and a vertical resolution of ~33 m using the LOS wind profiles (CPEX DAWN, 2019; Greco et al., 2020). DAWN profiles were severely attenuated when encountering opaque clouds (e.g., convective anvil cirrus), and data gaps frequently existed in the mid-troposphere due to low aerosol concentrations, as DAWN relies upon sufficient aerosol presence to measure wind velocity (Bedka et al., 2021). DAWN wind speed accuracy was < 0.05 m s −1 , while wind speed precision was < ~1.5 m s −1 (Greco et al., 2020). DAWN winds also showed a low bias of < 0.20 m s −1 with the dropsonde winds (Greco et al., 2020). Given the scales that are being explored for comparing mean-layer wind shear (differences hypothesized to be several m s −1 ), the DAWN wind accuracy and precision are adequate for the purposes of this study.
DAWN data was processed via methods described in Kavaya et al. (2014) and Greco et al. (2020).
In layers of low signal-to-noise ratio (SNR), the Adaptive Signal Integration Algorithm (ASIA) was employed to enhance vertical profile coverage (CPEX DAWN, 2019; Greco et al., 2020;Bedka et al., 2021). Wind data collected during aircraft banking that exceeded 1.5° was excluded, as was wind data sampled during aircraft ascents/descents greater than 7 m s −1 (CPEX DAWN, 2019). Both component wind values had to be present for the DAWN data to be included at a given height level (defined as above mean sea level). The lowest available integration index-with the lowest possible integration index being the baseline profile data and higher integration indices increasingly incorporating ASIA-for each height level was used, as suggested by CPEX DAWN (2019).

d) Third-Generation Airborne Precipitation Radar (APR-3)
Vertical radar reflectivity profiles of 3-D convective hydrometeor structure were collected using the APR-3 instrument ( Only profiles from the pseudo-nadir ray of each APR-3 cross-track scan were utilized in subsequent analysis because these profiles captured the vertical convective structure directly below the aircraft and thus better aligned with DAWN and dropsonde profiles. The pseudo-nadir ray is defined as the APR-3 scan's ray for which the difference between the ray incidence angle and the aircraft roll angle is minimized. This method thereby selects the ray that vertically scans closest to directly below the aircraft. Furthermore, only Ku-band reflectivity profiles were used for analysis, as the Ku-band captures precipitation structure better than the Ka-band, which is more quickly attenuated by precipitating hydrometeors. Doppler velocity datasets were corrupted for a majority of CPEX convective cases (i.e., cases prior to 16 June 2017). Therefore, Doppler velocity was not incorporated in subsequent analysis.

a) Convective Case Characterization
In order to investigate near-storm environmental relationships with TOC structure, each of the 20 CPEX(-AW) convective cases was categorized as either isolated, organized, or scattered based on horizontal precipitation extent and continuity provided by archived hourly Integrated Multi-satellitE Retrievals for GPM (IMERG) satellite data. Given the inherent small number of cases from the field campaigns, categorization of each convective system was manual. Isolated convective systems were defined as horizontally small, cellular precipitating regions, while organized convective systems were defined as broader, continuous, multi-core precipitating regions. Scattered convection was defined as broad, discontinuous precipitating regions. An example of each type of convection in the context of IMERG is depicted in Figure 7. The focus of this paper is on isolated and organized non-tropical cyclone TOC cases. As such, 12 out of the 20 total convective cases sampled during CPEX(-AW) underwent further analysis (Table 1). Two scattered convective cases were omitted due to non-distinct cellular nor multi-core horizontal precipitation structures. Additionally, six organized convective cases were omitted due to association with tropical cyclones (including tropical depressions). All but one (i.e., Case 15) of the qualifying 12 cases was sampled during a similar time of day (i.e., between 1800 UTC and 0000 UTC), so diurnal influences on convective structure are assumed to have been similar for each case.

b) Dropsonde Characterization
Only dropsonde profiles temporally collocated with their respective qualifying convective case's APR-3 data (i.e., within or near the anvil region) were considered for analysis, amounting to 111 dropsondes. Each dropsonde from the cases identified in Table 1 was characterized by the convective type of its respective case, along with the convective-relative environment it was deployed into. Using APR-3 plots overlaid with dropsonde and DAWN wind profiles (e.g., Figure   6), the three environmental categories were "Clear" (little to no reflectivity overlaid with the profile), "In Cloud" (deployed through a non-precipitating cloud layer(s)), and "In Precip" (deployed through an actively precipitating region). For example, the dropsonde around 2032 UTC in Figure 6a was characterized as "Clear", while the dropsonde near 2045 UTC was categorized as "In Precip". Similarly, all three dropsondes in Figure 6b were labeled as "In Precip". Skew-T diagrams of the dropsonde profiles also aided in validation of the environmental categorization.
For example, the environment of the dropsonde near 2039 UTC in Figure 6a appears to be "Clear".
However, the dropsonde's skew-T diagram in Figure 8 denotes a deep saturated layer from 1000 mb -650 mb, indicative of an "In Precip" environment. The distribution of dropsondes amongst the three environmental categories for each case is shown in Table 1, along with the distribution of full and partial (i.e., sparse data coverage in certain layers) dropsonde profiles for each case.
Due to the notable moisture biases of "In Precip" dropsondes, as previously mentioned in Section 2, only wind data from "In Precip" dropsondes was used for analysis in this paper.
Each dropsonde was also characterized by convective lifecycle stage at the time of deployment (see Table 1), as near-storm environments can vary with the maturity of convective systems. The convective lifecycle stage for each dropsonde was deemed to be "growing", "mature", or "weakening" based on APR-3 reflectivity features near the dropsonde profile (e.g., stratiform brightband strength, presence of embedded convective elements) and convective core evolution

c) Mean-layer, Near-storm Environmental Metrics
Mean relative humidity (RH) and vertical wind speed shear were calculated for each dropsonde profile for four distinct layers: the PBL, mid layer, upper layer, and deep layer ( Figure 9). The base of the PBL is defined as the profile height nearest to the surface, ranging from 6.5 m to 338.5 m because of dropsonde transmission issues. Metric values were not found to correlate with nearsurface height within that range. The top of the PBL is defined as the first height for which the virtual potential temperature exceeds its value nearest to the surface by 0.5 °C (e.g., Blumberg et al., 2017 Environmental metrics for layers that were not fully sampled by partial dropsonde profiles (see Table 1) were excluded from analysis.
Similar to the dropsonde data, only DAWN profiles that were temporally collocated with their respective convective case's APR-3 data-and thus near convection-were included in analysis.
The DAWN instrument provided much denser wind profiling (both spatially and temporally) in

d) Dropsonde Median Ku-band Reflectivity Profiles
To analyze potential links between near-storm, mean-layer environmental parameters and convective intensity, median APR-3 Ku-band reflectivity profiles were constructed for each dropsonde and matched with their associated dropsonde mean-layer metrics. These median reflectivity profiles were calculated from the collection of all Ku-band profiles within a 10-min temporal proximity (20-min total time delta) to a given dropsonde. The 5-, 7.5-, and 12.5-min time deltas were also considered for the median reflectivity profile calculation, but the 10-min time delta was ultimately chosen, as it captured brightband signatures while minimizing noise ( Figure   10). APR-3 data below 1.5 km was excluded to avoid inclusion of potentially spurious near-surface data (Sadowy et al., 2003;Durden et al., 2003). The In Precip dropsondes were omitted from the median reflectivity profile analysis with near-storm mean-layer RH and CAPE due to known dropsonde moisture biases in actively precipitating regions during the CPEX(-AW) campaigns (Greco et al., 2018).

e) Contoured Frequency by Altitude Diagrams
For analysis of convective intensity and the reflectivity features of each convective case, all APR-3 Ku-band reflectivity profiles for a given case were binned into 2-dimensional histograms with 5-dBZ and 0.5-km intervals to create Contoured Frequency by Altitude Diagrams (CFADs) (Yuter and Houze, 1995). The reflectivity bins extend from -20 dBZ to 60 dBZ. The height bins extend from 1.5 km to 8 km, so as to omit potentially attenuated near-surface data (Sadowy et al., 2003;Durden et al., 2003) and provide a uniform upper layer altitude cap to allow for case intercomparison. The heights from the reflectivity profiles, and thus in each CFAD, were not adjusted for brightband height, as brightband height across all the convective cases did not vary considerably. Each CFAD was normalized by the maximum bin count in any height interval, allowing for frequency comparisons across height levels (e.g., Zagrodnik et al., 2019). Each CFAD was separately normalized by the maximum bin count in each height interval, allowing for easier frequency comparisons at a given height level. Subsequent so-called "difference CFADs" were produced by subtracting one case's normalized CFAD from another. These difference CFADs allow for investigation of convective intensity and storm structure differences between cases with distinct mean-layer environmental metric differences (Yuter and Houze, 1995

a) Near-storm Environmental Relationships with Convective Type
To analyze potential near-storm environmental relationships with TOC type (i.e., isolated versus organized), mean-layer dropsonde metrics are presented as box-and-whisker plots (e.g., Figure   11a). Excluding the questionable In Precip dropsondes (i.e., owing to moisture biases), similar median PBL depths are observed for both convective types, while the upper and lower PBL depth quartiles are deeper (i.e., lower pressures) for organized convection (Figure 11a).   Figure 12a) and MLCAPE (570 J kg -1 vs.
524 J kg -1 , Figure 12b). More distinctly, isolated convection is also observed to have greater median upper layer MUCAPE (613 J kg -1 vs. 549 J kg -1 , Figure 12c) and MLCAPE (393 J kg -1 vs. 317 J kg -1 , Figure 12d) compared to organized convection. Similar to PBL depth, large CAPE variability exists within each convective type, motivating further observational grouping by case ( Figure 13). The further detail provided by the case grouping points to no clear relationship between convective type and MUCAPE/MLCAPE in either layer. However, some cases-even of similar convective type-do have distinctly different CAPE magnitudes compared to others (e.g., isolated cases 1 and 3 with upper layer CAPE), which will be investigated more in Section 4c and Section 4d. vs. 53.0%, Figure 14d), while isolated convection is associated with a slightly greater median deep layer RH (66.1% vs. 64.5%, Figure 14a). Both isolated and organized convection have similar mid layer median RH (71.0% vs. 71.7%, Figure 14c), which regionally differs from aforementioned western Pacific studies that consistently link greater mid layer RH to more organized convection.
As with PBL depth and CAPE, large mean-layer RH spreads exist for both convective types, particularly for organized convection. Further grouping the dropsonde observations by case highlights large mean-layer RH variability within many cases, especially cases 1, 13, and 16 ( Figure 15). Cases 13 and 16, in particular, record some of the lowest mean-layer RH values of any case and will be investigated further in Section 4e. When comparing the clear air dropsondes for each case across convective type, deep layer RH is generally greater for isolated convection (Figure 15a), while PBL RH is generally greater for organized convection (Figure 15b). Distinct mean-layer RH differences between cases of similar convective type do exist, however, and further motivate the case comparisons in Section 4c and Section 4d.
Organized convection is observed to have greater median PBL (5.0 kts vs. 3.5 kts, Figure 16b), mid (17.7 kts vs. 8.6 kts, Figure 16c), and deep layer (22.0 kts vs. 16.0 kts, Figure 16a) speed shears compared to isolated convection. As with prior metrics, a large degree of mean-layer speed shear variability exists within each convective type (especially for organized convection), motivating further dropsonde grouping by case ( Figure 17). Large speed shear spreads are evident for each layer for each case and are notably unrelated to the environment the dropsondes were deployed into (i.e., Clear, In Cloud, or In Precip). Similar to CAPE and RH, distinct mean-layer speed shear differences do exist amongst cases of similar convective type, which further warrants comparing individual cases in Section 4c and Section 4d. Ultimately though, organized convection is associated with generally greater PBL, mid, and deep layer near-storm speed shear than isolated convection (Figure 17a,b,c).
Incorporating the spatially and temporally denser DAWN observations provides an enhanced look into deep layer speed shear trends ( Figure 18). The addition of the DAWN observations reveals even larger deep layer speed shear variability within each convective type ( Figure 18a) and each convective case (Figure 18c) than the dropsonde observations previously showed. Separating the convective cases by field campaign (Figure 18b, see also  Figure 18a), which the less frequent dropsonde data did not capture. This result is highly influenced by Case 15 from CPEX-AW, which recorded some of the greatest deep layer shear values in either campaign, exceeding 45 knots (Figure 18c). When comparing all isolated case speed shear observations with the organized cases, however, the relationship between deep layer speed shear and convective type is unclear (Figure 18c).
Overall, the composite analysis of near-storm environmental relationships with convective type reveals broad near-storm environmental metric variability within each convective type. However, general mean-layer CAPE, RH, and speed shear trends do exist with regards to two-dimensional (2-D) TOC structure. As such, it raises the question whether similar (or perhaps even more pronounced) near-storm environmental trends apply to vertical convective structure (i.e., convective intensity) for each convective type.

b) Near-storm Environmental Relationships with Convective Intensity (Composite Analysis)
Median Ku-band reflectivity profiles in combination with associated dropsonde metrics were used to analyze composite near-storm environmental relationships with convective intensity. Median Ku-band reflectivity profiles for each qualifying dropsonde (see Section 3d) were constructed, grouped by convective type, and color-coded by their respective dropsonde's mean-layer environmental metrics to discern possible environmental metric influence on vertical reflectivity (i.e., a proxy for hydrometeor size and convective intensity). Dropsondes lacking Ku-band reflectivity data within the utilized 10-minute time delta were omitted from this analysis. Alternate time deltas were also tested (see Section 3d) and produced similar results to the 10-min time delta shown in subsequent figures.
The greatest reflectivities below 7 km (i.e., near the upper layer cap of 7622.5 m) for isolated convective cases, which exceed 20 dBZ, are associated with greater deep layer (600 -800 J kg -1 ) and upper layer (400 -500 J kg -1 ) MLCAPE, though the relationships are non-monotonic ( Figure   19e,g). This relationship could result from enhanced thermodynamic instability promoting stronger updrafts, which can suspend hydrometeors for longer periods of time and support continued hydrometeor growth. However, the lack of parallel reflectivity relationships with MUCAPE casts doubt on this suggestion. For the organized cases, no clear mean-layer CAPE trends with vertical reflectivity profile strength exist (Figure 19b,d,f,h).
For the isolated convective dropsondes, the greatest reflectivities below 7 km are also associated with the greatest mid layer RH values (75% -80%), though the relationship is non-monotonic ( Figure 20e). This relationship could be due to enhanced hydrometeor growth and less negative buoyancy introduction from less dry air entrainment into the single-core systems. Similar to CAPE, no clear mean-layer RH trends with vertical reflectivity profile strength are apparent for organized convective cases (Figure 20b,d,f,h).
For isolated cases, the greatest reflectivities below 7 km are additionally associated with greater deep layer (20 -30 kts) and mid layer (10 -20 kts) speed shear (Figure 21a,e). The relationships remain non-monotonic but are more distinct than the aforementioned CAPE and RH trends. This relationship could result from more effective separation of the convective updraft and downdraft in single-core systems by larger speed shear environments. For the organized cases, the greatest reflectivities below 7 km (exceeding 30 dBZ) are associated with generally lesser speed shears in each of the four layers (Figure 21b,d,f,h). The relationships are non-monotonic and contrast with the isolated convection speed shear relationships, which may point to differing dynamical relationships between single-core systems and multi-core systems with their near-storm environments. interval. However, similar analysis was performed with CFADs normalized by maximum bin count in any height interval, and each method produced the same conclusions.
Isolated cases 1 and 2 had many distinct metric differences that motivated convective intensity comparison. Case 1 had notably greater PBL RH than Case 2, but lesser average deep layer RH  For organized convection, distinct metric differences between cases were predominantly with mean-layer speed shear. When comparing Case 7 and Case 16, Case 16 had notably lesser mid layer speed shear ( Figure 17c) and was the more intense convection ( Figure 25). Meanwhile, Case 13 had distinctly lesser mid layer speed shear compared to Case 7 ( Figure 17c) and was less intense than Case 7 (Figure 26). Case 7 and Case 8 had many distinct metric differences that motivated convective intensity comparison. Case 8 was the more intense convection (Figure 27) with notably lesser mid layer speed shear, but notably greater PBL speed shear, upper layer RH, and deep layer MUCAPE (Figure 17b,c, Figure 15d, Figure 13a). Comparing Case 7 to Case 5, Case 5 had typically lesser deep layer and mid layer speed shear (Figure 18c, Figure 17c) and was the more intense convection (Figure 28). Meanwhile, Case 7 had typically greater deep layer speed shear compared to Case 4 (Figure 18c), along with lesser PBL speed shear and greater PBL RH ( Figure   17b, Figure 15b). Differencing Case 7 and Case 4 Ku-band reflectivity frequencies revealed Case 7 to be more intense than Case 4 ( Figure 29). Lastly, when comparing Case 7 and Case 14 ( Figure   30), the more intense Case 14 generally had lesser deep layer and mid layer speed shear, but notably greater deep layer and upper layer RH (Figure 18c, Figure 17c, Figure 15a,d).
In total, when comparing CPEX(-AW) organized convective cases with distinct mean-layer environmental metric differences, no consistent mean-layer speed shear, MUCAPE, nor MLCAPE relationships with convective intensity are found. However, when a notable difference in upper layer RH existed between two organized cases, the more intense convection was consistently associated with greater upper layer RH. Similar to a postulation in Section 4b, this relationship could be due to less dry air entrainment promoting enhanced hydrometeor growth and limiting negative buoyancy introduction. When intercomparing the isolated case comparison results with the organized case comparison results, it is noteworthy that there are no similar, consistent nearstorm mean-layer metric trends with convective intensity. This lack of similarity between convective types further supports a notion from Section 4b that single-core and multi-core systems interact differently with their near-storm environments. An additional lack of consistent near-storm environmental relationships with organized TOC intensity suggests multi-core systems also variably interact with their near-storm environments. The latter finding is likely attributed to diverse vertical organizational structures of multi-core TOC, which were observed during CPEX(-AW) and will be explored further in the following section.

e) Case 13 vs. Case 16 Organized Convection Analysis
As highlighted in Section 4a, large mean-layer RH variability existed within each sampled convective system, but especially within Case 13 and Case 16. Upon further examination, cases 13 and 16 containing many of the lowest mean-layer RH values of any sampled CPEX(-AW) convective system was responsible for their particularly large RH spreads (Figure 15). The relatively dry near-storm observations, coupled with the especially large RH variability, motivated further investigation into these two cases' convective environments.
The convective flight legs and observations of cases 13 and 16 each encompassed a synoptic-scale horizontal moisture gradient (Figure 31), providing the reason for large intra-case RH spreads and notably dry near-storm observations that were particularly influential on the RH results of Section 4a. No other observed convective systems during CPEX(-AW) were located near synoptic-scale moisture gradients, making Case 13 and Case 16 unique in that aspect. Further exploring the two cases, both had similar 2-D structures ( Figure 31) with an intensifying sector (e.g., Figure 32) and a matured sector (e.g., Figure 33) identified by decreasing and increasing infrared brightness being more linearly organized than Case 16, consistent with studies of linearly organized TOC and low-level shear in the western Pacific basin (Alexander and Young, 1992;LeMone et al., 1998;Chudler and Rutledge, 2021). However, all the Case 16 PBL shear observations within the moisture gradient were collected in actively precipitating regions, which may have been influenced by cold pools and thus not directly comparable to the Case 13 observations in non-precipitating regions.
Observations collected within the moisture gradient unsurprisingly tended to have greater meanlayer RH compared to observations collected beyond the moisture gradient ( Figure 38). With Case 16 only having quality (i.e., not In Precip) RH observations beyond the moisture gradient, meanlayer RH comparisons were only performed amongst data collected from the impinging dry air beyond the moisture gradient associated with each convective system. The deep layer and mid layer were notably drier for Case 16 compared to Case 13 (Figure 38a,c), with the 800 -650 mb jet layer (located entirely within the mid layer) in particular having dewpoint depressions exceeding three times those of Case 13 (Figure 34). The distinctly differing near-storm environmental moisture of cases 13 and 16 motivated further analysis on how the near-storm environments may have influenced convective intensity.
When analyzing both case's convective intensities in relation to near-storm environmental metrics using CFADs, Case 16 had the more intense intensifying sector ( Figure 39) with lesser average mid layer RH and lesser average deep layer, PBL, and upper layer speed shear. Case 16 also had the more intense matured sector (Figure 40), but with similar mean-layer RH and speed shear to the Case 13 matured sector. All four sectors (i.e., intensifying and matured for both cases) and their associated mean-layer environmental metrics were also compared and ranked against one another. The Case 16 embedded convection had the two most intense sectors overall, yet it had similar mean-layer speed shear to the two less intense sectors (i.e., the Case 13 sectors). However, the Case 16 sectors had the two lowest average deep layer and mid layer RHs and the two highest average upper layer RHs.
Ultimately, Case 16 was more intense than Case 13 despite similar near-storm, mean-layer speed shear and distinctly drier impinging air in the deep layer and mid layer. However, dropsonde observations from the northern and southern halves of Case 16 reveal the presence of synopticscale low-level convergence (Figure 41), which was not observed in Case 13. This broad low-level convergence was likely the reason for Case 16 being notably more intense than Case 13, despite much drier impinging air. Overall, the Case 13 and Case 16 analysis suggests that sufficient largescale low-level convergence can provide enough forcing for TOC to thrive in otherwise seemingly less favorable near-storm environmental conditions, and it may also influence the vertical structure of organized convection (e.g., increasing the number and intensity of convective elements).

Discussion a) Near-storm Environmental Relationships with Convective Type
The analysis presented in Section 4a investigated near-storm environmental relationships with 2-D TOC structure in the CPEX(-AW) observational domain. While each mean-layer near-storm environmental metric varied widely within a given convective type, notable environmental trends with convective type emerged amongst the variability. Median PBL depth was similar for both isolated and organized TOC (Figure 11a). However, organized TOC was associated with deeper lower and upper quartile PBL depth (Figure 11a), potentially implying stronger cold pool presence that would more significantly alter pre-existing PBL characteristics (e.g., depth). With this assumption, the PBL depth results of this study regionally differ from DYNAMO studies in the Indian Ocean, wherein stronger cold pools (and subsequently longer PBL recovery times, a proxy for PBL depth) were generally associated with more isolated convection (Savarin et al., 2014;Chen et al., 2016;Chandra et al., 2018). However, Chandra et al. (2018) acknowledges the two strongest cold pools in their DYNAMO study were associated with more organized convection. Despite apparent regional differences in PBL depth relationships with convective type, the CPEX(-AW) observations link PBL depth with mid-level dryness (RPearson  0.537, Figure 42). This result is consistent with the DYNAMO studies that relate drier mid-levels (via enhanced dry air entrainment promoting stronger cold pools) to longer PBL recovery times, and it further highlights the importance of mid layer moisture on near-storm PBL modification (Savarin et al., 2014;Chen et al., 2016;Chandra et al., 2018).
Median MUCAPE and MLCAPE in each layer was found to be slightly greater for isolated TOC (Figure 12) Investigations of near-storm, mean-layer RH revealed deep layer RH to be generally lesser for organized TOC (Figure 14a, Figure 15a) and PBL RH to be generally greater for organized TOC ( Figure 14b, Figure 15b). The deep layer RH result was unexpected and conflicts with the prevailing idea that a drier tropical troposphere inhibits TOC development through enhanced dry air entrainment and negative buoyancy introduction. The PBL RH result was expected and is consistent with CRM (Tompkins, 2001) and ERA-Interim reanalysis  studies, yet conflicts with KWAJEX observations in the west Pacific (Cetrone and Houze, 2006). Mid and upper layer RH relationships with convective type were unclear (Figure 15c,d), which was also unexpected and regionally differs from observational studies in the west Pacific (e.g., Brown and Zhang, 1997;LeMone et al., 1998;Cetrone and Houze, 2006). The observed mean-layer RH relationships of this study could be inconsistent with other studies due to legitimate regional variation, but also incorporation of observations from different convective regions and lifecycle stages (e.g., Mechem et al., 2002; see Section 1).
Organized convection was hypothesized to be associated with greater vertical speed shear in each of the four analyzed layers. The hypothesis was validated for the PBL and mid layer (Figure 16b,c), with the PBL findings consistent with prior ERA-Interim reanalysis and PISTON observational studies (e.g., Chen et al., 2017 andChudler andRutledge, 2021), along with prevailing cold pool convective initiation theory (Yuter and Houze, 1995;Houze, 2018). Unclear relationships between convective type and both deep ( Figure 18c) and upper layer (Figure 17d) speed shear existed, which was unexpected. In particular, the lack of deep layer speed shear trend with convective type differs from CRM, ERA-Interim reanalysis, and TOGA COARE studies that link stronger deep layer shear to more organized TOC (Tompkins, 2001;Igel and van den Heever, 2015;Saxen and Rutledge, 2000). These prior deep layer shear studies were predominately based on observations of quasi-linear TOC, however, and thus may not be directly comparable to the CPEX(-AW) observations that sampled few quasi-linear convective systems.

b) Near-storm Environmental Relationships with Convective Intensity
Motivated by large near-storm environmental metric variability within convective type and between cases of similar convective type, analysis of TOC vertical structure (i.e., convective intensity) found stronger isolated TOC to only be consistently associated with greater mean-layer MLCAPE. This result was consistent for both the median Ku-band reflectivity profile analysis in Section 4b (Figure 19e,g) and the CFAD analysis in Section 4c (Figure 13b,d; Figure 22c; Figure   23c). The relationship between isolated convective intensity and MLCAPE is in-line with observations from TOGA COARE (Kingsmill and Houze, 1999) and the concept of greater CAPE promoting hydrometeor growth through enhanced thermodynamic instability and buoyancy. The CPEX(-AW) results differ from KWAJEX (Cetrone and Houze, 2006) and PISTON (Chudler and Rutledge, 2021) observations, however, but the contrast could be attributed to negatively biased KWAJEX and PISTON CAPE measurements (see Section 1).
Meanwhile, no clear mean-layer CAPE nor speed shear relationships with organized TOC intensity were observed. However, cases with distinctly greater clear air upper layer RH were consistently more intense than their counterparts (Figure 15d; Figure 27c; Figure 30c), similar to KWAJEX observations in the west Pacific (Cetrone and Houze, 2006). The result is consistent with the theory of less dry air entrainment encouraging intensification through limiting negative buoyancy introduction into the convective system and enabling greater hydrometeor growth. However, this logic was expected to translate to moisture in other layers as well, which was not observed.
In comparison, a lack of similar trends between environmental metrics and convective intensity across convective type indicates that single-core and multi-core systems interact differently with their near-storm environments. As such, further observational investigation into distinguishing process-level features of single-core and multi-core TOC is needed to better understand and forecast TOC.

Structures of Organized TOC
Motivated by notably different vertical organizational structures despite similar 2-D structures and synoptic-scale setups, two organized TOC cases were analyzed against one another to determine potential links between multi-core TOC vertical structure and near-storm environments. The case with numerous embedded convective elements amongst prevailing stratiform (i.e., Case 16; Figure   36) was associated with similar speed shear and a distinctly drier impinging synoptic airmass compared to the leading line, trailing stratiform case (i.e., Case 13; Figure 35). Despite notably drier impinging air, Case 16 was more intense than Case 13 (Figure 39c; Figure 40c). This result was unexpected, based on prior observational (Brown and Zhang, 1997;LeMone et al., 1998;Cetrone and Houze, 2006), CRM (Tompkins, 2001), and ERA-Interim  studies that consistently link more intense TOC to greater mid-level RH via less dry air entrainment.
However, cases 13 and 16 were both located on the moist side of their respective synoptic-scale moisture gradient ( Figure 31). With Case 16 lacking quality RH observations within the moisture gradient, moisture observations could not be compared from the environments where both cases directly flourished. Therefore, evaluating the Case 13 vs. Case 16 RH analysis in the context of other studies may not be an apt comparison. Upon further investigation, synoptic-scale low-level convergence appeared to be the key force in driving the greater intensity of Case 16 ( Figure 41) and perhaps even the greater number of convective elements. Therefore, future observational investigation of synoptic-scale low-level convergence in relation to multi-core organizational structure and TOC intensity is warranted.

d) Caveats
The analysis and results of this study are valuable in addressing the regional gap in in situ research of non-tropical cyclone related TOC in the CPEX(-AW) observational domain. That being said, limitations exist with the data, as is inherently the case with field campaigns. The presented analysis is based on observations from a small sample size of 12 convective cases (4 isolated, 8 organized), and many dropsondes were omitted from analysis due to dysfunctionality (e.g., moisture biases). Additionally, the CPEX(-AW) science flights did not consistently sample convection in a similar region relative to the convection, so observations are often being compared from different regions of storms. Contextualizing observations by storm-relative region was attempted, but further dividing the already limited number of observations provided too small of sample sizes for noteworthy comparison. Given the regional variability of near-storm environmental relationships with TOC (e.g., Chen et al., 2017 and as previously discussed in this section), the results of this paper are also specific to the Gulf of Mexico, Caribbean Sea, and western Atlantic region and cannot be confidently translated to other tropical oceanic regions.

Summary and Conclusions
Using a unique suite of collocated, high-resolution airborne observations of non-tropical cyclone related TOC from the NASA 2017 CPEX and 2021 CPEX-AW field campaigns, this study presented an analysis of near-storm environmental relationships with 3-D TOC structure in the Gulf of Mexico, Caribbean Sea, and western Atlantic region. Large variability in near-storm meanlayer CAPE, vertical speed shear, and RH was observed amongst systems of similar convective type (isolated, organized) and also within individual convective systems. The unexpectedly large environmental metric ranges within each case necessitates future investigation into how near-storm environments coevolve with convection and why they vary considerably, taking into account convective lifecycle stage at the times of sampling.
Despite the large variabilities, notable trends emerged between near-storm environmental metrics and 3-D convective structure. The PBL was the layer most commonly related to 2-D TOC structure, with organized (i.e., multi-core) TOC being associated with generally greater PBL RH and speed shear compared to isolated (i.e., single-core) TOC. While prior studies, combined with this study, find inconsistent relationships between near-storm PBL environments and TOC, a majority of studies, including this study, denote the importance of the PBL and its environmental characteristics to the organization of TOC (e.g., PBL shear influencing TOC linearity, greater PBL RH providing greater convective buoyancy potential, lesser PBL RH providing enhanced low-level thermodynamic instability). Therefore, accurate PBL representation in weather and climate models is critical to improve TOC parameterization.
Meanwhile, the upper layer (i.e., above the freezing level) was the layer most consistently related to vertical convective structure (i.e., TOC intensity), with more intense isolated TOC being associated with greater upper layer CAPE and more intense organized TOC being associated with greater upper layer RH. A lack of similar environmental trends with TOC intensity across convective type suggests that single-core and multi-core TOC systems interact differently with their near-storm environments, thus necessitating distinguished process-level research on both types of TOC. Additionally, prior studies tend to not discuss upper layer environmental influences on TOC. The results of this study contend that relationships between TOC intensity and environments above the freezing level should be given more attention.
Lastly, a comparison analysis of two horizontally similar multi-core TOC systems with notably different vertical structures denotes the importance of synoptic-scale low-level convergence and its ability to foster intense TOC in otherwise seemingly less favorable near-storm environmental conditions. The analysis also postulates that synoptic-scale low-level convergence may influence the vertical organizational structure of multi-core TOC (e.g., by increasing the number and intensity of convective elements), though more research is needed.
With prior studies (e.g., Chen et al., 2017) showing TOC relationships with near-storm environments to vary regionally, this study helps address a notable regional gap in in situ analysis of relationships between TOC structure and near-storm environmental metrics. It also highlights the unique capabilities of the CPEX(-AW) remote sensing instrumentation in their ability to capture essential small-scale (both spatially and temporally) near-storm environmental features and variability. The CPEX(-AW) instrumentation, particularly DAWN, offer a glimpse into the potential of future spaceborne remote sensing, with higher resolution measurements capable of improving modeling efforts through improved process-level knowledge of tropical convection (e.g., Turk et al., 2020;Mazza and Chen, 2021), data assimilation (e.g., Cui et al., 2020;Hristova-Veleva et al., 2021;Minamide and Posselt, 2022), and model evaluation (e.g., Cui et al., 2020;Minamide and Posselt, 2022).
Future work will extend analysis of near-storm environmental relationships with 3-D TOC structure to the east Atlantic region using in situ observations from the recent 2022 NASA CPEX-Cabo Verde (CPEX-CV) field campaign. CPEX-CV sampled TOC with similar instrumentation to that of CPEX-AW. Unlike CPEX-AW, however, CPEX-CV science flights included a focus on repeated sampling of the same convective systems, providing the opportunity to better analyze and compare specific TOC regions that could not be adequately performed in this study. Furthermore, idealized TOC simulations using the NCAR Cloud Model 1 will also be executed, wherein input sounding moisture and winds (informed by CPEX(-AW) observations) will be altered to analyze their effects on convective structure and organization. TOC relationships with near-storm environments will be explored from satellite and reanalysis perspectives as well. Using collocated archived GPM-DPR, GOES, Atmospheric Infrared Sounder, and Megha-Tropiques remote sensing data, along with NASA MERRA-2 reanalysis, this analysis will help assess satellite and reanalysis capabilities to resolve fundamental TOC features and environmental relationships. In addition to the environmental metrics discussed in this paper, near-storm low-level convergence and moist static energy will be assessed in relation to TOC structure, as they have both been shown to influence TOC. Ultimately, future work will supplement the in situ analysis of this paper, provide an interregional analysis of TOC relationships with near-storm environments, and offer an intercomparison of results across data type (i.e., in situ vs. model vs. remote sensing vs. reanalysis).