1. Introduction
A sea breeze, or a wind that blows from water onto land due to a diurnal temperature gradient created between the warmer landmass and cooler water surface, affects nearly every coastal city and habitat in warm climates. The sea breeze often occurs as part of a sea-breeze circulation, where during the day onshore-moving air is lifted upward and advected back over the open water aloft before subsiding (Miller et al. 2003) and vice versa at night. Sea breezes are of key meteorological and climatological importance for their role in initiating deep convective thunderstorms (e.g., Zhang and Wang 2021). In cities such as Houston, Texas, with a large, sprawling urban center, the urban heat island effect is thought to amplify the strength of the sea breeze (Yoshik Ado 1992; Freitas et al. 2007). Morcrette et al. (2007) demonstrated that accurately predicting isolated thunderstorms requires an accurate representation of both favorable synoptic-scale dynamics and local (mesoscale) surface convergence boundaries. Depicting these fine-scale features has remained a challenge for many weather models, especially models that do not explicitly resolve convection or have poor horizontal resolution (e.g., Hock et al. 2022). With the exception of high-resolution large-eddy simulation (LES) models or similar mesoscale models (e.g., Nicholls et al. 1991; Abulikemu et al. 2016), the incredibly fine mesoscale structure of the sea-breeze circulation remains beyond many weather models’ ability to accurately predict when sea breezes can trigger deep convective thunderstorms (Crosman and Horel 2010; Wang et al. 2018).
The Experiment of Sea-Breeze Convection, Aerosols, Precipitation and Environment (ESCAPE) field project took place from 27 May through 2 July 2022 in the vicinity of Houston, Texas, focusing on how meteorological conditions, dynamics, and aerosols control the initiation, early growth stage, and evolution of coastal convective clouds. A companion paper (Kollias et al. 2025) fully describes the ESCAPE field campaign, objectives, instruments, and project goals. As part of ESCAPE, a study of sea-breeze formation, the influence of the sea breeze on daytime thunderstorm formation, and aerosol–cloud interactions in the context of sea-breeze-generated thunderstorms is critical. ESCAPE took place concurrently with the Tracking Aerosol Convection Interactions Experiment (TRACER; Jensen et al. 2022), with TRACER starting to collect observations in the Houston area in Summer 2021. Houston was selected given its ideal location for frequent daytime (diurnal) cloud cover, considerable presence of aerosols, and logistical considerations (e.g., Fan et al. 2007; Yuan et al. 2008; Zhang et al. 2021), all of which are critical to the broader ESCAPE science goals seeking to take measurements of the sea breeze, sea-breeze-driven convective clouds, and environmental aerosols during the entire sea breeze and convection life cycles, respectively. To meet ESCAPE’s required goals and needs, a dedicated forecasting and nowcasting team was assembled to provide short-term forecasts and real-time nowcasts, especially knowing that the quality of the collected datasets would be dependent on sampling the best available conditions during the campaign.
Accounting for the fact that ESCAPE was among the first field campaigns since the peak of the COVID-19 pandemic, the forecasting and nowcasting teams leaned heavily on a hybrid work structure of in-person and remote support to meet the objectives of the project, particularly for the deployment of airborne- and ground-based assets (Kollias et al. 2025). The hybrid nature of the forecasts/nowcasts provided unique opportunities and challenges. Given that hybrid work environments have continued to persist since the end of the COVID-19 pandemic, one of the goals of this article is to provide guidance for future campaigns based on the ESCAPE experience. The next section of this article describes the meteorological analysis performed and completed during each forecasting or nowcasting shift, while the following section describes the forecasting operations and hybrid work structure used during ESCAPE.
2. Meteorological analysis
Identifying atmospheric conditions favoring sea-breeze convection requires an innate understanding of both the synoptic-scale setup and mesoscale meteorology. To meet ESCAPE science goals, the forecasting team developed a decision matrix to determine the likelihood of sea-breeze-driven convection based on numerous criteria: The more criteria that were met (e.g., favorable humidity and low vertical wind shear), the higher the probability of sea-breeze-driven convection. Each criterion in this decision matrix was weighted equally and uniquely tailored for the ESCAPE field campaign. These criteria are highlighted in Table 1 and were determined from a variety of previous studies (e.g., Miller et al. 2003; Reddy et al. 2020) showing that these conditions favored the onset of sea-breeze-driven convection. One of the pros of using a decision matrix is that it allowed the forecasting team to organize all pertinent information related to sea-breeze convection forecasting into a single location, which helped facilitate efficient forecast discussions. Over the course of the campaign, no forecast “busts” occurred, which the forecasting team attributes to the use of this decision matrix for guiding forecast discussions and gauging overall forecast confidence. The next subsections describe in detail the rationale behind the criteria used in the forecasting team’s decision-making. These criteria were assumed to have equal importance in determining the likelihood of sea-breeze-driven convection during the forecast operation.
The decision matrix used by the ESCAPE forecasting team to evaluate atmospheric conditions conducive for sea-breeze-driven convection. Each criterion was evaluated in 3-h increments between 0600 and 1800 CDT for the target domain.
a. Moisture.
For deep convection to develop, sufficient moisture must be present especially in the low and midlevels of the atmosphere. The ESCAPE forecast/nowcast team checked whether model RH was at least 70% or greater at 1000, 850, and 700 hPa as well as greater than 50% at 500 hPa. Since each criterion in Table 1 was weighted equally for determining the probability of sea-breeze-driven convection, having moisture criteria represents 4 out of 15 total criteria ensured that humidity carried the greatest weight in forecast decision-making. Several models were examined (i.e., GFS, NAM, ECMWF, and HRRR) to evaluate the predicted presence of not only humidity but also surface and midlevel cloud cover development during the forecast period. If model disagreements existed, forecast participants would lean on the convection-allowing models (HRRR), since global models like the GFS or ECMWF do not explicitly resolve convection in the short term. Modeled cloud development represented another important marker that clouds were forecast to be present and, in conjunction with the humidity criteria, added confidence that enough moisture was indeed available to develop clouds through at least the midlevels of the atmosphere. High-altitude cloud cover was excluded from the decision-making criteria to ensure advected cirrus clouds did not result in “false positive” readings. Rather, with other metrics such as mixed-layer convective available potential energy (MLCAPE) and convection in the HRRR, high-level cloud cover resulting from the rapid deepening of an isolated thunderstorm was implicitly accounted for.
b. Synoptic-scale weather.
Another important component of the forecast decision involved evaluating the overall and evolving synoptic-scale meteorological pattern. The presence of an upper-level ridge east (E) of Texas [e.g., over the southeast (SE) United States], the presence of a surface high pressure system over the Gulf of Mexico, or even the expanse of the Bermuda high into the eastern Gulf of Mexico qualified as a positive criterion for sea-breeze-driven convection. Having a high pressure system or ridge in these regions implied large-scale atmospheric flow was conducive to onshore moisture advection as well as “kickstarting” an afternoon sea breeze over the SE Texas shoreline. In addition to favoring afternoon sea-breeze conditions, the presence of a ridge or high pressure system to the east also favors consistent onshore moisture advection, which is a key component for increasing buoyancy in the atmospheric boundary layer (Shin et al. 2021).
The forecast team also checked for the presence of nearby frontal systems and mesoscale convective systems (MCSs) that could leave residual frontal-like forcing along outflows. Organized lift along such a frontal system in SE Texas would imply convection along the front was the result of synoptic-scale organization rather than sea-breeze-driven convection. Frontal systems over the central and southern United States are not uncommon during early summer and late spring (Barth et al. 2015), and these systems often interact with the southerly flow to initiate convection beyond sea-breeze effects (Huang et al. 2019). The decision-making process involved verifying whether such MCS or frontal systems were forecast to approach within 1000 nm (i.e., as far away as central Nebraska from Houston) to assess whether convection was influenced by MCS outflow or convergent flow. Finally, global models such as the ECMWF and GFS were weighted more in the decision-making process here especially beyond 48 h due to their longer model run time and reliability (Rodwell and Wernli 2023).
c. Afternoon sea breeze.
After evaluating atmospheric moisture and large-scale dynamics, the probability of an afternoon sea breeze (usually between noon and 1500 local time) was examined. One criterion was to determine whether the afternoon land (i.e., the experiment region, either east or southwest of Houston) and respective sea surface temperature difference exceeded 5°C (9°F), consistent with accepted values in previous modeling and observational studies of coastal sea breezes (e.g., Wermter et al. 2022). Another criterion was to see whether the morning and afternoon coastal winds exceeded 10 kt (1 kt ≈ 0.51 m s−1). The purpose of checking both morning and afternoon wind speeds elucidated the role of large-scale atmospheric flow versus sea-breeze-driven flow. Optimum conditions required calm/variable wind speeds in the morning followed by a modest 5–10-kt increase in wind speed in the afternoon—a clear sign that the onshore wind was sea-breeze driven and not synoptically driven. This is especially important because wind (along with coastal morphology) represents the most important factor determining the inland extent of the sea breeze (Park et al. 2020; Hock et al. 2022)—a matter of crucial importance for deciding the positioning of mobile observing assets. The forecast team was also mindful of the fact that sea breezes can manifest earlier than models predict, owing to recent results showing that midmorning cold temperature biases (underestimated surface heat fluxes) frequently occur in models (Caicedo et al. 2019).
d. Model evaluation and sounding analysis.
Several forecast models were used for real-time nowcasting, short-term forecasting, and long-term forecasting of moisture, synoptic-scale weather, and sea breeze (see Table 2 for a detailed description of each of these models). The Hybrid Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT) (Draxler and Rolph 2010) was used with a variety of model outputs (including GFS and HRRR) to examine the evolution of the boundary layer wind field through the morning and afternoon hours. HYSPLIT has been used in many previous studies evaluating the evolution of daytime sea breezes (e.g., Han et al. 2022; Trošić Lesar and Filipčić 2022). HYSPLIT was also used to check the degree of midday mixing and provide a secondary evaluation of the likelihood that surface-based air parcels in a convectively favorable environment could develop into deep convection. In this regard, HYSPLIT proved to be an invaluable tool for the ESCAPE forecasting team.
The list of weather forecasting models used by the ESCAPE forecasting and nowcasting teams, including information on the type of model, horizontal resolution, number of model levels, primary use, and a reference for further detailed information on each model setup.
Global forecast models (GFS and ECMWF) and their ensemble systems provided valuable context about the large-scale synoptic pattern. One important synoptic pattern is the presence of an upper-level 500-hPa ridge. Figures 4–6 of Wang et al. (2022) show that a 500-hPa ridge is a common feature of the Houston summer months, including June, and that subtle shifts in the position of the ridge influence sea-breeze circulations and the frequency of convection. Hence, it is not only that a ridge was present but also specifically the variability in the position of the ridge that influenced the favorability in convection from day to day. Variability across the forecast models allowed for assessment of the uncertainty in synoptic-scale vertical velocity and moisture advection, giving confidence in the synoptic setting in which the mesoscale dynamics of interest to the project were taking place.
Convection-allowing models were also examined to determine the probability of convection. HRRR model forecasts were mainly used to check for the timing of sea-breeze initiation as well as to identify weak surface high pressure systems over the SE Texas coastline. Surface high pressure systems off the SE Texas coast were ideal for driving overnight offshore flow (land breezes) and provided favorable small-scale dynamics for midafternoon convergence zones across the SE Texas coastline. As one example, inland surface high pressure systems often produce offshore flow, but a sea breeze can still advance inland even if offshore flow exceeds 10 m s−1 (Arritt 1993). In addition to identifying regions of convergence along the coastline during the afternoon, the HRRR provided critical timing information regarding isolated shower or thunderstorm development. Knowing that convergence along a sea-breeze front or convective outflow/sea-breeze front interactions is not enough to trigger deep convection (Kingsmill 1995), using the HRRR in conjunction with other data sources helped the forecasting and nowcasting teams assess the likelihood these boundaries would trigger deep convection.
In addition to the HRRR, customized 1-km real-time model runs from the Weather Research and Forecasting (WRF) model, version 4.4 (Skamarock et al. 2021), provided additional guidance for assisting the forecast (and nowcast) teams in identifying times and locations for deploying airborne- and ground-based assets. The forecasting domain is shown in Fig. 1. During ESCAPE, the WRF simulations were automatically initialized twice daily at 0600 and 1200 UTC and run for a 24-h forecasting period. The simulations were automatically postprocessed to generate geospatial data interchange format based on JavaScript Object Notation (GeoJSON) files that were made publicly available on the internet using Mapbox. In total, more than 1600 real-time WRF simulations were conducted for the ESCAPE field campaign. Featuring a various combination of initial forcings, aerosol loadings, microphysical schemes, and planetary boundary layer schemes, the large ensemble of high-resolution WRF simulations not only aided forecasts and nowcasts but also allowed for statistically robust postcampaign analysis. Details regarding the WRF Model configurations and performance evaluations can be found in Hu and Lebo (2024a,b, manuscript submitted to Wea. Forecasting).
WRF Model domain overlayed with Convair and Learjet research flight tracks. The forecast operation center is indicated by the white dot and text. Note that the domain is not centered over Houston to better capture convection by sea-breeze and frontal systems for computational efficiency. Note that the western and eastern domains are, respectively, the land areas to the west-southwest (WSW) and E of Houston.
Citation: Bulletin of the American Meteorological Society 106, 3; 10.1175/BAMS-D-23-0015.1
Finally, an extensive multimodel sounding analysis was performed to finalize each forecast. Model-derived soundings were analyzed using the Sounding and Hodograph Analysis and Research Program in Python (SHARPpy) software package (Blumberg et al. 2017). SHARPpy provided multiple forecasting inputs such as CAPE, convective inhibition (CIN), lifting condensation level (LCL), level of free convection (LFC), wind shear (between 850 and 500 hPa), inversion identification, and moisture aloft, all of which provided a detailed picture of cumulus and thunderstorm formation likelihood. Wind shear was examined for each sounding: For isolated sea-breeze-driven convection, low wind shear (<20 kt) was considered ideal as it implied a vertically stacked atmosphere and limited influence by upper-level dynamical features such as nearby frontal systems or midlevel shortwave eddies. Atmospheric soundings also revealed the presence of subsidence inversions aloft, which were important for assessing the likelihood of deepening convection versus the probability of shallow cumulus or congestus cloud formation (Morcrette et al. 2007; Park et al. 2020). Inversion strength was not included in the decision-making criteria for a few reasons: Inversions were almost always associated with warm and dry layers (covered by the RH criteria), and slight inversions (<0.5°C) are ideal for temporarily suppressing convection and allowing enough CAPE to build up in the late morning/early afternoon, thereby increasing the probability of deep convection later in the afternoon. Cloud top and cloud base were also estimated from the large ensemble of model soundings, providing a crucial input for flight planning. Finally, forecast soundings were selected and analyzed at and around (e.g., within 25–50 km) areas expected to produce sea-breeze-driven convection.
3. Forecasting operations
The ESCAPE forecasting team included both an in-person team and a virtual team. The in-person forecasting team focused primarily on short-term forecasting for determining on which days the National Research Council of Canada Convair-580 and Stratton Park Engineering Company (SPEC Inc.) Learjet would fly, as well as identifying priority targets for sampling convection. The ESCAPE forecast was closely communicated with the other TRACER-related project teams during the TRACER intensive observation period (IOP). The virtual forecasting team supported the TRACER forecasting team and focused more broadly on meteorological evolution around the Houston area. A typical day-to-day schedule for both the in-person and virtual forecasting operations is given in Table 3.
A typical schedule for research flight days and nonflight days. The key difference in routine for flight and nonflight days was the need for nowcasting support and an early morning preflight forecast briefing. The main forecasting preparation and meeting times were otherwise the same every day of the campaign.
a. Precampaign workshop and training.
Prior to the start of the campaign, the forecasting and nowcasting team leaders hosted a virtual forecasting workshop. This workshop established a basic operational workflow that all participants could follow throughout the campaign. For both the seasoned and first-time forecasters on the team, the workshop was helpful in identifying the most important criteria for sea-breeze convection forecasting and development of the forecasting guide/checklist given in Table 1. Focusing on specific atmospheric conditions was also helpful in consolidating time and the number of resources needed to produce accurate forecasts. The workshop also allowed the forecasting and nowcasting teams to familiarize themselves with the NCAR Earth Observing Laboratory (EOL) field catalog (see Fig. 2 as an example). Finally, the forecasting workshop aided the forecasting coleaders in future decision-making and identifying areas of improvement, which were achieved quickly after the ESCAPE campaign formally started.
Example of the EOL catalog (http://catalog.eol.ucar.edu/escape) during RF05 on Wednesday, 8 Jun 2022, at ∼1015 CDT. The Convair had taken off and was headed for the southernmost triangle in the eastern domain to sample convective roll clouds (cloud streets) heading onshore.
Citation: Bulletin of the American Meteorological Society 106, 3; 10.1175/BAMS-D-23-0015.1
An important function of field campaigns is to train students and future scientists. Students who were in the field supporting instrumentation joined forecasting operations to supplement their instrument responsibilities. This is mutually beneficial in providing students with a deeper understanding of the meteorological conditions sampled by the instrumentation. As one example, three students from Texas Tech University played a significant role in ESCAPE forecasting operations while also supporting the Texas Tech University (TTU) Lightning Mapping Array (LMA) stations in collaboration with Texas A&M’s Houston LMA (Logan 2020; Chmielewski and Bruning 2016). As another example, two students supporting the Holographic Detector for Clouds (HOLODEC) airborne cloud probe (Spuler and Fugal 2011; Fugal and Shaw 2009) and one student supporting the National Research Council of Canada (NRC) cloud probe analysis were trained to provide crucial support to the forecasting and nowcasting teams. Students who traveled to the field to gain experience with mobile radar deployments also contributed to daily forecast reports. Students not directly involved with an instrument team but who will use project data also participated as members of the forecasting and nowcasting teams.
b. In-person field operation.
The in-person ESCAPE forecasting team focused their efforts on selecting targeted flight locations and sampling times given the goal to sample sea-breeze-generated convection.
The in-person team had three main foci:
- 1) Identify possible sea-breeze events and possible target domains. The in-person team met every morning at 0900 local time for a 3-h shift, examining a variety of weather model output, balloon-borne soundings, satellite data, radar data, and trajectory model analyses to gain an overarching understanding of the synoptic-scale environment. CAPE/CIN, presence of temperature inversions, and orientation of surface-level flow relative to the shoreline were all critical for establishing if—at minimum—a low-level cumulus field would form and if any probability of isolated thunderstorms forming from that cumulus field was possible. Using these criteria, the forecast team identified which of the two predetermined sampling areas (i.e., the western or eastern domain) was the optimum target for a prospective flight. At the ESCAPE forecast meeting, held at 1400 LT on nonflight days immediately after the general TRACER forecast meeting, the forecast team gave a recommendation of the possibility of flights for the next 3 days. The flight time and target domain were also recommended if a flight was favored for the next day. This meeting included the participation of the ESCAPE PIs and pilots so that flight plans could be finalized and filed.
- 2) Short-term forecast validation for possible flights the next day. If a highly probable sea-breeze event and flight were identified for the next day, a small group of forecasters (3–5 people) continued to monitor synoptic-scale and mesoscale weather conditions throughout the day and provided a short-term forecast update the morning of a potential flight. During this stage, forecasting efforts focused on sounding and trajectory analysis, such as identifying layers with sufficient CAPE as well as cloud height and base. Satellite data loops were also used for short-term forecast validation. A meeting was held with the PIs and pilots at least 3 h before the planned flight time, to assess whether conditions were still favoring flights and an updated flight time and domain were provided. These forecasting efforts allowed adjustment in flight plans if needed, providing better information on preferred flight altitudes.
- 3) Nowcasting support for research flights. During each flight, a small group of nowcasters provided live updates for the crew on board the NRC Convair-580 through its bandwidth-limited satellite communication platform. Although there was no direct communication with the crew on the SPEC Learjet, the pilots on the NRC Convair-580 were able to relay important updates to the SPEC Learjet pilots when the aircraft were flying at the same time. With the help of real-time radar, satellite, and flight location information, the nowcasting team gave important guidance to the flight direction with possible target clouds along the flight path.
c. Virtual operation.
A portion of the ESCAPE forecast team operated virtually and took an active role in coordinating and contributing to the TRACER forecasts throughout the aircraft operations period. The virtual methodology offered a route for field campaign participation to those who could not attend in person and ensured that ESCAPE senior personnel could represent campaign needs in the fully virtual TRACER forecast briefings.
The ESCAPE forecast was well coordinated with the TRACER daily forecast, which was performed virtually. The goal here was to provide 1) a forecast for the morning and afternoon of the next day and 2) a longer-term forecast for the next 2–7 days. The daily assigned forecast team had discussions using multiple aforementioned weather model outputs as well as a review of the latest National Weather Service forecast for SE Texas at around 1500 UTC. The team also reviewed the TRACER-operated daily high-resolution model simulations focused on the Houston area (Jensen et al. 2022). This discussion was virtual through an online conversation service (i.e., a Slack channel), which also involved the ESCAPE in-person forecast/nowcast team. This allowed for interactions between the two different forecasting teams which had common purposes. Although the discussion generally involved the daily forecast members, all ESCAPE and TRACER participants could join the discussion through the online conversation service.
The discussion was followed by the TRACER forecast briefing at 1300 local time using an online meeting platform (e.g., Zoom). The ESCAPE in-person forecast/nowcast team also participated in the briefing discussion to ensure that the ESCAPE forecast was consistent with the TRACER forecast. This ensured that both the ESCAPE and TRACER forecasting teams could resolve discrepancies in the short-term forecasts. This collaboration was paramount for ensuring forecast consistency and providing the highest amount of detail possible. Right after the TRACER virtual forecast briefing, the ESCAPE hybrid forecast/nowcast briefing was held in person at the designated ESCAPE operations center with an online meeting platform (e.g., Zoom) used so that remote participation was also possible.
Overall, the hybrid forecast operation of in person and virtual provided many advantages. First, this operation allowed all PIs, collaborators, and students to participate in the forecast. Another advantage of this operation was that it permitted wide insights and suggestions from both the in-person and virtual participants, allowing for easy facilitation between collaborations with other projects (specifically TRACER). The hybrid, collaborative forecast also supported the sharing of real-time observations from research instrumentation across the complex, multiagency field campaigns with the objective of cross-checking forecasts produced by the respective ESCAPE and TRACER teams as well as eventually enabling multiplatform studies on sea-breeze convection and related processes.
On the other hand, the hybrid structure brought out a few of the following challenges. While the TRACER forecast was scheduled at a fixed time every day (usually 1300 local time for the briefing), ESCAPE needed more flexible timing with the forecast/nowcast operation depending on the flight schedules and mobile truck deployments. The virtual TRACER forecast worked to supplement the ESCAPE forecast during the entire ESCAPE IOP, but sometimes it could not be well integrated into the ESCAPE in-person forecast due to the observation schedule. Another challenge involved the use of several online conversation platforms by various participants during the IOP (e.g., Slack, WhatsApp, emails, cell phone text/call, and Zoom). Not all participants registered on all platforms, causing some participants (especially virtual participants) to miss important information. The virtual participants missed nowcast information because the communication tools were not integrated. For future projects, it is recommended that communication tools should be integrated such that all onsite and virtual participants communicate immediately and smoothly.
d. Nowcasting operation.
Nowcasting operations were essential for capturing the best possible dataset from the campaign. In using the flight time most effectively, on-the-ground nowcasting gives aircraft operators real-time advice on where to target new convection, considering operational factors (distance, scientific goals, etc.) and using information that is not readily available on the plane (e.g., development of convection away from aircraft location). Nowcasting also contributes to safe operation by warning the aircraft operators of nearby lightning activity and watching for potential obstacles to landing. The lead nowcaster was responsible for sending concise recommendations to the mission scientist on board the NRC Convair-580 and closely monitoring the communication channel for requests/feedback from the mission scientist. These communications were carried out through a chat interface software that was accessible from both on board the NRC Convair-580 and on the ground. To streamline communication during operations and to avoid creating confusion, the communication between the nowcast team and the mission scientist was limited to only the lead nowcaster.
Nowcasting operations benefitted from having multiple team members focusing their analysis on different meteorological tools. Given that a significant focus of the lead nowcaster was directed toward communicating with the mission scientist, the additional nowcasters were essential for frequently checking satellite loops for outflow boundaries, convergence, cumulus organization, and other indications of high probability for convective cell development: Such loops could not be accessed on the aircraft due to limited bandwidth. Satellite data show the development of cumulus clouds into deepening congestus clouds and, in conjunction with radar reflectivity data, quickly reveal developing precipitation via rapidly increasing radar reflectivity and satellite-based evidence of convective organization. Radar velocity data are also useful for revealing the precise location of the sea-breeze and outflow boundaries from nearby thunderstorms (Morcrette et al. 2007). These boundary features represented ideal target locations for the aircraft to sample. Advantages of convection-allowing models such as HRRR include increased accuracy of convection mode (e.g., discrete versus organized) and better accuracy at resolving sea-breeze circulations (Cafaro et al. 2019). Cross-referencing these conditions on satellite against the most recent HRRR runs gave nowcasters increased confidence in recommending specific target region(s) during flight. Target areas of interest were discussed among the nowcast team before any recommendation was made to the mission scientist on the NRC Convair-580.
e. Pandemic and campaign management.
Undeniably, the pandemic created many challenges for the management of ESCAPE and for the people participating in the campaign. In response, many steps were taken to ensure ESCAPE went smoothly and safely. For example, masks were mandatory in the forecast room, social distancing was practiced, and the number of people in the forecast/nowcast room was restricted (Fig. 3). Different platforms and applications (e.g., Zoom, Google Drive/Slides, and WhatsApp) were used to maintain communication between those inside and outside the forecast/nowcast room and to optimize engagement in the campaign and discussions. The use of online tools also helped document the decision-making process because every forecast report was saved to cloud storage and every meeting was recorded through Zoom. As in many areas of society, a result of the pandemic was a sudden reliance on ubiquitous internet-enabled computing that fostered collaborative, multisite editing that naturally transitioned into a working reference for teams. With the collective efforts from every team member, the ESCAPE team was highly efficient during the campaign and collected a set of valuable data even with the challenges posed by the pandemic.
Photograph of the ESCAPE forecasting and nowcasting teams working together during a research flight while following COVID protocols. Pictured: (top left) Eric Bruning, (top right) Matt Miller, (bottom left) Raymond Shaw (HOLODEC principal investigator), and (bottom right) David Singewald.
Citation: Bulletin of the American Meteorological Society 106, 3; 10.1175/BAMS-D-23-0015.1
4. Campaign forecasting operations and select research flights
As is common during meteorological field experiments, the initial ESCAPE experiment design required constant adaptation to the present atmospheric conditions. Both forecasting and nowcasting team participants expected this, knowing that the mix of synoptic-scale weather (e.g., midtropospheric ridges and cold fronts) and mesoscale forcing (e.g., the sea breeze and occasional nearby MCS) would add complexity to the interpretation of weather model output and other tools used by the team. Two research flights highlighting a variety of the forecasting and nowcasting team’s experiences are selected, the 16 and 4 June 2022 research flights, and described in the next two subsections.
a. Convair research flight 11/Learjet research flight 08 (Los Angeles coastline).
A textbook case study of sea-breeze-driven convection took place on 16 June 2022 along the southwest (SW) Los Angeles coastline. A weak high pressure system was present off the SE Texas coast, helping aid an overnight land breeze evident in satellite imagery (not shown). The forecasting team identified these conditions during the previous day’s forecast development and discussion. During the morning, clear and calm conditions gave way to rapid daytime heating, creating a sea breeze that began advancing inland between 1100 LT and noon. Both the sea breeze and deepening congestus clouds were evident between these times in the satellite imagery and real-time HRRR weather model output used by the nowcasting team. As a result of this real-time development, both the forecasting and nowcasting teams identified this time as a high-probability event where further sea-breeze-driven convective initiation would occur hereafter. By 1300 CDT, with the NRC Convair-580 airborne, deepening congestus clouds were present along the Los Angeles coastline between Calcasieu Lake and Second Lake in SW Louisiana (Fig. 4). The NRC Convair-580 successfully sampled clear-air skies before cumulus began developing along the sea-breeze front while also capturing the initial deepening of the cumulonimbus into the upper free troposphere by 1300 CDT (represented in Fig. 4). The Learjet also penetrated this deepening convection several times.
Zoomed view of the primary ESCAPE operations domain (including county boundaries and coastlines), showing LMA detections (colored by time from blue–purple–orange–yellow; stations are white diamonds), aircraft positions (red: Convair; blue: Lear; 5 min highlighted), and research radar scan positions (gray dots) from 1822 to 1827 UTC, 26 Jun 2022.
Citation: Bulletin of the American Meteorological Society 106, 3; 10.1175/BAMS-D-23-0015.1
This research flight was successful for several reasons. Given that this flight took place near the end of the campaign, the forecasting team had gained significant experience learning the intricacies of sea-breeze timing and conditions necessary to produce sea-breeze-driven convection. Timing was especially important, given ESCAPE’s overarching goal to capture the initial development of convection prior to becoming too deep to sample in situ with aircraft due to a possible threat of hail and lightning. The nowcasting team learned from previous research flights the relative time the sea breeze was likely to begin advancing inland, allowing the nowcasting team to convey this information to the forecasting team and gauge this timing against the timing suggested by various forecast models. Furthermore, this timing also allowed the forecasting team to give better guidance for prospective thunderstorm development. One limitation of the aircraft was the limited flight duration (4 h for the Convair and 3 h for the Learjet). Therefore, optimizing the total amount of research hours required as precise information about timing as possible given that the ferry time from Sugar Land Airport to this domain had to be accounted for as well.
b. Convair research flight 04/Learjet research flight 03 (4 June).
The 4 June 2022 research flights took place under more complicated conditions compared to the 16 June research flights. Deep convection formed as the result of an interaction between an MCS and the sea breeze. The 3 June forecasting team identified an MCS moving through central Texas the day before, which arrived in the SE Texas domain (western experiment domain to the SW of Houston) during the afternoon of 4 June. Deep anvil cirrus was moving over the experiment domain near the start of the flight, with deep cumulonimbus developing dozens of miles to the west of Sugar Land, Texas (Fig. 5; near 29°N/96.2°W). Despite this, some deep congestus was measured about 10 nm inland from the SE Texas coastline and the forecasting/nowcasting teams determined that these clouds were isolated from the MCS to be considered a viable target for a research flight. The nowcasting team on 4 June verified this decision by watching the latest radar and satellite imagery throughout the morning as the sea breeze evolved and forced coastal convection away from the MCS but still well within the western experiment domain. This allowed for the 4 June flight to take on the “pizza slice” flight template where the aircraft flew from the Gulf of Mexico onshore, followed by an alongshore track to measure the sea-breeze/convergence boundary, and repeated these tracks until deep convection triggered.
As in Fig. 4, but from 2201 to 2206 UTC, 4 Jun 2022.
Citation: Bulletin of the American Meteorological Society 106, 3; 10.1175/BAMS-D-23-0015.1
This particular research flight highlights the complexities of forecasting localized convection when several synoptic-scale and mesoscale influences can affect the mode of convection. The flight was successful in measuring developing thunderstorms, but disentangling this development from the influence of the upper-level shortwave that triggered these deep MCS-style thunderstorms will require further research. The deep congestus along the sea breeze that ultimately developed into thunderstorms could make for an ideal case study. If this upper-level shortwave had not traversed SE Texas, would these thunderstorms have triggered under their own accord given the already ample surface-level moisture and conducive upper-level environment for deep convection? Would these storms have formed later in the afternoon or early evening? From a forecasting point of view, these scientific questions merit future research for the sake of guiding operational forecasters being able to discriminate MCS-style convection from sea-breeze convection. The sea breeze was apparent and clearly contributed to sustained convective development. Finally, student participants on the forecasting and nowcasting teams benefitted from this forecasting, especially being only the second flight of the campaign, by observing how distinct mesoscale features (i.e., the MCS and sea breeze) could be disentangled despite the complexity.
5. Conclusions
Overall, the ESCAPE forecasting and nowcasting teams succeeded in delivering high-quality forecasts throughout the field campaign, which helped the entire team meet their individual science objectives for the project (Kollias et al. 2025). Before the campaign began, a forecasting workshop focused on the fundamentals of sea-breeze-driven convection gave every team member a standardized review and set of knowledge. Every member of the team, from experienced scientists to new graduate students, benefitted from this workshop resulting in consistently accurate forecasts. Two such forecasts were highlighted, each illustrating the forecasting process, highlights, and challenges typical throughout the field campaign.
One aspect of forecasting operations that did not work well was integrating virtual participation during the actual forecasting analysis period. During the morning forecasting preparations, it was logistically difficult to manage virtual participants via Zoom/Microsoft (MS) Teams, where participants found this aspect too burdensome on their computers while also analyzing forecast model data/preparing forecast briefings. After a couple of forecasting shifts, this aspect of the hybrid approach was tabled, and hybrid discussions were saved until after everyone on both the in-person and virtual forecasting teams had a chance to complete their individual analyses. By the end of the campaign, this approach to the hybrid operation worked optimally well for all participants. Given this experience, it is recommended that whether working virtually or in-person, time should be allowed for all participants to perform analysis and develop a forecast before engaging in hybrid discussions and finalizing the forecast. Another recommendation is that forecasting teams should have a precampaign forecasting workshop to integrate new or potentially inexperienced personnel into the team as well as ensure a basic forecasting framework tailored to the project’s main objectives. The synergy of many forecasting activities demonstrably contributed to the overall quality of every forecast for the ESCAPE campaign. Multiple forecasters worked on specific aspects of the overarching forecast including HYSPLIT forward trajectories of air masses, model soundings for convective instability and likelihood of convection, sufficient isolation from any nearby MCS, short-range forecasting with highly detailed tentative flight suggestions, and long-range forecasting to assess pros and cons of assessing meteorological favorability and uncertainty for flights on upcoming days. This allowed the forecasting team to assess probabilities of sea-breeze convection forecasting in a precise manner, allowing the project PIs to make highly informed decisions regarding the use of aircraft flight hours. We believe this template for sea-breeze convection led to a richly successful forecasting operation and believe it will be a highly serviceable template for future field campaigns focusing on coastal convection.
Acknowledgments.
Michigan Tech participants acknowledge support from NSF Grant AGS-2019649. Texas Tech participants acknowledge support from DOE DE-SC0021247 and NSF AGS-2019939. OU participants acknowledge support from NSF AGS-2019968. We thank the staff of the Courtyard by Marriott Sugar Land/Stafford for their hospitality in support of crew morale in the operation center. We thank personnel from NCAR for the logistical organization of the forecast/nowcast center, from NRC and SPEC for the acquisition of the aircraft data, and from all the ground-based crews for the collection of unique datasets. Finally, we thank two anonymous reviewers and Mikael Witte for their independent, thorough, and constructive comments that greatly improved the quality and presentation of this manuscript.
Data availability statement.
All ESCAPE data are cataloged at the NCAR EOL ESCAPE data archive: https://data.eol.ucar.edu/project/ESCAPE.
References
Abulikemu, A., X. Xu, Y. Wang, J. Ding, S. Zhang, and W. Shen, 2016: A modeling study of convection initiation prior to the merger of a sea-breeze front and a gust front. Atmos. Res., 182, 10–19, https://doi.org/10.1016/j.atmosres.2016.07.003.
Arritt, R. W., 1993: Effects of the large-scale flow on characteristic features of the sea breeze. J. Appl. Meteor., 32, 116–125, https://doi.org/10.1175/1520-0450(1993)032<0116:EOTLSF>2.0.CO;2.
Barth, M. C., and Coauthors, 2015: The Deep Convective Clouds and Chemistry (DC3) field campaign. Bull. Amer. Meteor. Soc., 96, 1281–1309, https://doi.org/10.1175/BAMS-D-13-00290.1.
Blumberg, W. G., K. T. Halbert, T. A. Supinie, P. T. Marsh, R. L. Thompson, and J. A. Hart, 2017: SHARPpy: An open-source sounding analysis toolkit for the atmospheric sciences. Bull. Amer. Meteor. Soc., 98, 1625–1636, https://doi.org/10.1175/BAMS-D-15-00309.1.
Cafaro, C., T. H. A. Frame, J. Methven, N. Roberts, and J. Bröcker, 2019: The added value of convection‐permitting ensemble forecasts of sea breeze compared to a Bayesian forecast driven by the global ensemble. Quart. J. Roy. Meteor. Soc., 145, 1780–1798, https://doi.org/10.1002/qj.3531.
Caicedo, V., and Coauthors, 2019: Bay breeze and sea breeze circulation impacts on the planetary boundary layer and air quality from an observed and modeled DISCOVER-AQ Texas case study. J. Geophys. Res. Atmos., 124, 7359–7378, https://doi.org/10.1029/2019JD030523.
Chmielewski, V. C., and E. C. Bruning, 2016: Lightning mapping array flash detection performance with variable receiver thresholds. J. Geophys. Res. Atmos., 121, 8600–8614, https://doi.org/10.1002/2016JD025159.
Crosman, E. T., and J. D. Horel, 2010: Sea and lake breezes: A review of numerical studies. Bound.-Layer Meteor., 137 (1), 1–29, https://doi.org/10.1007/s10546-010-9517-9.
Dowell, D. C., and Coauthors, 2022: The High-Resolution Rapid Refresh (HRRR): An hourly updating convection-allowing forecast model. Part I: Motivation and system description. Wea. Forecasting, 37, 1371–1395, https://doi.org/10.1175/WAF-D-21-0151.1.
Draxler, R. R., and G. D. Rolph, 2010: HYSPLIT (HYbrid Single-Particle Lagrangian Integrated Trajectory) model access via NOAA ARL READY website. NOAA Air Resources Laboratory, 25 pp., https://www.ready.noaa.gov/HYSPLIT.php.
Fan, J., R. Zhang, G. Li, W.-K. Tao, and X. Li, 2007: Simulations of cumulus clouds using a spectral microphysics cloud-resolving model. J. Geophys. Res., 112, D04201, https://doi.org/10.1029/2006JD007688.
Freitas, E. D., C. M. Rozoff, W. R. Cotton, and P. L. Silva Dias, 2007: Interactions of an urban heat island and sea-breeze circulations during winter over the metropolitan area of São Paulo, Brazil. Bound.-Layer Meteor., 122, 43–65, https://doi.org/10.1007/s10546-006-9091-3.
Fugal, J. P., and R. A. Shaw, 2009: Cloud particle size distributions measured with an airborne digital in-line holographic instrument. Atmos. Meas. Tech., 2, 259–271, https://doi.org/10.5194/amt-2-259-2009.
Han, Z. S., J. E. González-Cruz, H. N. Liu, D. Melecio-Vázquez, H. Gamarro, Y. H. Wu, F. Moshary, and R. Bornsteine, 2022: Observed sea breeze life cycle in and around NYC: Impacts on UHI and ozone patterns. Urban Climate, 42, 101109, https://doi.org/10.1016/j.uclim.2022.101109.
Hock, N., F. Zhang, and Z. Pu, 2022: Numerical simulations of a Florida sea breeze and its interactions with associated convection: Effects of geophysical representation and model resolution. Adv. Atmos. Sci., 39, 697–713, https://doi.org/10.1007/s00376-021-1216-6.
Huang, Y., Y. Liu, Y. Liu, H. Li, and J. C. Knievel, 2019: Mechanisms for a record-breaking rainfall in the coastal metropolitan city of Guangzhou, China: Observation analysis and nested very large eddy simulation with the WRF model. J. Geophys. Res. Atmos., 124, 1370–1391, https://doi.org/10.1029/2018JD029668.
Jensen, M. P., and Coauthors, 2022: A succession of cloud, precipitation, aerosol, and air quality field experiments in the coastal urban environment. Bull. Amer. Meteor. Soc., 103, 103–105, https://doi.org/10.1175/BAMS-D-21-0104.1.
Kingsmill, D. E., 1995: Convection initiation associated with a sea-breeze front, a gust front, and their collision. Mon. Wea. Rev., 123, 2913–2933, https://doi.org/10.1175/1520-0493(1995)123<2913:CIAWAS>2.0.CO;2.
Kollias, P., and Coauthors, 2025: Experiment of Sea Breeze Convection, Aerosols, Precipitation, and Environment (ESCAPE). Bull. Amer. Meteor. Soc., 106, E310–E332, https://doi.org/10.1175/BAMS-D-23-0014.1.
Logan, T., 2020: An analysis of the performance of the Houston Lightning Mapping Array during an intense period of convection during tropical storm Harvey. J. Geophys. Res. Atmos., 126, e2020JD033270, https://doi.org/10.1029/2020JD033270.
Miller, S. T. K., B. D. Keim, R. W. Talbot, and H. Mao, 2003: Sea breeze: Structure, forecasting, and impacts. Rev. Geophys., 41, 1011, https://doi.org/10.1029/2003RG000124.
Morcrette, C., H. Lean, K. Browning, J. Nicol, N. Roberts, P. Clark, A. Russell, and A. Blyth, 2007: Combination of mesoscale and synoptic mechanisms for triggering an isolated thunderstorm: Observational case study of CSIP IOP 1. Mon. Wea. Rev., 135, 3728–3749, https://doi.org/10.1175/2007MWR2067.1.
Nicholls, M. E., R. A. Pielke, and W. R. Cotton, 1991: A two-dimensional numerical investigation of the interaction between sea breezes and deep convection over the Florida peninsula. Mon. Wea. Rev., 119, 298–323, https://doi.org/10.1175/1520-0493(1991)119<0298:ATDNIO>2.0.CO;2.
Park, J. M., S. C. van den Heever, A. L. Igel, L. D. Grant, J. S. Johnson, S. M. Saleeby, S. D. Miller, and J. S. Reid, 2020: Environmental controls on tropical sea breeze convection and resulting aerosol redistribution. J. Geophys. Res. Atmos., 125, e2019JD031699, https://doi.org/10.1029/2019JD031699.
Reddy, B. R., C. V. Srinivas, S. S. Raja Shekhar, R. Baskaran, and B. Venkatraman, 2020: Impact of land surface physics in WRF on the simulation of sea breeze circulation over southeast coast of India. Meteor. Atmos. Phys., 132, 925–943, https://doi.org/10.1007/s00703-020-00726-5.
Rodwell, M. J., and H. Wernli, 2023: Uncertainty growth and forecast reliability during extratropical cyclogenesis. Wea. Climate Dyn., 4, 591–615, https://doi.org/10.5194/wcd-4-591-2023.
Shin, H. H., and Coauthors, 2021: Large‐scale forcing impact on the development of shallow convective clouds revealed from LASSO large‐eddy simulations. J. Geophys. Res. Atmos., 126, e2021JD035208, https://doi.org/10.1029/2021JD035208.
Skamarock, W. C., and Coauthors, 2021: A description of the Advanced Research WRF Model version 4.3. NCAR Tech. Note NCAR/TN-556+STR, 165 pp., https://doi.org/10.5065/1dfh-6p97.
Spuler, S. M., and J. Fugal, 2011: Design of an in-line, digital holographic imaging system for airborne measurement of clouds. Appl. Opt., 50, 1405–1412, https://doi.org/10.1364/AO.50.001405.
Stein, A. R., R. R. Draxler, G. D. Rolph, B. J. B. Stunder, M. D. Cohen, and F. Ngan, 2015: NOAA’s HYSPLIT atmospheric transport and dispersion modeling system. Bull. Amer. Meteor. Soc., 96, 2059–2077, https://doi.org/10.1175/BAMS-D-14-00110.1.
Trošić Lesar, T., and A. Filipčić, 2022: Lagrangian particle dispersion (HYSPLIT) model analysis of the sea breeze case with extreme mean daily PM10 concentration in Split, Croatia. Environ. Sci. Pollut. Res., 29, 73 071–73 084, https://doi.org/10.1007/s11356-022-20918-3.
Wang, C.-C., N.-C. Su, J.-P. Hou, and D.-I. Lee, 2018: Evaluation of the 2.5-km cloud-resolving storm simulator in predicting local afternoon convection during the summer in Taiwan. Asia-Pac. J. Atmos. Sci., 54, 489–498, https://doi.org/10.1007/s13143-018-0054-7.
Wang, D., and Coauthors, 2022: Linking synoptic patterns to cloud properties and local circulations over southeastern Texas. J. Geophys. Res. Atmos., 127, e2021JD035920, https://doi.org/10.1029/2021JD035920.
Wermter, J., S. Noble, and B. Viner, 2022: Impacts of the thermal gradient on inland advecting sea breezes in the Southeastern United States. Atmosphere, 13, 1004, https://doi.org/10.3390/atmos13071004.
Yoshik Ado, H., 1992: Numerical study of the daytime urban effect and its interaction with the sea breeze. J. Appl. Meteor., 31, 1146–1164, https://doi.org/10.1175/1520-0450(1992)031<1146:NSOTDU>2.0.CO;2.
Yuan, T. L., Z. Q. Li, R. Y. Zhang, and J. W. Fan, 2008: Increase of cloud droplet size with aerosol optical depth: An observation and modeling study. J. Geophys. Res., 113, D04201, https://doi.org/10.1029/2007JD008632.
Zhang, N., and Y. Wang, 2021: Mechanisms for the isolated convections triggered by the sea breeze front and the urban heat Island. Meteor. Atmos. Phys., 133, 1143–1157, https://doi.org/10.1007/s00703-021-00800-6.
Zhang, Y., J. Fan, Z. Li, and D. Rosenfeld, 2021: Impacts of cloud microphysics parameterizations on simulated aerosol–cloud interactions for deep convective clouds over Houston. Atmos. Chem. Phys., 21, 2363–2381, https://doi.org/10.5194/acp-21-2363-2021.
