TBSs mainly consisted of two different balloons, a 35 m3 helikite constructed by Allsopp Helikites and a 79 m3 aerostat constructed by SkyDoc^™. The helikite is a balloon–kite hybrid that uses lighter-than-air principles to obtain its initial lift and a kite to achieve stability and dynamic lift, while the larger aerostat uses a skirt instead of a kite to achieve stability in flight. Lifts of both a helikite and an aerostat increase with increasing wind speed, so a relatively stable float altitude can be achieved even at elevated wind speeds. For POPEYE operations, both systems were operated using an electric winch integrated into a dedicated balloon trailer by Sandia National Laboratories. The payload and operating guidelines for the TBSs vary significantly with location and environmental conditions. Generally, the aerostat is operated for total payload weights of 8–27 kg, and the helikite is operated for total payload weights <27 kg. The helikite is not typically operated above 600 m a.g.l., because beyond this altitude the weight of the tether and payload exceed the maximum lifting force of the helikite. The aerostat can be operated at higher altitudes, but due to its larger size is not launched in sustained surface wind speeds >7 m s-1. The helikite is not launched in sustained surface wind speeds >11 m s-1. Operation of either platform is suspended, and the balloon is immediately retrieved if sustained wind speeds at the altitude of the balloon exceed 15 m s-1. In general, the strength of the wind is the main limiting factor governing the launch and final altitude of the TBSs, with rime accretion on the tether, instruments and balloon also contributing to altitude limitations.

POPEYE TBS operations involved a variety of sensors and payloads. To measure the thermodynamic properties of the atmosphere, the TBS team operated multiple different sensor packages from interMet. This includes the interMet iMet-1-RSB radiosonde package as well as the interMet XQ2 sensor packages developed for use on UAS. Additionally, a Silixa XT distributed temperature sensing (DTS) system was flown. This system, which includes a long fiberoptic cable suspended along the tether, provides a high-resolution, continuous measurement of air temperature based on Raman scattering (Keller et al., 2011; de Jong et al., 2015). Using this system, the temperature is typically measured along the length of the optical fiber every 30 to 60 s at 0.65 cm spatial resolution. To provide information on the winds aloft, vaned cup anemometers from APRS World were operated at specified intervals along the tether. It is important to note that while wind speed from these sensors appears to be relatively accurate when compared with Doppler lidar measurements, a variety of factors, including the high-latitude location, make the directional measurement inaccurate. Information on the aerosol particle population was provided using a combination of two Handix Scientific Printed Optical Particle Spectrometers (POPS) and a TSI Condensation Particle Counter (CPC) 3007. The two POPS provide information on the aerosol size distribution for particles between 140 and 3000 nm, while the CPC provides information on the total number of particles between 10 and 1000 nm. Additionally, vibrating wire sensors from Anasphere and the University of Reading provide information on the amount of supercooled liquid water in cloud. These sensors were collectively referred to as “Supercooled Liquid Water Content” (SLWC) sensors. Further details on all of these sensors and the expected level of accuracy (where available) are included in Table 1.

The main role of the TBS in POPEYE was to collect detailed information on the vertical structure of the lower atmosphere over the AMF-3. This provides information on stratification and the temporal evolution of the lower atmospheric structure. Additionally, the TBS is unique in that it is able to fly in and above cloud for extended time periods, providing an opportunity to collect in situ measurements of thermodynamic, aerosol and cloud microphysical properties on low-altitude Arctic clouds. To accomplish this, the TBS was flown as high as weather conditions would permit, conducting repeated profiles with sensors distributed along the tether. While the exact placement of the sensors would change from flight to flight to adapt to the present conditions, in general the system was operated with a cluster of sensors, including a POPS, CPC, iMet and SLWC near the top of the tether under the balloon, a DTS fiber along the entire length of the tether, and subsequent iMet sensors and anemometers below the main package as most desirable based on the meteorological conditions. When flying the aerostat, a second POPS would also be flown to get more detailed measurements of evolution of the aerosol profile in time. A schematic outlining this strategy is included in Fig. 2.

Another instrument platform used during POPEYE was the Datahawk2 sUAS, developed at the University of Colorado Boulder (description of the first version of the DataHawk can be found in Lawrence and Balsley, 2013). The DataHawk2 sUAS is a small (1.2 m wingspan, <1 kg take-off weight), robotic, pusher-prop aircraft designed to operate in a variety of conditions as a flexible and inexpensive measurement platform (see Table 2 for the specifications of the DataHawk2 UAS). The DataHawk2 has been used for a variety of purposes, including the study of turbulence (e.g., Kantha et al., 2017; Balsley et al., 2018) and high-latitude (e.g., de Boer et al., 2016, 2018) deployments. The relatively slow flight speed (14 m s-1, burst up to 22 m s-1) allows the platform to obtain measurements at high spatial resolution when compared to other aerial vehicles. Despite this relatively slow speed, the DataHawk2 has been operated in winds of up to 12 m s-1, making it a robust research platform for the harsh Arctic environment. DataHawk2 flights completed under POPEYE were generally autopilot guided except for during take-off and landing, when they were under the control of a local pilot through real-time telemetry. All flights were completed within radio communication range and within sight of the ground operators and were conducted within restricted airspace (R-2204; see Fig. 1, de Boer et al., 2016) controlled by the US DOE. This allowed operators to adjust the flight plan in real time to meet the needs of the science objectives and adapt to the changing environment. The ground controller and UAS communicate via 2.4 GHz radio with a range of approximately 10 km. Regulations limit DataHawk2 flight to within visual line of sight, meaning that it is not allowed to fly into clouds and must follow VFR weather minimums for operation (14 CFR 91.155). Additionally, winds hamper the operation of the DataHawk2, with DOE ARM guidelines restricting flight when winds top 7 m s-1.

The DataHawk2 carries a variety of sensors to make measurements of the atmospheric and surface states. Custom-built instrumentation includes a fine-wire sensor employing two cold wires and one hot wire. These provide high-frequency (800 Hz) information on temperature and fine-scale turbulence. High bandwidth is enabled by small surface-area-to-volume ratios of very thin (5 µm diameter) wires. In addition, the DataHawk2 carries a custom configuration that includes integrated-circuit slow response sensors (Sensiron SHT-31) for measurement of temperature through a calibrated semiconductor and relative humidity using a capacitive sensor. For POPEYE specifically, the DataHawk2 also carried an E+E EE03 digital temperature and humidity (capacitive) sensor that was externally mounted on the airframe. For information on surface and sky temperatures, DataHawk2s are equipped with up- and downward-looking thermopile sensors (Semitec 10TP583T with custom electronics). These sensors undergo a calibration using targets of a known temperature. Finally, DataHawk2s have also carried the commercially available iMet1 radiosonde package, providing comparative information on position (GPS), temperature (bead thermistor), pressure (piezoresistive) and relative humidity (capacitive), though these sensors were not installed during POPEYE.

The main objective for the DataHawk2 was to obtain as many profiles as possible of the lower atmosphere during daytime hours. To do this, the aircraft was programmed to climb from the surface to the maximum obtainable altitude. This maximum altitude was constrained by the pilot's ability to maintain visual contact with the aircraft (1000 m a.g.l.) or by the cloud ceiling. Because the endurance of the aircraft is approximately 50 min in Arctic operating conditions, the aircraft could generally complete between one and two full profiles before needing to land to change batteries. The turnaround time between flights can be as short as 10 min, but is generally on the order of 15–30 min. Because of the substantial interest in the interplay between thermodynamic and dynamic properties near cloud base, during cloudy conditions, the operators were requested to hold altitude around the cloud base height, as determined from the observatory ceilometer and visual tracking of the aircraft, for 10–15 min to collect statistics of that environment before descending back towards the surface. While the cloud base height is variable, ideally the altitude held by the aircraft would be within 25 m of the mean cloudbase level. Figure 3 provides an illustration outlining this flight pattern.

The DOE ARM program launched Vaisala RS-92 radiosondes on a regular schedule under POPEYE. Due to concerns about operator safety and fatigue, the number of radiosondes launched was scheduled at three per day, with requested launch times of 05:30, 17:30 and 23:30 UTC (21:30, 09:30, 15:30 AKDT) to match the 06:00, 18:00 and 00:00 UTC synoptic times. Radiosonde launches were at times suspended due to dangerous conditions, including the presence of bears on site, or high winds (>13.5 m s-1 sustained and gusting >18 m s-1) which could result in damage to the sensor package if the balloon does not achieve enough vertical lift due to the strong cross wind. Radiosondes are lifted using 350 g balloons with an average ascent rate target of 5.5 m s-1. Radiosonde data from the campaign are available through the ARM data archive (Atmospheric Radiation Measurement program, 2013a).

The presence of the ARM AMF-3, allows us to put the measurements from the radiosondes, TBS and UAS in a broader context. Figure 4 shows measurements from the AMF-3 surface meteorological instrumentation (Atmospheric Radiation Measurement Program, 2013b) over the 3-month POPEYE period. Synoptically, this period featured several driving features. For much of the campaign, there was a stationary area of high pressure positioned over the Gulf of Alaska, and Oliktok Point sat on the gradient between this area of high pressure and transient low pressure systems moving through the Chukchi and Beaufort seas. This generally resulted in west–northwesterly winds during this time period. Some of these cyclones passed closer to shore, thereby directly impacting the Oliktok Point area and creating precipitation events and shifting wind regimes (e.g., 7–10 July; 13 August; 16–17 August; 29–31 August). In late August there was a general shift in the pattern with high pressure beginning to set up over northern Alaska and eventually over the Beaufort Sea to the north. This resulted in a general shift towards easterly winds at the surface. The end of the POPEYE campaign featured a dominant area of high pressure over the area, resulting in weak easterly winds.

Considering the vertical structure of the lower atmosphere, the observations included measurements from a variety of stability regimes. While the presence of the sun in summer months generally results in more adiabatic lower atmospheric states than during other times of year in the Arctic, the data collected indicate sampling of both well-mixed and stratified conditions. This includes several stable boundary layer cases. Additionally, many of the completed flights were flown with some level of cloud cover in place. While the UAS did not sample through the cloud, the TBS was able to do so, providing insight into the thermodynamic and microphysical structure in and around these clouds. Based on ceilometer data from the AMF-3 (Atmospheric Radiation Measurement Program, 2013c), a cloud base was detected during 76 % of the campaign period. Of the times when clouds were detected, 73 % of the cloud bases occurred below 1 km altitude, 21 % occurred between 1 and 4 km altitude, and 6 % were found above 4 km.

In general, it is relevant and important to note that to some extent all of the POPEYE platforms were weather-limited in terms of their operations. Therefore, there is an element of selective sampling to consider when using the collected data sets. Most directly, the TBS and UAS systems were generally not operated during high winds. The UAS additionally had limitations related to visibility. The radiosondes were least impacted, though high winds did also prevent some launches.

Over the 3-month period, there were limited data outages and challenges related to the issues discussed in the previous sections. Figure 5 illustrates the operations completed under POPEYE. The most significant challenge to continuous operations was the electromagnetic interference (EMI) caused by a US Air Force radar station at Oliktok Point, located approximately 150–300 m from the DataHawk2 flight areas. Modifications made to this radar during the POPEYE time window unfortunately resulted in the grounding of the DataHawk2s for their planned second and third deployments. Additionally, this EMI resulted in some resets of the TBS instrumentation, and errors in the TBS GPS readings. In addition, there were some challenges associated with the Arctic weather. Despite it being summer, winds were a challenge to both TBS and UAS flights at times, and also resulted in the cancellation of some radiosonde launches. Wildlife also posed challenges, as the site is visited by both brown and polar bears during the summer months. The local presence of these large creatures generally required that operators ceased outdoor operations, impacting all three measurement platforms. Despite these challenges, the campaign totaled 238 radiosondes launched, 52 TBS flights (134.3 flight hours), and 59 DataHawk2 flights (64.6 flight hours). Figure 6 illustrates the completed flights in time–height space.

A map indicating the horizontal extent of the TBS flights is shown in Fig. 7 (top). The horizontal distances covered are governed by the positioning of the winch trailer for the system, the wind speed, and the amount of tether extended. The points drifting over the ocean surface are the result of erroneous GPS data, likely linked to EMI from the USAF radar system. The distribution balloon altitudes (the highest sampling height for any given TBS operation) is shown in Fig. 7 (bottom) and demonstrates that the balloon typically sampled the lowest 1 km of the atmosphere. Because the balloon can hover at a given altitude for extended time periods, there are multiple peaks in the altitude distribution, notably at around 150, 300, 700 and 1000 m. These altitudes correspond to altitudes chosen for extended sampling during the campaign. Also, a comparison of TBS altitudes with ceilometer-based cloud base measurements indicates that the TBS was operating at or above the lowest detected cloud base altitude 32 % of the time.

A map of the horizontal extent of the DataHawk2 flights is shown in Fig. 8 (top). All flights were conducted in close proximity to the AMF-3 instrumentation, within the restricted airspace outlined under R-2204. The flight patterns consisted of profiling of the lowest 1 km of the atmosphere, as indicated by the probability distribution of altitudes sampled in the lower panel. This distribution is binned by 20 m increments and based on this it becomes clear that most common altitude was between 20 and 40 m above ground level (a.g.l.). From this altitude, the frequency of visiting higher altitudes generally decreases slowly, resulting from limitations imparted by visibility and winds.

Figure 9 provides insight into the statistics of the radiosonde measurements. The right panel indicates the distance away from Oliktok Point that radiosondes traveled over the length of the POPEYE campaign. Within the troposphere (<10 km altitude), radiosondes generally remained within 20 km of the Oliktok Point facility. However, a few balloons traveled as far as 100 km away once in the stratosphere, with most staying within 50 km of the site all the way to the top of the profile. The temperature–height histogram (Fig. 9b) reveals a general cooling of the air with height through the depth of the troposphere, with most profiles cooling from temperatures of 0–10 ∘C near the surface to around -50 ∘C at the tropopause. Additionally, there are indicators of frequent low-level inversions in the lowest 1–2 km. There appear to be two modes of temperatures observed in the stratosphere, with a dominant mode between -40 and -50 ∘C, and a secondary mode at around -55 ∘C. Finally, a two-dimensional histogram of the winds with height (Fig. 6, middle panel) illustrates a broad range of measurements near the surface (0–20 m s-1), with winds generally increasing with height through the troposphere to values ranging between 5 and 50 m s-1. Winds in the stratosphere again decrease to less than 10 m s-1. Figure 10 illustrates time–height cross sections of radiosonde measurements of temperature, relative humidity and wind speed for the duration of POPEYE.

Finally, Fig. 11 provides an initial glimpse into measurements from the POPS sensor on the TBS. The top panel illustrates the cumulative number concentrations sampled, showing that the range of particle numbers measured tended to decrease with height, and that higher concentrations were typically sampled in the lowest parts of the atmosphere. This is likely a result of the numerous near-surface sources associated with oil production facilities in the vicinity of Oliktok Point. Additionally, there is some level of contamination very close to the surface from the diesel generator used to run the TBS winch system. The extent of this contamination is a function of wind speed and atmospheric mixing state and cannot be generally quantified. The second panel shows a similar two-dimensional histogram for particle sizes. This figure illustrates that most particles sampled were around 200 nm and that the balloon typically operated below 500 m. Again. The spread of diameters measured appears to increase with decreasing height. For both of the top panels it is important to keep in mind that white areas do not necessarily mean that zero samples were observed in that bin, as the lowest colorbar bin has some finite edges. The bottom panel shows the relationship between particle size and concentration as a function of altitude as observed over the campaign. There are numerous near-surface (dark blue) points where high concentrations of smaller particles were observed. These observations are complemented by measurements from the CPC for the majority of campaign flights (not shown).