1. Introduction
There is a growing need to understand the continuity of atmospheric processes on scales ranging from many kilometers down to scales as small as meters. The underlying impetus driving this need is the importance of understanding the far-reaching effects of multiscale turbulent energy cascading, wherein the energy in large-scale atmospheric processes (e.g., Kelvin–Helmholtz instabilities, convection) is transferred down to submeter scales via turbulent cascading. The consequences of this energy cascade are exceedingly important in all aspects of atmospheric dynamics, including weather prediction, pollution transport, diffusion, as well as for the propagation and dispersion of atmospheric gravity waves (Fritts et al. 2009; Fritts and Wang 2013; Fritts et al. 2013). Unfortunately, important details of these processes are poorly understood.
Appreciable inroads into our understanding of these atmospheric-scale interactions require a combination of access to larger volumes and higher spatial resolution observations than are possible with conventional remote or in situ sensing instruments. As discussed in some detail later, volume dimensions in this case can range up to several kilometers, both horizontally and vertically, with spatial resolutions needed down to meters. Specific critical atmospheric variables include temperature, pressure, humidity, wind (velocity and direction), ɛ (the turbulent energy dissipation rate), and 
Over the past few years, relatively large, unmanned aircraft have been employed to provide tropospheric and lower-stratospheric data over extended distances. Their use has been sporadic, however, because of several challenges with this technology as described herein. Recognition of these difficulties is just emerging, along with suitable measurement systems that address the challenges with practical designs. These emerging solutions take the form of smaller, lightweight, inexpensive, autonomous aircraft systems that are capable of collecting simultaneous multisensor data in and above the atmospheric boundary layer with very high resolution. In addition, these systems enable measurements over much smaller, localized regions compared to the larger unmanned vehicle systems, with greatly reduced costs and with much fewer operational difficulties. Examples of these smaller regions include documenting the dynamics around specific orographic features (mountains and narrow valleys), coastal boundaries, and the early evolution of local convection.
These promising capabilities are made possible by recent advances in low-cost sensing, computing, and communication technologies, which are driven by high-volume consumer markets for cell phones, GPS locators, and personal computers. Such developments enable a new class of small, low-cost aircraft with miniaturized sensors and unprecedented autonomy for their size.
Despite the availability of low-cost components, development of practical systems can require significant time and interdisciplinary expertise, so that the resulting development costs can easily fall outside typical scientific budget restraints. Even with a recognized need for improved measurement capabilities, next-generation commercial product development requires sufficiently large perceived markets and significant investment returns. It follows that civilian science applications rarely drive the commercial development of suitable products. On the other hand, engineering research funding is primarily focused on cutting-edge technologies perceived to have high future value; such research is seldom driven by the needs of low-cost near-term scientific applications. The combination of these two factors results in an unfortunate but persistent gap between immediate science needs and new technology development.
This paper describes a joint science–engineering effort to develop and field test the DataHawk small airborne measurement system (SAMS). The DataHawk system (aircraft, ground support, sensors, telemetry, operating modes, etc.) has been specifically designed for high-resolution, multiple-variable, state-of-the-art atmospheric sensing measurements coupled with the concomitant cost constraints of civilian atmospheric science applications.
Section 2 provides a brief outline of the current state of the art in atmospheric sensing, focusing on critical needs and the potential role of SAMS for satisfying these needs. Section 3 describes the development of the DataHawk SAMS specifically, while section 4 describes its current capabilities and operating modes. Section 5 presents typical results gathered from recent field campaigns that illustrate current measurement capabilities using the DataHawk. This paper focuses on the unique capabilities of the DataHawk: the capability to attain altitudes up to 10 km, high vertical resolution of all quantities (including vector wind), simultaneous measurement of ɛ and
2. The role of SAMS in atmospheric sensing
The atmospheric boundary layer (ABL) extends from the earth's surface up to 2–3 km and is important for a variety of reasons, particularly insofar as it serves as the interface between near-earth processes and the free atmosphere. Monitoring the ABL accurately over its entire height range is a formidable task. The important lowest levels are virtually inaccessible to piloted aircraft because of safety concerns. Instrumented towers measure only the first 100 meters or so above the ground and cannot provide high spatial resolution. Remote sensing systems (radars, lidars, sodars) are able to observe to higher altitudes but typically have limited height resolution and maximum altitude limitations. In addition, a given instrument typically measures only one or two atmospheric variables (e.g., winds, temperatures, reflected echo strengths, aerosol densities, or cloud properties). Tethered lifting systems (kites and balloons) can provide simultaneous, very high vertical resolution in situ measurements of a number of variables through most of the ABL under relatively benign atmospheric conditions, but they only provide samples along a near-vertical path. Nontethered balloonsondes and dropsondes provide only a single, nonrepeatable “snapshot” of a limited number of variables along a narrow, near-vertical trajectory through the entire height range, but they cannot be directed to specific regions of interest. Nontethered blimps provide some horizontal controllability but are restricted to operation under very calm conditions.
Thus, in order to improve our understanding of boundary layer processes, there is a clear need to 1) measure a variety of pertinent atmospheric quantities in situ, with close time synchronization; 2) make such measurements over a domain that includes the entire ABL height range; and 3) gather data over specified volumes of interest over reasonably extended horizontal distances. Furthermore, these measurements should be made with high-temporal and high-spatial resolution. Resolutions extending down to scales of a few seconds and meters are needed to understand the effects of small, rapid fluctuations associated with the nonlinear coupling over scales extending from many hundreds of meters down to meters (Cho et al. 2003; Balsley et al. 2008; Tjernström et al. 2009; Fritts et al. 2009; Fritts and Wang 2013; Fritts et al. 2013). Since these scale interactions typically involve turbulence cascading, the need to accurately estimate turbulence properties within the spatial and temporal limits requires even higher bandwidth measurements (i.e., at least 100-Hz temperature and velocity sensor bandwidths). Until recently, existing techniques have been incapable of making such high-resolution measurements throughout the height of the ABL over significant horizontal distances.
The advent of unmanned aerial vehicles (UAVs) has the potential to revolutionize the study of such ABL processes. Although no formal taxonomy exists to categorize the myriad sizes and capabilities of UAVs that are of interest here, they are within the parameters defined (e.g., Miller 2006; Spiess et al. 2007; Nonami 2007) by micro-UAVs (UAVs having gross weights less than about 500 g and wingspans smaller than 0.3–0.5 m), and mini-UAVs (UAVs having gross weights more than 500 g and wingspans between 0.5 and 3.8 m). These small sizes and weights relative to larger UAVs and manned aircraft result in lower costs, improved safety, and ease of operation, and enable access to more localized regions. Moreover, given the same sensor temporal bandwidths, the lower flight speed of these smaller platforms improves sensor spatial resolution.
With additional advances in autopilot electronics, these small aircraft are becoming capable of “supervised autonomy” that enables them to be directed to specific regions of interest, yet otherwise fly autonomously, thereby reducing demands on operator training. As sensors continue to become smaller and less expensive, these smaller platforms are becoming increasingly capable of making high-resolution measurements of many variables simultaneously. In addition, recent advances in electric propulsion using high-energy-density batteries, efficient electric motors, and switching motor controllers enable measurements to be made over extended horizontal and vertical distances.
It is also important to point out that these smaller vehicles are sufficiently inexpensive that they can be flown under marginal conditions that would be considered too challenging for the more expensive larger UAVs (for fear of loss or damage). Furthermore, they are rugged enough to be launched from—and landed on—virtually any terrain. Low costs and autonomous operation also lead to the possibility of flying a number of aircraft simultaneously along coordinated paths within a designated area. Such a capability can provide greatly improved estimates of local gradients, or area-averaged quantities needed to provide input for typical grid sizes in large-scale numerical models (Mahrt et al. 2001).
A 2006 survey (Miller 2006) of existing small UAVs reported 17 different countries using some 67 different platforms. The large majority of these platforms were classified as military surveillance and reconnaissance vehicles. A total of only four small UAVs in that report were designated as instrumented for atmospheric research, namely, the meteorological mini-unmanned aerial vehicle (M2AV) Carolo T200 (2-m wingspan, 6-kg gross weight) (Spiess et al. 2007; Martin et al. 2011); Manta (2.4-m wingspan, 27-kg gross weight); Aerosonde (33-m wingspan, 14-kg gross weight), (Curry et al. 2004); and Irkut 2F (2.0-m wingspan, 2.8-kg gross weight). Based upon the above definition, three of these four could be defined as being mini-UAVs. Subsequent to the Miller (2006) review, however, a number of additional atmospheric research UAVs have been reported, including the small Unmanned Meteorological Observer (SUMO; 0.8-m wingspan, 0.58-kg gross weight; Reuder et al. 2009); KALI (2-m wingspan, 3-kg gross weight; Egger et al. 2002); M2AV ANTEX-X02 (2.4-m wingspan, 10-kg gross weight); Kite Plane (3.08-m wingspan, 16-kg gross weight; Watai et al. 2006); and the CryoWing (3.8-m wingspan, 30-kg gross weight). The University of Colorado's DataHawk (1-m wingspan, 0.7-kg gross weight) described herein lies at the small end of the mini-UAV category. It is most closely comparable with the SUMO, but provides some significant new capabilities, as described in this paper. This size and weight range appears to be optimal for atmospheric measurement applications. Larger vehicles fly faster and are more efficient, owing to larger Reynolds numbers, but they suffer from cost, safety, and operations difficulties. Smaller vehicles have lower efficiency, hence limited flight duration, and their low airspeed leads to extreme susceptibility to typical wind conditions.
Given the proliferation of UAV technology, it is worthwhile to question why has there been so little “mainstream” success in utilizing these vehicles for atmospheric sensing applications. Despite the promising aspects mentioned above, there are at least five persistent challenges associated with this technology, namely, systems, cost, safety, sensors, and training. These challenges are articulated below to help frame a discussion of the difficulties and prospects for more widespread benefits from this technology.
a. Systems
The vehicle that carries the sensors is the most outwardly visible aspect. However, the vehicle is only part of an operational sensing system. UAV platforms are available from high-end commercial/military vendors all the way down to do-it-yourself radio control (RC) kits. Unfortunately, very few of these platforms have been designed with atmospheric sensing in mind. Suitable sensing systems can be very difficult to construct from an assembly of off-the-shelf airframes, autopilots, radios, antennas, sensors, software, ground stations, etc. Also, very few science programs have the wherewithal and engineering capability to put such a system together. The DataHawk project (section 3) provides an illustrative example of an integrated design of a SAMS for a specific sensing capability, whose success derives from a close and lengthy collaboration between engineers and scientists.
b. Cost
Although larger vehicles can carry more payload and fly longer and faster, they are subject to the insidious aerospace “upward cost spiral,” namely, where loss of the vehicle is so costly and dangerous that extreme measures are required to avoid failure, thereby making the vehicle more costly, etc. Even smaller vehicles can fall prey to these cost increases (albeit at lower cost levels), resulting in vehicles that are too valuable to lose (or, indeed, even acquire) unless explicit steps are taken in the specification of requirements and in the design of the system to reverse the cost drivers (see section 3).
c. Safety
The Federal Aviation Administration (FAA) quantifies safety by a two-variable system, relating the probability of an incident to the severity of the consequences. Current FAA rules (FAA 2013) require all aircraft in the United States civil airspace to have “an equivalent level of safety” to that of a manned aircraft. This results in platform airworthiness certification processes similar to manned aircraft, including requirements for a “see and avoid” capability in uncontrolled airspace, and transponder and human air traffic control communication capabilities in controlled airspace. No UAV has yet met these general requirements.
Public aircraft (those owned and operated by government entities or public universities) can apply for a Certificate of Authorization (COA) for specific vehicles, locations, and operating modalities. However, COAs are typically granted only in very limited circumstances, and typically require the pilot in command to maintain visual contact with the vehicle to provide “sense” deconfliction with other air traffic, and to satisfy the need for positive control by a certified pilot to safely “avoid” such incidents.
Alternative options exist for operation in 1) the United States within (military controlled) restricted airspace; 2) other countries, subject to their local airspace authority; and, in some instances, 3) international airspace beyond the 12 n mi offshore boundary. None of these options, however, can be characterized currently as easily permissible.
It is worthwhile pointing out that the FAA flight permission procedures are in flux, as a result of pressure being applied behind the scenes by commercial unmanned aircraft system (UAS) vendors, public safety organizations, scientific investigators, and other advocates. Specifically, there have been recent concrete signs of changing policy, for example, the FAA Modernization and Reform Act of 2012 (Pub. L. No. 112-95). This act stipulates that some UAVs (e.g., vehicles under 25 kg in first-responder public safety use) no longer require an individual COA. Although the constraints on access to U.S. airspace in the future are far from certain, this picture can be expected to improve as UAS technology itself evolves and as the benefits of UAV in civilian applications become more apparent.
d. Sensors
Airborne platforms pose a unique challenge for sensor technology. Size, weight, and power seldom drive ground-based sensor development, so many off-the-shelf sensors are either unsuitable or demand overly large vehicles. In addition, many sensors have inherently slow response times, resulting in poor spatial resolution when moving at typical flight speeds of a few tens of meters per second. There are additional challenges involving platform vibrations and wide ambient temperature variations in the ABL. In short, few sensors can be simply attached to a UAV; hence, the process of adapting and validating new sensors for this use can be time consuming and costly.
e. Training
Fielding a conventional UAV is difficult in many science applications. Specialized pilot training is usually needed to safely operate the vehicle, often necessitating a dedicated staff position. Additional support personnel are needed to spot other air traffic; and to assist in general transportation, preparation, launch, recovery; and to reduce the pilot workload by operating the payload sensor system during flight.
As discussed in the following section, the DataHawk SAMS was conceived specifically to address all the above-mentioned challenges in atmospheric measurement applications.
3. The DataHawk: Design approach and key attributes
The DataHawk system was developed from a fortuitous overlap of interests between the University of Colorado's (CU) Aerospace Engineering Sciences Department and CU's Cooperative Institute for Research in Environmental Sciences (CIRES). This overlap provided the opportunity to tailor the development of a small UAV prototype (Pisano et al. 2007; Pisano 2009) to address the emerging need for finescale measurements of atmospheric variables throughout the troposphere and lower stratosphere (Balsley et al. 2008; Reuder et al. 2009; Tjernström et al. 2009). The alignment of this engineering technology “push” and science application “pull” has resulted in a successful cooperation that has permeated all aspects of the DataHawk development, from establishing measurement objectives to sensor and platform and software design, and ultimately to successful field operations.
From the beginning, the DataHawk system was developed to address the measurement needs and challenges identified above, but cost was the largest driver, since few research budgets can accommodate the cost of a conventional UAV. As an example, the Aerosonde (Curry et al. 2004), which was designed specifically for atmospheric measurements, cost more than two orders of magnitude more than the design target for the DataHawk (~$500 per vehicle). This is also substantially less than the ~3k euros unit cost of the SUMO (Mayer et al. 2012). Furthermore, the proposed low cost of the DataHawk (including acquisition, maintenance, and replacement costs) would enable the fielding of a small fleet of vehicles. This was a desirable development for a variety of reasons: it would enable 1) simultaneous, spatially distributed measurements; 2) sequential relays for long-duration measurements; and 3) the availability of backup platforms in the event of loss or damage. The use of low-cost vehicles would also reduce concerns over operating the vehicles under the less-than-optimum conditions that would ground larger, more expensive aircraft. Inexpensive platforms could even enable operations that might involve expendable vehicles.
To reverse the “upward cost spiral,” it was necessary to take an unorthodox development approach: Instead of the traditional “flow down” procedures from science objectives to engineering design requirements, a highly collaborative process was followed, where science needs and emerging technology opportunities were jointly considered by the team. A design space resulted that was firmly constrained by cost, in addition to the additional constraints that derived from the other general challenges mentioned above. These additional constraints are listed below.
Vehicles must be small enough to be easily transported, set up, launched, and recovered by one person. Launch and landing must be possible from unimproved surfaces, using low-cost, lightweight, transportable ground support equipment.
The entire sensing system (SAMS) must be operable by a team of at most two people, without specialized UAV or RC piloting skills.
Vehicles must be inherently safe to operate, posing no danger to operational personnel, other personnel, or property. They must operate in a way that makes them unable to stray from predesignated flight boundaries, and visible to other air traffic that might encroach in the airspace. They must have a simple, predictable “deconfliction” scheme, should encroachment by other air traffic occur. Note these safety aspects generally are not sufficient (e.g., thus far for the FAA), although they significantly reduce the cost and risk to a level that is more manageable and they greatly assist in making the safety case for operation in any particular airspace.
The sensing and operating software must be completely under the control of the development team, so that system behavior can be precisely tailored as needed, and so that changes can be easily managed (e.g., to integrate new types of sensors or operating modes). Research programs tend to evolve rapidly, as new understanding is gained, and new questions come to the fore. It is therefore crucial that the sensing system evolve apace.
Within these constraints, the DataHawk design sought to maximize in situ measurement capabilities with particular focus on improving spatial and temporal resolution, and multisensor sampling (temperature, pressure, humidity, wind, and both
The resulting DataHawk SAMS (see Figs. 1 and 2) satisfies the above-mentioned requirements reasonably well, and has the following attributes:
Small size (0.7 kg, 1.0-m wingspan) for ease of handling, and a rugged, off-the-shelf, flying wing airframe, constructed of elastic foam (ParkZone F-27C Stryker). This basic $20 airframe is further reinforced by carbon fiber spars and fiber tape in key locations and the folding propeller is rear mounted using a custom aluminum frame for durability in rough field landing. Both bungee-launch and balloon-drop deployment modes employ a custom-made remote-release nylon line cutter. Ground support equipment consists of a laptop computer (2 kg), and a tripod-supported radio antenna (5 kg), and a 2.4-GHz transceiver radio link. Ground-based launches use a collapsible bungee launcher (2 kg), while balloon drops use 200-g helium-filled radiosonde-type balloons.
The DataHawk can operate semiautonomously or completely autonomously. The operator provides parameters via the ground station but does not pilot the airplane. Parameters can be modified in flight. The primary flight mode is an Auto-Helix, where the vehicle circles autonomously around designated GPS (latitude and longitude) coordinates, with a specified radius, and moves vertically at a specified rate between specified ceiling and floor altitudes. This is accomplished with a novel vector field guidance law (Lawrence et al. 2008) in the low-cost custom-built autopilot (Pisano et al. 2007) based on an 8-bit microcontroller. This guidance law causes the vehicle to be globally attracted to a circular helix, with prescribed center, radius, ascent and descent rates, and altitude ceiling and floor limits. The Auto-Launch (bungee) mode is triggered by a ground station command, as is the Auto-Drop (balloon) mode. An Auto-Land procedure glides the vehicle to a specified ground location. In the event of a problem, the flight-termination mode is either invoked automatically by an anomaly, or invoked manually on the ground station for airspace deconfliction. In either event, the vehicle is brought down rapidly in a tight spiral. All modes and mode parameters are controllable in real time from the ground station. All flight operations are monitored by a pilot in command (who issues flight mode and parameter changes as needed, based on the telemetered data available online from the ground station), and an observer (who monitors the flight range for potential hazards and controls the ground station antenna-pointing direction, as required).
Electric (battery powered) propulsion precludes flammable fuel problems. Low fight speed (11–15 m s−1) and a soft foam airframe eliminate injury to ground personnel or damage to property. Note that the Auto-Helix flight profile is easily detected from other aircraft because of the constantly varying aspect angle of a tightly banked turn (similar to raptors circling in a thermal). When activated, the flight termination mode yields right-of-way to all other air traffic, causing the DataHawk to rapidly descend (with higher bank angle, hence even more visibility) to avoid airspace conflict. This is a simple, predictable deconfliction technique that is similar to general aviation experience with birds: they typically drop to avoid striking an airplane.
The DataHawk and its sensors are managed by a custom autopilot (CUPIC) with custom software (in C) developed by the authors. This provides a flexible system with complete control over all aspects of sensing (signal conditioning, resolution, sampling rate) and semiautonomous vehicle behavior (low-level control, semiautonomous guidance, radio telemetry). The custom lightweight high-resolution sensors were developed as an integral part of this hardware/software system. Figure 3a shows a close-up of the custom high-bandwidth cold-wire temperature sensor. This is mounted on a boom above the airframe to isolate it from motor vibrations, place it in the free stream, and protect it from damage during rough-field landings. Figure 3b shows the custom Pitot-static probe that protrudes into the free stream from its mounting in the foam fuselage.
(a) The DataHawk vehicle, illustrating its foam construction, rear-mounted propeller, and the two turbulence sensors. (b) Typical field operation, with two laptop computers and two separate antenna/tripods (for redundancy). The 12-V battery provides one day of power for the computers via an alternating current (ac) inverter (not visible).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) The DataHawk vehicle, illustrating its foam construction, rear-mounted propeller, and the two turbulence sensors. (b) Typical field operation, with two laptop computers and two separate antenna/tripods (for redundancy). The 12-V battery provides one day of power for the computers via an alternating current (ac) inverter (not visible).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) The DataHawk vehicle, illustrating its foam construction, rear-mounted propeller, and the two turbulence sensors. (b) Typical field operation, with two laptop computers and two separate antenna/tripods (for redundancy). The 12-V battery provides one day of power for the computers via an alternating current (ac) inverter (not visible).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Examples of two existing launching procedures: (a) illustrates a computer-controlled bungee launch (just after release) for flights to altitudes of about 2 km AGL and (b) shows the technique for launching the DataHawk from beneath a standard meteorological balloon. The DataHawk is released upon command when the balloon reaches a desired altitude (up to at least 9 km MSL).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Examples of two existing launching procedures: (a) illustrates a computer-controlled bungee launch (just after release) for flights to altitudes of about 2 km AGL and (b) shows the technique for launching the DataHawk from beneath a standard meteorological balloon. The DataHawk is released upon command when the balloon reaches a desired altitude (up to at least 9 km MSL).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Examples of two existing launching procedures: (a) illustrates a computer-controlled bungee launch (just after release) for flights to altitudes of about 2 km AGL and (b) shows the technique for launching the DataHawk from beneath a standard meteorological balloon. The DataHawk is released upon command when the balloon reaches a desired altitude (up to at least 9 km MSL).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) Custom cold-wire turbulence sensor, mounted at the end of a carbon fiber boom above the DataHawk airframe. This sensor is calibrated in postflight analysis by comparison with a slower, collocated calibrated semiconductor temperature sensor. The 5-μm-diameter cold-wire sensing element has a bandwidth of 300 Hz, but this is currently sampled at 100 Hz. Protective aluminum shroud also houses the Honeywell humidity sensor. (b) Pitot-static probe based on a commercial differential pressure sensor (Freescale) and custom electronics.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) Custom cold-wire turbulence sensor, mounted at the end of a carbon fiber boom above the DataHawk airframe. This sensor is calibrated in postflight analysis by comparison with a slower, collocated calibrated semiconductor temperature sensor. The 5-μm-diameter cold-wire sensing element has a bandwidth of 300 Hz, but this is currently sampled at 100 Hz. Protective aluminum shroud also houses the Honeywell humidity sensor. (b) Pitot-static probe based on a commercial differential pressure sensor (Freescale) and custom electronics.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) Custom cold-wire turbulence sensor, mounted at the end of a carbon fiber boom above the DataHawk airframe. This sensor is calibrated in postflight analysis by comparison with a slower, collocated calibrated semiconductor temperature sensor. The 5-μm-diameter cold-wire sensing element has a bandwidth of 300 Hz, but this is currently sampled at 100 Hz. Protective aluminum shroud also houses the Honeywell humidity sensor. (b) Pitot-static probe based on a commercial differential pressure sensor (Freescale) and custom electronics.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
4. DataHawk sensing capabilities and operation
a. Sensing capabilities
A complete DataHawk small UAS (sUAS) consists of one or more aircraft platforms with onboard sensors, in addition to a single ground control station (laptop computer, data transceiver, and a tripod-mounted steerable antenna). One ground station is currently capable of controlling up to five aircraft (while only one aircraft has been used in the examples shown here, the vector field guidance technique is highly scalable).
The characteristics and capabilities of the system are quantified in Table 1, which also includes a description of the atmospheric sensors currently being used. The total sensor package includes a 5-Hz GPS; a Pitot-static tube (for airspeed and ɛ measurements) plus inexpensive, lightweight sensors for calibrated temperature, pressure, and relative humidity; as well as an in-house-designed high-bandwidth cold-wire turbulence probe used for both accurate wide-bandwidth temperature measurements as well as
DataHawk sensing system characteristics and capabilities.
The DataHawk airframe selection and propulsion system design were selected with low cost in mind, but also appropriate wing-loading and power efficiency for atmospheric measurement objectives. Specifically, a slower flight speed (lower wing area per unit weight) was desired compared to typical RC recreational aircraft (such as the SUMO airframe) to maximize spatial sensor resolution. Also, typical RC products have very large thrust capabilities that are optimized for high climb rates and short-duration aerobatic flight. The DataHawk propulsion system was desired for long duration (about twice that of the SUMO), with more limited climb rates—a combination that provides higher vertical-resolution measurements and comparably large altitude and lateral range.
Measurement of both
Although mean wind over several tens of seconds can be estimated over reasonable portions of a flight circle (Reuder et al. 2009; Balsley et al. 2013), higher-resolution estimates of the wind vector are currently obtained with the DataHawk in two ways: first, 1–3-s averaged wind speed and direction can be deduced via a postanalysis of GPS ground speed and heading, combined with Pitot airspeed (D. Lawrence and B. Balsley 2013, unpublished manuscript). Each such measurement establishes a circle of possible horizontal wind vectors. Two successive measurements, for example, over a 3-s interval, provide two solutions (the intersection of two circles), of which the smaller-magnitude one is the unique physical solution for the average wind vector over that interval. This solution is well conditioned provided the vehicle is turning (so the two solution circles are distinct), and it does not require the attitude of the vehicle to be known. See Fig. 4 for typical results, reported as horizontal wind magnitude and direction as a function of altitude. At the 2 m s−1 descent rate in this case, vertical resolution of the wind vector is 6 m, approximately 50 times better than provided by GPS-only (full circle) estimation, for example, with the SUMO (Reuder et al. 2009).
High-resolution profiles of the horizontal wind vector (magnitude in black; direction in gray) between 0 and 3500 m as the DataHawk was descending from an altitude of 9000 m (balloon-release mode) over the Jicamarca Radio Observatory in Peru on 12 Jan 2011. Finescale details visible on both the magnitude and direction profile illustrate the importance of high-resolution data collection throughout the ABL.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
High-resolution profiles of the horizontal wind vector (magnitude in black; direction in gray) between 0 and 3500 m as the DataHawk was descending from an altitude of 9000 m (balloon-release mode) over the Jicamarca Radio Observatory in Peru on 12 Jan 2011. Finescale details visible on both the magnitude and direction profile illustrate the importance of high-resolution data collection throughout the ABL.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
High-resolution profiles of the horizontal wind vector (magnitude in black; direction in gray) between 0 and 3500 m as the DataHawk was descending from an altitude of 9000 m (balloon-release mode) over the Jicamarca Radio Observatory in Peru on 12 Jan 2011. Finescale details visible on both the magnitude and direction profile illustrate the importance of high-resolution data collection throughout the ABL.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
A second approach (still in development) provides 3D inertial wind vector estimates with a temporal resolution of 0.1 s. This technique uses the weather-vaning characteristics of the small airframe that points the vehicle into the relative wind, together with the existing Pitot airspeed sensor to measure the magnitude of the relative wind. Its vector direction is given by the attitude of the vehicle (compass heading and pitch angle). This attitude relative to the local ground frame is determined at 10 Hz using a combination of a low-cost two-axis magnetometer and a custom suite of infrared horizon sensors (some of which are already used for flight control), together with a feedback inversion algorithm to estimate the attitude that produces the magnetic/optical sensor signals. Finally, the relative wind vector is combined with the GPS- and pressure–altitude-derived inertial vehicle velocity vector to produce a 10-Hz estimate for the instantaneous 3D inertial wind vector. Note that at a vertical descent rate of 2 m s−1, this provides wind profiles with submeter vertical resolution. Other approaches have used calibrated multihole Pitot probes to measure the relative wind on larger vehicles that cannot weathervane rapidly to point into the relative wind (van den Kroonenberg et al. 2008). This approach has also been used recently on the small SUMO vehicle (Reuder et al. 2012), although the airspeed data were not synchronized with vehicle attitude to produce wind estimates. It is an open question whether the relative wind fluctuations that are too fast for such a small plane to follow should be considered turbulent and isotropic. If so, only the longitudinal component of the relative wind would be necessary to measure for the estimation of inertial wind. Absent a clear need for 3D relative wind measurement, the DataHawk design has opted for the simplified longitudinal-Pitot approach in order to reduce costs.
b. Operations
In a typical operation, once a field operations site has been selected and both the ground station and the DataHawk have been powered on, the operator initializes the onboard sensors and verifies the onboard control system operation via the ground station link. The vehicle can then be Auto-Launched from the bungee-cord launcher (Fig. 2a) for flights up to about 2 km above ground level (AGL). This is initiated by operator command from the ground station. The plane can also be hand launched; however, for safety (prop clearance) and consistency (with minimal operator training), the autonomous launch is preferred. Other systems (the SUMO, M2AV, etc.) are typically RC piloted during takeoff, increasing demands on operator training and operations costs.
For high-altitude measurements (currently up to about 9 km AGL), the DataHawk can be Auto-Dropped (Fig. 2b) from a 200-g meteorological balloon. In the balloon-release mode, the release command can be manually initiated by the operator at any time, with a backup based on a specified ascent time or altitude limit in case of a communication failure. For additional security, a separate (timed) fail-safe release mechanism is inserted between the plane and balloon to cut the plane away, even if the release on the plane ceases to function. The DataHawk system was the first to demonstrate such an ability to make dynamics measurements over the entire height range of the ABL and into the lower troposphere. One subsequent balloon-drop example is known, although a much larger and more costly vehicle was used as a carrier for expendable micro-vehicles (Kahn and Edwards 2012). Also, the Skywisp balloon-dropped glider has been used for imaging (SWRI 2012) and greenhouse gas sampling (Favela et al. 2012) but apparently not for in situ atmospheric dynamics measurements.
Following deployment by either technique, the autopilot enters the nominal Auto-Helix mode, where flight path alterations to the vector field guidance law can be accomplished, if desired, at any time during the flight. Sensor data are downlinked in real time to the ground station for display and archiving. The display capability enables flight parameters to be modified as needed, in order to explore specific regions or details of interest that may be observed in the downlinked data. A variety of operating scenarios can be supported by such parametric changes, for example, high-vertical-resolution profiling, repetitive lateral sampling by circling at a fixed altitude, long-distance low-altitude transects between two loiter circle locations. Command modifications transmitted to the aircraft are automatically reflected on the ground station display for verification. A trace of the last 30 s of DataHawk motion is also displayed on a local map image to assess current flight behavior, along with other vehicle health and status data useful in conducting the flight. A separate “science screen” is used to display the real-time sensor data for monitoring and possible flight modifications by the flight science director.
Landing is initiated by the operator via the Land button on the ground station. This mode causes the vehicle to descend on the current helix to an altitude that will enable a gliding descent to the designated landing coordinates, and then automatically peel off the circle on a straight course at a 1 m s−1 glide slope toward the landing target. Landing can occur virtually anywhere, since the specially reinforced foam airframe is lightweight and resilient enough to land on unimproved terrain with minimal damage, and since the atmospheric sensors are located on the top of the vehicle for protection from contact during hard landings. This eliminates the need for an operator with piloting skills to mitigate landing damage (e.g., as used with most other vehicles, including the SUMO), greatly reducing operations costs.
5. Examples of typical results
In this section we present examples illustrating the overall capabilities of the DataHawk sUAS for atmospheric sampling. Specific examples have been chosen to show not only the overall data gathering capabilities but also to demonstrate the capability—and the importance—of simultaneous, finescale sampling of many variables in both time and space, and to heights extending to around 9 km.
All of the measurements made with the DataHawk sUAS described in this paper have been obtained with the permissions of the pertinent airspace authorities. The U.S. flight permission in R-3601A (restricted airspace) was obtained from the Kansas Air National Guard; flights in Peru (both Jicamarca and Paracas) were obtained from the Corporación Peruana de Aeropuertos y Aviación Comercial (CORPAC). It is worth noting that the flights made in Peru (shown in Figs. 6–8) were accomplished using the balloon-release technique, where the DataHawk release occurred at about 9 km AGL.
Prior to presenting results of the actual atmospheric data measurements, we provide examples of the two primary flight modes used for data collection. First, a height-versus-time plot of one typical DataHawk flight mode is shown in Fig. 5. This flight occurred over Paracas, Peru, on 17 July 2011. In this flight the DataHawk ascended from the surface up to 1000 m at 2 m s−1 and then descended at the same rate to the surface, landing some 20 min later. This vertical profile was made with the DataHawk flying a constant diameter helical pattern. Current flights have durations of 50 min in this mode. This flight time can be apportioned between higher-altitude single ascents, or multiple ascents/descents over a narrower altitude range for shorter revisit times, and vertical rates can be set at any desired value less than 5 m s−1.
Height vs time plot of a DataHawk flight over Paracas, Peru, during July 2011. This flight demonstrates the “profiling” flight mode in which the aircraft ascended from the surface to 1000 m before descending back to the surface at the same 2 m s−1 rate. During the entire flight, the DataHawk was flying in a helix of 50-m radius.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Height vs time plot of a DataHawk flight over Paracas, Peru, during July 2011. This flight demonstrates the “profiling” flight mode in which the aircraft ascended from the surface to 1000 m before descending back to the surface at the same 2 m s−1 rate. During the entire flight, the DataHawk was flying in a helix of 50-m radius.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Height vs time plot of a DataHawk flight over Paracas, Peru, during July 2011. This flight demonstrates the “profiling” flight mode in which the aircraft ascended from the surface to 1000 m before descending back to the surface at the same 2 m s−1 rate. During the entire flight, the DataHawk was flying in a helix of 50-m radius.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
A second operating mode is presented in Fig. 6a, wherein the DataHawk ascends/descends “stepwise” in a series of constant-height circular segments under autonomous control. The constant-altitude levels were modified 5 times during the flight, between 8 and 110 m, while the circle radius was held constant at 50 m. Note that the vehicle maintained its designated altitude within ±2 m during each constant-height segment. Lateral tracking accuracy depends on the winds. An example of the lateral accuracy of a series of 12 flight circles at constant height is illustrated in Fig. 6b, which plots the ground track extracted from the 5-Hz GPS signal. The helix diameter in this instance was about 230 m. The lateral accuracy of the actual flight path relative to the desired flight circle, based on this figure, was approximately ±6 m.
(a) Typical height vs time flight of the DataHawk during a 15-min flight at the Smoky Hill Air National Guard (ANG) site in Kansas during March 2011, where the “loiter” altitude was changed 4 times during the flight. Altitudes ranged between 8 and 110 m. (b) Plot of GPS latitude vs longitude recorded during a series of 12 constant-altitude circles during the flight shown in (a).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) Typical height vs time flight of the DataHawk during a 15-min flight at the Smoky Hill Air National Guard (ANG) site in Kansas during March 2011, where the “loiter” altitude was changed 4 times during the flight. Altitudes ranged between 8 and 110 m. (b) Plot of GPS latitude vs longitude recorded during a series of 12 constant-altitude circles during the flight shown in (a).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
(a) Typical height vs time flight of the DataHawk during a 15-min flight at the Smoky Hill Air National Guard (ANG) site in Kansas during March 2011, where the “loiter” altitude was changed 4 times during the flight. Altitudes ranged between 8 and 110 m. (b) Plot of GPS latitude vs longitude recorded during a series of 12 constant-altitude circles during the flight shown in (a).
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Turning to the data presentation, Fig. 7 shows temperature fluctuations recorded during a period when the DataHawk was flying a series of constant-altitude circles 46 m above the ground. This temperature-versus-time plot covers about 10 min in time and shows a small but measureable average temperature increase (dashed line) during this period, as well as quasi-ordered fluctuations that increase in intensity with time. In this example, the average temperature increased by about 0.008°C during the period, while the fluctuation levels were of the order of 0.02°C toward the end of the period. Taking into account the 14 m s−1 airspeed, the horizontal resolution of these measurements reveals structure in the fluctuations ranging from a few meters to many tens of meters.
Example of a temperature vs time plot using data gathered in Kansas, circling at a constant altitude of 46 m. Total flight time was roughly 20 min. Note the slight average temperature increase during the flight (dashed line) and the intense temperature fluctuations that appeared to increase with time.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Example of a temperature vs time plot using data gathered in Kansas, circling at a constant altitude of 46 m. Total flight time was roughly 20 min. Note the slight average temperature increase during the flight (dashed line) and the intense temperature fluctuations that appeared to increase with time.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Example of a temperature vs time plot using data gathered in Kansas, circling at a constant altitude of 46 m. Total flight time was roughly 20 min. Note the slight average temperature increase during the flight (dashed line) and the intense temperature fluctuations that appeared to increase with time.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
The velocity vector profiles shown in Fig. 4 provide clear evidence of the need for relative high-resolution wind measurements. These results, which were obtained using the medium-time-resolution (3 s) technique outlined in section 4, suggest vertical fluctuations of the horizontal wind on scales of a very few meters. Such results are consistent with earlier tethered balloon/kite results that examined vertical gradients in conjunction with dynamic finescale instability (Richardson numbers <0.25) (Balsley et al. 2008; Tjernström et al. 2009).
Two additional features that also show the need for high time and space resolution appear in Fig. 8. The first feature is shown in the potential temperature profiles (red curve) by the “stair step” shape of the profile, with steep gradients visible at about 2500, 2700, and 2900 m. Accurate documentation of these gradients requires a vertical resolution better than a very few tens of meters. The second feature concerns the lack of time resolution of the DataHawk humidity sensor. Note that, while the humidity profile (blue curve) has been “smoothed” by the slow response of humidity sensor (5-s time constant), the profile appears to exhibit hints of local minima at heights corresponding to the heights of the steep gradients in potential temperature. Specific examples are indicated by the double-ended horizontal arrows. Unfortunately, the inherent response time of the humidity sensor was much too slow to fully resolve the extent of these minima, illustrating the need for a high-time-resolution humidity sensor. Such a sensor is presently in development by the authors.
Profiles of the potential temperature (red) and relative humidity (light blue) between 2000 and 3500 m. These data were gathered during the same descent shown in Fig. 7. This plot shows details of the “stair step like” patterns in the potential temperature profile, particularly between 2400 and 2900 m. Heights of the steep gradients in potential temperature appear to correspond to local minima in the relative humidity profiles (horizontal two-headed arrows). Lack of detail in the humidity profile is a consequence of the poor time response of the humidity sensor. Both profiles, however, demonstrate the need for high-resolution measurements.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Profiles of the potential temperature (red) and relative humidity (light blue) between 2000 and 3500 m. These data were gathered during the same descent shown in Fig. 7. This plot shows details of the “stair step like” patterns in the potential temperature profile, particularly between 2400 and 2900 m. Heights of the steep gradients in potential temperature appear to correspond to local minima in the relative humidity profiles (horizontal two-headed arrows). Lack of detail in the humidity profile is a consequence of the poor time response of the humidity sensor. Both profiles, however, demonstrate the need for high-resolution measurements.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Profiles of the potential temperature (red) and relative humidity (light blue) between 2000 and 3500 m. These data were gathered during the same descent shown in Fig. 7. This plot shows details of the “stair step like” patterns in the potential temperature profile, particularly between 2400 and 2900 m. Heights of the steep gradients in potential temperature appear to correspond to local minima in the relative humidity profiles (horizontal two-headed arrows). Lack of detail in the humidity profile is a consequence of the poor time response of the humidity sensor. Both profiles, however, demonstrate the need for high-resolution measurements.
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Finally, the turbulence profiles in Fig. 9 are ideal for illustrating the need to document both the turbulence dissipation rate (ɛ) as well as the temperature turbulence structure constant (
Profiles of both epsilon (blue) and
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Profiles of both epsilon (blue) and
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
Profiles of both epsilon (blue) and
Citation: Journal of Atmospheric and Oceanic Technology 30, 10; 10.1175/JTECH-D-12-00089.1
6. Future potential and limitations
The capabilities of SAMS demonstrated above using the DataHawk provide a promising new method for observing the lower atmosphere over reasonable horizontal and vertical distances. New lightweight, inexpensive sensors are being developed to provide an even more extensive sensing capability. As with any emerging technology, these advantages and capabilities come with concomitant limitations. Both of these aspects are summarized briefly below.
a. Advantages
Both small size and electric propulsion provide a great advantage in SAMS by improving portability and safety. Autonomy is the key to reducing vehicle and operational costs. Continued system advances will certainly enable uses beyond current capabilities. As more SAMS capabilities become available and users apply them to different types of measurements, a wide range of developments can be anticipated. The following points are presented as near-term possibilities:
Documentation of a multiscale, nonlinear process involving gravity wave breaking, turbulence generation, and energy cascading; turbulence measurements require very high-frequency sampling of temperature and velocity to estimate
Three-dimensional measurements of atmospheric quantities over dimensions of a few kilometers (horizontally and vertically) via “formation flying” a number of small platforms simultaneously
Ultra-low-altitude (a few meters AGL) sensing, designed to examine the effects of heterogeneous heating/cooling at ground level as well as to monitor greenhouse gas emissions from thawing tundra
Expendable applications where the vehicle could potentially be lost, for example, severe storms, wildfires, polar science, volcanic emissions
In situ location of cloud margins (base height, top height, lateral extent) using longwave infrared sensors
Launches and landings from watercraft for extended access to marine environments via bungee launch with net recovery (or nearby water landing)
Flights along constant-temperature surfaces (isotherms) where the vehicle can be guided by the onboard sensor data in lieu of conventional GPS-derived flight paths
b. Limitations
Flight efficiency and sensor size are two limitations of SAMS. Smaller airborne platforms have more limited payload mass, size, and power, and are inherently lower in flight efficiency (as noted in section 2), thereby reducing flight duration. Sensor miniaturization and compatible measurement strategies are critical, although the low-cost of some SAMS already enables multiple vehicles to be used to extend sensing duration and spatial coverage significantly.
High-wind operation is an additional limitation. The inherent gust sensitivity of small conventional airframes (including designs similar to the DataHawk and SUMO) is a problem for flying in windy conditions. Strangely, gust sensitivity arises from their stability-enhancing wing sweep, wing dihedral, or high tail designs. Gust-insensitive airframes (Pisano 2009) provide an opportunity to expand operating capabilities in high-wind situations. A gust-insensitive version of the DataHawk is currently undergoing flight tests. In addition, the development of a Lagrangian guidance system (drifting with the mean wind and not being locked to geographic coordinates) enables operation in higher-wind situations, and will provide an opportunity to more thoroughly study the evolution of structures advected by the mean wind.
Although the kinetic energy of a small SAMS vehicle is low, collision with another air vehicle remains a safety concern. This concern is much reduced with the DataHawk class of aircraft because at typical airspeeds (~10–14 m s−1), these smaller vehicles are essentially stationary hazards (similar to birds and free balloons) for the higher-speed, manned aircraft. As this fact is recognized, U.S. airspace access by vehicles in the micro- and mini-classes of unmanned aircraft can be expected to improve, simplifying the current flight authorization processes for these types of vehicles,. Also, since many SAMS sensing applications occur in remote locations that have very little air traffic, flying at such locations could further ease the restrictions. Indeed, the National Aeronautics and Space Administration (NASA) Marginal Ice Zone Observations and Processes Experiment (MIZOPEX) project is currently pursuing expanded access for SAMS in the Arctic (Palo et al. 2012).
7. Conclusions
In broad terms, it is becoming increasingly apparent that small airborne measurement systems (SAMS) are poised to revolutionize atmospheric measurements in the ABL and lower troposphere. Access to this large volume with high-resolution measurements at low cost has been heretofore virtually impossible. The additional advantages of improved safety, ease of operation, and advanced sensing all contribute to this new class of capabilities. The resulting information promises myriad advances in understanding atmospheric dynamic processes, validating numerical models, and improving the accuracy of weather forecasting.
The DataHawk SAMS has demonstrated the feasibility of in situ sensing at spatial resolutions of ~1 m, over a horizontal scale >1 km, and over altitudes ranging from a few meters up to at least 9 km, with controllable horizontal positions and precise climb/descent rates. Cost levels for such platforms are low enough to enable a variety of new sensing schemes, ranging from coordinated multivehicle sensing of parameter gradients to sequential deployment of many vehicles for covering extended time periods to expendable sorties for sensing in dangerous or remote areas. The ability to access the height range above the top of the ABL, using the balloon-drop deployment mode, will provide important new insights into transport and diffusion through the top of the ABL and into the free atmosphere. Examples include the effects of orographically generated atmospheric gravity waves, convective activity, pollution transport and diffusion of ABL effluents, and finescale turbulence cascading at the higher levels.
Finally, the DataHawk's demonstrated ability to simultaneously estimate both
Acknowledgments
We are grateful for the assistance from Eric Payton, (the late) Dr. Rod Frehlich, Dr. Yannick Meillier, and Jon Rush (University of Colorado); Dr. Ronald Woodman (Instituto Geofísico del Perú); and Peru's Jicamarca Radio Observatory staff (in particular, Percy Condor). We also thank Dr. Scott Palo (CU's Aerospace Department) for his efforts in initiating this project. This research was supported by the Army Research Office under ARO Contract W911NF-10-C-0109, the National Science Foundation under Grant AGS 1041963, and EAGER Award 0933997 in addition to an Innovative Research Proposal (IRP) awarded by CIRES.
REFERENCES
Balsley, B., Svensson G. , and Tjernström M. , 2008: On the scale-dependence of the gradient Richardson number in the residual layer. Bound.-Layer Meteor., 127, 57–72, doi:10.1007/s10546-007-9251-0.
Balsley, B., Lawrence D. , Woodman R. , and Fritts D. , 2013: Fine-scale characteristics of temperature, wind, and turbulence in the lower atmosphere (0–1,300 m) over the south Peruvian coast. Bound.-Layer Meteor., 147, 165–178, doi:10.1007/s10546-012-9774-x.
Cho, J. Y. N., Newell R. , Anderson B. , Barrick J. , and Thornhill K. , 2003: Characterizations of tropospheric turbulence and stability layers from aircraft observations. J. Geophys. Res., 108, 8784, doi:10.1029/2002JD002820.
Curry, J., Maslanik J. , Holland G. , and Pinto J. , 2004: Applications of aerosondes in the Arctic. Bull. Amer. Meteor. Soc., 85, 1855–1861.
Egger, J., and Coauthors, 2002: Diurnal winds in the Himalayan Kali Gandaki Valley. Part III: Remotely piloted aircraft soundings. Mon. Wea. Rev., 130, 2042–2058.
FAA, 2013: Unmanned aircraft systems (UAS) operational approval. National Policy N 8900.207, Federal Aviation Administration, U.S. Department of Transportation, 43 pp. [Available online at http://www.faa.gov/documentlibrary/media/notice/n%208900.207.pdf.]
Favela, K., Tans P. , Jaeckle T. , and Williamson W. , 2012: Microcollection of gases in a capillary tube: Preservation of spatial and temporal resolution. Anal. Chem., 84, 8310–8316.
Frehlich, R., Meillier Y. , Jensen M. , and Balsley B. , 2003: Turbulence measurements with the CIRES TLS (tethered lifting system) during CASES-99. J. Atmos. Sci., 60, 2487–2495.
Fritts, D., and Wang L. , 2013: Gravity wave–fine structure interactions. Part II: Energy dissipation evolutions, statistics, and implications. J. Atmos. Sci.,in press.
Fritts, D., Wang L. , and Werne J. , 2009: Gravity wave–fine structure interactions: A reservior of small-scale and large-scale turbulence energy. Geophys. Res. Lett., 36, L19805, doi:10.1029/2009GI039501.
Fritts, D., Wang L. , and Werne J. , 2013: Gravity wave–fine structure interaction. Part I: Influences of fine-structure form and orientation on flow evolution and instability. J. Atmos. Sci.,in press.
Kahn, A., and Edwards D. , 2012: Navigation, guidance, and control of the CICADA expendable micro air vehicle. Proc. AIAA Guidance, Navigation, and Control Conf., Minneapolis, MN, AIAA, AIAA 2012-4536, doi:10.2514/6.2012-4536.
Lawrence, D., Frew E. , and Pisano W. , 2008: Lyapunov vector fields for autonomous UAV flight control. J. Guid. Control Dyn., 31, 1220–1229.
Mahrt, L., Vickers D. , Sun J. , and McCaughey J. , 2001: Calculation of area-averaged fluxes: Application to BOREAS. J. Appl. Meteor., 40, 915–930.
Martin, S., Bange J. , and Beyrich F. , 2011: Meteorological profiling of the lower troposphere using the research UAV “M2AV Carolo.” Atmos. Meas. Tech., 4, 705–716.
Mayer, S., Sandvik A. , Jonassen M. , and Reuder J. , 2002: Atmospheric profiling with the UAS SUMO: A new perspective for the evaluation of fine-scale atmospheric models. Meteor. Atmos. Phys., 116, 15–26, doi:10.1007/s00703-010-0063-2.
Miller, P., 2006: Mini, micro, and swarming unmanned aerial vehicles: A base line study. Federal Research Division, Library of Congress, 57 pp.
Nonami, K., 2007: Prospects and recent research and development for civil use of autonomous unmanned aircraft as UAV and MAV. J. Syst. Des. Dyn.,1, 120–128.
Palo, S., Lawrence D. , Weibel D. , LoDolce G. , Krist S. , Crocker I. , and Maslanik J. , 2012: UAV deployed sensor system for Arctic Ocean remote sensing. 2012 Fall Meeting, San Francisco, CA, Amer. Geophys. Union, Abstract C13E-0665.
Pisano, W., 2009: The development of an autonomous gust insensitive unmanned aerial vehicle. Ph.D. thesis, University of Colorado Boulder, 272 pp.
Pisano, W., Lawrence D. , and Gray P. , 2007: Autonomous UAV control using a 3-sensor autopilot. Proc. AIAA Infotech@Aerospace 2007 Conf. and Exhibit, Rohnert Park, CA, AIAA, AIAA 2007-2756, doi:10.2514/6.2007-2756.
Reuder, J., Brisset P. , Jonassen N. , Muller M. , and Mayer S. , 2009: The Small Unmanned Meteorological Observer SUMO: A new tool for atmospheric boundary layer research. Meteor. Z., 18, 141–147.
Reuder, J., Jonassen N. , and Olafsson H. , 2012: The Small Unmanned Meteorological Observer SUMO: Recent developments and applications of a micro-UAS for atmospheric boundary layer research. Acta Geophys., 60, 1454–1473.
Spiess, T., J. Bange, M. Buschmann, and P. Vorsmann, 2007: First application of the meteorological mini-UAV ‘M2AV.' Meteor. Z., 16, 159–169.
SWRI, cited 2012: Integrated Remote Sensing Imaging System using Skywisp. [Available online at http://www.swri.org/4org/d16/surveng/Skywisp/home.htm.]
Tjernström, M., Balsley B. , Svensson G. , and Nappo C. , 2009: The effects of critical layers on residual layer turbulence. J. Atmos. Sci., 66, 468–480.
van den Kroonenberg, A., Martin T. , Buschmann M. , Bange J. , and Vorsmann P. , 2008: Measuring the wind vector using the autonomous mini aerial vehicle M2AV. J. Atmos. Oceanic Technol., 25, 1969–1982.
Watai, T., Machida T. , Ishizaki N. , and Inoue G. , 2006: A lightweight observation system for atmospheric carbon dioxide concentration using a small unmanned aerial vehicle. J. Atmos. Oceanic Technol., 23, 700–710.
Werne, J., and Fritts D. , 1999: Stratified shear turbulence: Evolution and statistics. Geophys. Res. Lett., 26, 439–442.
