JANUARY 2021 HARRISON ET AL.

63

Demonstration of a Remotely Piloted Atmospheric Measurement and Charge Release

Platform for Geoengineering

R. GILES HARRISON,a KERI A. NICOLL,a,b DOUGLAS J. TILLEY,b GRAEME J. MARLTON,a STEFAN CHINDEA,b
GAVIN P. DINGLEY,b PEJMAN IRAVANI,c DAVID J. CLEAVER,c JONATHAN L. DU BOIS,c AND DAVID BRUSd
a
Department of Meteorology, University of Reading, Reading, United Kingdom
b
Department of Electronic and Electrical Engineering, University of Bath, Bath, United Kingdom
c
Department of Mechanical Engineering, University of Bath, Bath, United Kingdom
d
Finnish Meteorological Institute, Helsinki, Finland

(Manuscript received 26 June 2020, in final form 14 October 2020)

ABSTRACT: Electric charge is always present in the lower atmosphere. If droplets or aerosols become charged, their
behavior changes, influencing collision, evaporation, and deposition. Artificial charge release is an unexplored potential
geoengineering technique for modifying fogs, clouds, and rainfall. Central to evaluating these processes experimentally in
the atmosphere is establishing an effective method for charge delivery. A small charge-delivering remotely piloted aircraft
has been specially developed for this, which is electrically propelled. It carries controllable bipolar charge emitters (nominal
emission current 65 mA) beneath each wing, with optical cloud and meteorological sensors integrated into the airframe.
Meteorological and droplet measurements are demonstrated to 2 km altitude by comparison with a radiosonde, including
within cloud, and successful charge emission aloft verified by using programmed flight paths above an upward-facing surface
electric field mill. This technological approach is readily scalable to provide nonpolluting fleets of charge-releasing aircraft,
identifying and targeting droplet regions with their own sensors. Beyond geoengineering, agricultural, and biological
aerosol applications, safe ionic propulsion of future electric aircraft also requires detailed investigation of charge effects on
natural atmospheric droplet systems.
KEYWORDS: Atmospheric electricity; Cloud droplets; In situ atmospheric observations; Weather modification

1. Introduction We describe a remotely piloted aircraft [also known as an un-

manned aircraft system (UAS)] from which ions of either polarity
Electricity in the atmosphere has long been supposed to
can be released in a regularized manner, also providing an on-
influence clouds of water droplets. For example, Luke Howard
board measuring capability with which the local droplet, ther-
(1772–1864), whose cloud nomenclature system is still widely
modynamic and electrical conditions can be monitored.
used, stated that in nimbus (rain) clouds, water drops ‘‘are by a
A great advantage of charge release as a possible geo-
change in their electrical state made to coalesce, and descend in
engineering approach using airborne platforms is that large
drops of Rain’’ (Howard 1843). This assertion probably arose
volumes of modifying substance are not required to be carried
from the then fashionable interest in electrostatics rather than
aloft. It is consequently well suited to the capabilities of small
observations, but Lord Rayleigh subsequently reported direct
remotely piloted aircraft (RPA), equipped with charge emit-
experiments in which ‘‘instead of rebounding after collision, as
ters and monitoring instrumentation, as summarized in Fig. 1.
the unelectrified drops of water generally or always do, the
The technology developed is described here. Section 2 assesses
electrified drops coalesce’’ (Strutt 1879). More recent experi-
the requirements for an ion release system from which gener-
mental and theoretical work (e.g., summarized in Pruppacher and
ated ions ultimately become attached to water droplets,
Klett 1998) has confirmed that charge does indeed influence
charging them. Section 3 describes the charge emission and
droplet collisions and coalescence, and empirical findings indicate
meteorological sensing technology developed and section 4 the
that regional ionization release is associated with precipitation
integration of this technology with an aircraft. Addressing the
changes (Harrison et al. 2020). Highly charged droplets are also
practical difficulties of flying beyond visual line of sight into
known to disintegrate under intense electric forces (Rayleigh
clouds is a further essential aspect, to obtain good operating
1882; Duft et al. 2003). Here we demonstrate a new enabling
duration (tens of minutes) at significant altitude (to several
technology to modify droplet electrostatics as a potential geo-
kilometers). Section 5 describes trials of the system in specially
engineering technique, through releasing charge from a remotely
arranged airspace and section 6 evaluates the charge emission.
controlled platform capable of entering clouds or aerosol regions.
For such a widespread and fundamental influence as elec-
trostatics on droplet behavior there are many other associated
applications, including in biology, for which droplet charging is
Denotes content that is immediately available upon publica- recognized to enhance insect and foliage deposition (Gaunt
tion as open access.
et al. 2003; Inculet et al. 1981). Investigating the effect of charge
on the efficiency of airborne aerosol sampling provides a further
Corresponding author: R. Giles Harrison, r.g.harrison@reading. application. Beyond aerosol physics, biological systems and geo-
ac.uk engineering, additional motivation is provided by the need to

DOI: 10.1175/JTECH-D-20-0092.1
Ó 2021 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright
Policy (www.ametsoc.org/PUBSReuseLicenses).
Unauthenticated | Downloaded 10/10/22 02:28 PM UTC
64 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

FIG. 1. Conceptual picture of charge emission, droplet charging, and droplet growth in a
cloud (drawn to show droplet growth left to right). Corona ions released by an aircraft be-
come entrained into the cloud, charging the water droplets present by attachment of the ions.
Charging of the droplets modifies the droplet–droplet coalescence, influencing the growth
rate to large drops that ultimately fall out of the cloud as rain. (Droplets lost by evaporation,
or in the case of highly charged drops, charge-induced explosions which occur through
electrical instability, are indicated at the cloud boundaries.)

explore atmospheric consequences of future electric propulsion for artificial charge dispersal might arise from the practical
of aircraft by ion emission and the concept of using ion emission need to remove droplet or aerosol charge, such as in the case of
to neutralize the charge on aircraft to prevent the triggering of release of radioactive aerosol, which can become sufficiently
lightning (Xu et al. 2018; Ieta and Chirita 2019). The net elec- highly charged to be preferentially washed to the surface by
trostatic effects within natural aerosol systems, and their influ- water droplets (Tripathi and Harrison 2001).
ence on detailed microphysical droplet processes leading to rain, Release of corona ions into fogs and clouds has been con-
can be investigated with the new experimental capabilities templated previously and considered for possible hydrological
described. and electrical benefits. After observing a fog near a high volt-
age tower, the inventor and electrical engineer, Nikola Tesla
(1856–1943), said ‘‘I am positive . . . that we can draw unlimited
2. Charge release considerations
amounts of water for irrigation’’ (Cheney 2001). The most well-
Charging of water droplets can be achieved by release of air known artificial charge release work is probably that of
ions into the droplet region (e.g., Gunn 1954). The charge Vonnegut and Moore, in which corona ions were released from
modifies the behavior of the droplets, especially that concerned near-surface high voltage horizontal wires 14 km long (Vonnegut
with droplet–droplet collisions. This is now discussed further, et al. 1962a,b). With this apparatus, it was demonstrated that the
together with estimates of the charge required and generated. charge released modified the initial electrification of small cu-
mulus clouds. Later work (Phelps and Vonnegut 1970) estimated
a. Properties of charged droplets
the charging needed to influence the droplet growth.
When a charged water droplet approaches another water
b. Requirements for charge release
droplet, charged or uncharged, it induces a charge in the sec-
ond droplet, which induces a further charge in the original Introducing charge into an aerosol or cloud can be achieved
droplet, repeating indefinitely. Charged, colliding water droplets through surface or airborne release of air ions. Surface emis-
therefore experience an infinite system of electrostatic image sion systems require extensive installations, and depend on
charges between them, with associated electric forces (Thomson natural updrafts and entrainment processes to allow the gen-
1853; Russell 1922; Davis 1964). Formally, the net droplet– erated ions to reach and enter aerosols or clouds. As substan-
droplet force is always attractive at small separations regardless tial quantities of ions can be generated relatively easily, the
of the droplets’ relative polarities, unless the exact ratios of their inefficiency of the vertical transport process may not matter in
charges would make them an equipotential on contact (Lekner allowing some additional ions to ultimately reach and enter
2012; Banerjee and Levy 2015). With natural variability, this clouds, through following natural updraft routes. The disad-
unique equipotential condition is unlikely to occur, hence two vantage is that, even with large quantities of charge generation
colliding charged cloud droplets can be generally considered as at the surface, assessment of any consequent effects will be
being more likely to coalesce than two neutral droplets. complicated by the wide spatial dispersion of ions likely to be
Therefore, if cloud droplets can be charged artificially, the encountered. Using aircraft to provide targeted charge release
electrical influence on coalescence may, in turn, hasten the controlled from the surface provides a promising alternative,
generation of rain (Harrison et al. 2015). Another application allowing cloud regions to be located where small droplets,

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 65

which are those most likely to be influenced electrically, are efficiency of droplets to near 100%, an oppositely charged droplet
more abundant. In addition, because charge can be generated carrying an order of magnitude more charge than the surrounding
easily electrically, there are no substantial payload require- droplets would be needed. Takahashi (1973) showed that the
ments and hence small aircraft are particularly suitable. average charge on a droplet in a warm cloud was approximately
Although more detailed work at local scales is needed to 1 3 10217 C (;60jej) Thus, for enhanced coalescence, a charge of
fully evaluate the charge required to influence natural aerosols 1 3 10216 C would be needed on half of the cloud droplets.
and clouds, some bounding estimates can be made. The Assuming a cloud droplet concentration of 100 cm23 this would
regional-scale cloud and precipitation changes reported by require a charge delivery rate of 10 nC m23. With the typical RPA
Harrison et al. (2020) were associated with an approximate airspeed assumed of y 5 30 m s21, a charge delivery system would
doubling of the natural ion concentration. Over land surfaces, therefore need to provide 0.3 mA m22 which is similar to that
the typical volumetric ion production rate q0, is about 107 ion estimated above.
pairs m23 s21 (Chalmers 1967). This reduces with height, be- Releasing unipolar charge will also affect the electric po-
fore increasing from cosmic ray ionization above about 3 km. If tential of the aircraft compared with the local environmental
clear air is considered (i.e., neglecting ion removal to aerosol or potential, as the aircraft will develop an opposite charge equal
droplets), the steady-state mean ion number concentration n0 in magnitude to the charge released. The charging rate of the
is given by aircraft can be estimated as
rffiffiffiffiffi dV Ic
q0 5 , (5)
n0 5 , (1) dt C
a
where dV/dt is the rate of change of the potential of the aircraft
where a is the ion-ion recombination rate (1.6 3 10212 m3 s21).
and C is the aircraft’s capacitance. If the aircraft is considered
For q0 5 107 m23 s21, this gives n0 5 2500 3 106 m23 (Harrison
as an isolated spherical capacitor of radius 1 m, C ; 100 pF, and
and Carslaw 2003).
the associated dV/dt for Ic 5 10 nA will be 90 V s21. This is
For an air ion generator operating by corona emission, the
likely to overestimate the charging rate, as any loss of charge
associated unipolar ion production rate, neglecting recombi-
from the aircraft is neglected. This could occur by collision or
nation, is directly proportional to the current flowing to the
attraction of atmospheric space charge, which would act to
emitter tip. If the corona current is Ic, the corona ion produc-
reduce the charging rate.
tion rate Rc will be
The limitations on unipolar charge release implied by
Ic Eq. (5) are important, as if the charge emission continues in-
Rc 5 , (2) definitely, the electric field at the surface of the aircraft will
e
ultimately become dangerously large, leading to systems fail-
where e is the elementary charge (1.6 3 10219 C). If the aircraft ure through electrostatic discharge damage, and possible loss
is in level flight at a speed y, and air ions are emitted in a cy- of the aircraft. (In the case of ion thrusters for spacecraft,
lindrical beam of cross section area S, the instantaneous neutralizers are specifically included to avoid this; e.g., Kent
number of unipolar ions (nc) generated per unit volume due to et al. 2005.) This risk can be reduced by approximately bal-
corona is anced emission of positive and negative charge, as then the
aircraft charging will be less rapid, determined by the differ-
Rc I
nc 5 5 c . (3) ence in the emission currents which is likely to be smaller than
Sy Sye their absolute magnitude. A discharge wick, widely used on
The current required to generate an instantaneous corona ion traditional aircraft, provides another possibility. A conse-
concentration which is a multiple f of the steady-state back- quence of bipolar emission is, however, that the loss of corona
ground ion concentration n0 (i.e., f 5 nc/n0), is therefore ions by recombination will be increased.
rffiffiffiffiffi A controllable RPA charge emission system developed is
q0 now described (section 3) able to provide up to 65 mA of co-
Ic 5 fSye . (4)
a rona current, followed by considerations associated with its
integration into the aircraft (section 4). Flight tests evaluating
For a small aircraft (1 m wingspan) flying at y 5 30 m s21, the meteorological and electrical aspects are described in
emitting an ion plume into an area defined by the wingspan section 5.
(i.e., S 5 1 m2), Ic is found from Eq. (4) for f 5 1 as ;1028 A. If,
as observing smoke plume releases from small aircraft sug-
3. Aircraft charge emitters
gests, the emitted ion plume spreads vertically by up an order
of magnitude more, S ; 100 m2 and the associated Ic required is The charge emitters emit corona ions from a carbon fiber
;1026 A. Emission currents of at least 1026 A (i.e., 1 mA) are brush, raised to a high voltage. Two separate unipolar emitters
realizable, hence f
1 from a practical emission system is are used, controllable to release positive and negative charge
readily obtained. The total cloud volume into which ions are independently. These were designed to have a physical form
released is determined by the flight path and duration. (130 mm 3 40 mm 3 40 mm) and mass (100 g) suitable for
An alternative perspective was provided by Phelps and small aircraft, and to operate from a 12 V power supply. Each
Vonnegut (1970), who estimated that, to increase the coalescence emitter’s current varies with the operating voltage chosen,

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

66 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

FIG. 3. Conceptual diagram of the test system for a single charge

emitter. A controlled high voltage is applied to an emitting tip
FIG. 2. Charge emitter block diagram. A miniature high-tension (black arrow), and the corona current determined using the iso-
generator (EMCO A-series, A60P-5 for positive, A60N-5 for lated measurement system of Fig. 2. The current emitted is also
negative) is used to generate sufficient voltage to generate corona sampled at a nearby detection plate (shown by the black ellipse).
at the emitting tip. The HT voltage is set by the microcontroller, To achieve current balancing, a second opposite polarity emitter
using a 12-bit DAC to control a MOSFET-based op-amp regulator can be applied to the other side of the detection plate.
circuit. The HT voltage is sampled by the same microcontroller
using a 1000:1 resistive divider potential divider, at 10-bit resolu-
tion. The corona current flowing to the tip is sampled on the high- the module, the measurements are returned through an
voltage side (using an AD8293G160 instrumentation amplifier with optical link (e.g., Harrison 2002; Aplin et al. 2008), with its
gain of 160) and digitized at 10-bit resolution by a further micro- supply galvanically isolated from the low voltage section of
controller, with the values transmitted serial over an optically the board. A chain of three transformers (type PT6) with
isolated link to the main microcontroller. Control of the emitters is
their secondaries in series is used to provide a total isolation
achieved by data exchange with the aircraft systems, which also
of 9 kV, using a square wave oscillator drive, as shown in
provide data telemetry to the surface.
Harrison (1997).
The actual output currents from the charge emitters were
which is remotely controllable through the aircraft telemetry characterized using the experimental arrangement summa-
(Fig. 2). The currents supplied to the emitting tips can be rized in Fig. 3. For this, the emitting tip (a carbon fiber dis-
monitored, which allows the positive and negative currents to charge brush) was mounted on a PTFE stand-off, within a large
be balanced, to minimize the aircraft charging hazard. Because grounded diecast box. The discharge brush was pointed at a
the current measurement has to be obtained at the emission brass detector plate connected to a transresistance converter
(high voltage) part of the corona emitter circuit, an isolated (using a 1 MV feedback resistance) circuit, to measure the
system is required to provide the measurement at safe voltages corona ion current flowing to the brass plate at the local ground
for the aircraft’s data telemetry. Communication between the potential. The detector plate was mounted centrally within the
aircraft system and the emitters is therefore required in two box, to allow symmetrical operation of the two emitters either
directions, from the aircraft to the emitter to set the high ten- side of the plate. Through this arrangement, balancing of the
sion operating voltage (which is also confirmed by a signal in output currents from both emitters was achieved by indepen-
the opposite direction), and from the emitter to the aircraft to dently adjusting their operating high-tension (HT) voltages,
report the corona current. This information is recorded by the until the opposite currents were sufficiently similar that no net
aircraft data system. plate current was measured.
In each charge emitter, the operating high voltage is Figure 4a shows the current measured by the detection plate
requested by the aircraft’s datalogger, over a USB–UART as the operating voltage of the corona emitters was varied. The
serial link, and the resulting output current monitored. Within onset threshold for corona emission is also related to the pre-
each emitter, a microcontroller acts as the main control and cise emitter tip shape, but was found to be around 1200 V for
communication link between the aircraft and the device, pro- the negative emitter and 2300 V for the positive emitter. The
viding control of the high-voltage module and monitoring of absolute current measured by the detector plate depended
the output voltage, while another internal UART serial link on the geometry, as varying the distance between the dis-
communicates over an optical isolator with the current sensing charge wick and the plate changed the effectiveness of ion
section. capture and the associated detector plate current. Figure 4b
The ion production is determined by the current flowing shows the operation of the onboard isolated corona current
from the high-voltage module through the discharge brush into measurement circuit when the corona current was varied, by
the surrounding environment. This current is monitored by changing the HT voltage. The linear relationship found
measuring the voltage drop across a series resistor between the between the corona current and HT voltage demonstrates
module output and emitting tip. Since the current sensing cir- that, through adjusting the HT voltage, the emission current
cuit is elevated to the potential of the high-voltage output of can be altered in flight.

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 67

located on the left wing, the negative emitter on the right wing.
This positioning ensures that the corona ions are emitted into
the turbulent flow behind the aircraft, helping to disperse the
ions and ensuring they do not return to the aircraft, which
would modify its charge. The corona emitters can be switched
independently to provide positive, negative, or bipolar ion
emission, using optically isolated switches activated by the pi-
lot through the remote control (RC).

5. Flight tests of aircraft instrumentation

FIG. 4. Tests on the corona emitters. (a) Ion current measured at Separate series of flight tests were undertaken to evaluate
the detector plate of Fig. 3 as the high-voltage setting (HT voltage) the flight endurance and payload capability, meteorological
on the corona tip was varied, in separate experiments. (b) measurements, and charge emission. Calibration information
Relationship between current measured by the onboard corona on the sensors is provided in the appendixes.
current measurement circuit and HT voltage. (In both cases, red
circles are for the positive emitter and blue crosses for the negative a. Aircraft aspects
emitter.) To examine the flight capabilities of the extensively instru-
mented Skywalker airframe, test flights were conducted at the
Pallas Atmosphere-Ecosystem Supersite, in sub-Arctic Finland
4. Aircraft science equipment integration
during the Pallas Cloud Experiment (PaCE 2019) (latitude
The RPA platform chosen for this work is the commercially 68.018N, longitude 24.148E) during September 2019. This site
available Skywalker X8 fixed wing aircraft. Use of a standard had a designated temporary dangerous area (EFD527), per-
platform allows for possible scaling up to a fleet of aircraft. The mitting flights to a ceiling of 2 km above mean sea level (MSL).
X8 is capable of the long-range operations required to fly into Table 1 summarizes all the RPA flights undertaken, including
clouds, including an ability to climb to altitudes of 3 km. It is details of the eleven flights conducted at Pallas. The longest
a flying wing design made of expanded polyolefin foam, with a endurance flight path is shown in Fig. 6.
single folding propeller in a pusher configuration. It has a The maximum altitude reached in this flight was 2000 m
wingspan of 2.1 m and maximum takeoff mass of 5 kg, with MSL in a flight duration of 20 min 45 s. This consisted of an
capacity to carry scientific equipment in a small payload bay at 11 min climb at a 108 angle to 2000 m, followed by a 9 min glide
the front of the aircraft. The ‘‘pusher’’ configuration allows the to landing. The principal battery usage occurred during the
science instrumentation to be located far from the propellers climb, requiring a mean current of 16 A compared with 0.2 A
and noise generating components, reducing electrical interfer- during the descent. Over the entire flight, the total charge
ence on the measurements. In use, the RPA is flown autono- drawn from the propulsion battery was 3850 mA h, of the
mously using a Pixhawk 2.1 Cube autopilot with Arduplane 4000 mA h nominally available.
software (V3.9.6), propelled by a Cobra 3520 550 Kv motor with
b. Meteorological sensors
Aeronaut 13 in. (33 cm) 3 8 propellors and a FrSky Neuron 60
Electronic Speed Controller. UHF control links are made at The meteorological measurements made by the X8 RPA
868 MHz. Separate 3000 and 4000 mA h LiPo batteries are used during flight were compared with nearby meteorological mea-
to power the systems and propulsion motor, respectively. surements made using a balloon-carried instrument package,
The locations of the various science sensors installed on the employing an RS41 radiosonde augmented with additional sci-
aircraft are shown in Fig. 5. As the RPA is designed to fly ence sensors.
within, and sample, cloud properties, it has been instrumented The balloon payload consisted of a standard Vaisala RS41 ra-
with temperature (RSPRO 2.4-mm-diameter bead thermistor) diosonde with an OCS (Harrison and Nicoll 2014) and charge
and RH sensors (Honeywell HIH-4000), and an optical cloud sensor (Nicoll 2013) attached, of identical design to those on the
sensor (OCS) (Harrison and Nicoll 2014) located in the front of aircraft. The add-on sensors were housed in a 3D printed enclo-
the aircraft, pointing downward to minimize water ingress and sure. This enclosure had fixing spikes printed to grip into the
to provide shielding from solar radiation. Atmospheric space RS41’s polystyrene shell, firmly securing the add-on sensors with a
charge density sensors [with both linear (Nicoll 2013) and tensioned cable tie. Data from the sensors were relayed through
logarithmic (Harrison et al. 2017) responses] are also located in the RS41’s telemetry system using the ozone sensor (OIF411)
the front of each wing to monitor the charge environment port, following Harrison et al. (2012). The sensor data were in-
surrounding the aircraft. Data from all the science sensors are terleaved with the RS41’s data stream and recorded by the ground
logged at 1 Hz through a custom-made datalogging board station. The additional data packets were synchronized with the
based on a TinyDuino (an ATMEGA328-based device) as the standard meteorological data after the ascent. The RS41 carried
main processor, carrying its own GPS and data storage. standard temperature and humidity sensors, having a quoted ac-
Mounting positions for the corona emitters are also shown in curacy of 60.018C and 60.1%, respectively (Vaisala 2018).
Fig. 5, on the underside of the wings, approximately 20 cm from An intense observation period was undertaken at the Pallas
the propellor, facing backward. The positive corona emitter is site on 27 September 2019 to compare the balloon and aircraft

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

68 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

FIG. 5. (a) Instrumented Skywalker X8 aircraft in flight, with instrumentation labeled. (b) Arrangement of
sensors and systems on the X8 airframe (not to scale). (c) Detail of the individual science instruments: (c1)
optical cloud sensor, (c2) charge sensors, (c3a) thermodynamic (temperature and RH) sensors, (c3b) removable
protective housing for thermodynamic sensors, and (c4) corona emitter electrode.

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 69

TABLE 1. Summary of instrumented flights conducted.

Date Local time Flight duration (min) Max altitude (m) Location In-cloud duration (min)
24 Sep 2019 1130 15 1000 Pallas 0
24 Sep 2019 1630 15 1450 Pallas 0
25 Sep 2019 1415 16 1450 Pallas 1
25 Sep 2019 1530 17 1450 Pallas 1
26 Sep 2019 1210 22 1575 Pallas 5
27 Sep 2019 0915 17 1950 Pallas 5
27 Sep 2019 1045 21 2050 Pallas 1
28 Sep 2019 0920 17 1150 Pallas 0
28 Sep 2019 1210 15 1400 Pallas 1
28 Sep 2019 1250 20 1315 Pallas 2
1 Oct 2019 0925 18 815 Pallas 0
29 Nov 2019 1345 22 100 Bristol 0
29 Nov 2019 1455 17 100 Bristol 0

systems. For this, a fully instrumented RPA flight into a thin Figure 7c shows the charge density from the port wing-
stratiform cloud was made, followed by an RS41 special sensor mounted charge sensor, plotted alongside the charge density
balloon launch to provide reference data. The radiosonde and inferred from the charge sensor data from the radiosonde. The
aircraft data obtained are now compared. charge was calculated following Nicoll and Harrison (2016). In
Figures 7a and 7b show the standard thermodynamic mete- the cloud at 1700 m the wing mounted charge sensor detected a
orological quantities from the RS41 radiosonde in black, and the maximum positive charge density of approximately 50 pC m23;
X8 aircraft in red. From the temperature and RH data a cloud the radiosonde detected a similar maximum positive charge
layer 100 m thick at approximately 1700 m is apparent. The density of 60 pC m23. Such extensive layer clouds often show
cloud top is capped with a 58C inversion at 1800 m. Figure 7a charging associated with the upper and lower cloud boundaries
demonstrates a 228C cold bias of the temperature sensor on the (Nicoll and Harrison 2016). The two traces demonstrate similar
X8 when compared to the RS41 temperature sensor, which can charge profiles from two different measurement platforms
be corrected in future flights. The X8’s RH sensor tracked the which encountered the same cloud environment. The greater
RS41 sensor closely, except in the cloud-top region at 1800 m variability apparent in the X8 profiles may be due to additional
where it lagged the RS41, taking longer to adjust to the cloud electrical noise from the aircraft systems, or naturally gener-
features. As the radiosonde and the X8 each encountered the ated lateral charge variations.
cloud layer at different speeds, displaced in time, their lag times Finally, cloud droplet number concentrations derived from
cannot be uniquely identified. The response time of the RS41’s the OCS on the two measurement platforms are compared in
humidity sensor is given by the manufacturer (Vaisala 2018) as Fig. 7d. Both OCSs on the radiosonde and X8 aircraft recorded
less than 0.3 s at 208C and less than 10 s at 2408C. peak droplet concentrations of 150 cm23 within the cloud

FIG. 6. (a) Flight path and (b) altitude reached by the X8 during the longest endurance flight undertaken at Pallas at
1045 LT 27 Sep 2019.

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

70 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

FIG. 7. Comparison of radiosonde (RS41, released at 1045 UTC) and aircraft (X8, flown at 1009 UTC) profiles on 27 Sep 2019. These are
for (a) temperature, (b) relative humidity, (c) charge density, found from the portside charge sensor on the X8, and (d) droplet con-
centration, using a nose-mounted optical cloud sensor on the X8. (X8 data are in red and RS41 data in black).

layer. (The method of calculation of the droplet concentration three separate operational stages. Initially, a rectangular flight
from the raw sensor output is described in appendix B.) path conducted at 50 m altitude was used to ensure that the
In summary, the instrumented X8 airframe can provide aircraft was operating correctly. This was followed by level
thermodynamic, electrical, and optical measurements in cloud, flight operations above the EFM. Finally, a circular, unlimited
at up to 2000 m above the surface. loiter pattern was made above the EFM. The mean loiter speed
of the aircraft was 19 m s21 and the total flight time was 17 min.
c. Charge emission aspects
Figure 8a shows details of the flight path, demonstrating the
Further trials were undertaken to test the operation of the level flight operation legs, and the indefinite circular loiter
corona emitters in flight and quantify the emitted charge dur- pattern. The circular loiters were conducted at 15 and 20 m
ing low level flying over a surface electric field instrument. above ground level, with a 50 m radius. Each loiter was planned
Positive charge emission from the aircraft would result in a to position the edge of the flying circle above the EFM.
positive electric field perturbation beneath and a negative field Figure 9 shows the surface electric field E time series during
perturbation for negative charge emission. Flights were per- the X8’s second flight. Markers show when either corona
formed at the University of Bristol’s Fenswood Farm, Long emitter was switched on and off. Whenever one emitter was
Ashton, United Kingdom (51.4238N, 2.6718W). The site is a activated on the aircraft, a transient change in E was detected
large flat agricultural pasture without obstacles. Two flights beneath. For positive corona, E increased and for negative
were conducted on 29 November 2019, under fair weather corona events E decreased. When, however, both emitters
conditions with clear skies and no appreciable local charge were activated there was a negligible change in E, which in-
generation from meteorological processes. (Details of these dicates that the opposite polarities act to cancel the point
further flights are also provided in Table 1.) Detection of the charge, as perceived by the EFM.
aircraft’s charge emission was made using a Chubb JCI131 The densest region of charge emitted by the aircraft can be
electric field mill (EFM), to measure the vertical electric field considered quantitatively to be represented by an equivalent
at the surface. The EFM was mounted on a 3-m-high vertical point charge above the EFM. For a point charge Q, the electric
mast, separately calibrated to correct for the electric field dis- field E induced by the point charge at distance r is given by
tortion due to the presence of the earthed mast. The mea-
surement range of the EFM was 62 kV m21 with a resolution Q
E5 , (6)
of 0.1 V m21, and values logged at 1 Hz. 4p«0 r2
To detect charge emission from the aircraft, a stable and
reliable pattern of corona emission was required, which was where «0 is the permittivity of free space and r becomes the
achieved through conducting flight operations automatically to height of the aircraft above the EFM. By using a smoothing
maintain consistent flight paths. Each mission was divided into spline (with a unit smoothing parameter for a piecewise cubic

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 71

two emitters (calculated to be 5.3 and 22.8 mA for the positive

and negative emitters, respectively). The detected charge was
evidently much less (90%) than the instantaneous charge
emitted. For an emitter current of ;5 mA, a charge of ;5 mC
would be expected to be observed when the RPA passed di-
rectly over the electric field mill. These measurements, when
combined with the findings in Figs. 3 and 4, indicate that most
of the released ionic charge is rapidly dispersed in the atmo-
sphere, to be removed through ion recombination or attach-
ment to boundary layer aerosol (Harrison and Carslaw 2003).
The charge removal to droplets in a cloud situation can be
expected to be similar, with the mixing processes associated
with the turbulent air behind the aircraft acting to spread the
air ions released.
d. Summary
Taken together, the evaluations undertaken in sections 5a–c
show that the objectives stated in the opening paragraph have
FIG. 8. (a) Bird’s-eye view of the flight path of the aircraft been addressed and, specifically, that the instrumented aircraft
showing the square path (light blue) and circular loiter path (green is able to
and purple). (b) Three-dimensional view of the flight path with the
square pattern at 20 m altitude (light blue), 50 m radius loiter at 1) carry scientific payload to cloud-level altitudes, with an
20 m altitude (green), and the 50 m radius loiter at 15 m altitude endurance of 10 min,
(purple), centered on the surface field mill location (dashed red 2) provide thermodynamic meteorological profile information,
vertical line). 3) locate cloud regions through the combination of a rapid time
response relative humidity sensor and an optical cloud sensor
using backscattered light from the water droplets, and
spline interpolation), to detrend the electric field time series to
4) deliver charge in a controllable and monitored manner, of
retain only the transient changes (as shown in Fig. 10), the
either, or both polarities.
emitted charge from the aircraft can be calculated from Eq. (6).
Using this methodology, the mean of the inferred point charges Further, the commercial airframe employed and the stan-
for the five positive and five negative transients observed in dard devices and components used in construction of
Fig. 10 was found to be 0.43 mC when the positive emitter was the instrumentation make the production of multiple air-
activated, and 20.35 mC when the negative emitter was acti- craft readily achievable, to increase the volume of cloud
vated. The small magnitude difference is likely to be associated which can be intermittently sampled or continuously
with the different magnitudes of operating currents from the interacted with.

FIG. 9. Time series of electric field (E) from the Chubb JCI131 electric field mill at Fenswood
Farm on 29 Nov 2019, with the instrumented X8 aircraft flying overhead in different flight
patterns. The flight patterns were (a) loiter but no corona emitters activated and (b) 20 and
(c) 15 m radius loiter with corona emitters cycled. Crosses and asterisks mark when the charge
emission was switched on and off, respectively, with blue and red used to indicate the positive
and negative charge emitter, respectively.

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

72 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

FIG. 10. Detrended electric field (E) from the Chubb JCI131 electric field mill at Fenswood
Farm on 29 Nov 2019, from Fig. 9. (Red and blue crosses identify electric field transients from
which the charge released was calculated.)

6. Conclusions under the UAE Research Program for Rain Enhancement

Science (UAEREP). KAN acknowledges an Independent
The instrumented RPA platform described here generates a
Research Fellowship funded by the Natural Environment
new capability for cloud and aerosol investigations, and for
Research Council (NERC) (NE/L011514/1 and NE/L011514/2).
assessing effects on their electrical behavior following charge re-
Stefan Chindea’s involvement was supported by the NERC
lease. It successfully provided thermodynamic, optical, and elec-
Grant NE/N012070/1. Sam Hyams, James Male, and Freddie
trical properties of clouds at heights up to 2 km, allowing most
Sherratt made valuable contributions to building and flying the
boundary layer clouds to be accessed and studied, as well as mists,
aircraft. The meteorological sensor characterization was un-
fogs, and aerosol plumes. The novel combination of a controllable
dertaken by NERC SCENARIO summer students Heather
bipolar charge delivery system with integrated optical sensors
Jones and Charlie Bell. Test flights in Finland were made dur-
allows cloudy regions to be identified and targeted remotely or
ing the 2019 Aerosol, Clouds and Trace Gases Research
autonomously. Future use of electric aircraft by ionic propulsion,
Infrastructure (ACTRIS) project in Pallas coordinated by
or the neutralization of highly charged particle clouds presenting
the Finnish Meteorological Institute, a cofunded Institute of
electrostatic hazards, illustrate further environmental applications
UAEREP. (ACTRIS-2 received funding from the European
which may benefit from targeted charge release capability.
Union’s Horizon 2020 research and innovation program under
Atmospheric charge release has established biological and
Grant Agreement 654109; ACTRIS PPP is supported by the
agricultural applications and may ultimately have a new use in
European Commission under the Horizon 2020–Research and
geoengineering through providing cloud droplet charging. This
Innovation Framework Programme, H2020-INFRADEV-2016-
work shows that charge delivery into large atmospheric vol-
2017, Grant Agreement: 739530.) The University of Bristol
umes can be effectively achieved by small electrically powered
provided access to Fenswood Farm for flight tests.
aircraft. As the charge is only emitted from a single point,
further work is needed to establish the active area over which
Data availability statement. Data from the instrument tests
the charge is distributed. To achieve a greater effective release
are available from the corresponding author.
area, some alternative approaches could be considered. Fitting
a set of emitters on a larger airframe provides one possibility;
another, with greater volume coverage, would be through
APPENDIX A
implementing an aircraft ‘‘swarm,’’ with multiple aircraft fol-
lowing the same flight pattern and simultaneously releasing
Meteorological Sensor Calibration
charge across a range of altitudes.
Investigating geoengineering applications, whatever their ul- This section discusses the meteorological sensor package
timate societal value, is an increasingly urgent priority which is carried on the aircraft, and the associated calibration. The
directly addressed by this technology. For this new application, sensors consisted of a bead thermistor and an integrated rela-
electrically powered robotic aircraft provide adaptable delivery tive humidity sensor (Honeywell HIH4000).
platforms without combustion products, and the charge released The RSPRO 10 kV bead thermistor (type RS 151–237,
itself leaves no environmentally damaging residues. 2.4 mm diameter) was connected to a 10 kV precision resistor
to form a half-bridge, i.e., a potential divider from a regulated
Acknowledgments. This material is based on work supported supply, giving a voltage output VTHS. The thermistor was
by the National Center of Meteorology, Abu Dhabi, UAE, calibrated against a standard platinum resistance thermometer

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 73

FIG. B1. Comparison of OCS and LOAC devices. Time series showing the analog-to-digital
converter (ADC) voltage counts from the two IR cloud sensor channels (black and red) and the
integrated droplet count (blue) across the 10–30-mm-diameter bins from the LOAC, between
0600 and 1300 UTC 15 Feb 2019.

(PRT), TPRT over a temperature range of 2208 to 408C in an photodiode signal is bandpass filtered to eliminate fluctuating
environmental chamber. Through this, the thermistor bridge daylight, so that only the modulated backscattered signal from
was found to have a first-order linear response of the cloud droplets is retained. The independent square wave
signals driving the LEDs are also used for phase-sensitive de-
TPRT 5 20:397VTHS 2 25:960, (A1) tection of the individual channels, to allow the photodiode
signal to be decomposed into separate responses associated
for TPRT in Celsius and VTHS in volts.
with each LED. Each of the four recovered signals is separately
The HIH4000 humidity sensor was calibrated by placing it in
low-pass filtered and amplified to yield a DC voltage output
an environmental chamber at 208C. The HIH4000s voltage
which is proportional to the backscatter, from which the size
output VRH was calibrated against a Michell dewpoint sensor
and concentration of water droplets are found by calibration.
in the chamber. The relative humidity (RH) within the cham-
The OCS was calibrated against a light optical aerosol
ber was increased from 30% to 100% in 5% steps. A first-order
counter (LOAC), described in Renard et al. (2015). The
response was found between the HIH4000s voltage output
LOAC measures the concentrations of aerosol, dust, and water
VRH and the RH measured from the dewpoint hygrometer
droplets in the size range 0.1 to 50 mm. It operates by pumping
RHD of
air through a laser chamber, with photodiodes mounted at 128
RHD 5 30:547VRH 2 24:607, (A2) and 608 from the laser path to receive light scattered by
droplets and particles. The number of forward scattered
for RH measured in percent and VRH in volts. pulses received at each photodiode gives the concentration.
By comparing the nature of the scattered light at each pho-
todiode, information about the size and type of particle, e.g.,
APPENDIX B
carbon, mineral, ice, or water can also be recovered. The
LOAC returns concentrations in 17 size bins at 1 min time
Cloud Droplet Sensor Calibration
resolution.
The calibration of the optical cloud sensor (OCS) is pre- In a calibration experiment, two OCS devices were mounted
sented here. Its operation is described in Harrison and Nicoll alongside the LOAC above the surface on a 2.5 m mast, ap-
(2014), but for this application it was extended to provide four proximately 500 m from the River Thames in a large flat arable
channels. It consists of four high-power light-emitting diodes field on the University of Reading’s Sonning farm (51.478N,
(LEDs) in an open-path arrangement, with their backscattered 0.898W). This site experiences fog and river mists. The OCS
light sensed by a photodiode mounted behind the LEDs. Two devices were logged by an Arduino microcontroller operating
of the four LEDs are infrared devices (peak emitting wave- in a similar manner to that used on the aircraft logging system.
length: 850 nm), one cyan (505 nm), and one orange (590 nm), This arrangement was deployed in January 2019 for two
each of which is driven by a square wave at a unique frequency months. During 14, 15, and 17 February 2019, fog events lasting
in the range between 1.1 and 1.5 kHz. Any cloud droplets in the several hours occurred at the site.
optical path from the LEDs will backscatter the modulated Figure B1 shows data from a fog event on 15 February 2019.
light, some of which is received by the photodiode. The The fog formed at approximately 0700 UTC and dissipated at

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

74 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 38

REFERENCES
Aplin, K. L., K. L. Smith, J. G. Firth, B. J. Kent, M. S. Alexander,
and J. P. W. Stark, 2008: Inexpensive optically isolated
nanoammeter for use with micro-Newton electric propulsion
technology. J. Propul. Power, 24, 891–895, https://doi.org/
10.2514/1.30994.
Banerjee, S., and M. Levy, 2015: Exact closed-form solution for the
electrostatic interaction of two equal-sized charged conduct-
ing spheres. J. Phys. Conf., 646, 012016, https://doi.org/
10.1088/1742-6596/646/1/012016.
Chalmers, J. A., 1967: Atmospheric Electricity. 2nd ed. Pergamon
Press, 515 pp.
Cheney, M., 2001: Tesla: Man Out of Time. 1st ed. Simon and
Schuster, 396 pp.
Davis, M. H., 1964: Two charged spherical conductors in a uniform
electric field: Forces and field strength. Quart. J. Mech. Appl.
Math., 17, 499–511, https://doi.org/10.1093/qjmam/17.4.499.
Duft, D., T. Achtzehn, R. Müller, B. A. Huber, and T. Leisner,
2003: Rayleigh jets from levitated microdroplets. Nature, 421,
128, https://doi.org/10.1038/421128a.
Gaunt, L. F., J. F. Hughes, and N. M. Harrison, 2003: Electrostatic
deposition of charged insecticide sprays on electrically iso-
FIG. B2. Comparison of OCS and LOAC devices. Infrared lated insects. J. Electrost., 57, 35–47, https://doi.org/10.1016/
channel ADC counts (IR_ADC) of the OCS plotted against the S0304-3886(02)00094-3.
LOAC integrated droplet count (10–30 mm size range) for all fog Gunn, R., 1954: Diffusion charging of atmospheric droplets by
events during 14, 15, and 17 Feb 2019. ions and the resulting combination coefficients. J. Meteor.,
11, 339–347, https://doi.org/10.1175/1520-0469(1954)011,0339:
DCOADB.2.0.CO;2.
1200 UTC. Only the infrared channels of the OCS are con-
Harrison, R. G., 1997: An antenna electrometer system for at-
sidered here. The voltage outputs from the OCS’s two in-
mospheric electrical measurements. Rev. Sci. Instrum., 68,
frared channels are plotted in red and black, with LOAC 1599–1603, https://doi.org/10.1063/1.1147932.
droplet count (integrating across the several size bins that ——, 2002: A wide-range electrometer voltmeter for atmospheric
span the 10–30 mm range) in blue. The time series from the measurements in thunderstorms and disturbed meteorological
two instruments track well, showing the OCS response to fog conditions. Rev. Sci. Instrum., 73, 482–483, https://doi.org/
droplets. 10.1063/1.1435840.
Figure B2 shows the raw ADC counts (IRADC), from the ——, and K. S. Carslaw, 2003: Ion-aerosol-cloud processes in the
infrared channel of one of the cloud sensors plotted against the lower atmosphere. Rev. Geophys., 41, 1012, https://doi.org/
integrated droplet count ND from the LOAC. A least squares 10.1029/2002RG000114.
fit to the data allows ND to be calculated from the OCS’s ——, and K. A. Nicoll, 2014: Active optical detection of cloud
from a balloon platform. Rev. Sci. Instrum., 85, 066104, https://
IRADC, as
doi.org/10.1063/1.4882318.
——, ——, and A. G. Lomas, 2012: Programmable data acquisition
ND 5 (0:47 6 0:03)IRADC 1 (27:05 6 3:01). (B1)
system for research measurements from meteorological ra-
Uncertainties in the fit are given by 95% confidence intervals, diosondes. Rev. Sci. Instrum., 83, 036106, https://doi.org/
10.1063/1.3697717.
implying that the fitted line is robust despite the scatter.
——, ——, and M. H. P. Ambaum, 2015: On the microphysical
Mature fogs often have fairly consistent droplet sizes, hence
effects of observed cloud edge charging. Quart. J. Roy. Meteor.
the scatter evident may indicate changes in the droplet size Soc., 141, 2690–2699, https://doi.org/10.1002/qj.2554.
distribution during the fog evolution. ——, G. J. Marlton, K. A. Nicoll, M. W. Airey, and P. D. Williams,
The derived calibration was applied to both the balloon- 2017: A self-calibrating wide range electrometer for in-cloud
borne and aircraft OCS, as described in the main text. To measurements. Rev. Sci. Instrum., 88, 126109, https://doi.org/
reduce the effects of instrumental drift, the drive signal to 10.1063/1.5011177.
each LED was made steady (i.e., without square wave ——, K. A. Nicoll, M. H. P. Ambaum, G. J. Marlton, K. L. Aplin,
modulation) every 4 min for 10 s, to effectively provide a and M. Lockwood, 2020: Precipitation modification by ioni-
zero for that channel without changing the balance of cur- zation. Phys. Rev. Lett., 124, 198701, https://doi.org/10.1103/
PhysRevLett.124.198701.
rents flowing in the overall device. This reference value was
Howard, L., 1843: Lecture fifth. Seven Lectures on Meteorology,
subsequently subtracted from the observed signal. As noise
2nd ed, Harvey and Dalton, 100–121.
was also present on the OCS channels, the calibration was Ieta, A., and M. Chirita, 2019: Electrohydrodynamic propeller for in-
only applied when the mean backscattered signal from a atmosphere propulsion; rotational device first flight. J. Electrost.,
10 s moving window was greater than the mean and one 100, 103352, https://doi.org/10.1016/J.ELSTAT.2019.05.004.
standard deviation of the background noise from the Inculet, I. I., G. S. P. Castle, D. R. Menzies, and R. Frank, 1981:
whole flight. Deposition studies with a novel form of electrostatic crop

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC

JANUARY 2021 HARRISON ET AL. 75

sprayer. J. Electrost., 10, 65–72, https://doi.org/10.1016/ Atmos. Meas. Tech. Discuss., 8, 1261–1299, https://doi.org/
0304-3886(81)90024-3. 10.5194/amtd-8-1261-2015.
Kent, B. J., K. L. Aplin, L. Wang, S. E. Huq, R. Stevens, Russell, A., 1922: The problem of two electrified spheres. Proc.
A. Malik, and D. Nicolini, 2005: A field effect spacecraft Phys. Soc. London, 35, 10–29, https://doi.org/10.1088/
neutralizer for the LISA Pathfinder mission. Classical 1478-7814/35/1/302.
Quantum Gravity, 22, S483–S486, https://doi.org/10.1088/ Strutt, J. W., 1879: The influence of electricity on colliding water
0264-9381/22/10/047. drops. Proc. Roy. Soc. London, 28, 406–409, https://doi.org/
Lekner, J., 2012: Electrostatics of two charged conducting spheres. 10.1098/RSPL.1878.0146.
Proc. Roy. Soc. London, 468A, 2829–2848, https://doi.org/ Takahashi, T., 1973: Measurement of electric charge of cloud
10.1098/RSPA.2012.0133. droplets, drizzle, and raindrops. Rev. Geophys. Space Phys.,
Nicoll, K. A., 2013: A self-calibrating electrometer for atmo- 11, 903–924, https://doi.org/10.1029/RG011i004p00903.
spheric charge measurements from a balloon platform. Thomson, W., 1853: On the mutual attraction or repulsion between
Rev. Sci. Instrum., 84, 096107, https://doi.org/10.1063/ two electrified spherical conductors. Reprint of Papers on
1.4821500. Electrostatics and Magnetism, Macmillan, 86–97.
——, and R. G. Harrison, 2016: Stratiform cloud electrification: Tripathi, S. N., and R. G. Harrison, 2001: Scavenging of electrified
Comparison of theory with multiple in-cloud measurements. radioactive aerosol. Atmos. Environ., 35, 5817–5821, https://
Quart. J. Roy. Meteor. Soc., 142, 2679–2691, https://doi.org/ doi.org/10.1016/S1352-2310(01)00299-0.
10.1002/qj.2858. Vaisala, 2018: RS41 datasheet. Vaisala Doc., 2 pp., https://www.
Phelps, C. T., and B. Vonnegut, 1970: Charging of droplets by vaisala.com/sites/default/files/documents/RS41-SG-Datasheet-
impulse corona. J. Geophys. Res., 75, 4483–4490, https:// B211321EN.pdf.
doi.org/10.1029/JC075i024p04483. Vonnegut, B., C. B. Moore, G. E. Stout, D. W. Staggs, J. W.
Pruppacher, H. R., and J. D. Klett, 1998: Microphysics of Cloud and Bullock, and W. E. Bradley, 1962a: Artificial modification of
Precipitation. 2nd ed. Kluwer Academic Publishers, 954 pp. atmospheric space charge. J. Geophys. Res., 67, 1073–1083,
Rayleigh, L., 1882: On the equilibrium of liquid conducting masses https://doi.org/10.1029/JZ067i003p01073.
charged with electricity. London Edinburgh Dublin Philos. Mag. ——, ——, ——, ——, ——, and ——, 1962b: Effect of atmo-
J. Sci., 14, 184–186, https://doi.org/10.1080/14786448208628425. spheric space charge on initial electrification of cumulus
Renard, J. B., F. Dulac, G. Berthet, T. Lurton, D. Vignelle, clouds. J. Geophys. Res., 67, 3909–3922, https://doi.org/
F. Jégou, and R. Akiki, 2015: LOAC: A small aerosol optical 10.1029/JZ067i010p03909.
counter/sizer for ground-based and balloon measurements of Xu, H., and Coauthors, 2018: Flight of an aeroplane with solid-
the size distribution and nature of atmospheric particles—Part 2: state propulsion. Nature, 563, 532–535, https://doi.org/
First results from balloon and unmanned aerial vehicle flights. 10.1038/s41586-018-0707-9.

Unauthenticated | Downloaded 10/10/22 02:28 PM UTC