remote sensing

Article
Radiological Surveillance Using a Fixed-Wing UAV Platform
Stef Geelen 1,2, * , Johan Camps 1,2 , Geert Olyslaegers 1 , Greet Ilegems 3 and Wouter Schroeyers 2

1 SCK CEN Belgian Nuclear Research Centre, Boeretang 200, B-2400 Mol, Belgium
2 CMK, NuTeC, Nuclear Technology—Faculty of Engineering Technology, Hasselt University,
Agoralaan Building H, B-3590 Diepenbeek, Belgium
3 SABCA BRUSSELS, Chaussée de Haecht 1470, B-1130 Bruxelles, Belgium
* Correspondence: stef.geelen@sckcen.be

Abstract: A drone–detector system was designed, developed, and tested for radiological monitoring.
The system was tailored to perform measurements during the threat, release, and post-release phases
of a nuclear or radiological event. This allows the surveillance of large areas, with an autonomy of up
to 12 h, in a large range of altitudes above ground level. The detector system was optimized for gamma
spectroscopy, taking into account the available payload for maximum endurance and maximum
detection efficiency using ‘PENELOPE (2018)’ Monte Carlo simulations. A generic methodology was
used to derive quantitative information on radioactivity levels from the raw measured gamma-ray
spectra at different altitudes. Based on the methodology, it was demonstrated that the drone–detector
system can measure the concentration of potassium-40 (K-40) that is naturally present in the soil.
These measurements complied within 30% of the soil sampling results taking into account the
uncertainties. The functioning of the system was tested during test flights, which demonstrated that
radionuclide identification and quantification of radioactivity concentrations are possible.

Keywords: UAV; drones; radiological monitoring; fixed-wing; detector optimization

Citation: Geelen, S.; Camps, J.;
Olyslaegers, G.; Ilegems, G.;
Schroeyers, W. Radiological
Surveillance Using a Fixed-Wing 1. Introduction
UAV Platform. Remote Sens. 2022, 14, Unmanned aerial vehicles (UAVs) have recently seen increased use in the field of
3908. https://doi.org/10.3390/ surveillance and monitoring. Depending on the payload, they can be used for environ-
rs14163908 mental monitoring, fire control, search and rescue, visual inspections, or other applications.
Academic Editors: Pablo Rodríguez- Most systems require a high-resolution or thermal camera. Although it might be useful
Gonzálvez and Eben Broadbent to have visual information, such systems are in principle not required for environmental
monitoring of radioactivity. It is, after all, not possible to detect, identify. and characterize
Received: 27 June 2022
radioactive material with this equipment [1–4].
Accepted: 6 August 2022
Most radioactive materials emit gamma and/or X-rays as they decay. The fact that
Published: 12 August 2022
this type of radiation can be detected over large distances in the environment, due to
Publisher’s Note: MDPI stays neutral the limited attenuation in air, makes gamma spectroscopy the most common detection
with regard to jurisdictional claims in technique of radioactivity using UAVs [5]. While the environmental monitoring of gamma
published maps and institutional affil- and X-ray radiation with UAVs creates new opportunities for radiological protection in
iations. routine and emergency scenarios, as well as for the monitoring of radioactivity in the
environment, it also poses new challenges. A first advantage of the use of UAVs is the
possible autonomy, i.e., not requiring an operator to perform the measurements. With most
radiological surveys, an operator has to systematically map the area of investigation with a
Copyright: © 2022 by the authors.
hand-held detector. Operators could be exposed to increased levels of ionizing radiation,
Licensee MDPI, Basel, Switzerland.
which is a safety concern. Additionally, operators could possibly influence measurements
This article is an open access article
by shielding the detector with their bodies. By using a UAV, both issues are resolved.
distributed under the terms and
conditions of the Creative Commons
No dose is absorbed by operators and the UAV does not shield the detector from the
Attribution (CC BY) license (https://
ground [6,7]. A second advantage is accessibility. People are restricted in their movement
creativecommons.org/licenses/by/ because of terrain or vegetation. Although UAVs must also be operated cautiously near
4.0/). high structures and trees, UAVs are less restricted in their movement and can easily reach

Remote Sens. 2022, 14, 3908. https://doi.org/10.3390/rs14163908 https://www.mdpi.com/journal/remotesensing

Remote Sens. 2022, 14, 3908 2 of 17

difficult-to-access places. A third advantage is the reproducibility of UAV surveys. The

flight plans can be repeated to see the evolution of a contaminated site over time or to
follow up remedial actions [8]. A fourth advantage is the speed of UAVs, making them
able to map large areas in a relatively short timeframe. This is especially important in
emergency scenarios where decisions on protective measures need to be made quickly.
An alternative for UAVs are helicopters. However, helicopters require both a pilot and an
operator, which puts two people at risk of exposure to ionizing radiation.
A downside of using UAVs as an instrument for surveys is the minimal flight altitude
of the UAV, which causes a decreased efficiency of the measurement. Depending on
the geometry of the radioactive source, the reduction in efficiency can be caused by a
combination of the inverse-square law and the attenuation in air. For point sources, the
inverse-square law will cause the most significant part of the reduction, while for a surface
source the attenuation will be the main parameter that influences the efficiency. Besides the
reduction in efficiency, an increase in height also increases the field of view of the detector
and therefore decreases the spatial resolution. The decrease in spatial resolution due to the
field of view can be improved by using a collimator. However, this makes detection more
complex and adds weight to the payload. A second problem specific to battery-powered
drones is their power consumption. On the one hand, if a UAV would fly too slowly over
an area of investigation, it will not be able to carry out the full site survey due to the limited
battery power. On the other hand, by flying too fast, the measurements will be integrated
over a larger area, resulting in a decrease in spatial resolution. A final issue is the presence
of high structures that have to be avoided. This requires pre-flight visits to acquire the
position of the obstacle, and pilots might have to intervene during an automated flight
when there is a risk of collision.
In recent years, new experiences have been reported regarding radiological monitoring
and surveillance using drones. Several studies were performed with both fixed-wing and
multicopter drones incorporating a detector system as payload [9–15]. Depending on the
task, payloads were optimized in regard to their efficiency, detection method, or mapping.
In this paper, we will discuss the development and first results of a new fully optimized
drone–detector system using a fixed-wing UAV. For the intended applications, namely
radiological monitoring in accident scenarios with gamma-ray-emitting nuclides, we do
not require data that are highly spatially resolved. Therefore, the minimal height does not
pose a problem. However, during accidents, it is beneficial to use a system that can be
airborne for multiple hours and that has the ability to identify and quantify the release
or contamination based on gamma spectroscopy. The optimization and development of
the new drone–detector system described in this paper is focused on these parameters
and requirements.
In principle, if the geometry is known, the gamma-ray flux can be calculated. Com-
bined with the characterization of the detector this allows to quantify the source. For
point sources, the gamma-ray flux at the detector location can be easily determined. For
surface contaminations, however, the procedure is slightly more involved. To determine
the gamma-ray flux in this case, the historic work of Beck et al. [16] on in-situ gamma-ray
spectroscopy or the IAEA technical document 1092 [17] can be extended to UAV measure-
ments. Their theoretical frameworks allow for calculating the flux at the detector that
originates from either contamination or naturally occurring radioactivity in the soil. This
requires extensive knowledge of the detector response as well as a good approximation of
the distribution of radioactivity in the soil as a function of the depth.
A spectroscopic system allows us to detect increased radiation levels either as an
increase in the total count rate or as the presence of certain lines in the gamma-ray energy
spectrum. The latter allows for the identification of radioisotopes. However, the geometry
of a source has a significant influence on the measurements. This can make the quantifica-
tion of a radioactive source a challenging task. The relation between the detector signal
(e.g., the net gamma peak area count rates in gamma spectroscopy) as measured during
Remote Sens. 2022, 14, 3908 3 of 17

the drone flight and the radiation source (i.e., the calibration factor Cf , which is specific for
every gamma-ray energy), can in general be written as:

Nf N N ϕ
Cf = = 0 f (1)
I ϕ N0 I

where Nf is the net peak area counting rate obtained from the measured spectrum, I
is the surface or soil concentrations expressed in Bq/m2 or Bq/m3 respectively, N0 is
the net peak area counting rate from a point source measurement with the point source
positioned normal to the detector face, and ϕ the flux of incoming gamma rays expressed in
gammas/cm2 s. Beck et al. and IAEA split the detector response in three factors as shown
in Equation (1). The first factor, N0 /ϕ, is the estimate of the counts per second under the net
peak counting rate due to a unit flux incident normal to the detector face. The second factor,
Nf /N0 , is the angular correction factor for the detector and depends on the gamma energy
incident on the detector. The angular response is integrated over the spatial distribution in
the ground. The last factor, ϕ/I, is the ratio of the incident flux over the concentration.
In this paper, we describe the development and testing of an optimized drone–detector
combination for monitoring during the threat, release, and post-release phases of a nuclear
or radiological event. Further, it is demonstrated that the system at different flying altitudes
is capable of quantifying a radioactive source. This is done for naturally present potassium-
40 in the soil by comparing the UAV results with soil sampling results. This experiment
demonstrates that the system can also be used for other applications outside nuclear and
radiological emergency response.
In Section 2, the fixed-wing UAV parameters are described as well as the design and
hardware structure of the detector system. Additionally, Equation (1) will be explained in
more depth, especially regarding the quantification of radioactive concentration in soil.
In Section 3, we will present the data obtained during two separate flights. The flights
were performed above agricultural fields that contain only natural radioactivity typical for
the region and soil type. The gathered data were then used to calculate the potassium-40
(K-40) concentration in the soil. K-40 is a natural radioactive material that emits a gamma
ray with an energy of 1460.8 keV that can be measured by a gamma spectroscopy device.
To test the detector system, we compared the measured concentrations with concentrations
obtained by taking soil samples at the surrounding fields.

2. Materials and Methods

2.1. Penguin C
The Penguin C is a fixed-wing drone with a gas-powered engine with electronic fuel
injection. It can fly with a maximum payload of 2 kg and has a maximum take-off weight
(MTOW) of 23 kg. With a flight endurance of up to 12 h, it can easily manage to stay airborne
during accidental radiological releases to give live updates on the situation. The wingspan
of the Penguin C is 3.3 m, and it has a length of 2.3 m. A pressurized catapult is used to
launch the drone with an acceleration of 6 G. To cope with these forces, it is necessary to
have a robust detector system. To land, the drone uses a parachute in combination with
an airbag. The landing procedure is started a couple of minutes in advance, taking into
account wind, location, and heading to make sure the drone lands in a clear zone without
trees or other obstacles. The airbag is used to reduce the impact and minimize the risk of
damaging the drone and payload. The drone and ground control station communicate
with each other over long-range data transmission using a tracking antenna. The antenna
has a range of 100 km and can send data at a speed of 12 Mbps. The altitude range of
the fixed wing is between 120 m (400 ft) and 5000 m (FL160). For radiological mapping
using drones, the lowest altitude is determined by the automatic deployment altitude of
the parachute, which Sabca set to a height of around 150 m for safety reasons. This range
is ideal to survey the area of a nuclear site and its surroundings. In the threat and release
phases of a radiological accident, the UAV can fly at high altitudes to avoid contamination
by potentially released radioactive material. In the post-release phase, the UAV can fly at
Remote Sens. 2022, 14, x FOR PEER REVIEW 4 of 18

Remote Sens. 2022, 14, 3908 4 of 17

phases of a radiological accident, the UAV can fly at high altitudes to avoid contamination
by potentially released radioactive material. In the post-release phase, the UAV can fly at
lower altitudes,
lower altitudes, typically
typically just
just above
above common
common structures
structures in
in the
the landscape,
landscape, to
to map
map the
the area
area
in detail. The drone has a nominal cruise speed of 19–22 m/s and a maximum
in detail. The drone has a nominal cruise speed of 19–22 m/s and a maximum speed ofspeed of 32
m/s. Figure 1 shows the fixed-wing drone mounted on the pressurized catapult.
32 m/s. Figure 1 shows the fixed-wing drone mounted on the pressurized catapult.

Figure 1.
Figure 1. Penguin
Penguin C
C UAV
UAVfrom
fromSABCA
SABCAmounted
mountedon
onthe
thepressurized
pressurizedcatapult.
catapult.

The
The flight
flightplan
plancan
canbebecreated
createdand
andadapted
adaptedon on
thethe
spotspot
withwith
the ‘Piccolo’ software
the ‘Piccolo’ [18].
software
The
[18].Piccolo software
The Piccolo can adapt
software the flightpath
can adapt during
the flightpath the the
during flight. ForFor
flight. accident
accidentscenarios,
scenar-
adapting the height
ios, adapting is useful
the height to avoid
is useful tocontamination by the released
avoid contamination by the radiological plume or
released radiological
to improve the spatial resolution and efficiency for monitoring contaminated
plume or to improve the spatial resolution and efficiency for monitoring contaminated sites. For the
performed flights, we used two flightpaths. The first path was a circle with
sites. For the performed flights, we used two flightpaths. The first path was a circle a 300-m radius
with
at multiple
a 300-m heights,
radius while the
at multiple second
heights, paththe
while was a scanning
second path was grid.a scanning
The path grid.
can be closely
The path
monitored since the drone transmits all parameters each 0.1 s. Afterwards, these
can be closely monitored since the drone transmits all parameters each 0.1 s. Afterwards, parameters
can
thesebeparameters
used for post-processing.
can be used for For example, this involves
post-processing. incorporating
For example, the tiltincorporat-
this involves angle into
the angular dependence and correcting the altitude due to topography changes.
ing the tilt angle into the angular dependence and correcting the altitude due to topogra-
phy changes.
2.2. Overal Requirements and Limits
The design
2.2. Overal of the payload
Requirements and Limits has certain limits and requirements. Firstly, the weight
limit of the payload is 2 kg. In case of additional weight, the flight performance of the
The design of the payload has certain limits and requirements. Firstly, the weight
Penguin C is significantly worsened, and adaptations are required for stabilizing the drone.
limit of the payload is 2 kg. In case of additional weight, the flight performance of the
Secondly, to achieve a decent spatial resolution, the detector has to take spectra every
Penguin C is significantly worsened, and adaptations are required for stabilizing the
second. However, to obtain better counting statistics in the spectra, it is possible to combine
drone. Secondly,
multiple to achieve
spectra, which reducesa decent spatial
the spatial resolution,
resolution the detector
but improves has to take
the counting spectra
statistics of
every second. However, to obtain better counting statistics in the spectra,
the spectra. Thirdly, data can be transmitted with a maximum bandwidth of 12 Mbps. The it is possible to
combine multiple spectra, which reduces the spatial resolution but improves
bandwidth is meant to be used for camera footage and therefore it is certainly possible to the counting
statisticstext
transmit of the
filesspectra. Thirdly,
that contain data
time, can be
height, GPS transmitted
coordinates, with
andaspectrum
maximum bandwidth
data, which add of
12 Mbps. The bandwidth is meant to be used for camera footage and therefore
up to approximately 24 kbps. Fourthly, the efficiency of the payload has to be optimized, as it is cer-
tainly
the possible
height to transmit
will cause text filestothat
the efficiency dropcontain
due totime, height, GPS coordinates,
the inverse-square law for point and spec-
sources
trum data, which add up to approximately 24 kbps. Fourthly, the efficiency
and due to attenuation for surface contaminations. Fifthly, because the drone–detector of the payload
has to be
system optimized,
is subject as the
to strong height willand
accelerations cause the efficiency
vibrations, a securetoand
drop due
safe to the inverse-
connection has to
square
be law forbetween
established point sources and due
the payload andtotheattenuation for surface
drone. Therefore, contaminations.
we use Fifthly,
a micro-D connector.
because
These the drone–detector
connectors are mainlysystem
used inisthe subject to strong
aerospace accelerations
industry and complyand vibrations,
with our needs a se-
cure and safe connection has to be established between the payload
regarding vibration resistance because of their shape and fastening screws between bothand the drone. There-
fore, of
ends wetheuseconnectors.
a micro-D connector. These connectors are mainly used in the aerospace in-
dustry and comply with our needs regarding vibration resistance because of their shape
2.3.
andDetector
fastening Design
screws between both ends of the connectors.
One of the main limitations of using a fixed-wing drone is the minimum height
required to fly safely. For the Penguin C, this is approximately 150 m. Because of the
inverse-square law and attenuation effects, the efficiency of a detector system is reduced
significantly compared to a measurement at or near ground level. This can be compensated
by increasing the detector volume. This, however, cannot be done irrespective of the
 2.3. Detector Design
One of the main limitations of using a fixed-wing drone is the minimum height re-
quired to fly safely. For the Penguin C, this is approximately 150 m. Because of the inverse-
square law and attenuation effects, the efficiency of a detector system is reduced signifi-
Remote Sens. 2022, 14, 3908 5 of 17
cantly compared to a measurement at or near ground level. This can be compensated by
increasing the detector volume. This, however, cannot be done irrespective of the payload
of the UAV system. Consequently, to maximize the efficiency of the detector system it is
payload
necessaryofthatthe the
UAV system.
detector Consequently,
geometry to maximize
and material the efficiency
be optimized. This was of done
the detector
by per-
system it is necessary that the detector geometry and material be
forming Monte Carlo simulations using the Monte Carlo code PENELOPE (2018),optimized. This was more
done
by performing Monte Carlo simulations using the Monte Carlo code PENELOPE
specifically the main program penmain [19–23]. Four types of detector material were se- (2018),
more
lectedspecifically the main
(CsI, NaI, CeBr, and program
LaBr) andpenmain
tested by[19–23]. Fouratypes
simulating of detector
cobalt-60 (Co-60)material were
point source.
selected (CsI, NaI, CeBr, and LaBr) and tested by simulating a cobalt-60 (Co-60) point
Three other materials, CZT, BGO, and HPGe, were also taken into consideration initially
source. Three other materials, CZT, BGO, and HPGe, were also taken into consideration
but respectively discarded because of production volume, spectroscopic resolution, and
initially but respectively discarded because of production volume, spectroscopic resolution,
cooling requirements. Each simulated detector was enclosed in an aluminum housing
and cooling requirements. Each simulated detector was enclosed in an aluminum housing
with a thickness of 1 mm that was separated from the detector crystal by another 1 mm of
with a thickness of 1 mm that was separated from the detector crystal by another 1 mm
air. An example of the simulated detector geometry can be seen in Figure 2, although
of air. An example of the simulated detector geometry can be seen in Figure 2, although
please note that it is not to scale. We constrained the weight of the crystal to 1 kg and
please note that it is not to scale. We constrained the weight of the crystal to 1 kg and
varied the thickness and radius of the crystal to keep the same volume and weight of
varied the thickness and radius of the crystal to keep the same volume and weight of
detector crystal. This allowed us to add 1 kg of additional hardware materials, such as
detector crystal. This allowed us to add 1 kg of additional hardware materials, such as
detector housing, electronics, cables, etc. The detection efficiency is strongly dependent
detector housing, electronics, cables, etc. The detection efficiency is strongly dependent
on the gamma-ray energy emitted by the radionuclide. Co-60 was used because of its main
on the gamma-ray energy emitted by the radionuclide. Co-60 was used because of its
gamma
main energies
gamma at 1173
energies at keV
1173and
keV1332
and keV.
1332 These energies
keV. These are onare
energies theonhigh
the end
highwhen
end whencom-
pared to gamma energies encountered in the decay of the most common
compared to gamma energies encountered in the decay of the most common artificial artificial radio-
nuclides. By optimizing
radionuclides. By optimizingthe the
detector for for
detector higher
higherenergies, we
energies, wemake
makesuresurethat
that the lower
the lower
gamma ray energies will also be detected. Besides the gamma-ray energy,
gamma ray energies will also be detected. Besides the gamma-ray energy, Co-60 is also Co-60 is also
easily available as a point source for calibrations.
easily available as a point source for calibrations.

Figure 2.
Figure 2. Schematic
Schematic of
of the
the simulated
simulated detector
detector geometry
geometry (not
(not to
to scale).
scale).

The
The simulations
simulations were performed on the supercomputer of VSC (Vlaams Supercom-
puter
puter Centrum)totoperform
Centrum) performparallelization
parallelization to to
increase
increasethethe
number
numberof counts in the
of counts in spectrum
the spec-
and
trumreach a statistical
and reach uncertainty
a statistical of less
uncertainty of than 0.2%.0.2%.
less than Parallelization waswas
Parallelization performed
performed by
running
by running multiple
multiple instances
instances of of
the
thesimulations
simulationson ondifferent
differentcores
cores of
of the supercomputer.
For
For each instance, a different seed was selected for the pseudo-random pseudo-random numbernumber generator.
generator.
Afterwards,
Afterwards,the theresults
resultsofofthe
theseparate
separate simulations
simulations were
werecombined
combinedto obtain
to obtainoneone
simulation.
simula-
tion. In Table 1, all simulations are listed together with the geometries of the crystals and
their In
fullTable
energy peak efficiencies calculated from the combined Co-60 peaks
1, all simulations are listed together with the geometries of the crystals and corrected for
the gamma ray intensity. The simulations with the square slab geometry
their full energy peak efficiencies calculated from the combined Co-60 peaks corrected for were slightly
worse in regard
the gamma raytointensity.
solid angleThe than their cylindrical
simulations with thecounterparts.
square slabAgeometry
last reasonwere
to prefer the
slightly
cylindrical geometry is the symmetry: the cylindrical geometry makes it easier
worse in regard to solid angle than their cylindrical counterparts. A last reason to prefer to determine
the angular dependence and implement corrections for this effect. Therefore, we chose to
investigate the cylindrical geometry further.
Remote Sens. 2022, 14, 3908 6 of 17

Table 1. List of geometric parameters for all Penelope Monte Carlo simulations and the full energy
peak efficiency with their respective percentile error. The simulations were done with the detector
crystal and a 1-mm-thick aluminum housing. The Co-60 point source was placed 5 m below the
detector and the two main gamma peaks were combined for the full energy peak efficiency.

Detector Geometry Material Thickness (cm) Side/Radius (cm) FEP Efficiency Error %
1.00 14.00 6.24 × 10−6 0.10
CeBr
2.00 9.90 2.99 × 10−6 0.15
1.00 14.06 6.05 × 10−6 0.10
LaBr
2.00 9.94 2.90 × 10−6 0.14
Square Slab
1.00 16.51 5.98 × 10−6 0.10
NaI
2.00 11.67 7.54 × 10−6 0.09
1.00 14.89 7.26 × 10−6 0.09
CsI
2.00 10.53 8.56 × 10−6 0.08
0.50 11.17 4.50 × 10−6 0.11
1.00 7.90 6.29 × 10−6 0.10
CeBr 2.00 5.58 7.68 × 10−6 0.09
3.00 4.56 8.07 × 10−6 0.15
4.00 3.95 7.68 × 10−6 0.09
0.50 11.21 4.37 × 10−6 0.11
1.00 7.93 6.15 × 10−6 0.10
LaBr 2.00 5.61 7.47 × 10−6 0.09
3.00 4.58 7.94 × 10−6 0.15
4.00 3.97 7.58 × 10−6 0.09
Cylindrical Disk
1.00 9.31 6.00 × 10−6 0.10
2.00 6.59 7.59 × 10−6 0.09
NaI
3.00 5.38 8.03 × 10−6 0.09
4.00 4.66 7.96 × 10−6 0.08
1.00 8.40 7.23 × 10−6 0.09
2.00 5.94 8.69 × 10−6 0.08
CsI 3.00 4.85 8.91 × 10−6 0.08
3.50 4.49 8.90 × 10−6 0.13
4.00 4.20 8.68 × 10−6 0.12
3.00 5.50 1.17 × 10−6 0.10
Final design CsI
3.50 5.00 1.12 × 10−6 0.11

In the simulations, we worked with both an energy deposition detector and an energy
impact detector defined to the detector crystal. Energy deposition detectors keep track of
the distribution of the amount of energy that is absorbed by the entire crystal per decay
while the energy impact detectors keep track of the distribution of the gamma rays that
reach the detector crystal. Using these two types of detectors allowed us to calculate the
solid angle, the full energy peak efficiency, and the detector efficiency of the detector crystal.
The solid angle, full energy peak efficiency, and detector efficiency calculated for
different detector thicknesses are given respectively in Figures 3–5. The solid angle is
defined as the ratio of the number of gamma-rays (with energy E) that reach the detector
and the number of gamma rays (with energy E) emitted by the simulated source. The
absolute full energy peak efficiency is defined as the ratio of counts in the simulated full
energy peak and the number of gamma rays (with energy E) emitted by the simulated
source. The detector efficiency is defined as the ratio of counts in the simulated full energy
peak and the number of gamma rays (with energy E) that reach the detector. Figure 3
demonstrates that CsI with a thickness of 3.0 cm gives the highest full energy peak efficiency.
Remote Sens. 2022, 14, x FOR PEER REVIEW 7 of 18
Remote Sens. 2022, 14, x FOR PEER REVIEW 7 of 18

2.00 6.59 7.59 × 10−6 0.09

Remote Sens. 2022, 14, 3908
2.00
3.00 6.59
5.38 7.59
8.03 ×× 10
10 −6
−6 0.09
0.09 7 of 17
3.00
4.00 5.38
4.66 8.03 × 10
7.96 × 10 −6
−6 0.09
0.08
4.00
1.00 4.66
8.40 7.96
7.23 ×× 10
10−6−6 0.08
0.09
1.00
The source–detector 8.40
2.00distance is 5 m in the
5.94 7.23 ×
8.69the
simulations because 10 −6 0.09
× 10simulation efficiency
−6 0.08 at a
CsI 2.00 5.94 8.69 × 10 −6
distance of 200 m is too low. For the extrapolation to larger distances, the detector0.08
3.00 4.85 8.91 × 10 −6 0.08
efficiency,
CsI
shown in Figure3.00 4, can be used as a 4.85
3.50 4.49
decision parameter 8.91
for××the
8.90 10
10−6−6selection of0.08
0.13
a suitable
detector material3.50
and size or the inverse4.49
4.00 square law can be8.90
4.20 used××to
8.68 10 −6
10apply
−6 0.13
a correction
0.12 factor
to account for the4.00
reduction in efficiency.
3.00 In both cases, CsI8.68
4.20
5.50 comes
1.17 10out
×× 10 −6 as an optimal
−6
0.10 option.
0.12
Final design CsI CsI because of these simulations and because it is less expensive
We chose than CeBr and
3.00
3.50 5.50
5.00 1.17
1.12 ×× 10 −6
10−6 0.10
0.11
Final design CsI
LaBr. Furthermore,3.50aerial measurements 5.00do not require high
1.12 ×spectral
10−6 resolution, 0.11making
CsI a valid candidate [24].

Figure 3. Results of Monte Carlo simulations using Penelope (2018). Different detector materials and
Figure
Figure 3. 3.
geometries Results
with of
Results of
a massMonte
Monte 1 Carlo
ofCarlo
kg (allsimulations
except Final
simulations using
using Penelope
Design)
Penelope (2018).
compared
(2018). toDifferent detector
the detector
Different materials
efficiency
detector ofand
materials and
the
geometries
combination
geometries with
of the
with a mass of
two main
a mass 1
of 1 kg kg
gamma(all except Final
peaksFinal
(all except Design)
of Co-60 (1173
Design) compared
keV and to
compared to
1332 the detector
thekeV). efficiency of
detector efficiency of thethe
combination
combination ofof the
the twotwo main
main gamma
gamma peaks
peaks ofof Co-60
Co-60 (1173
(1173 keV
keV and1332
and 1332keV).
keV).

Figure 4. Results of Monte Carlo simulations using Penelope (2018). Different detector materials and
geometries with a mass of 1 kg (all except Final Design) compared to the full energy peak efficiency
of the combination of the two main gamma peaks of Co-60 (1173 keV and 1332 keV).
Remote Sens. 2022, 14, x FOR PEER REVIEW 8 of 18

Figure 4. Results of Monte Carlo simulations using Penelope (2018). Different detector materials and
Remote Sens. 2022, 14, 3908 geometries with a mass of 1 kg (all except Final Design) compared to the full energy peak efficiency
8 of 17
of the combination of the two main gamma peaks of Co-60 (1173 keV and 1332 keV).

Figure 5. Results of Monte Carlo simulations using Penelope (2018). Different detector materials and
Figure 5. Results of Monte Carlo simulations using Penelope (2018). Different detector materials and
geometries with a mass of 1 kg (all except Final Design) compared to the solid angle of an isotropic
geometries with a mass of 1 kg (all except Final Design) compared to the solid angle of an isotropic
Co-60
Co-60 point
point source.
source.

In the final design, it was noticed that the weight of the complete detector system was
In the final design, it was noticed that the weight of the complete detector system was
still below the predefined limit of 2 kg. Therefore, we adapted the detector shape slightly
still below the predefined limit of 2 kg. Therefore, we adapted the detector shape slightly
by increasing the radius of the detector to 5.5 cm. We chose to adapt the radius instead of
by increasing the radius of the detector to 5.5 cm. We chose to adapt the radius instead of
the thickness by running a final set of Monte Carlo simulations investigating both options.
the thickness by running a final set of Monte Carlo simulations investigating both options.
The results of the simulations are shown in Figures 3–5. The final detector shape was set to
The results of the simulations are shown in Figures 3–5. The final detector shape was set
a CsI disk with 3.0-cm thickness and a radius of 5.5 cm. This results in a detector weight of
to a CsI disk with 3.0-cm thickness and a radius of 5.5 cm. This results in a detector weight
approximately 1.4 kg, including the aluminum housing with a thickness of 0.8 mm at the
of approximately 1.4 kg, including the aluminum housing with a thickness of 0.8 mm at
entrance window of the detector.
the entrance
Usually,window of the
scintillation detector.require a photomultiplier tube (PMT) to amplify the
detectors
Usually, scintillation detectors
signal caused by the light-output from require a photomultiplier
the crystal. However, tubea PMT (PMT) to amplify
is large the
and heavy.
signal caused by the light-output from the crystal. However, a PMT is large
Therefore, it is better to use silicon photomultipliers (SiPM) that are smaller and lighter. and heavy.
Therefore,
Their it is better
downside is thatto use
the silicon photomultipliers
resolution of the spectra will(SiPM) that are
be slightly smaller
worse, and lighter.
but since we are
Their downside is that the resolution of the spectra will be slightly worse,
already using a CsI detector, the reduced resolution will have a limited influence, as shown but since we
are already
in [25–28]. using a CsI detector, the reduced resolution will have a limited influence, as
shown in [25–28].
Figure 6 gives a schematic overview of the detector system and the necessary connec-
tions Figure
between 6 gives
all theacomponents.
schematic overview
In Figureof thethe
7a,b, detector system can
final product andbethe necessary
seen mounted con-
on
nections between all the components. In
the fiberglass interface plate to mount it in the drone. Figure 7a,b, the final product can be seen
mounted on the fiberglass interface plate to mount it in the drone.
2.4. Topaz Sipm Multi Channel Analyzer
The Brightspec Topaz-SiPM multichannel analyzer (MCA) is developed specifically
for scintillation detectors that use SiPMs. The MCA uses 1024 channels, but the number
of channels can be increased to 4096 channels if necessary. However, since spectra of
approximately 1-s integration time are recorded, most of the channels will contain zero
counts, and therefore 1024 channels are sufficient. The MCA is controlled by the Raspberry
Pi using Python and the C++ library from Brightspec [29,30].
Remote Sens. 2022,
Remote
Remote 14,2022,
Sens.
Sens. x FOR
2022, 14,PEER
14, x FOR
3908 REVIEW
PEER REVIEW 9 of 18 9 9ofof 18
17

Figure Figure
6. Schematic
Figure 6. of the detector
6. Schematic
Schematic ofof
thethesystem
detector and connection
system
detector andand
system to the Penguin
connection to the
connection C UAV
to Penguin
the Cfor
Penguin remote
UAV com-
for remote
C UAV com-
for remote
munication.
munication.
communication.

(a) (a) (b) (b)

Figure Figure
7. (a) Detector
Figure 7.
7. (a) system system
(a) Detector
Detector electronics.
system Brightspec
electronics.
electronics. SiPM multi-channel
Brightspec
Brightspec SiPM analyzer
SiPM multi-channel
multi-channel (blue), (blue),
analyzer
analyzer Zubax
(blue), Zubax
Zubax
GNSS 2.0 (red),
GNSS Raspberry Pi 3 B+ Pi
(green), BEC 2.0BEC
DC-DC converter (grey). (b) Scionix CsI detector
GNSS 2.0 (red), Raspberry Pi 3 B+ (green), BEC 2.0 DC-DC converter (grey). (b) Scionix CsI
2.0 (red), Raspberry 3 B+ (green), 2.0 DC-DC converter (grey). (b) Scionix CsI detector
detector
with SiPMs
with and
SiPMsaluminum housing.
and aluminum housing.
with SiPMs and aluminum housing.
Remote Sens. 2022, 14, 3908 10 of 17

2.5. Raspberry Pi and Peripherals

The Raspberry Pi 3B+ is a single-board computer with the Raspbian operating system.
When the payload/detector is turned on, a Python script is initiated automatically on this
computer. The script reads GPS coordinates from the Zubax GPS module together with the
date and time in UTC, the above-mean-sea-level (AMSL) height, and the pressure. From the
MCA, the spectrum together with the live and real-time of the measurement are obtained.
With these data, a text file is created and saved in the Raspberry Pi’s memory. In addition
to being stored on the Raspberry Pi, the text file is also sent to the ground control station to
provide data for a live feed.
The Zubax GNSS 2.0 [31] is a GPS module specifically designed to be used in combina-
tion with UAVs or drones. The outputs are divided into multiple lines that show the GNSS
data (Global Navigation Satellite System). Among these data, there are latitude, longitude,
AMSL height, UTC time, date, magnetic heading, and pressure. The data obtained from
the GPS module are then read by the Raspberry Pi and added to the text file containing the
spectrum. The BEC 2.0 DC-DC converter is used to reduce the 24 V coming from the drone
to the 5.1 V that is required to power the Raspberry Pi.

2.6. Detector Integration

The detector system is located in the front of the drone in a fiberglass casing of 0.85-mm
thickness to protect it from wind and rain. It also has dampers to counteract vibrations
and shocks. These measures are necessary to avoid damage to the detector system. We
verified that the possible attenuation due to the encasement will be limited to less than 1%
for the gamma energy of K-40 (1460.8 keV) by Monte Carlo simulations using Penelope.
The detector and the drone are connected by a micro-D connector to make sure they will
not detach due to vibrations.

2.7. Theoretical Framework

Following Beck et al. [16], we find that Equation (1) requires three factors: (1) the
angular dependence, (2) the efficiency, and (3) the theoretical distribution of the source
in the soil. For K-40 measurements, the theoretical distribution is homogeneous in depth.
However, for contaminations that originate from incidents, it is possible to define the
distribution as an exponential function.
The first two factors have to be determined experimentally using point sources and
measurements under different angles. Monte Carlo calculations, as used during the opti-
mization process, can be used to determine the first two factors in Equation (1). However,
for a real detector with its housing, exact position of crystal, and SiPMs, it was decided to
determine these factors experimentally. The results from these experiments are shown in
Figures 8 and 9. For the angular dependence and the full energy peak efficiency η, we used
a europium-152 (Eu-152) point source. The full peak energy efficiency is given as
c
η= (2)
Atε
where c is the number of net counts in the gamma peak, A the activity of the point source at
the time of the measurement, t live time of the measurement, and ε the gamma-ray intensity.
For the angular dependence, multiple measurements were taken with a constant distance
between the source and the detector but under different angles ranging from 0◦ to 90◦ ,
with 0◦ normal to the detector surface. For the full energy peak efficiency, we used the
measurement at 0◦ . Europium-152 was chosen because it has multiple gamma-ray energies,
mainly between 122 and 1408 keV, allowing for the determination of these factors for a
wide range of gamma energies. Additionally, the peak at 1408 keV is relatively close to the
1460-keV peak coming from K-40 which is the nuclide used for testing the methodology for
quantifying the concentration in soil.
Remote
RemoteSens.
Sens.2022,
2022,14,
14,x xFOR
FORPEER
PEERREVIEW
REVIEW 1212ofof1818
Remote Sens. 2022, 14, 3908 11 of 17

Figure
Figure8.8.
Figure 8.Full
Fullenergy
Fullenergy
energypeak
peak
peakefficiency
efficiencyexperimentally
efficiency experimentallydetermined
experimentallydeterminedwith
determinedwitha aEu-152
with aEu-152point
Eu-152pointsource
pointsourceposi-
sourceposi-
posi-
tioned
tioned50
tioned50cm
50cm under
cmunder the
under thedetector.
the detector.
detector.

Figure Relative
9.Relative angular dependence to ◦ for the full energy peak efficiency of 1408 keV from a
0for
Figure
Figure9.9. Relativeangular
angulardependence
dependencetoto0°0° forthe
thefull
fullenergy
energypeak
peakefficiency
efficiencyofof1408
1408keV
keVfrom
froma a
Eu-152
Eu-152 point
point source.
source.
Eu-152 point source.

2.8. The last factor can be determined by calculating the theoretical flux at the detector
2.8.Topography
Topography
location. We start with the equation for the theoretical flux coming from a point source at a
One
Oneofofthe thefactors
factorsthat
thatinfluences
influencesthe theresults
resultsobtained
obtainedby bythe
thetheoretical
theoreticalframework
framework
distance R (m) and an activity S0 (Bq), as can be seen in Equation (3). Herein, we adapt the
ofofBeck
Beck etetal.
al.isisthe
the topography
topography ofofthe
the area
area ofofinterest.
interest. InIncase
case ofofterrain
terrain that
thatcannot
cannotbe be
source term by incorporating the exponential distribution in the soil and the self-attenuating
assumed
assumed flat,
flat,the
the results
results need
need totobebecorrected
corrected for
for the
the roughness
roughness and
and irregularities
irregularities found
found
effects. However, the point source used Cartesian coordinates while polar coordinates
ininthe region
thebe region [32].
[32]. Using
Using height
heightmaps
maps from the theregion obtained from Geopunt Vlaanderen
will of more use to incorporate thefrom
angular region
dependencyobtained from
of the Geopunt
detector. Vlaanderen
Therefore, we
[33],
[33],ititisisclear
clear that
that there
there are
areonly
only slight
slight differences
differences ininheight
height over
over the
the measurement
measurement site.
site.
substitute the Cartesian integrals with their polar counterparts. After integrating over the
The
The changes
changes have
haveathis
amaximum
maximum value ofof2.3 2.3mmoror1.1%
1.1%ofofthetheflight
flightheight
heightofof209.9
209.9m.
m.There-
azimuthal angle, results invalue
Equation (4). There-
fore,
fore,ititisisassumed
assumedthat thatthe
thetopography
topographyhas hasnonosignificant
significantinfluence
influenceon onthethemeasurements
measurements
over
overthisthisregion
regionregarding
regardingthe theattenuation
attenuationofofsoil. 0 However,
Ssoil. However,
−µ a R the
thechanges
changesininthe theheight
height
above ϕ point ( R) = ethe (3)
aboveground
grounddo dohave
haveaasignificant
significant influence
influencedue due
4πR to
2 to the attenuation
attenuation ofofair
airand
andtherefore
therefore
have
havetotobe beaccounted
accountedfor forbybyaaheight
heightcorrection.
correction.This Thisisisdifferent
differentininthe thesense
sensethat
thatthe
the
Remote Sens. 2022, 14, 3908 12 of 17

Z Z∞
π/2
S0 −( αρ )ρz 2 h h
ϕsoil (r, θ ) = 2π e r sin θ e−µs (r− ω ) e−µa ( ω ) dr dθ (4)
4πr2
0 h/ω

In Equation (4), r is the distance from each element of differential volume to the
detector position in m, h is the height above ground in m, ω = cos θ with θ the angle of the
photon incident on the detector, z is the depth beneath the surface in cm, S0 is the surface
activity in photons/cm2 s, α is the reciprocal of the relaxation length of the approximately
exponentially distributed source activity with depth in cm−1 , ρ and ρa are respectively the
soil and air density expressed in g/cm3 , and µa and µs are respectively the air and soil
mass attenuation coefficients in cm−1 . By integration over r and moving from θ to ω, we
obtain Equation (5).
 
( ρ aa ) (hρ a )
Z 1 µ
S0 /ρ  1 (− )
ϕ=     e  ω dω (5)
0 2 α
ω+ s
µ
ρ ρ

In the work of Beck et al., an exponential function is used to define the activity as a
function of depth with an exponent α/ρ. However, for naturally occurring radioactivity,
such as K-40 in soil, this exponent is equal to zero since K-40 can be assumed to be
homogeneously distributed with respect to depth.

2.8. Topography
One of the factors that influences the results obtained by the theoretical framework
of Beck et al. is the topography of the area of interest. In case of terrain that cannot be
assumed flat, the results need to be corrected for the roughness and irregularities found in
the region [32]. Using height maps from the region obtained from Geopunt Vlaanderen [33],
it is clear that there are only slight differences in height over the measurement site. The
changes have a maximum value of 2.3 m or 1.1% of the flight height of 209.9 m. Therefore,
it is assumed that the topography has no significant influence on the measurements over
this region regarding the attenuation of soil. However, the changes in the height above
ground do have a significant influence due to the attenuation of air and therefore have to
be accounted for by a height correction. This is different in the sense that the topography
correction would correct for radiation coming from higher areas or areas that attenuate the
contamination. Meanwhile, the height correction is only correcting for the above ground
level (AGL) height. The AGL height is calculated by obtaining the AMSL height from the
drone and subtracting the average AMSL height of the topography.

3. Results
The test flights were performed above Droneport Sint-Truiden (50◦ 470 31”N 005◦ 120 06”E) [34].
Droneport is a small airfield that combines both the flight control of small airplanes (e.g.,
logistics of personal aircraft) and UAVs (e.g., logistics of multicopter and fixed-wing
drones). The test flights had different goals; firstly, to demonstrate the good functioning
of the detector–drone system (data collection, transmission, etc.), and secondly to test
the reproducibility of results collected during the flight, and, finally, to demonstrate that
quantitative results can be obtained with the system if a source geometry can be assumed.
A first flight to evaluate the performance of the Penguin C UAV and CsI detector took
place at Droneport in Sint-Truiden, Belgium on 7 October 2021. After a calibration and
check-up of both the UAV and the detector system, the Penguin C drone was launched.
The detector system was powered off during the launch to avoid any interference. After the
take-off, the detector system was turned on and a connection to the ground control station
was established. Thereafter, a circular flight pattern was executed at multiple heights.
On a second flight on 22 January 2022, we used the same flight plan as in the first
to verify the reproducibility of the flights and measurements and to collect more data as
a function of the detector height above ground. In the second part of this flight, a grid
 Remote Sens. 2022, 14, 3908 13 of 17

pattern flight path was used to investigate the implications on the measurements regarding
tilt angle and turning radius. The tilt angle is influenced by the turning radius and alters
the geometry from a detector parallel to the ground to a geometry with a tilted detector
relative to the ground. This can be corrected with the use of the tilt angle coming from
the UAV log files and the angular dependence. However, for grid patterns, the tilt angle
will be 0◦ for the straight flight lines. Only during the changes in direction will it have an
influence on the detector geometry and since these changes happen outside the border of
the area of investigation, their impact on the overall results will be limited.
Remote Sens. 2022, 14, x FOR PEER REVIEW 14 of 18
Figure 10a,b show the total counts per second for every measurement at different
heights and illustrates how the changes in height influence the number of counts per second.
Remote Sens. 2022, 14, x FOR PEER REVIEW 14 of 18
We can clearly see the inverse connection between the height of the drone and the total
number of counts per second. However, in Figure 11a,b, the only clear K-40 peak is visible
for the AGL height of 209.9 m. The K-40 peak at 311.4 m is still measureable but at higher
altitudes it was not possible to measure a net peak count.

(a) (b)

Figure 10. Total number of counts per second compared to the height above ground level in m for
(a) (b)
flight 1 (a) and flight 2 (b).
Figure10.
Figure Totalnumber
10.Total numberofofcounts
countsper
persecond
secondcompared
comparedtotothe
theheight
heightabove
aboveground
groundlevel
levelininmmfor
for
flight 1 (a) and flight 2 (b).
flight 1 (a) and flight 2 (b).

(a) (b)

Figure 11. (a) Average number of counts per second at different heights shown on a logarithmic scale.
(a)
Figure
(b) 11. (a)
Average Average
number number
of counts perof counts
second at per second
different at different
heights zoomed(b)
heights shown
in on the K-40 on a logarithmic
peak (1460 keV)
scale. (b) Average number of counts per second at different heights zoomed in on the K‐40 peak
on a linear scale.
(1460 keV) on a linear scale.
Figure 11. (a) Average number of counts per second at different heights shown on a logarithmic
In these
scale. (b) Averagespectra,
number theofgamma
counts peak of K-40
per second at at 1460 keV
different is visible
heights zoomed andinison
used
the to calculate
K-40 peak
(1460
In these
the K-40 spectra, the gamma peak of K‐40 at 1460 keV is visible and is used to calculate
concentration.
keV) on a linear scale.To verify our calculated results, three samples were taken near the
the K‐40 concentration. To verify our calculated results, three samples were taken near the
siteInofthese
measurement.
spectra, theThe results
gamma from
peak theseatsamples
of K-40 1460 keV are shown and
is visible in Table 2. to
is used The obtained
calculate
theresults were verified To
K-40 concentration. by verify
using the
our samples
calculatedtaken by FANC
results, (Federalwere
three samples Agency
takenfor Nuclear
near the
Control) at multiple sites in Belgium [35]. Figure 12a,b show the flight paths
site of measurement. The results from these samples are shown in Table 2. The obtained of both flights.
Remote Sens. 2022, 14, 3908 14 of 17

site of measurement. The results from these samples are shown in Table 2. The obtained
results were verified by using the samples taken by FANC (Federal Agency for Nuclear
Control) at multiple sites in Belgium [35]. Figure 12a,b show the flight paths of both flights.
For the heights between 209.9 m and 311.4 m, we were able to find a K-40 peak. In Table 3,
the K-40 peaks are used to calculate the K-40 concentrations and compare them with the
concentrations of the ground samples. The efficiency is corrected from 1408 keV to 1460.8
keV by Monte Carlo simulations. Two simulations were executed with a point source and
the final detector geometry, but one with an energy of 1408 keV and another one with an
energy of 1460.8 keV. The detector efficiencies of both simulations were then taken, and the
ratio of both efficiencies used to correct for the efficiency used in Equation (1). At higher
altitudes, the gamma rays of K-40 are more strongly attenuated and therefore they cannot
be seen in these spectra.

Table 2. Table with sampled concentration data and their uncertainties.

ID K-40 Concentration (Bq/kg) Uncertainty (Bq/kg) (1 σ)

Remote Sens. 2022, 14, x FOR PEER REVIEW 15 of 18
1 500 30
2 460 30
3 460 25
(1). At higher altitudes, the gamma rays of K‐40 are more strongly attenuated and there‐
be seen in these spectra. 473
Average
fore they cannot 16

(a) (b)

Figure 12.
Figure (a) Flight
12. (a) Flight path
path of
of the
the first
first flight
flight with
with the
the circular
circular pattern. (b) Flight
pattern. (b) Flight path
path of
of the
the second
second
flight with
flight with the
the circular
circular pattern
pattern and
and the
the grid
grid pattern.
pattern.

TableUsing Equation
2. Table (5), the
with sampled theoreticaldata
concentration fluxes
andfrom a homogeneously distributed source
their uncertainties.
were calculated at different flight heights. The theoretical flux can then be used together
with theID full energy peakK‐40 Concentration
efficiency (Bq/kg)per seconds
and the counts Uncertainty
from the(Bq/kg) (1 σ)to
K-40 peak
1 500 30
calculate the concentration at the ground using Equation (1). Results are shown in Table 3.
The uncertainty
2 in the concentration is460
mainly caused by the limited number
30 of counts in
the K-40 peak.
3 460 25
AAverage
measurement at the site of the473 sampling was executed and in the 16 same way as
with the drone measurements a K-40 concentration was calculated. At a height of 1 m, we
measured a net peak counting rate of 4.945 ± 0.052 counts per second and calculated a
Using Equation (5), the theoretical fluxes from a homogeneously distributed source
theoretical flux of 0.83 ± 0.04 gammas/cm2 s. These values give us the corresponding K-40
were calculated at different flight heights. The theoretical flux can then be used together
concentration of 492.6 ± 28.4 Bq/kg. All uncertainties are for one standard deviation.
with the full energy peak efficiency and the counts per seconds from the K‐40 peak to
calculate the concentration at the ground using Equation (1). Results are shown in Table
3. The uncertainty in the concentration is mainly caused by the limited number of counts
in the K‐40 peak.
A measurement at the site of the sampling was executed and in the same way as with
the drone measurements a K‐40 concentration was calculated. At a height of 1 m, we meas‐
ured a net peak counting rate of 4.945 ± 0.052 counts per second and calculated a theoret‐
ical flux of 0.83 ± 0.04 gammas/cm² s. These values give us the corresponding K‐40 con‐
Remote Sens. 2022, 14, 3908 15 of 17

Table 3. Table comparing K-40 concentration measurements at multiple heights with the sampled
concentration of K-40.

Theoretical Flux
Net Counts per Uncertainty Net Calculated with K-40 Uncertainty K-40 Relative Difference
Height (m) Second in K-40 Counts per Second Equation (5) Concentration Concentration (%) from Sampled
Peak (cps) in K-40 Peak (%) (Bq/kg) (1 σ) Concentration (%)
(Gammas/s.cm2 )
209.9 0.5994 4.6 0.083 604.4 7.2 27.8
239.1 0.4660 6.2 0.065 600.0 8.3 26.8
271.7 0.3666 6.5 0.049 626.1 8.6 32.4
288.4 0.3172 8.2 0.043 617.4 9.9 30.5
311.4 0.2927 6.5 0.036 680.4 8.6 43.9

4. Discussion
The drone–detector system for radiological monitoring as presented here started from
the pre-requisite to be a system with a long flight autonomy (more than 8 h, ideally 12 h)
and an extended range of operation. The range should allow for operating the system
remotely from a safe location in case of a nuclear or radiological emergency. Autonomy
is important to guarantee continuous surveillance and minimal downtime to recharge or
refuel. During the Fukushima accident, for example, releases to the environment continued
for two weeks [36]. Having remote access to data during this period with flights over
different areas and different altitudes, depending on, e.g., the meteorological situation,
with a drone–detector system as described here, would have been an important asset in the
radiation protection strategy. Further, the system has been designed to be flexible: the UAV
can be launched by catapult and recuperated with a parachute from nearly all locations
with limited resources needed. The used flight patterns are chosen for specific reasons. The
circular flight patterns can be used for monitoring an installation and detecting possible
anomalies, while the grid pattern can be used to map contaminated site or to search for a
lost source.
The restrictions defined in the certificate of conformity of the UAV put limitations on
the payload available for the radiation detection system. The radiation detection system
was fully optimized to yield the highest efficiency for the available payload. Both the type
of detector as well as its size were optimized. To allow radionuclide identification, only
detection methods allowing gamma spectroscopy have been included in this optimization
process. The optimization using the Penelope Monte Carlo code (2018) allowed us to
further investigate the detector efficiency and solid angle for four detector materials: LaBr,
CeBr, CsI, and NaI. The best detector option was a CsI using SiPMs with a miniature MCA.
Other detector materials, such as HPGe or CZT, have drawbacks, e.g., cooling or limited
volumes, and were not included in the simulations.
The detector takes a spectrum every second. The one-second spectra create a flexible
data flow in which we can combine several spectra for increased statistics since the statistics
for the one-second spectra are very limited.
The results in Table 3 are within the range as stated by FANC [36]. The calculated
concentrations from the UAV measurements can differ somewhat from the samples due to
the differences in sampled area, as is also the case for measurements at different heights
due to a different field of view. However, for the measurement at 1 m height, the result
agrees well with the ground sample concentrations.
The sensitivity of the system is dependent on the altitude. We can use this to our
advantage to obtain a required sensitivity for a specific case or during radiological releases
by flying at higher altitudes to avoid contamination. If sensitivity is important, we can use
the repeatability of the flight controller software to accumulate statistics, similar to what
was done for K-40. The drone–detector combination can also be used for other applications,
such as the screening for radiation and/or radioactivity during public events in the context
of security, locating orphan radioactive sources, or radiological mapping of large areas
such as legacy uranium mines. For the location of orphan radioactive sources, the system
could scan large areas to obtain an area of interest which can be scanned more accurately
Remote Sens. 2022, 14, 3908 16 of 17

using a multicopter drone. This application will require more extensive knowledge on the
minimum detectable activities of the detector. The radiological mapping of large areas can
be done in a similar way to what was done for K-40 in this paper. However, to obtain a
correct calibration, it might be more useful to work with calibration pads, especially for
uranium or thorium mines. Their energies are significantly higher, and it would be difficult
to get a full energy peak efficiency for these energies using point sources.

Author Contributions: Conceptualization, S.G., J.C. and G.I.; methodology, S.G., J.C. and G.I.;
software, S.G. and G.O.; validation, S.G., J.C. and G.I.; formal analysis, S.G.; data curation, S.G.;
writing—original draft preparation, S.G.; writing—review and editing, S.G., J.C., G.O., G.I. and W.S.;
visualization, S.G.; supervision, J.C. and W.S.; funding acquisition, J.C. and G.I. All authors have read
and agreed to the published version of the manuscript.
Funding: This research was funded by the Belgian Federal Public Service Economy, SMEs, Self-
employed and Energy, under the Energy Transition Fund within the BUDDAWAK (Belgian unmanned
aerial systems for Decommissioning, Detection and AWAreness of radioactive risK) project.
Data Availability Statement: All data can be requested by contacting the first author.
Acknowledgments: Computer services used in this work were provided by the VSC (Flemish
Supercomputer Center), funded by the Research Foundation—Flanders (FWO) and the Flemish
Government. We would also like thank Gaetan Verpoorten, Federico Bus, Alex Shaposhnik, Vincent
Francois, Katrijn Vandersteen and Johan Paridaens for their insights and help with the measurements.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.

References
1. Radoglou-Grammatikis, P.; Sarigiannidis, P.; Lagkas, T.; Moscholios, I. A Compilation of UAV Applications for Precision
Agriculture. Comput. Netw. 2020, 172, 107148. [CrossRef]
2. Lu, K.; Xu, R.; Li, J.; Lv, Y.; Lin, H.; Liu, Y. A Vision-Based Detection and Spatial Localization Scheme for Forest Fire Inspection
from UAV. Forests 2022, 13, 383. [CrossRef]
3. Stolfi, D.H.; Brust, M.R.; Danoy, G.; Bouvry, P. CONSOLE: Intruder detection using a UAV swarm and security rings. Swarm.
Intell. 2021, 15, 205–235. [CrossRef]
4. Liew, C.F.; DeLatte, D.; Takeishi, N.; Yairi, T. Recent Developments in Aerial Robotics: A Survey and Prototypes Overview. arXiv
2017, arXiv:1711.10085.
5. Gabrlik, P.; Lazna, T.; Tomas, G. ScienceDirect Simulation Simulation of of Gamma Gamma Radiation Radiation Mapping
Mapping Simulation of Gamma Radiation Mapping Using an Unmanned Aerial System Simulation of Gamma Radiation
Mapping Using an Unmanned Aerial System Using an Unmanned Aerial. IFAC-PapersOnLine 2018, 51, 256–262. [CrossRef]
6. Buchanan, E.; Cresswell, A.J.; Seitz, B.; Sanderson, D.C.W. Operator Related Attenuation Effects in Radiometric Surveys. Radiat.
Meas. 2016, 86, 24–31. [CrossRef]
7. Martin, P.G.; Hutson, C.; Payne, L.; Connor, D.; Payton, O.D.; Yamashiki, Y.; Scott, T.B. Validation of a Novel Radiation Mapping
Platform for the Reduction of Operator-Induced Shielding Effects. J. Radiol. Prot. 2018, 38, 1097–1110. [CrossRef]
8. Connor, D.T.; Martin, P.G.; Smith, N.T.; Payne, L.; Hutton, C.; Payton, O.D.; Yamashiki, Y.; Scott, T.B. Application of Airborne
Photogrammetry for the Visualisation and Assessment of Contamination Migration Arising from a Fukushima Waste Storage
Facility. Environ. Pollut. 2018, 234, 610–619. [CrossRef]
9. Aleotti, J.; Micconi, G.; Caselli, S.; Benassi, G.; Zambelli, N.; Calestani, D.; Zanichelli, M.; Bettelli, M.; Zappettini, A. Unmanned
Aerial Vehicle Equipped with Spectroscopic CdZnTe Detector for Detection and Identification of Radiological and Nuclear
Material. In Proceedings of the 2015 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), San Diego,
CA, USA, 1 October–7 November 2015. [CrossRef]
10. Connor, D.T.; Wood, K.; Martin, P.G.; Goren, S.; Megson-Smith, D.; Verbelen, Y.; Chyzhevskyi, I.; Kirieiev, S.; Smith, N.T.;
Richardson, T.; et al. Corrigendum: Radiological Mapping of Post-Disaster Nuclear Environments Using Fixed-Wing Unmanned
Aerial Systems: A Study From Chornobyl. Front. Robot. AI 2020, 6, 149, Erratum in Front. Robot. AI 2020, 7, 28–30. https:
//doi.org/10.3389/frobt.2020.00030. [CrossRef]
11. Sanada, Y.; Torii, T. Aerial Radiation Monitoring around the Fukushima Dai-Ichi Nuclear Power Plant Using an Unmanned
Helicopter. J. Environ. Radioact. 2015, 139, 294–299. [CrossRef]
12. Connor, D.; Martin, P.G.; Scott, T.B. Airborne Radiation Mapping: Overview and Application of Current and Future Aerial
Systems. Int. J. Remote Sens. 2016, 37, 5953–5987. [CrossRef]
Remote Sens. 2022, 14, 3908 17 of 17

13. Royo, P.; Pastor, E.; Macias, M.; Cuadrado, R.; Barrado, C.; Vargas, A. An Unmanned Aircraft System to Detect a Radiological
Point Source Using RIMA Software Architecture. Remote Sens. 2018, 10, 1712. [CrossRef]
14. van der Veeke, S.; Limburg, J.; Koomans, R.L.; Söderström, M.; de Waal, S.N.; van der Graaf, E.R. Footprint and Height Corrections
for UAV-Borne Gamma-Ray Spectrometry Studies. J. Environ. Radioact. 2021, 231, 106545. [CrossRef]
15. van der Veeke, S.; Limburg, J.; Koomans, R.L.; Söderström, M.; van der Graaf, E.R. Optimizing Gamma-Ray Spectrometers for
UAV-Borne Surveys with Geophysical Applications. J. Environ. Radioact. 2021, 237, 106717. [CrossRef] [PubMed]
16. Beck, H.L.; DeCampo, K.; Gogolak, C. In Situ Ge(Li) and NaI(Tl) Gamma-Ray Spectrometry; Health and safety Laboratory (Atomic
Energy Commision): New York, NY, USA, 1972.
17. International Atomic Energy Agency. Generic Procedures for Medical Response during a Nuclear or Radiological Emergency; International
Atomic Energy Agency: Vienna, Austria, 2005.
18. Collins Aerospace Piccolo. Available online: https://www.cloudcaptech.com/products/detail/piccolo-ii (accessed on
27 April 2022).
19. Salvat, F.; Fernández-Varea, J.M. Overview of Physical Interaction Models for Photon and Electron Transport Used in Monte
Carlo Codes. Metrologia 2009, 46, S112–S138. [CrossRef]
20. Sandev, T. Monte Carlo Simulation of Buildup Factors for Single and Multi-Layer Shields by Using PENELOPE Code. In
Proceedings of the Second Conference on Medical Physics and Biomedical Engineering, Skopje, Macedonia, 5–6 December 2010;
pp. 24–28.
21. Salvat, F. {PENELOPE} 2018: A Code System for Monte Carlo Simulation of Electron and Photon Transport. In Workshop
Proceedings; NEA: Barcelona, Spain, 2019; ISBN 978-92-64-99066-1.
22. Sempau, J.; Fernández-Varea, J.M.; Acosta, E.; Salvat, F. Experimental Benchmarks of the Monte Carlo Code PENELOPE. Nucl.
Instrum. Methods Phys. Res. Sect. B Beam Interact. Mater. Atoms 2003, 207, 107–123. [CrossRef]
23. Baró, J.; Sempau, J.; Fernández-Varea, J.M.; Salvat, F. PENELOPE: An Algorithm for Monte Carlo Simulation of the Penetration
and Energy Loss of Electrons and Positrons in Matter. Nucl. Inst. Methods Phys. Res. B 1995, 100, 31–46. [CrossRef]
24. Bell, S.J.; Aitken-Smith, P.; Beeke, S.; Collins, S.M.; Regan, P.H.; Shearman, R. A Comparison of Emerging Gamma Detector
Technologies for Airborne Radiation Monitoring. J. Phys. Conf. Ser. 2016, 763, 012010. [CrossRef]
25. Ghassemi, A.; Sato, K.; Kobayashi, K. MPPC. In Hamamatsu Photonics K.K., Solid State Division; Hamamatsu Photonics K.K., Solid
State Division: Hamamatsu, Janpan, 2017; pp. 1–65.
26. Grodzicka-Kobylka, M.; Szczesniak, T.; Moszyński, M.; Swiderski, L.; Szawłowski, M. Silicon Photomultipliers in Scintillation
Detectors Used for Gamma Ray Energies up to 6.1 MeV. Nucl. Instrum. Methods Phys. Res. Sect. A Accel. Spectrometers Detect.
Assoc. Equip. 2017, 874, 137–148. [CrossRef]
27. Kim, J.; Park, K.; Hwang, J.; Kim, H.; Kim, J.; Kim, H.; Jung, S.H.; Kim, Y.; Cho, G. Efficient Design of a ∅2×2 Inch
NaI(Tl)Scintillation Detector Coupled with a SiPM in an Aquatic Environment. Nucl. Eng. Technol. 2019, 51, 1091–1097.
[CrossRef]
28. Can Silicon Photomultipliers Replace Vacuum PMTs? A Practical Approach; Scionix: Bunnik, The Netherlands, 2020; pp. 1–12.
29. TOPAZ-SiPM—A Miniature Digital; MCA Brightspec: Niel, Belgium, 2020; pp. 1–2.
30. Brightspec. BrightSpec Software Reference Manual SDK Reference Manual; Brightspec: Niel, Belgium, 2017; pp. 1–65.
31. Zubax GNSS 2 Datasheet; Zubax Robotics: Tallinn, Estonia, 2019; pp. 1–49.
32. Laedermann, J.-P.; Byrde, F.; Murith, C. In-Situ Gamma-Ray Spectrometry: The Influence of Topography on the Accuracy of
Activity Determination. J. Environ. Radioact. 1998, 38, 1–16. [CrossRef]
33. Vlaamse Overheid Geopunt Vlaanderen. Available online: https://www.geopunt.be/ (accessed on 4 May 2022).
34. Skeyes Droneport Skeyes AIP. Available online: https://ops.skeyes.be/html/belgocontrol_static/eaip/eAIP_Main/html/eAIP/
EB-AD-2.EBST-en-GB.html (accessed on 22 June 2022).
35. FANC. Radiologisch Toezicht Op Het Grondgebied En de Effluenten van Nucleaire Inrichtingen Klasse 1; Federal Agency for Nuclear
Control: Brussel, Belgium, 2021; pp. 1–91.
36. IAEA Fukushima Daiichi Accident—World Nuclear Association. Available online: https://world-nuclear.org/information-
library/safety-and-security/safety-of-plants/fukushima-daiichi-accident.aspx (accessed on 17 June 2022).