Characterisation of a New Lightweight Lorawan Gps Biologger and Deployment on Griffon Vultures Gyps Fulvus

Abstract

1. Information provided by tracking studies using satellite telemetry is providing ecologists with invaluable new insights into animal behaviour and movement strategies. Here we describe a new type of GPS (global positioning system) tracking device which makes use of a growing network of LoRaWAN (long range, low power wide area network) gateways. These tags have the potential to be a low weight and power consumption solution.
2. We characterise the GPS accuracy and data transmission range, including uplinks and downlinks, for the tracker using a series of standardised tests. Data transmission range was tested by visiting locations with line of sight to the LoRa gateway at distances up to 75km and recording whether data transmission was completed successfully from each location. These tests were complemented by a trial deployment of six devices on Griffon Vultures Gyps fulvus.
3. These LoRa tags reliably provided accurate GPS location estimates, particularly on shorter location acquisition cycles. At one-minute intervals the GPS location bias was 4.71m in the horizontal plane and 5m in the vertical plane while GPS precision, measured by standard deviation, was 3.9m in horizontal space and 7.7m in vertical space. GPS locations were less accurate on a longer acquisition cycle but still comparable with other commercially available tags. Ground based range tests revealed reliable transmission of multiple data payloads was recorded from a maximum distance of 40.7km. Initial results from a deployment on Griffon Vultures Gyps fulvus yielded useful information about flight speeds, height and transmission range during the first two weeks after tag deployment.
4. The LoRa tags demonstrated potential for effective use in the study of animal movement. The small size and power needs allows for flexibility in which combination of battery, solar panel, and housing they are paired with meaning that fully assembled tags can weigh less than 5g. The LoRaWAN gateway network does not yet allow global coverage, hence at present, this solution is best suited to the study of colonial species, where gateways can be installed on site, or species with consistent migration routes.

Full text

Page 1/14
Characterisation of a New Lightweight Lorawan Gps Biologger and
Deployment on Griffon Vultures Gyps Fulvus
Jethro Gauld (  j.gauld@uea.ac.uk )
University of East Anglia
Philip W. Atkinson
University of East Anglia
João P Silva
University of East Anglia
Andreas Senn
University of East Anglia
Aldina M. A. Franco
University of East Anglia
Method Article
Keywords: GPS Tracking, Biologging, LoRa, LoRaWAN; LPWAN, Animal Movement
Posted Date: October 14th, 2022
DOI: https://doi.org/10.21203/rs.3.rs-2146211/v1
License:   This work is licensed under a Creative Commons Attribution 4.0 International License.  Read Full License

Page 2/14
Abstract
1. Information provided by tracking studies using satellite telemetry is providing ecologists with invaluable new insights into animal behaviour and movement
strategies. Here we describe a new type of GPS (global positioning system) tracking device which makes use of a growing network of LoRaWAN (long range,
low power wide area network) gateways. These tags have the potential to be a low weight and power consumption solution.
2. We characterise the GPS accuracy and data transmission range, including uplinks and downlinks, for the tracker using a series of standardised tests. Data
transmission range was tested by visiting locations with line of sight to the LoRa gateway at distances up to 75km and recording whether data transmission
was completed successfully from each location. These tests were complemented by a trial deployment of six devices on Griffon Vultures
Gyps fulvus
.
3. These LoRa tags reliably provided accurate GPS location estimates, particularly on shorter location acquisition cycles. At one-minute intervals the GPS
location bias was 4.71m in the horizontal plane and 5m in the vertical plane while GPS precision, measured by standard deviation, was 3.9m in horizontal
space and 7.7m in vertical space. GPS locations were less accurate on a longer acquisition cycle but still comparable with other commercially available tags.
Ground based range tests revealed reliable transmission of multiple data payloads was recorded from a maximum distance of 40.7km. Initial results from a
deployment on Griffon Vultures
Gyps fulvus
yielded useful information about ight speeds, height and transmission range during the rst two weeks after tag
deployment.
4. The LoRa tags demonstrated potential for effective use in the study of animal movement. The small size and power needs allows for exibility in which
combination of battery, solar panel, and housing they are paired with meaning that fully assembled tags can weigh less than 5g. The LoRaWAN gateway
network does not yet allow global coverage, hence at present, this solution is best suited to the study of colonial species, where gateways can be installed on
site, or species with consistent migration routes.
1. Background
Advances in biologging technology are allowing ecologists to gain an unprecedented understanding, with high spatial and temporal resolution information, of
the movement behaviour of animals using GPS (Global Positioning System), accelerometers and other sensors (Ossi, Urbano, and Cagnacci 2019; Perona,
Urios, and López-López 2019). In turn the data from tracking studies provide insights into animal movement ecology (e.g. Rotics et al. 2021), which is key to
understand where conicts exist between wildlife and human activities (e.g. Gauld et al 2021) and to identify conservation solutions (Schaub et al. 2020;
Frankish et al. 2021). Cost and weight of tracking devices have been major factors limiting their wider use (Kauth et al. 2020). For animal welfare reasons,
tags should not exceed 3–5% of the animal’s weight (Rodríguez et al. 2012) and lower weights are recommended to make sure the devices do not affect
animal behaviour or tness (Rodríguez et al. 2012). The cost of the devices and data transmission and consumption places a major constraint upon sample
sizes, limiting the inference capability of results, and biasing tracking studies towards larger species. A variety of technologies are available. To reduce energy
requirements, tracking devices suitable for smaller species often rely on the physical recovery of the device from the tracked animal or, as is the case with
devices which utilise UHF download of location data from within a few hundred metres of the animal (Ossi, Urbano, and Cagnacci 2019; Evens et al. 2018;
Stienen et al. 2016). The requirement to physically retrieve the tag or get close to the animal to receive the data can be labour intensive and losing data is a
signicant risk if the animal dies or the device falls off before retrieval. Transmitting data remotely, over long distances, is a better solution since it can provide
near real time understanding of an animal’s movements and ensure that no data are lost. Most remote tracking systems send data via GSM (Global System
for Mobile communications), ARGOS or Iridium satellites (Ossi, Urbano, and Cagnacci 2019) but with increased costs and power needs (Ossi, Urbano, and
Cagnacci 2019), hence new low-cost, light weight technologies with the capability to remotely send data are needed.
We describe the characteristics of a new type of GPS logger based on LoRa (long range communications) which is a type of LPWAN (low-power wide area
network system) communications protocol within the IOT (Internet of Things) architecture. Under this architecture the devices, often referred to as Nodes, are
connected to the internet by transmitting data to a LoRaWAN (long range wide area network) gateway which in turn forwards the data to a server. The low
energy requirements mean that Nomad™ GPS-LoRa loggers have a key potential advantage over other devices because they can be deployed on smaller
species while still being able to send data over long distances. However, the low energy requirement does have a drawback in terms of the rate of data
transmission. LoRa achieves low-energy long distance transmission by restricting each payload to less than 243 bytes and the maximum data rate is 50kbps
which is roughly one third of the speed of 3G GSM transmission (Muteba, Djouani, and Olwal 2019). As such it is important when planning the deployment of
LoRa devices to understand if the animals being studied are likely to spend a signicant amount of time away from transmission range to a LoRa gateway
and plan data acquisition and transmission rates that account for connectivity availability. As more LoRa gateways are deployed, coverage will improve, and
more reliable data transmission will be possible. This includes proposals to launch a network of LoRa satellites such as (Lacuna 2022) or (OneWeb 2022).
One other important feature is the capability to recording and sending high frequency and high accuracy GPS locations alongside high-resolution acceleration
(up to 200hz), temperature, gyroscope and other measurements over long periods of time. These additional measurements can provide important insights into
the movement behaviour of the animal and can also detect mortality events such as when a bird collides with a wind turbine. This study aims to describe the
characteristics of a new LoRa miniaturised device (GPS accuracy, data transmission range, power consumption, impact of GPS acquisition cycle on power
consumption and post deployment performance) and assess its viability for animal tracking studies.
2. Methods
2.1 Description of the Nomadtm GPS -LoRaWAN devices
The NOMAD™ LoRa GPS-LoRaWAN logger (Fig.1), referred to as “NOMAD” hereafter, used in this study has been developed under a NERC Proof-of-concept
project led by University of East Anglia in a collaborative project between Movetech Telemetry and Miromico (UK Research and Innovation 2016). The device

Page 3/14
has been developed to enable animal tracking solutions but can also be used for other applications, the rst version of this device has been used to track the
movement of boulders in slopes and landslides in Nepal (Dini et al. 2021). The devices are capable of recording GPS measurements up to every second (1hz)
independently of the other sensors which include two accelerometers, a gyroscope, a magnetometer, a barometer and thermometer The 9-axis sensor can
collect acceleration measurements up to 200hz and forces up to 16g and can be programmed alongside the gyroscope and magnetometer and each sensor
can collect data at the same or different rates. The device also measures battery voltage and temperature so battery health can be monitored and
reprogrammed remotely as needed. Data is sent via a LoRa gateway. The LoRa gateway forwards the data via the internet to an Internet of Things network
server such as LORIOT (LORIOT 2021) or The Things Network (TTN 2021) server which in turn can forward the data to a data repository such as Movebank or
IoT Wonderland (Movebank 2019; IoT Wonderland 2022) (Fig.1).
The NOMAD module measures 23mm by 13mm, weighs less than 1.5g can be paired with a range of different LoRa and GPS antennas to suit the study
species’ needs. The software allows users to tailor the data collection to the research question depending on study species and battery size. LoRa devices can
be set to sample GPS and acceleration at regular intervals or to trigger recording of higher intensity data based on trigger parameters detected by one of the
accelerometers, for example when the force measured by the accelerometer exceeds a pre-programmed force threshold. This allows for live detection of
collisions between birds and human infrastructure and more intensive sampling when the bird is moving. The tags can store up to 60,000 data records in the
onboard memory, users have the option of deciding whether to transmit this data in chronological order or to receive the most recent data rst by changing the
data buffer mode. With a migratory bird it might be useful to ensure the most recent locations are sent rst whenever the bird is in range of the gateway
whereas for a more sedentary species which is likely to remain within gateway range most of the time sending the data in chronological order may be
preferable. When a device is in range of a gateway it will try to transmit on a user-specied duty cycle (usually one payload every 15–20 seconds or 180–220
payloads per hour). The device detects when it is in range of a gateway because an acknowledgement message is sent by the LoRa gateway once the data
has been successfully forwarded to the server. The acknowledgement ensures that each message stored on the payload buffer on the device is only deleted
from memory after it has been successfully forwarded to the server by the LoRa gateway. This conrmation feature also facilitates the remote programming
of device settings and ensures data not received by a gateway, is not lost. If an acknowledgement message is not received, data is kept in the buffer and the
device switches to a User-Specied longer interval forced transmission duty cycle (usually 30–60 minutes) once the payload is sent and an acknowledgement
is received the device restarts transmitting data more frequently according to the standard duty cycle. Animals may move away from the range of existing
gateways (or may be in landscapes with signal obstructions) and in these cases the data transmission alternates between TX and TXF cycles to save energy.
Weaker signals are associated with a high spreading factor and reduced data transmission speed meaning the acknowledgement is less likely to be received
by the device within the required timeframe, designated as the RX1 (usually 1-second) or RX2 (usually 2-seconds) window (TTN 2021). Transmission speed
decreases by a factor of two as spreading factor (SF) increases i.e. transmission speed is four times slower at a SF of 12 compared with a spreading factor of
10. SF also inuences the data rate. This is because the devices have an adaptive data rate (ADR) feature to ensure compliance with the LoRa fair use policy
(TTN 2016; Miromico AG 2022). The NOMAD devices do this by measuring the spreading factor (SF), the further the tag is from the gateway the higher the
spreading factor will be. In practice this means that at a spreading factor of 7 or less the user specied transmission cycle will be closely followed whereas
when the device is further from the gateway or there’s interference resulting in a high spreading factor the data rate may drop to just a few messages per hour
(Kim, Lee, and Jeon 2020).
The NOMAD software allows users to activate the onboard tri-axis gyroscope and magnetometer to allow for recording of the angle of the bird and direction of
travel independently of the GPS and battery saving options which switch off the GPS at low voltages. The NOMAD module can be integrated with a solar
powered solution and a harvester. The NOMAD can also be programmed remotely using a downlink function allowing settings to be adjusted post-
deployment. In this study, data was forwarded by the LoRa gateway (Fig.1b) to a private server such as LORIOT (LORIOT 2021) however the devices could
instead be programmed to access open networks such as The Things Network™ (TTN 2021).
2.2 Quantifying GPS Accuracy
The accuracy of GPS measurements was tested under different x acquisition rates (1, 30 and 60 minutes) by leaving a tag in position on a geographical
marker (Supplementary Material 1, Fig.1) with known co-ordinates (37.7481317, -8.0403338), a clear view of the sky and altitude above sea level (221m) for
sucient time to acquire at least 300 locations (358 locations at 60-minute intervals, 367 locations at 30-minute intervals and 767 locations at one-minute
intervals). The device was set up with a wire-type GPS antenna (typically used for deployments on smaller species) and the GPS sampling interval was
adjusted remotely via the downlink feature in LORIOT. Accuracy for each location estimate was calculated by calculating the distance between the horizontal
and vertical co-ordinates with the known location of the geographic marker. These horizontal distances were calculated using the Geosphere package in R
version 4.0.5 (R. J. Hijmans, Williams, and Vennes 2015). Accuracy encompasses two components, bias and precision (Walther and Moore 2005). Bias was
calculated as the mean location error relative to the true location to inform us about the magnitude of systematic over or underestimation of the true position
resulting from the GPS measurements and the standard deviation was used to quantify the precision i.e. the random spread of the error relative to the true
value (Walther and Moore 2005). A one-way ANOVA and post-hoc Tukey’s HSD signicance test was used to determine whether there was a signicant
difference in accuracy between different GPS acquisition cycles in terms of horizontal and vertical location estimation.
2.3 Quantifying LORA Transmission Range
The documented estimates of maximum transmission range of long-range wireless vary considerably in the available literature. Data transmission for a
mobile gateway model (MultiTech Conduit MTCAP-LEU1-868-001A) can range from 17km with clear line-of-sight and 2-3km in more cluttered environments
(Multi-Tech Systems Inc. 2021) but it has also referred to be up to 40km with clear line of sight for LPWAN data transmission (Mekki et al. 2019). The world
record for data transmission between a node and a Lora Gateway with LoRaWAN within the Earth’s atmosphere of 766km was set in 2019 using a weather
balloon at high altitude (The Things Network 2019). The record highlights the potential of this technology to send data long distances however, this kind of
transmission distance is not realistic under terrestrial usage scenarios.

Page 4/14
The main focus of the range tests was on data transmission from a NOMAD node to a Multi-Tech IP67 outdoor gateway (Concept 13 Limited 2022) located at
co-ordinates 37.731°, -8.029°, mounted on a mast 5.5m above the ground (Fig.1). Four NOMAD devices with different antenna types (exi, gold-plated wire,
brass wire and silver-plated wire) were set to record and transmit GPS locations and status messages at least every 15 minutes. The devices were taken to
locations of known distance away from the gateway ranging from < 1km to ~ 70km. Test locations were identied using viewshed analysis in QGIS (QGIS
Development Team 2019) to determine locations with line of sight to the gateway. The viewshed analysis assumed that both device and gateway were at 5
metres above ground level. To achieve this, the devices were elevated using a telescopic pole to a height of 5 metres at all test locations (Fig.2). At each
location, the data stream from the gateway was monitored using a web browser interface on a smartphone and the devices left in position until either all four
devices had successfully transmitted data, or half an hour had elapsed. The range test was also repeated at the locations in (Fig.2c) with two Nomad™
devices using a portable Multitech™ Mobile Gateway (MTCAP2-L4E1-868-002A-POE) up to 17km. All range tests were conducted under eld conditions in
Portugal during May and June of 2021 on calm, dry days with minimal cloud cover.
2.4 Power Consumption
It is dicult to assess power consumption in a standardized manner under eld conditions because of daily variations in solar and temperature conditions
and variations between batteries which make it dicult to fairly compare devices using different settings. As such a spreadsheet for calculating power
consumption and battery longevity under different settings has been provided by the manufacturer in Supplementary Material A2.
2.5 Case Study with Griffon vultures
In October of 2021, six Nomad devices were deployed on Griffon Vultures
Gyps fulvus
as a trial deployment to test the GPS and Accelerometer features of the
loggers. The GPS-LoRa Modules were assembled in a solar powered conguration weighing 83g (Fig.1). The devices were programmed to record a GPS
location every 30 minutes and a one second burst of Acceleration measurements at a frequency of 50hz whenever a force exceeding 3.2g was detected with a
view to detecting avoidance or collision events. The loggers were deployed in Southern Portugal under licence from CNF - Instituto da Conservação da
Natureza e das Florestas, on the 23rd and 24th of October 2021 using a backpack style harness with a weak link to ensure the loggers fall off after a few
years (typically 1–3 years) without harming the bird. Four of the tagged birds were caught and tagged near Bensafrim, north of Sagres in Southwest Portugal
on the 23rd of October and the remaining two birds were rehabilitated birds. The rehabilitated birds were released near Mertola, in Southeast Portugal. This
study deployed four xed position LoRaWAN gateways within important areas for bird migration in Southern Portugal and Spain (Fig.4e) and there are plans
to deploy more to cover the main migration routes. Data were produced in R version 4.0.5 (R Core Team 2019) and maps were produced using the ggmap and
patchwork packages (Wickham 2016; Pedersen 2020)
Here we report the initial results from tracking data received during the rst two weeks after tag deployment allowing us to understand the performance of the
tags under real world conditions including data transmission range and GPS performance. Data transmission range was assessed using the pointDistance
function in the raster package in R (Robert J. Hijmans 2019) to measure the distance between the nearest LoRa gateway and the GPS location of the bird.
Transmission success was coded as a binary variable with 0 representing locations where transmission was not possible and 1 representing locations where
the GPS position was successfully sent to a gateway within 1 minute of the GPS location being recorded. Transmission success was then modelled in a
binomial generalised linear model (GLM) with a logit link to identify whether any factors aside from distance to the nearest gateway signicantly inuenced
transmission success. The initial model included distance to the nearest LoRa gateway (km), height of the bird above ground (m), roughness of the terrain as
measured by the terrain roughness index (TRI) (Evans;, Murphy;, and Ram 2021; Riley, DeGloria, and Elliot 1999) and landcover associated with each GPS
location (ESRI 2021). Landcover was investigated because features in the landscape such as buildings and trees may impact upon transmission, however
including landcover in the model was found to introduce signicant bias. As such landcover was not included in the nal model. Co-linearity was checked
using the ggpairs function (Schloerke et al. 2021); non-signicant variables were sequentially eliminated from the model in a stepwise fashion until the most
parsimonious model with the lowest AIC value was established. Model outputs were then plotted using the ggeffects function (Lüdecke et al. 2022).
3. Results
3.1 GPS ACCuracy
Location bias relative to the true location in horizontal space (Fig.3a, Table1) ranged between 4.71m for the 1-minute cycle, 6.63m for the half-hourly cycle
and 8.44m for the 60-minute cycle. Signicant differences between cycles were detected in terms of horizontal precision (Fig.6a), ANOVA (F (2, 1486) = 19.62,
p < 0.01). Location precision was highest in the 1-minute cycle (3.88m precision) and lowest the 60-minute cycle (18m precision) (2.79, p = < 0.01, 95% C.I. =
[1.68, 3.90]). Errors greater than 100m only occurred on three occasions representing 0.2% of recorded locations and were associated with the hourly GPS
position cycle.

Page 5/14
Table 1
GPS position error (bias) and precision (standard deviation) in horizontal and vertical space for different transmission sch
Cycle Total
Locations Horizontal
Bias Min
Horizontal
Error
Max
Horizontal
Error
Horizontal
SD Horizontal
2.5
Percentile
Horizontal
97.5
Percentile
Horizontal
CI 95% Vertical
Bias Min
Vertical
Error
Max
Vertical
Error
Ver
SD
1
minute 767 4.71 0.35 21.9 3.88 0.53 15 14.4 5 -18 43.1 7.77
30
minute 367 6.63 0.35 39.5 6.91 0.53 27.5 27 2.03 -78.3 102 15.3
60
minute 358 8.44 0.35 224 18 0.35 34.7 34.3 2.99 -138 349 27.4
Location bias relative to the true position in vertical space (Fig.5b, Table1) ranged between 2-5m for the three acquisition cycles used. Vertical precision was
signicantly different between GPS acquisition cycles (Fig.6b), ANOVA (F(2, 1486) = 12.72, p < 0.01), with the post hoc TUKEY-HSD test conrming the
signicant differences between the 60-minute cycle (28m precision) and the 1-minute cycle (7m precision) (-3.44, p < 0.01, 95% C.I. = [-5.37, -1.52]) as well as
the 30-minute (15.3m precision) and 1-minute cycles (-3.25, p < 0.01, 95% C.I. = [-5.16, -1.34]) but not the 60-minute and 30-minute cycles (-0.19, p = 0.98, 95%
C.I. = [-2.43, 2.04]). These results indicate a slight overestimation of altitude relative to the true position across cycles with reduced position bias observed in
the 30-minute and 60-minute cycles than with the 1-minute cycle. A small but signicant reduction in both horizontal and vertical precision at lower frequency
cycles (30-minute and 60-minute) was also detected which manifests in a greater spread of position estimates compared to the 1-minute cycle (Fig.3a and
Fig.3c, Table1).
3.2 Data Transmission Range
Standardised range tests in locations with theoretical line of sight to the “Castro Verde” gateway (Fig.2) conrmed the ability of these tags to transmit data
reliably at all locations at distances less than 40km from the gateway (Fig.2B). The successful transmission of a single data payload from over 62km to the
North of the LoRa Gateway Mendro (Test location 22, Fig.2B) suggests that the devices can send data over that distance. However, the latency may be too
great for the tag to receive acknowledgement of the payload, or the gateway detects that the signal is weak and therefore fails to send the acknowledgement
back to the device. Without an acknowledgement the tag will not send further data. The tags used during these tests had different types of wire antenna
namely, exible plastic-coated wire, brass wire, silver wire and gold-plated wire but it was not part of the objectives of this study to determine whether this
affected transmission range, transmission success was conrmed for all antenna types. This is broadly in line with ndings by another study which
investigated transmission range and data transmission speed in low power wireless networks which indicated a maximum transmission range of between
20km and 40km (Mekki et al. 2019). Furthermore, we performed range tests using a LoRa device and a mobile gateway. These tests were performed in areas
which did not have line of sight to the xed position Castro Verde gateway and ranged from 2km to 17km (Fig.2c). These tests demonstrated successful data
transmission at 17km from the gateway which is in line with the manufacturer claiming a range of 800m in cluttered environments, 15km with line of sight
(Multi-Tech Systems Inc. 2022). The tests also highlighted how transmission can be impeded by terrain and vegetation because data transmission was not
successful at 7km, 9km and 14km where there was higher tree cover.
3.3 Performance during deployment
Initial data from the deployed tags provided some useful insights into their performance; location data was obtained for two of the six birds which stayed
within the vicinity of the LoRa gateways allowing full data download to occur, the other four tags provided accelerometer data only. This provided 2,208 GPS
locations which allowed us to plot the post tagging movements along with daily height and speed for two of the six vultures (Fig.4a). A daily summary of the
movements is provided in Table2 for each bird. Both birds exhibited a high daily variation in movement behaviour in terms of mean daily speed (0–2.65 m/s),
height (-24.9m – 450.9m) and total daily displacement of (0.06km – 209.3km). The low average step-speed across both birds of 0.26 ± 0.52 SD m/s during
this period suggests they may have spent a signicant proportion of their ight time circling on thermals. The wild caught bird, 9012_Eduardo, ew to the
Extremadura region of Spain near the Portuguese border prior to returning to the area around Sagres, southwest Portugal. The tag on this bird was last seen by
a LoRa gateway on 08/11/2021 as it was travelling west to east along the southern coast of Portugal toward Spain. The rehabilitated bird, 7DA6_Marta,
stayed close to the release site near Mertola, Portugal before heading to the Southwest point of Portugal (Fig.4b). The nal location received from this tag
was over the Atlantic to the south of Lagos, Portugal at 15:40 on the 6th of November (36.88323, -9.01409). This vulture lost altitude over the hour preceding
the nal GPS record obtained (1474m down to 583m above sea level). Had the individual returned to the Algarve, one of the gateways would have picked up
the signal from the tag suggesting that, most likely, this individual failed to return to land and likely drowned.
Table 2
Locations obtained within range of a gateway (Y) and their distance in kilometres to the nearest LoRa gateway and locations which were recorded by the
device while out of gateway range (N) which were transmitted a posteriori.
Bird ID Gateway
Range Count Minimum Distance
(Km) Mean Distance
(Km) Max Distance
(Km) Standard Deviation Distance
(Km) Percent
Eduardo_9012 N 283 14.8 81.8 106 32.4 94.6
Eduardo_9012 Y 16 8.6 29.6 50.2 9.21 5.4
Marta_7DA6 N 1625 9.3 34.5 54.2 13.9 85.1
Marta_7DA6 Y 284 4.1 21.5 53.4 6.99 14.9

Page 6/14
Table 3
summary of nal binomial GLM relating the
probability of successful data transmission to the
height above ground (m) and distance of the logger
(km) from the gateway.
(Intercept) 1.53 ***
 (0.26)
Minimum_Distance_to_Gateway_km -0.13 ***
 (0.01)
Height_Above_Ground 0.003 ***
 (0.00)
AIC 1282.66
BIC 1299.85
Log Likelihood -638.33
Deviance 1276.66
Num. obs. 2277
The devices of the other four individuals only sent acceleration but no location data. Despite this it was possible to follow the birds’ movements by monitoring
which gateways received data from them and when. Data were most recently received from two of the birds on 08/11/2021 and 07/11/2021 by the
3C_F4_Tarifa gateway near the southern tip of Spain suggesting that two birds attempted to migrate to Africa. This is a total minimum distance of
approximately 480km from where they were tagged in southwest Portugal. Two of the six tags deployed appear to have stopped sending data within 48 hours
after deployment and it is unclear whether this is because the birds moved out of range of the gateway or some other issue. In both cases, the payload buffer
on the tag was clearly lled by acceleration recording being erroneously triggered numerous times suggesting an issue with the user dened accelerometer
settings. We do not summarise acceleration data further as it is beyond the scope of this paper.
Locations within range of a gateway (n = 300) obtained at the time they were transmitted, ranged between 4.1–53.4km from the nearest LoRa gateway.
Whereas locations which were recorded but not immediately transmitted (n = 1,908) ranged from 9.3–106km from a gateway, (Table2, Fig.4). Distance from
the nearest gateway was found to be a signicant negative relationship with transmission success (-0.135 + 1.58, DF = 2277, P < 0.001, AIC = 1282) and height
above ground in metres was found to have a signicant positive effect on transmission success (-0.003 + 1.58, DF = 2277, P < 0.001, AIC = 1282). No
signicant difference was found between tags and there was no effect of terrain roughness. Plotting the output of the binomial GLM (Fig.5) suggests the
probability of successful transmission drops below 50% at approximately 15km from the nearest gateway and that transmission range from these devices
during deployment is limited to approximately 50km.
4. Discussion
Standardised tests of tag performance revealed the GPS of the Long-range wireless NOMAD™ is suciently accurate for animal tracking studies with
horizontal bias of < 9m and precision of < 18m and vertical bias of < 5m and precision of < 28m (Fig.6) on up to one hour location acquisition cycles.
Accuracy was improved at higher frequency of GPS position acquisition. Long-range wireless is a promising technology for animal movement studies.
Especially for colonial species that frequently return to the same locations and for smaller species because of the possibility to assemble devices weighing
less than 5g. Ground based tests suggest reliable data transmission up to 40km from the gateway and data from deployed tags indicates a maximum data
transmission range of approximately 53km. For species weighing more than 180g, this transmission range provides a clear advantage over alternative
lightweight GPS loggers using short range transmission (e.g. UHF download). While these other devices perform similarly in terms of GPS accuracy, they either
require manual recovery or for the animal to pass within a few hundred metres of a receiver (Evens et al. 2018; Stienen et al. 2016; Ripperger et al. 2020).
4.1 GPS Accuracy
Under all GPS position acquisition cycles tested, horizontal position bias was less than 9m (4.71–8.44) relative to the true position of the tag with precision of
18 or less (± 2.88 – ±18.0) as measured by the standard deviation from the mean (Fig.3, Table1). This is comparable with other, commercially available GPS
devices and bio-loggers (Forin-Wiart et al. 2015; Acácio et al. 2022; Evens et al. 2018) and well within the accuracy required for the majority of tracking studies
(Forin-Wiart et al. 2015; Katzner and Arlettaz 2020). The relationship between the GPS position acquisition interval and accuracy indicates that where high
position accuracy is a concern, a shorter interval between GPS recording can be adopted. The vertical position bias varied between + 2 and + 5m relative to the
true position. Across all cases these results suggest a slight bias towards over-estimating altitude relative to the true position of the tag in vertical space
(Fig.3d). Precision, as measured by the standard deviation from the mean (Table1), was lowest for the 1-minute acquisition cycle (± 7.77m) compared with
the 30-minute (± 15.3m) and 60-minute (± 27.4m) cycles (Fig.4). The variation in accuracy between GPS acquisition cycles is likely due to the GPS switching
off or going to sleep between xes whereas during the 1-minute cycle, the GPS remains switched on constantly meaning it can maintain contact with a larger
number of GPS satellites. This relationship between GPS accuracy and sampling interval in the Nomad™ tags is comparable to that observed in other GPS
tags (Evens et al. 2018). It is important to be aware of these errors when planning deployment of these tracking devices, particularly where height data is used
to assess the behaviour of the animal relative to anthropogenic hazards such as planes, powerlines or wind turbines (Katzner and Arlettaz 2020).

Page 7/14
4.2 LoRa Data Transmission Range
Our ground-based tests conrmed data transmission up to distances of approximately 40km is reliable (Fig.2). Maps produced using QGIS (QGIS
Development Team 2019). While tags deployed on Griffon Vultures
Gyps fulvus
highlighted that successful data transmission up to approximately 53km is
possible (Fig.8; Table5). Although the probability of successful data transmission declines signicantly in relation to distance from the gateway and
proximity to the ground. Our results suggest that placing gateways approximately every 30–60km should provide sucient coverage for tracking studies for
birds, particularly when paired with the use of mobile gateways which may be temporarily deployed in the eld at colonies, nest sites or known migratory
stopover areas to complement the xed position outdoor gateways. This technology seems to outperform other remote download devices including VHF and
UHF download with typical ranges of a couple of kilometres (Ripperger et al. 2020) and the high frequency Zigbee data transmission protocol that can reach
8.5km with line of sight (Bouten et al. 2013).
Our tests with the mobile gateway included visits to locations with tree cover or large boulders which highlighted how obstructions to line of sight clearly
inhibited data transmission. Therefore, signal coverage is likely to be more of a constraint for tracking mammals with these tags because of the low height
above ground compared with birds. Depending on how mobile the study species is, to facilitate studies of mammal movements, the viewshed would be more
limited, gateways would need to be placed within ~ 5-10km of each other to provide sucient coverage of the study area. Alternative solutions such as regular
drone ights or vehicle transects with a mobile gateway to download the data or deployment of gateways near den sites or known feeding areas could also
help in areas where perfect coverage is not possible.
4.3 Deployment performance
Preliminary results from the trial deployment of the NOMAD™ tags on Griffon Vultures
Gyps fulvus
highlighted tracking data can be obtained over large areas.
This study demonstrated this technology can be used to successfully follow the birds’ daily movements (Fig.4), obtain information on the daily variability in
ight speed, ight height relative to ground level and measure the daily displacement. This included the detection of a failed sea crossing attempt of one bird
(Fig.4b). However, to date, we have only successfully received post-deployment location data for two of the six birds tagged.
Acceleration data was obtained for the remaining four vultures. Although we do not know the route taken by 9DF1_Jethro (total of 2,156 data payloads sent)
and 6923_Carlos (2,047 data payloads sent) between southwest Portugal and Tarifa, the minimum possible distance travelled is approximately 480km. The
data transmission pattern around the Tarifa gateways suggests that the birds were likely thermaling in vicinity of the gateway to gain altitude and then moved
further away. Hence these birds likely attempted to cross the strait of Gibraltar but this has not been conrmed. The tags for FB45_Benoit and 7E32_Aldina
most likely moved to areas out of transmission range. Issues related to animals moving beyond transmission range will ease as the number and density of
LoRaWAN gateways continues to increase. The vulture 9012_Eduardo was not detected by the LoRa gateway in Tarifa which, suggesting that it did not
migrate to Africa. Provided the tag was still on the bird, had it migrated, data would have been received by the gateway in Tarifa.
The capability of the NOMAD™ tags to record high resolution accelerometer data means they have the potential to record when birds collide with infrastructure
such as wind turbines or if a bird is shot. To test this feature, the devices were programmed to trigger acceleration acquisition at 50hz when forces exceeding a
3.2g threshold were detected. Using this feature, the GPS-LoRa tags have been successfully used to study movement of boulders where the accelerometer
triggered events can be used to detect movement of objects in response to environmental factors (e.g. oods or heavy precipitation) (Dini et al. 2021). This
threshold was exceeded for 4 birds during transport and deployment of the loggers resulting in large quantities of acceleration measurements being collected
prior to the release of the birds, so clearing the memory buffer prior to the birds’ release is recommended. Further work is required to rene the appropriate
settings to avoid the 9-axis sensor from being erroneously triggered and preventing location data from being sent.
Despite this issue with the accelerometer data preventing location data being sent, our tests and trial deployment with these tags have demonstrated them to
be a viable alternative to GSM, Iridium or UHF download tags for high resolution GPS tracking studies, particularly as the network of LoRaWAN gateways
improves. As with other gateway or antenna reliant systems, an important consideration for the settings on these tags is how often the bird or other study
species is likely to be in range of a gateway. This will inform the sampling regime and the number of gateways deployed to collect the data. Using the upper
limit of ight speeds recorded during the trial deployment on Griffon Vultures
Gyps fulvus
as a guide (Table4). A bird ying at 6 m/s (21.6 Km/h) will pass
through the area in range of a mobile gateway (17km radius) in approximately 2,833 seconds (47 minutes) allowing up to a maximum of approximately 188
payloads to be sent. This is assuming perfect coverage and no obstacles to transmission, in reality high SF at long transmission ranges is likely to reduce the
transmission rate but we were not able to quantify the impact of this. For a more powerful xed position gateway, depending on terrain, the potential
transmission radius is up to 53km meaning the bird could be in range for 17,666 seconds (4.9 hours) allowing for up to approximately 1,177 payloads to be
sent depending on SF. For transient species who may not be in contact with a gateway very often, lower intensity settings may therefore be more desirable.
This is less of a concern for a site faithful species like White Stork
Ciconia ciconia
because the birds typically return to the same location and spend a
signicant amount of time at the nest allowing considerable time for the data to be sent to a gateway in range of the nest location.
5. Conclusions
This is an exciting time in the eld of movement ecology, a diverse range of technologies are available to monitor the movements and behaviours of animals
(Ripperger et al. 2020). Our tests of GPS accuracy, transmission range and performance of the GPS-LoRa tags have demonstrated their potential as a viable
alternative to other tracking technologies currently available, especially when seeking for gps/sensorial information that can be transmitted from long
distance. Their key advantage over tags which transmit via satellite or GSM is the substantial weight saving because LoRa uses less energy to send the data
over long distances. The ability to use smaller batteries and solar panels will help increase the range of species we are able to track in near real time. The tags
can be deployed in form factors weighing from 5g, for birds with a body mass of 180g or more, up to 80g for birds such as Griffon Vultures
Gyps fulvus
with a
body mass in excess of 8kg. Data can be received from long distances (tens of km), reducing the number of gateways and eld visits required to retrieve the

Page 8/14
data. Currently, these devices are particularly suited to site faithful species. As gateway coverage improves, including plans to launch LoRa gateways into
space (Lacuna 2022), there will be less need for researcher to invest in their own gateway systems to receive data and the range of species that can effectively
be tracked will increase.
Declarations
Acknowledgements
We thank all those who have contributed to this research. Special thanks go to Carlos Pacheco from CIBIO, along with Eduardo Realinho and Hugo Blanco for
their invaluable assistance with eldwork to capture and deploy tags on the vultures. We also thank Miguel González for arranging for us to mount a LoRa
gateway at the Cazalla Observatory in Tarifa, Spain. The team at STRIX (https://strixinternational.com/) kept us informed about the movements of large
vulture ocks. The wildlife rescue organisations RIAS (https://riaswildliferescuecentre.wordpress.com/), CARAS (LPN, Alentejo) and SEPNA/GNR
(https://www.gnr.pt/atrib_SPENA.aspx) for allowing us to deploy tags on rehabilitated birds. Nick Grin and David Ciplic from the University of East Anglia
electronics workshop for their assistance. For technical assistance with the devices and LoRa gateways, we thank Marcel Wapper and Simon Galli from
Miromico-AGTM, Benedetta Dini from the University of Grenoble, Steven Drewett from Concept13TM and Miles Clark from the University of East Anglia.
Conict of Interest
AMAF, PA and JPS obtained proof of concept funding to co-develop this tracking device with Miromico as a Movetech project. The devices assembled for the
vultures were developed for research purposes only and are not available commercially. Aside from this, all authors declare that they have no conict of
interest with the content of this work.
Authors’ Contributions
J.G.G. is the principal author of the work; A.M.A.F. suggested the initial idea for the paper and provided advice throughout the analysis and writing stages of
the work along with P.W.A., J.P.S. and A.S.; All authors contributed critically to data collection, drafting the manuscript, gave nal approval and consent for
publication of the manuscript.
Availability of data and materials
All data and R-scripts used to produce this work are available for download from: https://drive.google.com/drive/folders/1gvgRK2TnU1SuGhYkAyoDlVTu-
QBSDLHf?usp=sharing
Ethics Approval
All eldwork with vultures was performed under licence from CNF - Instituto da Conservação da Natureza e das Florestas (the Portuguese Government agency
responsible for wildlife and orests) and Ethics approval was granted by the ethics committee of the University of East Anglia.
References
1. Acácio, Marta, Philip W. Atkinson, João Paulo Silva, and Aldina M. A. Franco. 2022. “Performance of GPS/GPRS Tracking Devices Improves with
Increased Fix Interval and Is Not Affected by Animal Deployment.” Plos One 17 (3): e0265541. https://doi.org/10.1371/journal.pone.0265541.
2. Bouten, Willem, Edwin W. Baaij, Judy Shamoun-Baranes, and Kees C.J. Camphuysen. 2013. “A Flexible GPS Tracking System for Studying Bird Behaviour
at Multiple Scales.” Journal of Ornithology 154 (2): 571–80. https://doi.org/10.1007/s10336-012-0908-1.
3. Concept 13 Limited. 2022. “Concept 13: Sensors-Gateways-LoRaWAN.” 2022. https://www.concept13.co.uk/.
4. Dini, Benedetta, Georgina L. Bennett, Aldina M.A. Franco, Michael R.Z. Whitworth, Kristen L. Cook, Andreas Senn, and John M. Reynolds. 2021.
“Development of Smart Boulders to Monitor Mass Movements via the Internet of Things: A Pilot Study in Nepal.” Earth Surface Dynamics 9 (2): 295–315.
https://doi.org/10.5194/esurf-9-295-2021.
5. ESRI. 2021. “Sentinel-2 10-Meter Land Use/Land Cover.” 2021. https://livingatlas.arcgis.com/landcover/.
. Evans;, Jeffrey S., Melanie A. Murphy;, and Karthik Ram. 2021. “SpatialEco: Spatial Analysis and Modelling Utilities.” cran-r. https://cran.r-
project.org/web/packages/spatialEco/index.html.
7. Evens, Ruben, Natalie Beenaerts, Eddy Ulenaers, Nele Witters, and Tom Artois. 2018. “An Effective, Low-Tech Drop-off Solution to Facilitate the Retrieval of
Data Loggers in Animal-Tracking Studies.” Ringing and Migration 33 (1): 10–18. https://doi.org/10.1080/03078698.2018.1521116.
. Forin-Wiart, Marie Amélie, Pauline Hubert, Pascal Sirguey, and Marie Lazarine Poulle. 2015. “Performance and Accuracy of Lightweight and Low-Cost GPS
Data Loggers According to Antenna Positions, Fix Intervals, Habitats and Animal Movements.” PLoS ONE 10 (6): 1–21.
https://doi.org/10.1371/journal.pone.0129271.
9. Frankish, Caitlin K., Cleo Cunningham, Andrea Manica, Thomas A. Clay, Stephanie Prince, and Richard A. Phillips. 2021. “Tracking Juveniles Conrms
Fisheries-Bycatch Hotspot for an Endangered Albatross.” Biological Conservation 261 (January): 109288. https://doi.org/10.1016/j.biocon.2021.109288.
10. Hijmans, R. J., E. Williams, and C. Vennes. 2015. “Geosphere: Spherical Trigonometry.” R. url: https://CRAN. R-project. org/package = geosphere.
11. Hijmans, Robert J. 2019. “Raster: An R Package for Working with Raster Data.” https://www.rspatial.org/.
12. IoT Wonderland. 2022. “IoT Wonderland: Geo-Spacial Data Visualization.” 2022. https://www.iotwonderland.com/.

Page 9/14
13. Katzner, Todd E., and Raphaël Arlettaz. 2020. “Evaluating Contributions of Recent Tracking-Based Animal Movement Ecology to Conservation
Management.” Frontiers in Ecology and Evolution 7 (January). https://doi.org/10.3389/fevo.2019.00519.
14. Kauth, Hilary R., Robert C. Lonsinger, Adam J. Kauth, and Andrew J. Gregory. 2020. “Low-Cost DIY GPS Trackers Improve Upland Game Bird Monitoring.”
Wildlife Biology
2020 (2). https://doi.org/10.2981/wlb.00653.
15. Kim, Seungku, Heonkook Lee, and Sungho Jeon. 2020. “An Adaptive Spreading Factor Selection Scheme for a Single Channel Lora Modem.” Sensors
(Switzerland) 20 (4). https://doi.org/10.3390/s20041008.
1. Lacuna. 2022. “Lacuna Space.” 2022. https://lacuna.space/.
17. LORIOT. 2021. “LORIOT Internet of Things LoRaWAN Server.” 2021. https://www.loriot.io/.
1. Lüdecke, Daniel, Frederik Aust, Sam Crawley, and Mattan S. Ben-Shachar. 2022. “Ggeffects: Create Tidy Data Frames of Marginal Effects for ‘ggplot’ from
Model Outputs.”
19. Mekki, Kais, Eddy Bajic, Frederic Chaxel, and Fernand Meyer. 2019. “A Comparative Study of LPWAN Technologies for Large-Scale IoT Deployment.” ICT
Express 5 (1): 1–7. https://doi.org/10.1016/j.icte.2017.12.005.
20. Miromico AG. 2022. “Miromico Docs: Tracking Solutions Documentation.” Zurich. https://docs.miromico.ch/tracker/dev/index.html.
21. Movebank. 2019. “Movebank Data Repository.” 2019. https://www.movebank.org/.
22. Multi-Tech Systems Inc. 2021. “MultiTech Conduit IP67 Base Station 16-Channel v2.1 Geolocation EU686 for Europe.” 2021.
https://www.multitech.com/documents/publications/data-sheets/86002223.pdf.
23. ———. 2022. “Multi-Tech DeviceHQ: Cloud-Based Application Store and IoT Device Management.” 2022. .
24. Muteba, Franck, Karim Djouani, and Thomas Olwal. 2019. “A Comparative Survey Study on LPWA IoT Technologies: Design, Considerations, Challenges
and Solutions.” Procedia Computer Science 155: 636–41. https://doi.org/10.1016/j.procs.2019.08.090.
25. Ossi, Federico, Ferdinando Urbano, and Francesca Cagnacci. 2019.
Biologging and Remote-Sensing of Behavior
.
Encyclopedia of Animal Behavior
.
Second Edi. Vol.3. Elsevier. https://doi.org/10.1016/B978-0-12-809633-8.90089-X.
2. Pedersen, Thomas Lin. 2020. “Patchwork: The Composer of Plots.” https://patchwork.data-imaginist.com.
27. Perona, Arturo M., Vicente Urios, and Pascual López-López. 2019. “Holidays? Not for All. Eagles Have Larger Home Ranges on Holidays as a Consequence
of Human Disturbance.”
Biological Conservation
231 (July 2018): 59–66. https://doi.org/10.1016/j.biocon.2019.01.010.
2. QGIS Development Team. 2019. “QGIS Geographic Information System.” Open Source Geospatial Foundation Project.
29. R Core Team. 2019. “R: A Language and Environment for Statistical Computing.” http://www.r-project.org/.
30. Riley, S.J., S.D. DeGloria, and R. Elliot. 1999. “A Terrain Ruggedness Index That Quanties Topographic Heterogeneity.” Intermountain Journal of Sciences
5: 23–27.
31. Ripperger, Simon P., Gerald G. Carter, Rachel A. Page Supervision, Niklas Duda, Alexander Koelpin, Robert Weigel, Markus Hartmann, et al. 2020. “Thinking
Small: Next-Generation Sensor Networks Close the Size Gap in Vertebrate Biologging.” PLoS Biology 18 (4): 1–25.
https://doi.org/10.1371/journal.pbio.3000655.
32. Rodríguez, Airam, Juan J. Negro, Mara Mulero, Carlos Rodríguez, Jesús Hernández-Pliego, and Javier Bustamante. 2012. “The Eye in the Sky: Combined
Use of Unmanned Aerial Systems and GPS Data Loggers for Ecological Research and Conservation of Small Birds.”
PLoS ONE
7 (12).
https://doi.org/10.1371/journal.pone.0050336.
33. Rotics, Shay, Sondra Turjeman, Michael Kaatz, Damaris Zurell, Martin Wikelski, Nir Sapir, Wolfgang Fiedler, et al. 2021. “Early-Life Behaviour Predicts First-
Year Survival in a Long-Distance Avian Migrant.”
Proceedings of the Royal Society B: Biological Sciences
288 (1942).
https://doi.org/10.1098/rspb.2020.2670.
34. Schaub, Tonio, Raymond H.G. Klaassen, Willem Bouten, Almut E. Schlaich, and Ben J. Koks. 2020. “Collision Risk of Montagu’s Harriers Circus Pygargus
with Wind Turbines Derived from High-Resolution GPS Tracking.” Ibis 162 (2): 520–34. https://doi.org/10.1111/ibi.12788.
35. Schloerke, Barret, Di Cook, Joseph Larmarange, Francois Briatte, Moritz Marbach, Edwin Thoen, Amos Elberg, et al. 2021. “GGally: Extension to ‘Ggplot2.’”
cran-r. https://cran.r-project.org/package=GGally.
3. Stienen, Eric W.M., Peter Desmet, Bart Aelterman, Wouter Courtens, Simon Feys, Nicolas Vanermen, Hilbran Verstraete, et al. 2016. “GPS Tracking Data of
Lesser Black-Backed Gulls and Herring Gulls Breeding at the Southern North Sea Coast.”
ZooKeys
2016 (555): 115–24.
https://doi.org/10.3897/zookeys.555.6173.
37. The Things Network. 2019. “LoRaWAN® Distance World Record Broken, Twice. 766 Km (476 Miles) Using 25mW Transmission Power.” 2019.
https://www.thethingsnetwork.org/article/lorawan-distance-world-record.
3. TTN. 2016. “The Things Network Fair Usage Policy.” 2016. https://www.thethingsnetwork.org/forum/t/fair-use-policy-explained/1300.
39. ———. 2021. “The Things Network LoRaWAN Server.” 2021. https://www.thethingsnetwork.org/.
40. UK Research and Innovation. 2016. “Long Range Wireless Devices for High-Resolution Monitoring of Animal Movement.” 2016.
https://gtr.ukri.org/projects?ref=NE%2FP003907%2F1.
41. Walther, Bruno A., and Joslin L. Moore. 2005. “The Concepts of Bias, Precision and Accuracy, and Their Use in Testing the Performance of Species
Richness Estimators, with a Literature Review of Estimator Performance.” Ecography 28 (6): 815–29. https://doi.org/10.1111/j.2005.0906-7590.04112.x.
42. Wickham, H. 2016. “Ggplot2: Elegant Graphics for Data Analysis in R.”
Figures

Page 10/14
Figure 1
An overview of the LoRa system. The data is sent via LoRa [A] to a gateway which in turn forwards the data [B] to a network server such as LORIOT [C] via an
internet connection such as GSM, WiFi or Ethernet. The network server then forwards the data onto one or multiple application servers [D] which decode the
data into a format suitable for a data base such as Movebank [E] where the data can be downloaded for analysis by the user [F]. Fixed position gateways can
be indoors, mounted to a building or standalone solar powered systems. Mobile gateways can be powered by a portable power bank and carried in a car, on
foot or own on a drone to maximise coverage. The lower panel highlights how these tags are used in practice to track animals.

Page 11/14
Figure 2
A: Pole used to elevate the NOMAD devices to 5m at each location. B: Ground based range test locations for the xed position LoRa gateway. C. Ground based
range test locations for the mobile LoRa gateway. D. the mobile gateway used. E. The location in Portugal where the mobile gateway was tested. Hollow
circles represent locations where data transmission from the devices to the gateway was not conrmed, lled circles represent locations where data
transmission was conrmed.

Page 12/14
Figure 3
A: Distribution of horizontal GPS locations relative to the true position under different GPS acquisition cycles. The mean position in all instances is similar,
however the precision of the altitude is best under a 1-minute GPS acquisition cycle. B: The distribution of GPS position estimates relative to the true position
in vertical space. C: Horizontal position error under a 1-minute, 30-minute and 60-minute GPS acquisition cycle, error bars represent the standard deviation
from the mean. D: Vertical position error under a 1-minute, 30-minute and 60-minute GPS acquisition cycle, error bars represent the standard deviation from
the mean. As expected, both bias and precision was best under the 1-minute cycle.

Page 13/14
Figure 4
Panels A and B show the movements of the two tracked vultures. All locations shown are in ight. Panel C: mean height of the birds above ground and the
95% condence interval. Panel D: mean speed of the birds along with the 95% condence interval and panel E: location of the four LoRa gateways deployed
for this project.

Page 14/14
Figure 5
Relationship of the likelihood of successful data transmission with distance from a LoRa gateway (A) and height above ground(B).
Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.
Chapter3bSupplementaryMaterialA1.docx
SupportingInformationA2batteryconsumptionestimatelora.xlsx