Impact of Temperature Variations on the Reliability of LoRa

An Experimental Evaluation

Carlo Alberto Boano, Marco Cattani and Kay R

omer

Institute of Technical Informatics, Graz University of Technology, Austria

Keywords:

Long-range technology, Networks, PHY settings, Temperature, Reliability, RSSI, Wireless.

Abstract:

Temperature variations are known to affect the performance of wireless sensor networks deployed outdoors.

Whilst the impact of temperature on IEEE 802.15.4 transceivers has long been investigated by the research

community, still little is known about how temperature affects the performance of increasingly popular long-

range wireless technologies such as LoRa. To ﬁll this gap, this paper presents an experimental evaluation

of the reliability of LoRa in the presence of temperature variations. First, we highlight that temperature can

have a signiﬁcant impact on LoRa’s communication performance and show that an increase in temperature

can be sufﬁcient to transform a perfect LoRa link into an almost useless one. We then carry out a detailed

investigation on the performance of different LoRa physical settings with ﬂuctuating temperatures and show

that an optimal selection can help in increasing the probability of packet reception and is hence key to mitigate

temperature-induced effects. We believe that our results will serve as a reference to orient researchers and

system designers employing LoRa to build large-scale low-power wide area networks.

1 INTRODUCTION

Outdoor environmental conditions are known to dras-

tically affect the performance of wireless sensor net-

works (WSN). Networks deployed outdoors, indeed,

often experience a reduction of the delivery rate and

the connectivity between nodes in the presence of

fog (Anastasi et al., 2004), rain (Boano et al., 2009),

snow cover (Stoianov et al., 2007), thick vegeta-

tion (Marﬁevici et al., 2013), and temperature varia-

tions (Bannister et al., 2008). Especially the latter can

have a severe impact on the performance of outdoor

WSN systems (Wennerstr

om et al., 2013), as they can

affect clock drift, battery capacity and discharge, as

well as radio efﬁciency, hence directly affecting fun-

damental aspects such as time synchronization, net-

work lifetime, and reliability of transmissions.

Most of the existing work investigating the im-

pact of temperature on low-power wireless networks

has focused on IEEE 802.15.4 transceivers. The

impact of temperature was shown to be platform-

speciﬁc (Boano et al., 2010) and to lead to se-

vere consequences on the functionality of duty-cycled

medium access control (Boano et al., 2014a) and rout-

ing (Keppitiyagama et al., 2013) protocols.

Despite the comprehensive studies carried out on

IEEE 802.15.4 transceivers, however, still little is

known about the impact of temperature on other low-

power wireless technologies, especially on the in-

creasingly popular low-rate and long-range transmis-

sion technologies used to realize low-power wide area

networks (Centenaro et al., 2016).

Long-range wireless technologies such as Sig-

fox (Sigfox, 2017), Weightless (Weightless SIG,

2017), and LoRa (LoRa Alliance, 2017) are very

promising for the realization of large-scale Internet of

Things (IoT) applications, as they allow city-wide de-

ployments and large outdoor installations in remote

areas. Among others, LoRa has been proposed to re-

alize a number of smart city applications (Kartakis

et al., 2016), outdoor parking guidance systems (KSK

Developments, 2017), as well as smart water manage-

ment infrastructures (Cattani et al., 2017b).

As a few preliminary works have shown a possi-

ble temperature-dependent performance of LoRa net-

works deployed outdoors (Iova et al., 2017), (Cattani

et al., 2017a), it is important to investigate in detail

the impact of temperature on LoRa communications

and understand how it can be possibly mitigated.

Contributions. In this paper we experimentally study

the impact of temperature on the performance of

LoRa using a temperature-controlled testbed. We ﬁrst

show that temperature has a strong impact on the per-

formance of LoRa, and that an increase in temper-

Boano, C., Cattani, M. and Römer, K.

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation.

DOI: 10.5220/0006605600390050

In Proceedings of the 7th International Conference on Sensor Networks (SENSORNETS 2018), pages 39-50

ISBN: 978-989-758-284-4

ature similar to the daily and seasonal ﬂuctuations

of temperature outdoors may transform a good link

(i.e., a link with high packet reception rate) into an

almost useless one (i.e., a link sustaining a packet re-

ception rate close or equal to 0). In particular, we

show that a gradual increase in temperature leads to a

higher number of corrupted and lost packets, as well

as to a progressive reduction of the received signal

strength (RSS). We show that such RSS attenuation

is platform-dependent and that it varies depending on

the distance between nodes.

We then thoroughly analyze the performance of

different LoRa’s physical layer (PHY) settings in the

presence of varying temperatures and show that an

optimal selection is key to minimize temperature-

induced effects. In particular, we show that select-

ing a lower bandwidth and a higher spreading factor,

as well as a more robust coding rate can signiﬁcantly

help in maintaining a reliable communication despite

an increase of the on-board temperature.

This paper proceeds as follows. In the next sec-

tion, we introduce the reader to the increasingly pop-

ular LoRa long-range technology and to the physi-

cal layer settings that can be conﬁgured to ﬁne-tune

the transceiver’s operations. After describing related

work in Sect. 3, we show the results of a series of

experiments conducted in a temperature-controlled

testbed in Sect. 4, highlighting how an increase in

temperature can transform a perfect link into an al-

most useless one. We then thoroughly analyze the

performance of each PHY setting in LoRa at varying

temperature in Sect. 5 and show how an optimal se-

lection can help minimizing temperature-induced ef-

fects. After elaborating on our ﬁndings in Sect. 6, we

conclude the paper and outline possible future work

in Sect. 7.

2 LoRa TECHNOLOGY

LoRa is a proprietary radio modulation technology

developed by Cycleo and acquired by Semtech in

2012 that is very promising for building wide access

networks with star topology (often referred to as low-

power wide area networks). The latter provide long-

range communication to thousands of devices at min-

imal cost and limited energy expenditure, therefore

allowing to realize large-scale urban IoT networks.

LoRa has the ability to improve the signal-to-noise

ratio (SNR) at the receiver by spreading the energy of

the signal over a wider frequency band, effectively re-

ducing the spectral power density of the signal (Kar-

takis et al., 2016). The core of LoRa technology is

its chirp spread spectrum (CSS) modulation: the car-

rier signal of LoRa consists of chirps, signals whose

frequency increases or decreases over time (Cattani

et al., 2017a). LoRa’s chirps allow the signal to travel

distances up to several kilometers (Centenaro et al.,

2016) and to be demodulated even when its power is

up to 20 dB lower than the noise ﬂoor.

One of the key advantages of LoRa is the re-

duced complexity of networking protocols, as the

long communication range allows to form star topolo-

gies where the low-power end devices directly com-

municate with a more powerful orchestrator. This al-

lows to design asymmetric communication schemes

and to shift the load to a powerful central device,

keeping the design of end devices simple and cheap.

LoRa transceivers communicate using the sub-

GHz unlicensed industrial, scientiﬁc, and medical

(ISM) bands. Among others, LoRa exploits for its

communications the 433 MHz and 868 MHz ISM

bands in Europe, and the 915 MHz ISM band in North

America. LoRa radios also allow to adjust the trans-

mission power and hence to control the energy neces-

sary to transmit a packet: common transceivers sup-

port transmission powers between -4 and +20 dBm.

Besides the selection of carrier frequency and

transmission power (supported by most low-power

transceivers used to build IoT applications), the com-

munication performance of LoRa can be ﬁne-tuned by

varying a number of PHY settings, such as bandwidth,

spreading factor, coding rate, and carrier frequency.

Among others, these speciﬁc PHY settings allow to

trade receiver sensitivity (and hence a longer commu-

nication range and a more robust communication) for

a higher data-rate, as we describe next.

Bandwidth (BW). By varying the range of frequencies

over which LoRa’s chirps spread (i.e., by varying the

bandwidth), one can trade radio air-time against radio

sensitivity, thus choosing between a higher energy-

efﬁciency (lower air-time) and a higher communica-

tion range and robustness. The use of a narrow band-

width maximizes sensitivity, but increases air-time.

Increasing the bandwidth, instead, allows for faster

transmissions (and hence a lower air-time), but re-

duces sensitivity (Semtech Corporation, 2013).

Spreading Factor (SF). LoRa “spreads” each symbol

(information bits) over several chips to increase the

receiver’s sensitivity. LoRa’s spreading factor can be

selected between 6 and 12, resulting in a spreading

rate ranging from 2

to 2

chips/symbol. Please note

that LoRa modulation employs orthogonal spreading

factors. This enables multiple spread signals to be

transmitted at the same time and on the same channel

with minimal degradation of the receiver sensitivity

SENSORNETS 2018 - 7th International Conference on Sensor Networks

Table 1: Summary of the main conﬁgurable physical settings in LoRa and their impact on communication performance.

Setting Typical values Impact on communication performance

Bandwidth [kHz] 125 ... 500 A higher bandwidth allows to transmit packets at a higher data rate. How-

ever, a higher bandwidth also reduces the receiver sensitivity and hence the

communication range.

Spreading Factor 6 .. . 12 A high spreading factor increases the signal-to-noise ratio and hence the ra-

dio sensitivity / communication range. However, a higher spreading factor

increases the length of packets, hence causing a higher energy expenditure.

Coding Rate 4/5 .. . 4/8 A larger coding rate increases the resilience to interference bursts and de-

coding errors. However, a large coding rate implies the transmission of

longer packets and hence increases the energy expenditure.

(Semtech Corporation, 2015), i.e., packets transmit-

ted with different spreading factors appear as noise to

the target receiver.

Coding Rate (CR). LoRa makes use of forward er-

ror correction to increase the resilience to packet cor-

ruption. In particular, one can specify the number of

redundant bits, ranging from 1 to 4, where a higher

number should be used when transmitting in con-

gested radio environments (i.e., one should select a

higher coding rate to maximize the probability of suc-

cessful packet reception). Transceivers operating with

different coding rates can still communicate to each

other, since the packet header (transmitted using the

maximum coding rate of 4/8) includes the code rate

used for the payload.

Table 1 summarizes the typical values and the

impact of each PHY setting on data rate, re-

ceiver sensitivity, communication range, and energy-

efﬁciency (Semtech Corporation, 2015).

3 RELATED WORK

The impact of temperature on the performance of

wireless sensor networks has been largely analyzed

by the research community, especially in the context

of IEEE 802.15.4 radios. After showing the correla-

tion between temperature and received signal strength

(RSS) in a deployment in the Sonoran desert, Bannis-

ter et al. have conﬁrmed in a temperature-controlled

chamber that the RSS of the TI CC2420 radio atten-

uates at high temperatures. The authors have further

identiﬁed that such RSS attenuation is due to the im-

pact of temperature on the CC2420 transceiver’s low-

noise and power ampliﬁers (Bannister et al., 2008).

Based upon this work, a number of researchers

have conﬁrmed that these ﬁndings also apply in a sim-

ilar way to other IEEE 802.15.4 platforms such as the

TI CC1020 and CC2520 (Boano et al., 2010), and

have highlighted how diurnal and seasonal temper-

ature variations can cause a complete disruption of

an IEEE 802.15.4 link (Boano et al., 2013). Wen-

nerstr

om et al. have also presented results from a

long-term outdoor deployment of TelosB nodes and

have shown that packet reception ratio and RSS are

highly correlated with temperature, whereas their cor-

relation with other factors such as absolute humid-

ity and precipitation is less pronounced (Wennerstr

et al., 2013). Other authors have also shown how

the impact of temperature variations on low-power

radio transceivers may strongly affect the operations

of duty-cycled medium access control (Boano et al.,

2014a), (Oppermann et al., 2015), as well as routing

protocols (Keppitiyagama et al., 2013).

The impact of environmental conditions on long-

range radios, instead, has not yet been investigated in

detail. In previous work (Cattani et al., 2017a) we

have reported that one can observe a certain correla-

tion between temperature ﬂuctuations and variations

in packet reception rate and received signal strength in

outdoor and underground LoRa deployments. How-

ever, we did not study the problem in depth and we

did not investigate the impact of temperature on com-

munication when using different LoRa PHY settings.

Iova et al. (Iova et al., 2017) have deployed a num-

ber of LoRa networks in urban and mountain environ-

ments, and reported that environmental factors such

as the presence of vegetation and temperature vari-

ations can negatively affect communication perfor-

mance. The authors, however, did not quantify the

impact of these environmental factors (especially the

one of temperature) in detail.

The remaining body of work on LoRa has

focused primarily on the characterization of

packet loss (Marcelis et al., 2017), signal atten-

uation (Pet

arvi et al., 2015), sensitivity (Augustin

et al., 2016), channel utilization (Georgiou and Raza,

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation

100

0 10 20 30 40 50 60 70 80 90

Rec. correctly Rec. corrupted Lost

0 10 20 30 40 50 60 70 80 90

Temp. [°C]

Time [min]

(a) Link 0–1 (SF = 7)

100

0 10 20 30 40 50 60

Rec. correctly Rec. corrupted Lost

0 10 20 30 40 50 60

Temp. [°C]

Time [min]

(b) Link 0–2 (SF = 12)

Figure 1: Disruption of two LoRa links caused by an increase of the on-board temperature. A very good link (i.e., sustaining

almost 100% successful receptions) at low temperature experiences an increasing corruption and loss as soon as temperature

increases, up to a point in which the link becomes unusable. The two links make use of spreading factor 7 and 12, respectively.

2017), (Voigt et al., 2017), and energy consump-

tion (Bor and Roedig, 2017). Other works have also

studied the ability of LoRa to penetrate buildings (Bor

et al., 2016a) and to receive packets from concurrent

transmissions (Bor et al., 2016b).

In this paper, we speciﬁcally study the impact of

temperature on LoRa’s communication performance

and answer two key questions: (i) how severe is such

an impact, and (ii) whether it can be minimized with a

proper selection of PHY settings. Towards this goal,

we carry out several experiments in a temperature-

controlled testbed and analyze the communication

performance of LoRa links using different PHY set-

tings, as we describe in the next sections.

4 IMPACT OF TEMPERATURE

ON LORA COMMUNICATIONS

We ﬁrst study whether variations in temperature can

lead to a signiﬁcant packet loss or a visible RSS at-

tenuation, similar to the one shown in the literature

for IEEE 802.15.4 radios. Towards this goal, we make

use of an experimental setup where the temperature of

several nodes can be controlled in a ﬁne-grained fash-

ion (Sect. 4.1), and analyze the evolution of packet

reception rate (Sect. 4.2) and RSS (Sect. 4.3).

4.1 Experimental Setup

We make use of TempLab (Boano et al., 2014b), a

temperature-controlled testbed, to expose a number

of LoRa nodes to repeatable temperature variations

Please note that the temperature variations to which

LoRa transceivers are exposed when using TempLab

are within the supported temperature range of LoRa

Infra-red heating lamps and enclosures are distributed

across a 50m

area and can be controlled wirelessly.

We build a star network of LoRa nodes composed

of a sink (node 0) periodically broadcasting messages

with 5 bytes payload to four nodes (1 to 4). Two

of the nodes (1 and 2) are at the edge of their com-

munication range (signal strength of received pack-

ets close to -100 dBm), whereas the remaining nodes

(3 and 4) are located closer to the sink and hence re-

ceive packets with a higher signal strength. All nodes

are based on the Moteino MEGA platform (LowPow-

erLab, 2016) that embeds a HopeRF RFM95 LoRa

transceiver (Hope RF Microelectronics, 2017). We

employ the highest transmission power available (+5

dBm), a coding rate of 4/5, a bandwidth of 125 kHz

and a spreading factor of 7 or 12.

All nodes measure temperature using a Bosch

BME280 sensor and log this information via USB on

a central testbed PC, along with the content of the re-

ceived messages, the results of the cyclic redundancy

check, the received signal strength, and the SNR. De-

pending on the employed spreading factor, the sink

transmits messages at 0.5 or 2 Hz (SF=12 or SF=7,

respectively). We do not make use of any radio duty

cycling and initially do not connect an SMA antenna

to the nodes in order to limit their range. We have

later repeated some of the experiments with an SMA

antenna or cable and obtained similar results.

4.2 Impact on Packet Reception

In a ﬁrst experiment, we focus on the two receivers

that are located at a distance from the sink that is close

to the edge of their communication range, but that still

guarantees a high packet reception rate (i.e., nodes 1

transceivers, i.e., between -55 and +115

◦

C for the Hop-

eRFM95 and the Semtech SX1272 transceivers.

SENSORNETS 2018 - 7th International Conference on Sensor Networks

-64

-63

-62

-61

-60

-59

-58

0 10 20 30 40 50

RSSI [dBm]

Temperature [°C]

Fitting line: f(x)=-0.073x -58.7 (r

=+0.94)

RSSI value

(a) Node 3 (SF = 7)

-86

-85

-84

-83

-82

-81

-80

-79

-78

0 10 20 30 40 50

RSSI [dBm]

Temperature [°C]

Fitting line: f(x)=-0.090x -79.7 (r

=+0.95)

RSSI value

(b) Node 4 (SF = 7)

Figure 2: RSSI attenuation as a function of temperature on the Moteino MEGA employing a HopeRF RFM95 transceiver.

and 2). We then instruct the temperature-controlled

testbed to slowly increase the temperature of the sink

node and to then quickly cool down its temperature to

the initial value. We repeat the test multiple times and

make use of different temperature proﬁles.

Link Disruption. Fig. 1 shows the distribution of

lost, corrupted, and successfully received packets

over time for the two different links. In both cases, we

can observe that what was a perfect link at low tem-

peratures, i.e., a link sustaining a 100% packet recep-

tion rate (PRR), slowly becomes unusable at higher

temperatures. An increase in temperature at the trans-

mitter, indeed, causes node 1 to experience a signif-

icant corruption and loss (Fig. 1(a)), up to a point at

which the link becomes unusable when the transmit-

ter’s on-board temperature is higher than 55

◦

C. Simi-

larly, also node 2 does not receive any packet at high

temperatures (Fig. 1(b)): what was an almost perfect

link at 0

◦

C becomes unusable when the transmitter’s

on-board temperature is higher than 55

◦

It is worth highlighting that the two links shown in

Fig. 1 exhibit a visibly different amount of corruption

and react to temperature variations differently. In-

deed, link 0–2 experiences more corruption than its

counterpart, and its transitional region (i.e., the re-

gion in which the nodes do not communicate reli-

ably (Z

niga and Krishnamachari, 2004)) covers ap-

proximately a temperature variation of 60

◦

C, whereas

the latter is 30

◦

C for link 0–1. As the two links make

use of largely different spreading factors (SF=7 for

link 0–1, whilst SF=12 for link 0–2), our results seem

to hint that the behavior of LoRa links exposed to tem-

perature variations depends on the PHY settings used.

4.3 Impact on Received Signal Strength

In a second experiment, we focus on the two Moteino

MEGA that are located in closer proximity to the

sink (i.e., nodes 3 and 4), conﬁgure them with identi-

cal settings (i.e., CR=4/5, BW=125 kHz, and SF=7),

-54

-52

-50

-48

0 10 20 30 40 50

RSSI [dBm]

Temperature [°C]

Fitting line: f(x)=-0.111x -48.2 (r

=+0.93)

RSSI value

Figure 3: RSSI attenuation as a function of temperature

recorded on ST Microelectronics Nucleo L073RZ platform

employing a Semtech SX1272 transceiver.

and study the evolution of received signal strength

while temperature increases. We instruct the testbed

to slowly increase the on-board temperature of all

nodes in the range between 0 and 50

◦

C during ﬁve

hours. We then plot the relationship between the RSS

of received packets and the average temperature of the

communicating nodes. Fig. 2 shows the results.

RSS Attenuation at High Temperatures. A ﬁrst

observation is that the received signal strength on

a Moteino MEGA board (HopeRF RFM95 radio)

exhibits an attenuation at high temperatures simi-

lar to the linear trend observed on IEEE 802.15.4

transceivers (Boano et al., 2013). Both nodes 3 and

4 exhibit indeed a linear decrease of the RSS in dis-

crete steps for a total of about 3-4 dB in the temper-

ature range of [0 − 50]

◦

C. Bannister et al. argue that

the reason for this attenuation in the TI CC2420 radio

is due to the fact that, for a given voltage, a higher

temperature increases the resistance of conductors,

while reducing the pass-through current. For radio

transceivers, this implies that higher temperatures re-

duce the received signal strength and SNR (Bannister

et al., 2008). We conjecture that LoRa transceivers

are affected in a similar way.

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation

Table 2: PHY settings used in our experiments ordered by decreasing bit-rate.

Setting ID 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

SF 7 7 7 7 7 7 9 9 9 9 9 9 12 12 12 12 12 12

CR 4/5 4/5 4/5 4/8 4/8 4/8 4/5 4/5 4/5 4/8 4/8 4/8 4/5 4/5 4/5 4/8 4/8 4/8

BW (kHz) 125 250 500 125 250 500 125 250 500 125 250 500 125 250 500 125 250 500

RSS Attenuation Varies at Different Distances.

Another observation from our results is that the ﬁt-

ting line shown in Fig. 2(a), interpolating the received

RSS values of node 3 (whose RSS is about -60 dBm),

has a smaller slope than the one shown in Fig. 2(b)

for node 4 (whose RSS is in the area of -80 dBm).

This hints that the impact of temperature is speciﬁc

to each node and may depend on the distance to the

transmitter.

Platform-speciﬁc Attenuation. We further investi-

gate the RSS attenuation using a different LoRa plat-

form. We employ the same experimental setup de-

scribed in Sect. 4.1 and expose ST Nucleo L073RZ

boards (ST Microelectronics, 2016) equipped with a

Semtech SX1272 radio (Semtech Corporation, 2017)

to the same change in temperature. We employ

the same settings described previously, i.e., CR=4/5,

SF=7, BW=125 kHz. Fig. 3 shows the relationship

between RSS and temperature: also in this case we

can observe a linear decrease of RSS in discrete steps

by a total of about 5-6 dB in the temperature range

of [0 − 50]

◦

C. The absolute decrease is higher than

that observed on the Moteino MEGA nodes, which

hints that the RSS attenuation – as for IEEE 802.15.4

transceivers – may be platform-speciﬁc.

5 THE ROLE OF PHY SETTINGS

Next, we explore whether changing LoRa’s PHY set-

tings helps in minimizing the impact of temperature

on the reliability of communications. In our cur-

rent investigation, we speciﬁcally focus on the role

of bandwidth, spreading factor, and coding rate.

5.1 Explored Settings

We create an application that periodically reboots all

nodes and switches to a different combination of PHY

settings over time. Using the time-stamp provided by

a Maxim DS3231 real-time clock, we reboot transmit-

ter and receivers every 6 minutes and iterate over a list

of 18 hard-coded combinations of bandwidth, spread-

ing factor, and coding rate, as summarized in Table 2.

Please note that the settings are ordered by decreasing

bit-rate (i.e., setting ID 21 is the slowest one) and that

they have been selected to enable an easy comparison

with earlier works studying LoRa’s performance as a

function of its PHY settings (Cattani et al., 2017a).

We ﬁrst let the temperature-controlled testbed heat

transmitter and receiver nodes to 30

◦

C – a tempera-

ture that we use as a baseline. After running each

combination of PHY settings at least three times,

hence allowing the transmitter to broadcast about

5000 packets, we let the testbed vary the tempera-

ture of the transmitter and the receiver to 70

◦

C and

re-iterate over the 18 combinations of settings. In

order to study whether an increase in temperature at

the transmitter or at the receiver affects performance

differently, we let the testbed heat to 70

◦

C ﬁrst only

the transmitting node, then only the receivers (while

keeping the transmitter at the 30

◦

C baseline), and then

both transmitter and receivers.

The rest of the experimental setup is the same as

described in Sect. 4.1, with each Moteino MEGA log-

ging via USB all statistics about packet reception,

RSS, and SNR to the main testbed PC. We analyze

the impact of temperature on PRR and RSS/SNR as

a function of the employed PHY settings, describing

our ﬁndings in Sect. 5.2 and 5.3, respectively.

5.2 Packet Reception

Figs. 4 and 5 summarize the impact of bandwidth,

spreading factor and coding rate on the packet re-

ception of nodes 2 and 1, respectively. For each

combination of PHY settings, four bars depict the

packet reception when none of the nodes was heated

(None), when only the transmitter (TX), only the re-

ceiver (RX), or when both transmitter and receiver

(TX+RX) were heated. The green portion of each bar

shows the ratio (in percent) of packets that have been

received correctly; the black and red portions of each

bar, instead, indicate the percentage of corrupted and

lost packets, respectively. Independent from the qual-

ity of the two links, we note four consistent effects

that LoRa’s PHY settings have on packet reception.

1. Impact of heated device(s). When both trans-

mitter and receivers are at our baseline temperature

(None-labeled bars), the amount of packets received

correctly is the highest. As soon as we increase the

temperature of the transmitter (TX bars), the percent-

ages of corrupted and lost packets increase (black

and red areas). The effect worsens if the receiver

SENSORNETS 2018 - 7th International Conference on Sensor Networks

100

(Setting 16) (Setting 17) (Setting 18)

Lost Rec. corrupted Rec. correctly

100

(Setting 10) (Setting 11) (Setting 12)

100

(Setting 4) (Setting 5) (Setting 6)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

(a) Coding Rate = 4/5

100

(Setting 19) (Setting 20) (Setting 21)

Lost Rec. corrupted Rec. correctly

100

(Setting 13) (Setting 14) (Setting 15)

100

(Setting 7) (Setting 8) (Setting 9)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

(b) Coding Rate = 4/8

Figure 4: Number of packets that are lost, corrupted, and received correctly by node 2 when employing different PHY settings.

For each combination of PHY settings, four bars depict the packet reception when none of the nodes was heated (None), when

only the transmitter (TX), only the receiver (RX), or when both transmitter and receiver (TX+RX) were heated.

100

(Setting 16) (Setting 17) (Setting 18)

Lost Rec. corrupted Rec. correctly

100

(Setting 10) (Setting 11) (Setting 12)

100

(Setting 4) (Setting 5) (Setting 6)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

(a) Coding Rate = 4/5

100

(Setting 19) (Setting 20) (Setting 21)

Lost Rec. corrupted Rec. correctly

100

(Setting 13) (Setting 14) (Setting 15)

100

(Setting 7) (Setting 8) (Setting 9)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

(b) Coding Rate = 4/8

Figure 5: Number of packets that are lost, corrupted, and received correctly by node 1 when employing different PHY settings.

For each combination of PHY settings, four bars depict the packet reception when none of the nodes was heated (None), when

only the transmitter (TX), only the receiver (RX), or when both transmitter and receiver (TX+RX) were heated.

is heated instead of the transmitter (RX bars) and

when both receiver and transmitter are heated simul-

taneously (TX+RX bars). While this effect is coher-

ent with the observations made in previous work on

IEEE 802.15.4 transceivers (Boano et al., 2013), we

cannot explain the reason why heating the receiver

has a higher impact compared to the case in which

only the transmitter is heated. We plan to further an-

alyze this aspect in future work. Nevertheless, inde-

pendent from the selected PHY setting, we note a con-

sistent degradation of the link quality whenever one or

more devices are heated.

2. Bandwidth. In the presence of message loss due to

heat, it is possible to improve the reliability of LoRa

links by lowering the bandwidth (left region of the x-

axis in Figs. 4 and 5). Whilst it is expected that a

lower bandwidth results in a more robust communi-

cation, we are – to the best of our knowledge – the

ﬁrst to show that such robustness extends also to the

effects of temperature.

3. Coding rate. Using an higher coding rate allows

to recover corrupted packets and to increase the suc-

cess ratio. We can see this phenomenon by comparing

Figs. 4(a) and 5(a) against Figs. 4(b) and 5(b). The

former make use of less redundant bits and thus expe-

rience a higher corruption (areas in black). In spite of

that, we notice that only in a few cases, when nodes

experience heterogeneous temperatures, the number

of corrupted packets is high enough to justify the use

of an higher coding rate.

4. Spreading factor. Using a lower spreading fac-

tor drastically reduces the reliability of LoRa links at

high temperatures. A higher spreading factor, indeed,

increases the receiver sensitivity (i.e., the radio’s abil-

ity to receive weaker signals) at the cost of longer

transmission times. Figs. 4 and 5 show that a LoRa

link that cannot sustain a reliable packet delivery rate

when using a spreading factor of 7 (i.e., a spreading

rate of 2

=128).

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation

5.3 RSS and SNR

Figs. 6, 7, and 8 show the impact of temperature on

the received signal strength and signal-to-noise ra-

tio measured at receiving nodes 2 and 3. As for the

previous experiment, for each combination of PHY

settings, four bars depict the packet reception when

none of the nodes was heated (None), when only the

transmitter (TX), only the receiver (RX), or when both

transmitter and receiver (TX+RX) were heated. Inde-

pendent from the receiving node and the quality of

its connection with the transmitter, we note four con-

sistent effects that LoRa’s PHY settings have on the

received signal strength and the SNR.

1. Impact of heated device(s). Consistent to what

we observed for the PRR, the RSS decreases with the

number of heated devices, with the worst effect being

recorded when both transmitter and receiver nodes are

heated (see Fig. 6). This is however not the case when

nodes receive packets with a signal strength that is

very close to the sensitivity threshold (≈ -100 dBm).

This is indeed the case for node 2, which is at the

edge of its communication range (see Fig. 7). Also

the measured signal-to-noise ratio exhibits a trend that

is temperature-dependent, with the lowest SNR being

recorded when both transmitter and receiver nodes are

heated (see Fig. 8).

2. Temperature-independent variations. Whilst we

can clearly notice that the measured RSS and SNR

values vary depending on the temperatures of the

transmitter and receiver, we can observe that changes

in RSS and SNR occur also when varying bandwidth,

spreading factor, and coding rate, independently of

the nodes’ temperature.

Bandwidth. We note that the absolute RSS value,

measured in dBm at the transceiver after the recep-

tion of a packet, changes depending on the employed

LoRa setting. In particular, increasing the bandwidth

results in higher measured RSS. In the case of node 2,

this is because at higher bandwidths LoRa sensitivity

worsens, allowing only the packets with high signal

strength to be received (see Fig. 4 and 7). On the

contrary, at higher bandwidths we observe the SNR

decreasing (see Fig. 8), indicating that LoRa is more

sensitive to noise when higher bandwidths are used.

Spreading factor. A higher spreading factor increases

the receiver sensitivity and hence the reliability of

LoRa communications at high temperatures. The in-

creased sensitivity when using a higher spreading fac-

tor can be conﬁrmed by comparing Figs. 6 and 7

against Fig. 8. We can observe that, whilst there is

no signiﬁcant difference in SNR when using differ-

ent spreading factors, the RSS decreases signiﬁcantly

when using a SF of 9 or 12, hinting that the sensitivity

of the radio increases proportionally.

Coding rate. By comparing Figs. 6(a) and 7(a) against

Figs. 6(b) and 7(b), we can observe that a higher cod-

ing rate results in higher measured RSS. A higher

coding rate also results in a slightly lower SNR (ap-

proximately 0.7 dB lower), regardless of temperature

variations. A more thorough analysis of these phe-

nomena is out of the scope of this paper and will be

carried out in future work.

6 DISCUSSION

The experimental results described in Sects. 4 and 5

show that the effects of temperature on LoRa’s packet

delivery can be quite severe, in line with earlier stud-

ies carried out on IEEE 802.15.4 transceivers (Bannis-

ter et al., 2008), (Boano et al., 2013), (Wennerstr

et al., 2013). We derive next a number of insights

and recommendations for developers and network en-

gineers that may be useful to minimize the impact of

temperature on LoRa communication performance.

Monitor Temperature Variations. Having know-

ledge about the evolution of the on-board tempera-

ture at run-time is important to be able to understand

whether a decrease in RSS is due to heat or to other

environmental effects such as multi-path fading or in-

terference. Furthermore, as our experiments show

that the impact of temperature is platform and node-

speciﬁc, each node needs to derive its own strategy to

mitigate the effects of temperature.

Adapt PHY Settings Accordingly. A thorough un-

derstanding of the evolution of the on-board tem-

perature also allows to predict trends and, if neces-

sary, proactively switch PHY settings to increase the

chances of packet reception. In case temperature in-

creases to a point at which it starts affecting commu-

nication performance, one should, if possible, switch

to a lower bandwidth, a higher spreading factor, as

well as a higher coding rate.

Do not Always Prefer Faster PHY Settings. Ear-

lier studies have shown that selecting the fastest PHY

settings and the highest transmission power – where

possible – is more efﬁcient than selecting slower set-

tings that maximize the link quality (Cattani et al.,

2017a). Our results in Sect. 5, however, highlight that

selecting a higher bandwidth and lower spreading fac-

tor makes the LoRa nodes less robust to temperature

variations. When selecting PHY settings that mini-

mize the air time and the radio’s energy expenditure,

one should hence carefully monitor ﬂuctuations of the

SENSORNETS 2018 - 7th International Conference on Sensor Networks

-86

-84

-82

-80

(Setting 16) (Setting 17) (Setting 18)

RSSI [dBm]

-86

-84

-82

-80

(Setting 10) (Setting 11) (Setting 12)

RSSI [dBm]

-86

-84

-82

-80

(Setting 4) (Setting 5) (Setting 6)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

RSSI [dBm]

(a) Coding Rate = 4/5

-86

-84

-82

-80

(Setting 19) (Setting 20) (Setting 21)

RSSI [dBm]

-86

-84

-82

-80

(Setting 13) (Setting 14) (Setting 15)

RSSI [dBm]

-86

-84

-82

-80

(Setting 7) (Setting 8) (Setting 9)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

RSSI [dBm]

(b) Coding Rate = 4/8

Figure 6: RSS measured on node 3 when using different PHY settings. For each combination of PHY settings, four bars

depict the packet reception when none of the nodes was heated (None), when only the transmitter (TX), only the receiver

(RX), or when both transmitter and receiver (TX+RX) were heated.

-105

-102

-99

-96

(Setting 16) (Setting 17) (Setting 18)

RSSI [dBm]

-105

-102

-99

-96

(Setting 10) (Setting 11) (Setting 12)

RSSI [dBm]

-105

-102

-99

-96

(Setting 4) (Setting 5) (Setting 6)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

RSSI [dBm]

(a) Coding Rate = 4/5

-105

-102

-99

-96

(Setting 19) (Setting 20) (Setting 21)

RSSI [dBm]

-105

-102

-99

-96

(Setting 13) (Setting 14) (Setting 15)

RSSI [dBm]

-105

-102

-99

-96

(Setting 7) (Setting 8) (Setting 9)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX RX TX+RX None TX RX TX+RX None TX RX TX+RX

RSSI [dBm]

(b) Coding Rate = 4/8

Figure 7: RSS measured on node 2 when using different PHY settings. For each combination of PHY settings, four bars

depict the packet reception when none of the nodes was heated (None), when only the transmitter (TX), only the receiver

(RX), or when both transmitter and receiver (TX+RX) were heated.

-24

-18

-12

-6

(Setting 16) (Setting 17) (Setting 18)

SNR [dB]

-24

-18

-12

-6

(Setting 10) (Setting 11) (Setting 12)

SNR [dB]

-24

-18

-12

-6

(Setting 4) (Setting 5) (Setting 6)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX+RX None TX+RX None TX+RX

SNR [dB]

(a) Coding Rate = 4/5

-24

-18

-12

-6

(Setting 19) (Setting 20) (Setting 21)

SNR [dB]

-24

-18

-12

-6

(Setting 13) (Setting 14) (Setting 15)

SNR [dB]

-24

-18

-12

-6

(Setting 7) (Setting 8) (Setting 9)

125 250 500

Bandwidth

Spreading Rate

128 512 4096

Heat: None TX+RX None TX+RX None TX+RX

SNR [dB]

(b) Coding Rate = 4/8

Figure 8: SNR measured on node 2 when using different PHY settings. For each combination of PHY settings, four bars

depict the packet reception when none of the nodes was heated (None), when only the transmitter (TX), only the receiver

(RX), or when both transmitter and receiver (TX+RX) were heated.

on-board temperature and make sure that these do not

affect communication performance. If, despite the se-

lection of a lower bandwidth and a higher spreading

factor (and coding rate), communication performance

is still insufﬁcient, one can only resort to a higher

transmission power, where applicable.

Careful Deployment of Nodes. An important take-

away message from our experimental study is that

LoRa nodes employing the radio transceivers used

in our experiments (i.e., the Semtech SX1272 and

the HopeRF RFM95 transceivers) should be deployed

during the warmest time of the day or year, to en-

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation

sure that network performance is sufﬁcient through-

out the system lifetime despite temperature varia-

tions. Deployments of wireless sensor networks have

indeed observed that on-board temperature ﬂuctua-

tions across different times of the year can be as high

as 85

◦

C (Wennerstr

om et al., 2013), (Beutel et al.,

2009), (Boano, 2014).

Avoid Sun Exposure. When deploying a network,

one should also pay attention to shield nodes from

sunlight as much as possible. Especially the orches-

trator should be shielded from sunlight as, in a star

network, an increase of its temperature would affect

all other nodes (see Fig. 1). Sunshine may indeed

easily heat a packaged sensor node up to 70

◦

C – espe-

cially if the enclosure absorbs infra-red radiation (Po-

lastre et al., 2004), (Szewczyk et al., 2004). Further-

more, airtight and waterproof enclosures may protect

the node from corrosion, humidity and atmospheric

contaminants (Barrenetxea et al., 2008), (Beutel et al.,

2009), but may also increase the temperature of the in-

ner casing (Boano et al., 2009), (Polastre et al., 2004).

7 CONCLUSIONS

Large temperature ﬂuctuations are typical of outdoor

scenarios where most low-power wide-area networks

are deployed. Hence, it is important to understand the

impact of temperature on the performance of long-

range wireless technologies.

This paper presents an experimental evaluation of

the reliability of LoRa in the presence of tempera-

ture variations. First, using a temperature-controlled

testbed we have shown that an increase in tempera-

ture can be sufﬁcient to transform a perfect LoRa link

into an almost useless one. Second, we carried out a

detailed investigation on the performance of different

LoRa’s PHY settings with ﬂuctuating temperatures

and shown that an accurate selection helps in increas-

ing the probability of packet reception and is hence

key to mitigate temperature-induced effects. Finally,

we distilled a list of recommendations to help the suc-

cessful deployment of low-power wide-area networks

in outdoor scenarios.

Future work includes a more detailed investiga-

tion of the problem on a hardware-level, in order to

suggest an improved transceiver layout.

ACKNOWLEDGEMENTS

The authors would like to thank the anonymous re-

viewers for their thorough and valuable feedback.

This work has been supported by the Sino-Austrian

Electronic Technology Innovation Center and was

partially performed within the LEAD-Project “De-

pendable Internet of Things in Adverse Environ-

ments”, funded by Graz University of Technology

(Graz, Austria).

REFERENCES

Anastasi, G., Falchi, A., Passarella, A., Conti, M., and Gre-

gori, E. (2004). Performance Measurements of Motes

Sensor Networks. In Proceedings of the 7

ACM

International Symposium on Modeling, Analysis and

Simulation of Wireless and Mobile Systems (MSWiM),

pages 174–181.

Augustin, A., Yi, J., Clausen, T., and Townsley, W. M.

(2016). A Study of LoRa: Long Range & Low Power

Networks for the Internet of Things. Sensors, 16(9).

Bannister, K., Giorgetti, G., and Gupta, S. K. (2008). Wire-

less Sensor Networking for Hot Applications: Effects

of Temperature on Signal Strength, Data Collection

and Localization. In Proceedings of the 5

Inter-

national Workshop on Embedded Networked Sensors

(HotEmNets).

Barrenetxea, G., Ingelrest, F., Schaefer, G., and Vetterli, M.

(2008). The hitchhiker’s guide to successful wireless

sensor network deployments. In Proceedings of the

ACM International Conference on Embedded Net-

worked Sensor Systems (ACM SenSys), pages 43–56.

Beutel, J., R

omer, K., Ringwald, M., and Woehrle, M.

(2009). Deployment techniques for sensor networks.

In Sensor Networks, Signals and Communication

Technology, pages 219–248. Springer Berlin Heidel-

berg.

Boano, C. A. (2014). Dependable Wireless Sensor Net-

works. PhD thesis, Graz University of Technology,

Graz, Austria.

Boano, C. A., Brown, J., He, Z., Roedig, U., and Voigt, T.

(2009). Low-power radio communication in industrial

outdoor deployments: The impact of weather condi-

tions and ATEX-compliance. In Proceedings of the

International Conference on Sensor Networks Ap-

plications, Experimentation and Logistics (SENSAP-

PEAL), pages 159–176.

Boano, C. A., Brown, J., Tsiftes, N., Roedig, U., and Voigt,

T. (2010). The Impact of Temperature on Outdoor In-

dustrial Sensornet Applications. IEEE Transactions

on Industrial Informatics, 6(3):451–459.

Boano, C. A., R

omer, K., and Tsiftes, N. (2014a). Miti-

gating the Adverse Effects of Temperature on Low-

Power Wireless Protocols. In Proceedings of the 11

IEEE International Conference on Mobile Ad hoc and

Sensor Systems (MASS), pages 336–344.

Boano, C. A., Wennerstr

om, H., Z

niga, M. A., Brown,

J., Keppitiyagama, C., Oppermann, F. J., Roedig, U.,

Nord

en, L.-

A., Voigt, T., and R

omer, K. (2013). Hot

Packets: A Systematic Evaluation of the Effect of

Temperature on Low Power Wireless Transceivers. In

SENSORNETS 2018 - 7th International Conference on Sensor Networks

Proceedings of the 5

Extreme Conference on Com-

munication (ExtremeCom), pages 7–12.

Boano, C. A., Z

niga, M. A., Brown, J., Roedig, U., Kep-

pitiyagama, C., and R

omer, K. (2014b). TempLab: A

Testbed Infrastructure to Study the Impact of Temper-

ature on Wireless Sensor Networks. In Proceedings

of the 13

ACM/IEEE International Conference on

Information Processing in Sensor Networks (IPSN),

pages 95–106.

Bor, M. and Roedig, U. (2017). LoRa Transmission Param-

eter Selection. In Proceedings of the 13

IEEE In-

ternational Conference on Distributed Computing in

Sensor Systems (DCOSS).

Bor, M., Roedig, U., Voigt, T., and Alonso, J. M. (2016a).

Do LoRa Low-Power Wide-Area Networks Scale? In

Proceedings of the 19

ACM International Confer-

ence on Modeling, Analysis and Simulation of Wire-

less and Mobile Systems (MSWiM), pages 59–67.

Bor, M., Vidler, J., and Roedig, U. (2016b). LoRa for

the Internet of Things. In Proceedings of the 1

In-

ternational Workshop on New Wireless Communica-

tion Paradigms for the Internet of Things (MadCom),

pages 361–366.

Cattani, M., Boano, C. A., and R

omer, K. (2017a). An ex-

perimental evaluation of the reliability of LoRa long-

range low-power wireless communication. Journal of

Sensor and Actuator Networks (JSAN), 6(2).

Cattani, M., Boano, C. A., Steffelbauer, D., Kaltenbacher,

S., G

unther, M., R

omer, K., Fuchs-Hanusch, D., and

Horn, M. (2017b). Adige: An Efﬁcient Smart Wa-

ter Network based on Long-Range Wireless Technol-

ogy. In Proceedings of the 3

International Workshop

on Cyber-Physical Systems for Smart Water Networks

(CySWATER).

Centenaro, M., Vangelista, L., Zanella, A., and Zorzi, M.

(2016). Long-Range Communications in Unlicensed

Bands: the Rising Stars in the IoT and Smart City Sce-

narios. IEEE Wireless Communications, 23(5):60–67.

Georgiou, O. and Raza, U. (2017). Low Power Wide Area

Network Analysis: Can LoRa Scale? IEEE Wireless

Communications Letters, 6(2):162–165.

Hope RF Microelectronics (2017). RFM95/96/97/98(W) –

Low Power Long Range Transceiver Module, v1.0.

Iova, O., Murphy, A. L., Ghiro, L., Molteni, D., Ossi, F., and

Cagnacci, F. (2017). LoRa from the City to the Moun-

tains: Exploration of Hardware and Environmental

Factors. In Proceedings of the 2

International Work-

shop on New Wireless Communication Paradigms for

the Internet of Things (MadCom).

Kartakis, S., Choudhary, B. D., Gluhak, A. D., Lambrinos,

L., and McCann, J. A. (2016). Demystifying Low-

Power Wide-Area Communications for City IoT Ap-

plications. In Proceedings of the 10

ACM Workshop

on Wireless Network Testbeds, Experimental Evalua-

tion, and Characterization (WiNTECH), pages 2–8.

Keppitiyagama, C., Tsiftes, N., Boano, C. A., and Voigt, T.

(2013). Poster abstract: Temperature hints for sensor-

net routing. In Proceedings of the 11

ACM Confer-

ence on Embedded Networked Sensor Systems (ACM

SenSys), poster session.

KSK Developments (2017). KSK Outdoor Parking

Guidance System. http://www.ksk-dev.com/wp-

content/uploads/2017/06/KSK-Magnetic-ENG.pdf.

LoRa Alliance (2017). LoRa: Wide Area Net-

works for IoT. http://www.lora-alliance.org/What-Is-

LoRa/Technology.

LowPowerLab (2016). Moteino MEGA LoRa.

http://lowpowerlab.com/shop/product/119.

Marcelis, P., Rao, V., and Prasad, R. V. (2017). DaRe:

Data Recovery through Application Layer Coding for

LoRaWAN. In Proceedings of the 2

International

Conference on Internet-of-Things Design and Imple-

mentation (IoTDI), pages 97–108.

Marﬁevici, R., Murphy, A. L., Picco, G. P., Ossi, F., and

Cagnacci, F. (2013). How Environmental Factors

Impact Outdoor Wireless Sensor Networks: A Case

Study. In Proceedings of the 10

IEEE International

Conference on Mobile Ad-Hoc and Sensor Systems

(MASS), pages 565–573.

Oppermann, F. J., Boano, C. A., Z

niga, M. A., and R

omer,

K. (2015). Automatic protocol conﬁguration for de-

pendable internet of things applications. In Proceed-

ings of the 10

International Workshop on Practi-

cal Issues in Building Sensor Network Applications

(SenseApp), pages 742–750.

Pet

arvi, J., Mikhaylov, K., Roivainen, A., H

anninen, T.,

and Pettissalo, M. (2015). On the Coverage of LP-

WANs: Range Evaluation and Channel Attenuation

Model for LoRa Technology. In Proceedings of the

IEEE International Conference on ITS Telecom-

munications (ITST), pages 55–59.

Polastre, J., Szewczyk, R., Mainwaring, A., Culler, D., and

Anderson, J. (2004). Analysis of wireless sensor net-

works for habitat monitoring. In Wireless Sensor Net-

works, pages 399–423. Kluwer Academic Publishers.

Semtech Corporation (2013). SX1272/3/6/7/8: LoRa Mo-

dem, Designer’s Guide – Application Note 1200.13,

Revision 1.

Semtech Corporation (2015). LoRa Modulation Basics –

Application Note 1200.22, Revision 2.

Semtech Corporation (2017). SX1272/73 – 860 MHz to

1020 MHz Low-Power Long-Range Transceiver, Re-

vision 3.1.

Sigfox (2017). Sigfox Technology. http://www.sigfox.com.

ST Microelectronics (2016). STM32 Nucleo Pack for LoRa

Technology (P-NUCLEO-LRWAN1), DocID029505

Rev. 2.

Stoianov, I., Nachman, L., Madden, S., and Tokmouline,

T. (2007). PIPENET: a wireless sensor network for

pipeline monitoring. In Proceedings of the 6

In-

ternational Conference on Information Processing in

Sensor Networks (IPSN), pages 264–273.

Szewczyk, R., Polastre, J., Mainwaring, A., and Culler, D.

(2004). Lessons from a sensor network expedition

wireless sensor networks. Wireless Sensor Networks,

2920:307–322.

Voigt, T., Bor, M., Roedig, U., and Alonso, J. (2017). Mit-

igating Inter-network Interference in LoRa Networks.

In Proceedings of the 2

International Workshop on

Impact of Temperature Variations on the Reliability of LoRa - An Experimental Evaluation

New Wireless Communication Paradigms for the In-

ternet of Things (MadCom).

Weightless SIG (2017). Weightless open standard.

http://www.weightless.org/.

Wennerstr

om, H., Hermans, F., Rensfelt, O., Rohner, C.,

and Nord

en, L.-A. (2013). A Long-Term Study of

Correlations between Meteorological Conditions and

802.15.4 Link Performance. In Proceedings of the

International Conference on Sensing, Communi-

cation, and Networking (SECON), pages 221–229.

niga, M. A. and Krishnamachari, B. (2004). Analyzing

the transitional region in low-power wireless links. In

Proceedings of the 1

IEEE International Conference

on Sensor, Mesh and Ad Hoc Communications and

Networks (SECON), pages 517–526.

SENSORNETS 2018 - 7th International Conference on Sensor Networks