The use of wireless sensor networks to measure soil parameters eliminates the need to remove sensors for field operations, such as tillage, thus allowing long-term measurements without multiple disturbances to soil structure. Wireless sensors also reduce above-ground cables and the risk of undesired equipment damage and potential data loss. However, implementing wireless sensor networks in field studies usually requires advanced and costly engineering knowledge. This study presents a new underground, wireless, open-source, low-cost system for monitoring soil oxygen, temperature, and soil moisture. The process of system design, assembly, programming, deployment, and power management is presented. The system can be left underground for several years without the need to change the battery. Emphasis was given on modularity so that it can be easily duplicated or changed if needed and deployed without previous engineering knowledge. Data from this type of system have a wide range of applications, including precision agriculture and high-resolution modelling.
A remaining challenge in vadose zone monitoring is the measurement of soil parameters, such as water content, without the need to remove sensors between field operations, such as tillage, which often causes damage to wires connecting below-ground sensors to above-ground data loggers. In addition, the standard setting of using cables to connect underground sensors to an above-ground data logger and power source can change soil structure by causing macro-pores and fractures. The altered structure can potentially cause unwanted experimental artifacts, such as preferential water flow or higher aeration rates. Also, above-ground cables can be subject to undesired damage resulting from weather events, agricultural machinery, pests, and animals (Hardie and Hoyle, 2019; Vuran et al., 2018). The use of wireless underground sensor networks (WUSNs), instead of cables, can solve these issues (Cardell-Oliver et al., 2019; Zhang et al., 2017).
While above-ground wireless sensor networks (WSNs) have been readily studied and implemented (DeBell et al., 2019; García et al., 2021), the use of WUSNs is still in its early stages (Hardie and Hoyle, 2019; Wan et al., 2017). A WUSN is defined as a system in which all sensors and communication components are buried underground (node) while wirelessly transmitting data through the soil to an above-ground hub (also referred to in the literature as a gateway) (Huang et al., 2020). The definition of an underground node is study-dependent: here a node is defined as a single transmitting system located at a single depth; different sensors can be connected to one node via underground cables. The most basic WUSN configuration includes a single underground node and an above-ground hub. Advanced WUSNs can consist of several underground nodes connected to a single above-ground hub (Liedmann and Wietfeld, 2017; Tiusanen, 2013) or a wide-area network combining multiple underground nodes with multiple above-ground hubs (Froiz-Míguez et al., 2020).
Ritsema et al. (2009) were the first to
monitor soil moisture in a golf course over several square kilometres using a WUSN consisting of multiple locations. They used a complex array of 18
underground nodes, each installed at 0.1 m below the soil surface, and a single above-ground hub. To bridge the distance between underground nodes and the single above-ground hub, they installed 24 additional above-ground nodes.
Private engineering companies provided the network architecture (hardware
and software), and no further information was provided to allow
reproducibility. More recently, Zhang et al. (2017)
presented the Thoreau network as the first long-term WUSN for agricultural
and environmental sensing. The network consists of a single hub installed at
41 m above ground, which receives data from 27 underground nodes. Each node was buried at
The extensive development of Internet of Things (IoT) hardware and
IoT-related wireless communication protocols provides new opportunities for
implementing communication solutions for WUSNs (Salam
and Raza, 2020). García et al. (2020) and
Vuran et al. (2018) present a comprehensive review of
the most utilized wireless communication protocols, including Bluetooth,
cellular, Wi-Fi, Sigfox, and ZigBee. Out of these, one of the most suitable
for WUSNs is the low-power long-range (LoRa) network protocol (referred to here as LoRa-WUSN). LoRa is a relatively new, open technology designed for
small data rates up to 50 kbps
(Abrardo and Pozzebon, 2019;
García et al., 2020) over above-ground distances of 10
Radio signal attenuation is a primary consideration for any type of WUSN, with the total attenuation strongly dependent on the length of the signal path in the soil (Bogena et al., 2009). According to Tiusanen (2013), there are four components affecting the signal attenuation between an underground node and an above-ground hub: signal loss due to soil medium attenuation, due to partial reflection from the soil surface, due to angular defocusing, and free air path loss. The first two components are associated with the soil medium and the latter two with the air above. Signal quality, defined by the received signal strength index (RSSI; expressed using negative dBm units, closer to zero means greater signal strength) and signal-to-noise ratio (SNR; positive values indicate more signal than noise), primarily depend on the transmitter operating frequency, burial depth, transmitter power, distance between the underground node and the above-ground hub, antenna type, data rate, soil moisture, and to a lesser extent soil texture and electrical conductivity (Bogena et al., 2009; Hardie and Hoyle, 2019). The limit of acceptable RSSI for wireless communication is subjective and determined based on experimental needs and the RSSI-to-SNR ratio (Hardie and Hoyle, 2019).
Alongside the advances made in IoT-related wireless communication, open-source hardware is an additional emerging field of interest in environmental research (Concialdi et al., 2020; Fisher and Gould, 2012; Froiz-Míguez et al., 2020). Open-source hardware consists of electronics that can be freely replicated or assembled using openly available instructions, such as schematics, drawings, and layouts (Chan et al., 2020). Arduino microcontrollers and Raspberry Pi microcomputers, with their software platforms, are perhaps the most common examples of open-source hardware. The widespread adoption of open-source hardware is led by hobbyists and the public and to a lesser extent by the academic community (e.g. the OPEnS Lab, Oregon State University). However, the available online information of tutorials, forums, and ongoing developments minimizes the learning curve and can help bridge the gap toward a higher implementation rate of open-source hardware in academic research (e.g. Levintal et al., 2020, 2021b; Reck et al., 2019; Weissman et al., 2020). Additional benefits of utilizing open-source hardware are no prior experience with electronics or coding being needed, though it can help (Chan et al., 2020), lower costs than existing commercial hardware (Levintal et al., 2021a), and the option for customized solutions. Specifically, harnessing open-source hardware for LoRa-WUSN lowers such sensor networks' cost and complexity, thus making them accessible for researchers.
The use of LoRa-WUSN in soil studies has not been comprehensively investigated (Hardie and Hoyle, 2019). The majority of studies on LoRa-WUSN can be found in engineering disciplines and focus on in-soil signal propagation (Wan et al., 2017) and antenna optimization (Salam and Raza, 2020), making it challenging to adapt their conclusions to other disciplines, such as hydrology and soil science. Moreover, there is a lack of studies showing the performance of LoRa-WUSN for long-term measurements (Cardell-Oliver et al., 2019). Most studies previously published on LoRa-WUSN are either proof-of-concept studies or short-lived laboratory experiments (e.g. Huang et al., 2020; Wan et al., 2017). In addition, to the best of our knowledge, there is only one study published to date that tested and presented results from LoRa-WUSN for soil measurements (Cardell-Oliver et al., 2019). None of the aforementioned studies, however, focused on providing step-by-step instructions for the design, assembly, and installation of WUSN by the end-user. Despite the rapid technological advancement of WUSNs, it seems the technology itself (assembly, programming) remains a major challenge to utilizing WUSNs more widely in environmental and academic research. By sharing detailed instructions on the design, assembly, and installation of WUSNs, we think these systems can be more widely used by other scientists and adapted to individual research needs.
The aim of this study is to present a new open-source, low-cost, LoRa-WUSN
system for measuring soil moisture and oxygen levels at multiple depths in
an agricultural soil and to provide in detail the technical information for
the system design, assembly, programming, deployment, power management, and
data analysis so that other researchers can adapt the system to their needs.
Emphasis is placed on modularity to allow the end-user to duplicate or change, if needed, and deploy without previous engineering knowledge.
Therefore, hardware was limited to readily available components only. Eight
sensors consisting of four digital soil moisture sensors and four analogue oxygen sensors were connected to an underground open-source LoRa transmitter
node, and an above-ground LoRa hub logged the received data. For validation, the system was deployed in a field planted with young poplar trees (
Hardware was limited to readily purchasable products in order to assemble a
low-cost LoRa-WUSN that can be easily replicated. The LoRa-WUSN installed in
the field included two segments: a single underground node (to which the
sensors were connected via underground cables) and a single above-ground hub combined with a data logger (Fig. 1). Table 1 summarizes the details of each sub-segment component. The core of the LoRa-WUSN is the
Adafruit Feather M0 with RFM95 LoRa Radio (Adafruit, USA), hereafter called
LoRa-Feather, which utilized a non-licensed 900 MHz radio band (or 433 MHz in Europe). This is an open-source microcontroller with an embedded LoRa
radio module, which is light and affordable (USD
The underground node (Fig. 1c) included a LoRa-Feather connected to an external omnidirectional antenna (900 Mhz Antenna Kit, Adafruit, USA) and a battery (lithium-ion cylindrical battery – 3.7 V, 2200 mAh, Adafruit, USA). A power relay extension board was mounted on the LoRa-Feather to optimize the sensors' power consumption (Adafruit Non-Latching Mini Relay FeatherWing, Adafruit, USA). The relay provides power to four digital soil moisture sensors (5TM, METER Group, USA) and to an analogue-to-digital converter (ADC) (ADS1015 12 bit ADC, Adafruit, USA). The ADC was used to convert the data from four analogue oxygen sensors (KE-25 Figaro Engineering Inc., Japan). We choose the Decagon 5TM combined soil moisture and temperature sensor and the Figaro KE-25 oxygen sensor due to their low power requirements (5TM – 0.03 and 3 mA during sleep mode and measurement, respectively; KE-25 – no external power supply required for sensor operation), low cost, and long-term use in soil monitoring (Oroza et al., 2018; Turcu et al., 2005; Weitzman and Kaye, 2017). All components, excluding the antenna and sensors, were placed in a waterproof enclosure and sealed using rubber coating (Performix 12213, Plasti Dip International, USA) to protect against expected soil water (Fig. 1c).
Underground nodes for WUSNs need to be highly energy-efficient because the battery cannot be recharged without excavation
(Hardie and Hoyle, 2019). To reduce the
underground node's power consumption, which in our case measured and
transmitted sensor data every 1 or 2 h, depending on the scenario tested, we used two independent methods simultaneously. The first method was putting
the LoRa-Feather into a low-power consumption sleep mode. During the sleep mode, tested power requirements were reduced from
The above-ground hub (Fig. 1b) includes a LoRa-Feather connected to an external omnidirectional antenna (900 Mhz Antenna Kit, Adafruit, USA) and
a battery (lithium-ion polymer battery – 3.7 V, 1200 mAh, Adafruit, USA). Received data were logged on an SD card using an extension board mounted on
the LoRa-Feather (Adalogger FeatherWing – RTC
The total cost of the system, apart from the sensors and solar panel, was
USD 150 (Table 1). The sensors' cost amounted to USD
Detailed assembly instructions for the underground node, sensors, and above-ground hub are provided in the Supplement (Sect. S1).
Scheme of the LoRa-WUSN experimental set-up in the field
Summary of hardware components and materials used in this study.
The LoRa-Feather microcontrollers were programmed using C
Every measurement cycle conducted by the underground node included 5 s of
sensor readings followed by the transmission of four data packets. Splitting
the data into four packets was necessary because each packet is constrained
to 20 chars (a char is a data type used in C
The above-ground hub stays constantly in the receiver mode. At the end of a receiving cycle, once the four data packets from the underground node are received, the above-ground hub will send a reply to the underground node with instructions on the new measurement interval or new transmission power before data packets are written onto the SD card, together with the RSSI, SNR, and a timestamp. If no reply was sent from the above-ground hub to the underground node, then the node will use its current measurement interval and transmission power for the next cycle.
Flow chart for underground node
A field experiment was conducted to validate the proposed LoRa-WUSN
(Fig. 1a). The system was installed in a poplar orchard irrigated
with surface drip located near Davis, CA, USA. The soil type at the site is a reiff very fine sandy loam (SoilWeb, 2021), and the climate is Mediterranean with a total annual precipitation of about
500 mm and a mean annual temperature of 16.9
To test the performance of the LoRa-WUSN, three experimental scenarios with
different sleep modes and signal strengths during data transmission were
tested in sequence between 11 November 2020 and 31 March 2021. The three
scenarios include (1) 2 h measurement intervals and low transmission power of 5 dBm (2 h, low transmission power) (10 November 2020–7 January 2021 and
27 February–31 March 2021), (2) 1 h measurement intervals and low transmission
power of 5 dBm (1 h, low transmission power) (8–28 January 2021), and (3) 2 h measurement intervals and high transmission power of 20 dBm (2 h,
high transmission power) (29 January–25 February 2021). In addition to these three
scenarios, we also tested a 1 min measurement interval and high transmission power of 20 dBm (26 February 2021) to assess the effect of the distance between the
above-ground hub and the underground node on the wireless communication signal strength. In this scenario, the above-ground hub was positioned at
different distances from the underground node ranging from 1, 10, 20, 30, to
50 m. At each distance, the above-ground hub measured for 10 min (1 min intervals) at a constant height of 2 m above ground with the same antenna orientation. RSSI and SNR values from each location were used to assess
communication strength. During the scenarios, the default LoRa-Feather
parameters were used (bandwidth
Because the focus of this study is on the design and performance of an open-source LoRa-WUSN for measuring soil parameters, our results and discussion will mainly concentrate on the LoRa-WUSN's capabilities, such as battery and wireless communication performance, and not on the interpretation of the actual soil data collected in the field. The field data shown in Fig. 3 are mainly used to highlight and validate performance aspects of the LoRa-WUSN and to stimulate possible future applications.
Time series data from the field experiment. Orange
asterisks in plot
Soil moisture below the drip emitter (measured with the LoRa-WUSN) increased
following each precipitation event and then after the irrigation was restarted on 22 February 2021 (Fig. 3b, color lines). Soil moisture
between tree rows increased only after major precipitation events in
mid-December and at the end of January (Fig. 3b, black line).
Oxygen concentrations in the soil were approximately 2 %–5 % lower than
atmospheric concentrations with higher concentrations observed in the dry
area between the tree rows than below the drip emitter (Fig. 3c). A
general decrease trend in soil oxygen was observed during periods when soil
moisture increased. All four oxygen sensors transmitted very low voltages
(1–2 mV) corresponding to 0 % soil oxygen content after several weeks of
operation. This was likely due to clogging of the sensors' membranes, yet it
was unexpected as we used a common sensor type
(Kallestad et
al., 2008; Turcu et al., 2005). The problem was solved by embedding the
sensor in a customized enclosure that contained an additional hydrophobic
membrane (polytetrafluoroethylene – PTFE – type). The added membrane blocked the water while still allowing gas exchange with the surrounding soil (see the enclosure design in the Supplement, Fig. S5). We note that only the oxygen sensor at 0.15 m below the drip emitter was replaced with a new oxygen
sensor and enclosure (10 January 2021) due to limited sensor availability; the
clogged sensors were not reused due to the uncertainty regarding their
performance and accuracy after clogging. Soil temperatures measured at the
shallow depths (0.15 and 0.30 m) showed a typical diel pattern as well as a
seasonal trend of decreasing temperatures until the end of January
(Fig. 3d). Temperatures were stable during February, followed by a
sharp increase of
The underground node's battery voltage was 4 V at the start of the field
experiment and decreased to 3.77 V after 5 months of continuous operation (Fig. 3e). The battery decline rate was linear but varied depending
on the measurement interval and transmission power used. Figure 4
presents the battery decrease rates for the three main tested scenarios:
2 h between measurements with 5 dBm transmission power, 1 h with 5 dBm
transmission, and 2 h with 20 dBm transmission. Unexpectedly, the fastest
voltage decrease rate was during the 2 h intervals with 5 dBm transmission
(
Decrease rate of the underground node battery for three operational scenarios. The slopes were calculated using linear regressions (solid lines) and represent the average voltage decrease rate for each scenario. Temperature values represent the average soil temperature surrounding the node at 0.3 m below the soil surface during that specific scenario.
Fitting a linear regression to the battery voltage decrease rate allows
estimation of a total lifetime of the underground node's battery, assuming a 0.5 V range (i.e. from an initial 4.1 V charged battery to 3.6 V). The average battery decrease rate over the entire experiment was
As power consumption is critical for WUSNs, we measured the duration and current of one complete operation cycle for four different transmission levels (5, 12, 17, and 20 dBm). The test was conducted in the lab using the same setting as in the field with the addition of the INA260 (Adafruit, USA) to measure the current at 100 ms intervals. The results (Table 2 and Figs. S10 and S11 in the Supplement) indicate that the main parameter affecting power consumption is the duration of the sensor measurements (63 mA for 5 s), with a smaller contribution from the transmission (e.g. 129 mA for 0.3 s during 20 dBm transmission). We note that the power consumption during sleep mode was below the detection limit, and therefore we used the value provided by the manufacturer of 0.035 mA. The end-user can use the power consumption values presented in Table 2 to optimize system performance according to specific needs and batteries.
Power consumption.
Two power management methods were used in the underground node, a power
relay for the sensors (hardware type) and a sleep mode between measurement
and transmission cycles (software type). If longer battery life is needed,
there are additional power conservation methods available, such as
cancelling a data transmission if measured values have not changed above a
defined threshold compared to the previous measurement
(Tiusanen, 2013) or reducing the number of data packets being sent by implementing an algorithm that reduces overall data size
(Cardell-Oliver et al., 2019). Nevertheless, battery prices
are relatively low, and therefore the best solution to extend battery life without complicating the system is to purchase (if needed) a battery with a
larger capacity. In cases where extended battery life is needed, it is
recommended to use battery technologies with lower self-discharge rates,
such as non-rechargeable lithium-thionyl batteries with self-discharge rates
lower than 1 % per year. A comparison between different battery
technologies is detailed in Callebaut et
al. (2021). For instance, using a non-rechargeable lithium-thionyl battery
with
Average RSSI and SNR throughout the experiment were
The effective communication range between the underground node and
above-ground hub was tested for 2 h on 26 February 2021. This relatively short period was chosen to ensure similar ambient conditions throughout the test
(similar relative humidity, temperature, soil moisture, etc.). The
underground node was set to transmit at 1 min intervals and 20 dBm via a
command from the above-ground hub. RSSI and SNR decreased with increasing distance between the above-ground hub and the underground node (Fig. 5). At the maximum distance tested (50 m), data packages from the
underground node were still received and logged with an RSSI and SNR of
RSSI and signal-to-noise ratio (SNR) as a function of the linear distance between the underground node and the above-ground hub. The above-ground hub was positioned at five distances from the underground node (1, 10, 20, 30, and 50 m). At each distance, the hub recorded one reading per minute for 10 min at a constant height of 2 m above ground with the same antenna orientation. Transmission of the underground node was set to 20 dBm. Error bars indicate the standard deviation from the average of 10 measurements at each distance.
The chosen transmission power and radio band should also follow legal restrictions derived from local regulation. In Europe, for instance, the maximum approved transmission power is 14 dBm (for 433 MHz) compared to 30 dBm in the USA (915 MHz) (Fraga-Lamas et al., 2019; Froiz-Míguez et al., 2020; Haxhibeqiri et al., 2018). The results of our study show that even 5 dBm provided sufficient power for transmitting data from the underground node to the above-ground hub located at a horizontal distance of 2 m (Fig. 3). The relationship between transmission power, underground node depth, distance between an underground node and an above-ground hub, and soil texture is discussed in Hardie and Hoyle (2019). We note that the authors used a radio band of 433 MHz compared to 915 MHz used in this study, and therefore some differences are expected; lower radio band frequency will result in lower radio propagation losses (i.e. a larger range) (Froiz-Míguez et al., 2020). To the best of our knowledge, there is no published comparison between the two radio bands for LoRa-WUSN, and thus we cannot conclude which of the two is preferable (e.g. in terms of RSSI or SNR). Another regulative limitation is the duty cycle for an on-air time. In Europe, it is 1 %, which means that for a 1 s LoRa transmission, this specific node cannot transmit during the following 99 s (Haxhibeqiri et al., 2018). The 0.3 s transmission duration presented in this study (Table 2) translates to a minimum interval time of 29.7 s before the subsequent transmission can be made.
Apart from RSSI and SNR, another critical parameter is the data packet's receiving ratio, defined as the number of received data packets at the above-ground hub divided by the packets that were sent from the underground node; a ratio of 100 % represents ideal conditions under which all sent packets were also received at the first attempt. The average ratio during the experiment was 75 %, with higher ratio values observed at the start of
the winter (
In this study, we present a LoRa-WUSN that was built to measure soil moisture, soil temperature, and soil oxygen content. Nevertheless, the modular nature of open-source hardware allows the end-user various options for system configurations without adding substantial complexity. For example, instead of on-site data logging as conducted here, it is possible to add a Wi-Fi component to the above-ground hub to get online real-time data (DeBell et al., 2019). If no Wi-Fi is in range, a cellular modem can replace the Wi-Fi component (Spinelli and Gottesman, 2019), which will be more costly due to the cellular service charges, but it would provide greater flexibility and range. An alternative solution is sending the data from the above-ground hub at the experimental site to another above-ground LoRa station situated several kilometres from the site in an area with a Wi-Fi signal or Ethernet connectivity. Combined solutions, such as on-site data logging and Wi-Fi, Ethernet, or cellular communication, are also possible (Levintal et al., 2021a).
Embedding a feedback mechanism within the code is possible if a packet receiving ratio of 100 % is desired (i.e. all the sent data are also logged on the above-ground hub). The underground node continuously sends the same data packet until a reply from the above-ground hub is received stating that the packet was logged. The tradeoff of this modification is the increase in power consumption of the underground node due to the potentially greater number of transmission cycles. Power consumption can be managed within the code by implementing a predefined threshold voltage below which the feedback mechanism will be disabled.
Installing multiple underground nodes at different locations is also feasible. This requires a simple software modification, in which every data packet (i.e. every singular transmission) is labelled at the start of the packet with another identifier specifying the underground node that sent the packet, and accordingly the above-ground hub knows from which node the packet was received. A similar method was presented by DeBell et al. (2019) for above-ground LoRa networks. We tested and validated this method in the lab using three nodes and a single hub. Using this approach simplifies system assembly for the end-user; however, it increases the risk of data packet loss in the cases of two nodes transmitting simultaneously. To quantify this risk, we conducted a test in which three nodes transmitted data packets at 1 min intervals for 20 h (i.e. 20 data packets per node). Data packet receiving ratios were 100 %, 95 %, and 100 % for nodes 1, 2, and 3, respectively. These ratios indicate a low probability of transmission collisions between nodes, yet if a significantly larger number of nodes is required, it is recommended to use more complex solutions like the LoRa wide area networking technology (LoRaWAN). LoRaWAN is an open-source protocol that uses the LoRa protocol to enable communication between multiple nodes and hubs (also referred to as gateways), with additional benefits such as adaptive data rates that can reduce power consumption (Froiz-Míguez et al., 2020; Haxhibeqiri et al., 2018). There is also an emerging use of LoRaWAN solutions commercialized by private companies, yet they are still costly and, in most cases, target big end-users, such as cities, and therefore are less relevant for field-scale research. A review of the LoRaWAN technology is provided by Haxhibeqiri et al. (2018), and a more detailed focus on the limitations is provided by Adelantado et al. (2017).
This study presents a novel, low-cost wireless underground sensor network
(WUSN) for soil monitoring using the relatively new, open-communication protocol named low-power long-range (LoRa). A field test, conducted for 5 months in an agricultural field, allowed assessment of the system's capabilities. Soil moisture content, temperature, and soil oxygen
concentrations were measured at three depths (0.15, 0.3, and 0.5 m), and data were transmitted from an underground node (0.3 m) to an above-ground receiving and logging hub. Communication tests showed an effective range of
at least 50 m is possible between the underground node and above-ground hub. Using power management methods, battery life was estimated at
The complete technical guide is available in the Supplement file below and in the GitHub repository (
The supplement related to this article is available online at:
EL conceptualized and conducted the study and wrote the first manuscript draft. HED provided the resources. All the authors (EL, YG, GT, PFS, KS, and HED) contributed to the final version.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher’s note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors would like to thank Cristina Prieto Garcia for her help with the field and laboratory experiments and the two anonymous reviewers who helped improve this paper.
This research has been supported by the Gordon and Betty Moore Foundation (grant no. 7975), the United States–Israel Binational Agricultural Research and Development Fund (Vaadia-BARD Postdoctoral Fellowship no. FI-605-2020), the National Institute of Food and Agriculture (grant no. CA-D-LAW-2513-H), and the U.S. Department of Energy, Office of Environmental Management (grant no. DE-SC0020164).
This paper was edited by Jan Vanderborght and reviewed by two anonymous referees.