WSN Application for Sustainable WaterManagement in Irrigation Systems

Andre Gloria(1,2), Carolina Dionisio(1,2), Goncalo Simoes(1,2), Pedro Sebastiao(1,2), Nuno Souto(1,2)(1) ISCTE-IUL, Av. das Forcas Armadas, Lisbon, Portugal

(2) Instituto de Telecomunicacoes, Av. Rovisco Pais, 1, Lisbon, PortugalEmail: afxga@iscte-iul.pt, cadoo@iscte-iul.pt, garss@iscte-iul.pt, pedro.sebastiao@iscte-iul.pt, nuno.souto@iscte-iul.pt

Abstract—This paper introduces a new way of maintaining asustainable irrigation system, applied to gardens or agriculturalfields, replacing human intervention with Wireless Sensor Net-works, in order to achieve a better efficiency in the process thatcan lead to savings for the final user, not only monetary butalso in natural resources, such as water and energy, leading to amore sustainable environment. The system can retrieve real timedata and use them to determinate the correct amount of waterto be used in the garden in order to keep it healthy. Besidesthe system architecture, this paper includes a real case scenarioimplementation and result discussions.

Index Terms—Internet of Things, Wireless Sensor Networks,Sustainability, LoRa, ESP32, Water Savings

I. INTRODUCTION

Internet of Things (IoT) research and implementation hasbeen growing in the last years, focusing in topics such asTransportation, Smart Cities, Retail, Agriculture, Smart Fac-tories, Healthcare, Culture and Tourism [1]. Connecting realworld elements and adding intelligence and communicationsfor smart process and autonomous decisions, IoT is enablingdifferent types of beneficial applications and services that cansustain our day-by-day in ways that we never expected before[2]. For this, IoT monitors and automate through the useof sensors and actuators networks and intelligent processingunits, that share informations between devices and allows themto work together to improve user experience.

In a world where natural resources, such as water, arestarting to vanish and material goods prices, like energy, arein all times highs, the need to create more efficient processesis a must in order to improve sustainability. To achieve this,innovations in a large scale are required and IoT can takea major role. Combining IoT, sustainability and MachineLearning (ML) algorithms is possible to achieve a disruptiveinnovation capable of saving energy, cutting costs, fewer fuels,waste reduction and time saving.

All around the world, water shortage is becoming a biggerissue and with a big amount of fresh water being usedin agriculture or related activities, the need to control theamount of water in those activities is essential. An irrigationsystem wastes on average 30% of water due to environmental

This work was supported in part by ISCTE - Instituto Universitrio de Lisboafrom Portugal under the project ISCTE-IUL-ISTA-BM-2018

conditions [3] or from the cycles not being optimized for thetypes of plants.

Using a Wireless Sensor Network (WSN) in the fields, thatanalyze temperature, humidity and soil moisture, combinedwith the location, weather forecast and type of plants, thesystem will be able to automatically and dynamically predictthe real amount of water to be applied in order to keep thefields healthy and obtain the best automatized and sustainable(economic, social and environment) decision.

In this paper, a system based on WSN for retrieving realtime environmental data in order to optimize irrigation timingsis explained. The test case, followed by its experimental resultsand comparison with a traditional irrigation system, are alsorepresented. Finally, the main conclusions and future work arealso presented.

II. RELATED WORK

The literature shows a variety of projects [4]–[7] imple-menting IoT systems for irrigation control, using single nodeswith Arduinos, combined with a set of environmental datasensors, using large amounts of wire to spread them acrossthe fields, in which data is used to increase the garden oragricultural fields efficiency, including user comfort and watersavings. Also, some full WSN projects were found, such asMaksudjon Usmonov, who developed a low-cost WSN basedon LoRa [8], capable of controlling an existing drip irrigationsystem with the aim of improving it in order to automate it bytotally or partially eliminating human interaction, or Dr. M. N.Rajkumar et al. who develop a smart irrigation system [9] thatautomatically turn on or off the valves based on soil moisturesensor data. Our proposal stands out from the others by the useof algorithms capable of analyzing the environmental data andunderstand the correct amount of water needed for the field,instead of watering only when moisture is low. Also a morereliable WSN, based on modular nodes using LoRa, allows theability to perform in any conditions or specifications withoutthe need of wires, spreading multiple sensor nodes within theenvironment.

III. SYSTEM ARCHITECTURE

The system architecture consists on a WSN composedby a set of nodes with different features. The network iscontrolled by a single gateway, the broker, that is in constantcommunication with an online server, using Message Queuing978-1-5386-4980-0/19/$31.00 © 2019 IEEE

842

Telemetry Transport (MQTT), a messaging protocol suitablefor small, low-power, low-bandwidth devices [10], via a WiFiconnection, making it responsible for exchanging the messagesbetween the server and the multiple nodes in the network,that communicate with the broker using a LoRa peer-to-peerconnection.

The chosen communication standard to connect the WSNwas LoRa, since it is a long range low power wirelesstechnology that uses unlicensed radio spectrum [11] and aimsto eliminate repeaters, reduce device cost, increase batterylifetime on devices, improve network capacity and support alarge number of devices. All these features, when comparedto other major communication protocols in IoT systems [12]–[16], as can be seen in Table I, make LoRa the ideal solutionto connect our WSN.

TABLE IMAJOR COMMUNICATION WIRELESS PROTOCOLS CHARACTERISTICS

Feature Wi-Fi Bluetooth ZigBee LoRaData Rate [kbps] 11 x 10 3 1 x 10 3 250 110Frequency [GHz] 2.4 2.4 2.4 0.868Range [m] 1-100 10-100 10-100 2000Nodes/Master 32 7 65540 15000Power Consumption [mA] 100-350 1-35 1-10 1-10Security WPA/WPA2 128 bit 128 bit 128 bit

In order to gather information from the environment thenetwork is complemented with sensor nodes. A high level viewof the system architecture can be seen in Figure 1.

Fig. 1. System Architecture

Each node is composed by two main elements, an ESP32chip, a dual core 32bit ultra-low-power consumption mi-crocontroller with built-in WiFi and BLE and 34 pro-grammable GPIOs, including 18 12-bit Analogic Digital Con-verters (ADCs) for Analog Sensors [17], and a RFM95W,a transceiver LoRa radio module capable of creating multi-point network, with individual node addresses and encryption,in a range of approx 2000 meters [18]. Despite of the samehardware, each node has different characteristics. The brokermakes use of the on-chip WiFi connection, while the sensornode has this feature turn-off as well as uses the deep-sleepfunctionality in order to be a low-power device, allowing it torun on batteries. Also an array of sensors are attached to thesensor node, composed by:

• DHT22, a low cost temperature and humidity sensor thatallow temperature readings between -40 to +80 degreesCelsius and humidity between 0 to 100%, with a pre-cision of 0.5°C and 2-5% for temperature and humidityrespectively [19];

• Soil moisture sensor, a simple breakout, composed bytwo pads that work like probes, that uses the conductivitybetween them for measuring the moisture in soil andsimilar materials [20].

The proposed system is capable of sustain up to 250 nodeswithout compromising the system efficiency. Creating a biggernetwork does not involve extra work for the users, since thenodes are configured automatically inside the network whenbooting for the first time.

In terms of security, all the messages are encrypted using aprivate key. This includes the radio messages, as well as theMQTT messages. Also each user has a private MQTT topic,that only we can listen to, so no other user can receive or sendthose messages.

IV. RESULTS AND DISCUSSIONS

In order to test the proposed solution, there was a needto choose a garden with a convectional irrigation systemalready implemented. A small garden in AUDAX-ISCTE -Entrepreneurship & Innovation Centre of ISCTE - InstitutoUniversitario de Lisboa, was chosen due to known irrigationcharacteristics as well as the possibility to keep track of thesystem implementation. The garden, as seen in Figure 2, has3 irrigation zones so, to achieve the proposed goal, a networkof 3 sensor nodes, each one in a irrigation zone, plus a brokerwere included in the garden.

Fig. 2. Satellite Image of the Garden

The proposed scenario does not replace the convectionalirrigation system, being the goal only to retrieve environmentaldata in order to comprehend the efficiency of the currentsystem and how, with this data, will be possible to improve it.

Therefore each sensor node was running for a week, retriev-ing air temperature and relative humidity and soil moistureevery 10 minutes and sending the data to the servers via thebroker. The first thing in order to prove our system is reliablewas to guarantee that the sensor data is correct. For this, theretrieved data was compared with daily data from the nationalPortuguese weather institute - IPMA - that through their API[21] provide hourly temperature and the higher and lowestvalues for relative humidity for each day in the garden location.

843

Figure 3 shows that the retrieved temperature informationfrom the sensor network follows the online temperature forthat region, with some variations.

Fig. 3. IPMA vs Sensor data comparison

To further evaluate the reliability of the sensor information,the margin of error between the obtained sensor data andthe IPMA API data for the hourly temperature, maximumand minimum day temperature and relative humidity werecalculated, as can be seen in Table II.

TABLE IITEMPERATURE SENSOR DATA ERROR MARGIN

Parameter Error marginHourly Temp [°C] ± 0.77Max Temp [°C] ± 0.57Min Temp [°C] ± 1.14Max Hum [%] ± 6.85Min Hum [%] ± 1

Regarding the temperature parameters is possible to see anaverage margin of error of about 0.82°C and for humidityof about 4%. This is justified by the sensor precision, asdescribed before, of 0.5°C and 2-5% for temperature andhumidity respectively.

Knowing that the retrieved data is reliable is now possibleto understand if the current system is efficient. For this, theair relative humidity, soil moisture, rain and irrigation timeswere compared, as can be seen in Figure 4.

Fig. 4. Sensor data gathered by the network

The first thing to notice is that some irrigation times occurduring rain moments or in situations were the soil moisture

was above 65%. These situations are the main responsible forthe waste of water in irrigation systems. It is also possible toretain that when the air humidity is higher the soil humiditydecreases more slowly.

These results shows that the real time information obtainedfrom the environment can help improve the efficiency of theirrigation system. To calculate the ideal amount of water isnecessary to take into consideration the type of plants, theevapotranspiration, garden area, type of valves and tubing,distance between valves and the number of irrigation’s per day.All of these parameters go along as displayed in Equation 1

T =A× [7.2×A× (ET × 0.001)]× 60

V ×N × 1000/P (1)

where T is the irrigation time, V is the volume of water pervalve, N the number of valves, P is the number of irrigationperiods, A is the garden area, given by Equation 2

A = [(0.5×N)− 1]×D2 (2)

and ET is the evapotranspiration, given by Equation 3

ET = 0.0023×(Tmed+17.78)R0×(Tmax−Tmin)×0.5 (3)

where Tmed is the average temperature, Tmax is the max-imum temperature, Tmin is the minimum temperature andR0 is the incident solar radiation which, based on latitudeand time of the year, for Lisbon is about 17.5 mm/day.These temperature values can be the maximum and minimumretrieved from IPMA for that day or those values adapted fromthe retrieved sensor data, changing the values if any higher orlower values exists.

To have a more efficient system and use as much sensordata as possible, the irrigation times formula (Equation 1), wasadapted to take in consideration the last air humidity and soilmoisture values. With this is possible to obtain a more reliableirrigation time, taking into consideration all the environmentaldata.

Table III shows the irrigation times for each of the followingscenarios: a) Irrigation without controller, the one alreadyapplied to the garden; b) Irrigation system controlled by oursolution while using Equation 1 to calculate the irrigationtimes; c) Irrigation system controlled by our solution whileusing the optimized Equation 1 to calculate the irrigationtimes. Both solutions b) and c) are prediction based on thesensor values retrieved from the environment.

TABLE IIIAVERAGE IRRIGATION TIME PER SCENARIO

Scenario Average Irrigation Time [min]a) 10b) 8.20c) 7.49

Is possible to verify that by applying a solution based on thesensor data retrieved from the environment the amount of time

844

the valves are open are reduced meaning that he amount ofwater used will be lower while maintaining the garden healthy.

Considering the Equation 4 the amount of water used bythe system can be calculated based on the time the irrigationsystem is turned on, allowing to compare the consumption’sof the normal irrigation system and a system based on oursolution.

C =(V ×N × 1000)× T

60(4)

Table IV shows the amount of water used in the scenariosdescribed above.

TABLE IVWATER USED PER SCENARIO

Scenario Water used [L/h] Savingsa) 1890 -b) 1550 17.9%c) 1415 25.1%

Using only temperature data is possible to achieve a 18%improvement in the water used in the system whereas whenthe humidity is taken into consideration a 25% improvementis achievable.

Watering at exactly the same times, but with the idealamount of water, we get identical results in terms of soilmoisture and garden health, but with a much lower waterconsumption.

V. CONCLUSIONS AND FUTURE WORK

In this article, a system based on WSN for retrieving realtime environmental data in order to optimize irrigation timingsas well as the results from a real case implementation, werepresented.

It is possible to conclude that the retrieved data was reliableand capable of being introduced in an intelligent and efficientalgorithm in order to predict a more accurate irrigation timing.Also possible to understand that by keeping track of envi-ronmental status, such as rain or high humidity values, it ispossible to detect if it is really necessary to irrigate.

The real case scenario results show that a reduction of25% can be achieved with the application of a WSN in analready existing environment. Although the theoretical valueis 30%, this can be achieved and even improved with somemodifications to the irrigation system, namely the valves andtubing used as well as the application layout.

By combining the sensor data gathered by the WSN withAI algorithms it possible to configure, in real time, the WSNto adapt to new conditions or specifications, but also achieve abetter efficiency in the process leading to savings for the finaluser, not only monetary but also in natural resources, such aswater and energy, leading to a more sustainable environment.

In order to further test the system, in the future we willimplement the system in a larger garden, with as least 20irrigation zones. Also an application for agriculture, where abig amount of crops exists, each with different water needs,that affect the irrigation times.

REFERENCES

[1] J. A. Stankovic, “Research directions for the internet of things,” IEEEInternet of Things Journal, vol. 1, no. 1, pp. 3–9, 2014.

[2] R. P. Kumar and D. S. Smys, “A Novel Report on Architecture,Protocols and Applications in Internet of Things (IoT),” 2018 2ndInternational Conference on Inventive Systems and Control (ICISC),no. Icisc, pp. 1156–1161, 2018.

[3] United States Environmental Protection Agency, “Why save wa-ter? Statistics and Facts.” [Online] Available: https://www.epa.gov/watersense/statistics-and-facts, (visited 26/11/2018).

[4] A. R. Roselin and A. Jawahar, “Smart agro system using wireless sensornetworks,” in Proceedings of the 2017 International Conference onIntelligent Computing and Control Systems, ICICCS 2017, pp. 400–403,2018.

[5] A. Mondal, I. S. Misra, and S. Bose, “Building A Low Cost Solutionusing Wireless Sensor Network for Agriculture Application,” in 2017 In-ternational Conference on Innovations in Electronics, Signal Processingand Communication (IESC), pp. 60–65, 2017.

[6] T. Savic and R. Milutin, “WSN Architecture for Smart IrrigationSystem,” 2018 23rd International Scientific-Professional Conference onInformation Technology, pp. 1–4, 2018.

[7] R. K. Math and N. V. Dharwadkar, “A Wireless Sensor Network BasedLow Cost and Energy Efficient Frame Work for Precision Agriculture,”International Conference on Nascent Technologies in the EngineeringField (ICNTE-2017), 2017.

[8] M. Usmonov and F. Gregoretti, “Design and Implementation of a LoRaBased Wireless Control for Drip Irrigation Systems,” 2nd InternationalConference on Robotics and Automation Engineering, pp. 248–253,2017.

[9] M. N. Rajkumar, S. Abinaya, and V. V. Kumar, “Intelligent irrigationsystem An IOT based approach,” in 2017 International Conference onInnovations in Green Energy and Healthcare Technologies (IGEHT),pp. 1–5, 2017.

[10] N. Tantitharanukul, K. Osathanunkul, K. Hantrakul, P. Pramokchon, andP. Khoenkaw, “MQTT-Topics Management System for sharing of OpenData,” in 2nd Joint International Conference on Digital Arts, Media andTechnology 2017: Digital Economy for Sustainable Growth, ICDAMT2017, pp. 62–65, 2017.

[11] A. J. Wixted, P. Kinnaird, H. Larijani, A. Tait, A. Ahmadinia, andN. Strachan, “Evaluation of LoRa and LoRaWAN for wireless sensornetworks,” in Proceedings of IEEE Sensors, pp. 1–3, 2017.

[12] A. Lavric and A. Petrariu, “LoRaWAN Communication Protocol : TheNew Era of IoT,” 14th International Conference on DEVELOPMENTAND APPLICATION SYSTEMS, pp. 74–77, 2018.

[13] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, “A comparative studyof LPWAN technologies for large-scale IoT deployment,” ICT Express,2018.

[14] A. Dementyev, S. Hodges, S. Taylor, and J. Smith, “Power ConsumptionAnalysis of Bluetooth Low Energy, ZigBee and ANT Sensor Nodes ina Cyclic Sleep Scenario,” 2013 IEEE International Wireless Symposium(IWS), pp. 1–4, 2013.

[15] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-rangecommunications in unlicensed bands: The rising stars in the IoT andsmart city scenarios,” IEEE Wireless Communications, vol. 23, no. 5,pp. 60–67, 2016.

[16] A. Gloria, F. Cercas, and N. Souto, “Comparison of CommunicationProtocols for Low Cost Internet of Things Devices,” 2017 South East-ern European Design Automation, Computer Engineering, ComputerNetworks and Social Media Conference (SEEDA-CECNSM), pp. 1–6,2017.

[17] Espressif Systems, “ESP32 Series Datasheet.” [Online] Avail-able: https://www.espressif.com/sites/default/files/documentation/esp32/datasheet/en.pdf, (visited 28/11/2018).

[18] Adafruit, “Adafruit RFM95W LoRa Radio Transceiver Breakout - 868or 915 MHz - RadioFruit.” [Online] Available: https://www.adafruit.com/product/3072, (visited 27/11/2018).

[19] SparkFun, “Humidity and Temperature Sensor - DHT22.” [On-line] Available: https://www.sparkfun.com/products/10167, (visited29/11/2018).

[20] SparkFun, “SparkFun Soil Moisture Sensor.” [Online] Available: https://www.sparkfun.com/products/13322, (visited 29/11/2018).

[21] IPMA, “IPMA - API.” [Online] Available: http://api.ipma.pt/, (visited06/12/2018).

845