The Pennsylvania State University

The Graduate School

INTERNET OF THINGS (IOT)-BASED PRECISION IRRIGATION

WITH LORAWAN TECHNOLOGY APPLIED TO VEGETABLE PRODUCTION

A Thesis in

Agricultural and Biological Engineering

by

Haozhe Zhang

© 2021 Haozhe Zhang

Submitted in Partial Fulfillment of the Requirements

for the Degree of

Master of Science

August 2021

ii

The thesis of Haozhe Zhang was reviewed and approved by the following:

Long He Assistant Professor of Agricultural and Biological Engineering Thesis Co-Advisor

Francesco Di Gioia Assistant Professor of Vegetable Crop Science Thesis Co-Advisor

Paul Heinemann Professor of Agricultural and Biological Engineering Head of the Department of Agricultural and Biological Engineering Daeun Choi Assistant Professor of Agricultural and Biological Engineering

iii

ABSTRACT

Precision irrigation with sensors has proven to be effective for water saving in crop production.

Internet of things (IoT) system is necessary for monitoring real-time data from sensors and

automating irrigation systems. Long-range wide-area network (LoRaWAN), a type of low-power

wide-area network (LPWAN), is low-cost and easy to be implemented in IoT systems that can be

used for precision crop irrigation. In this study, an IoT-based precision irrigation system with

LoRaWAN technology was developed and evaluated as a precision management tool on fresh-

market tomato production in an open field. Four irrigation scheduling treatments were designed

and tested, including ET (ETc), MP60 (Watermark 200SS-5 soil matric potential sensors, -60

kPa), MP40 (Watermark 200SS-5 soil matric potential sensors, -40 kPa), and GesCoN (decision

support system). The treatments were arranged based on a randomized complete block design

(RCBD) with four replications. System feasibility, yield, quality, and irrigation water use

efficiency (iWUE) were evaluated during the experiment. The results indicated that treatment

MP60 and GesCoN had a marketable yield 15.2% and 22.1% higher than ET, respectively.

MP40 had a marketable yield 12.5% lower than ET. GesCoN had a significantly higher yield

than ET and MP40. However, MP60 did not produce significantly different results from GesCoN

and ET but had higher yield than MP40. MP40 received relatively low water via irrigation

because of unproper installation and positioning of the soil moisture sensors, which caused a

higher incidence of blossom end-rot and thus lower marketable yield. These results suggest the

importance of installing sensors in multiple sections of a given field to ensure soil moisture

reading are representative of the actual field conditions. Results from MP60 were not

significantly different from GesCoN and ET on marketable fruit numbers, fruit dry matter, fruit

total soluble solids (TSS), and iWUE. Nevertheless, the LoRaWAN-based IoT system worked

iv

well in terms of power consumption, communication, sensors reading and valve control. It can

be potentially implemented for precision and automatic irrigation operation in vegetable fields.

v

Table of contents

List of Figures ............................................................................................................................... vii

List of Tables ................................................................................................................................. ix

Acknowledgments........................................................................................................................... x

1. Introduction ................................................................................................................................. 1

2. Literature Review........................................................................................................................ 6

2.1 Introduction ....................................................................................................................... 6

2.2 Precision irrigation strategies ............................................................................................ 7

2.3 Automatic irrigation control strategies ............................................................................ 10

2.4 Application of IoT in irrigation ....................................................................................... 11

3. Methodology ............................................................................................................................. 14

3.1 Experimental setup .......................................................................................................... 14

3.2 Irrigation system setup..................................................................................................... 18

3.3 Sensor system setup ......................................................................................................... 19

3.4 Valve control ................................................................................................................... 21

3.5 IoT system and data collection ........................................................................................ 22

3.6 Harvest, yield components, biometric assessment, and fruit quality............................... 25

3.6.1 Yield ...................................................................................................................... 25

3.6.2 Fruit quality ........................................................................................................... 25

3.6.3 Biometric assessment ............................................................................................ 26

3.6.4 Irrigation water use efficiency............................................................................... 26

3.7 Weather ............................................................................................................................ 27

vi

3.8 Statistic analysis .............................................................................................................. 27

4. Results and Discussion ............................................................................................................. 28

4.1 Weather data .................................................................................................................... 28

4.2 Feasibility of the IoT system ........................................................................................... 30

4.2.1 Data loss in communication .................................................................................. 30

4.2.2 Battery ................................................................................................................... 31

4.2.3 Valve control ......................................................................................................... 32

4.3 Soil moisture monitoring ................................................................................................. 33

4.4 Fruit yield components .................................................................................................... 39

4.5 Fruit numbers ................................................................................................................... 42

4.6 Fruit Quality .................................................................................................................... 45

4.6.1 Mean fruit weight .................................................................................................. 45

4.6.2 Total soluble solids (TSS) and dry matter ............................................................. 45

4.7 Tomatoes plant dry biomass ............................................................................................ 46

4.8 Irrigation water use efficiency (iWUE) ........................................................................... 47

4.9 Future Improvement ........................................................................................................ 47

5 Conclusion ................................................................................................................................. 49

Bibliography ................................................................................................................................. 50

vii

List of Figures

Figure 1. View of the open field irrigation experiment. ........................................................ 15

Figure 2. Experimental setup and the locations of the sensors and dataloggers. .................. 17

Figure 3. The schematic of the overall irrigation system setup with one treatment irrigation

pipeline as an example. .................................................................................................. 19

Figure 4. Connection of soil MP sensors with the data logger. (A) schematic illustration. (B)

connection with the datalogger ...................................................................................... 20

Figure 5. Connection of pressure sensors with the data logger. (A) schematic illustration. (B)

connection with the datalogger ...................................................................................... 21

Figure 6. Connection of valves with the data logger. (A) schematic illustration. (B)

connection with the datalogger ...................................................................................... 22

Figure 7. Structure of the experimental IoT system. ............................................................. 22

Figure 8. Sensor data display and irrigation valve control on the IoT platform. ................... 24

Figure 9. The notification of the IoT platform on its mobile application. ............................. 25

Figure 10. Daily and cumulative rainfall during the experiment. ......................................... 29

Figure 11. Mean, minimum, and maximum daily air temperature, and growing degree days

during the experiment. ................................................................................................... 29

Figure 12. Daily and cumulative solar radiation during the experiment. .............................. 30

Figure 13. Signal loss rate of six data loggers during the experiment. .................................. 31

Figure 14. Battery voltage of six data loggers during the experiment .................................. 32

Figure 15. Daily average soil MP sensor readings, rainfall, and irrigation of treatment ET

during the experiment .................................................................................................... 34

Figure 16. Daily average soil MP sensor readings, rainfall, and irrigation of treatment MP60

during the experiment. ................................................................................................... 35

viii

Figure 17. Daily average soil MP sensor readings, rainfall, and irrigation of treatment MP40

during the experiment. ................................................................................................... 37

Figure 18. Daily average soil MP sensor readings, rainfall, and irrigation of treatment

GesCoN during the experiment. .................................................................................... 38

Figure 19. Soil MP sensors and pressure sensor readings of treatment MP60 during an

irrigation event on 77 DAT. .......................................................................................... 39

ix

List of Tables

Table 1. Fertigation dates, fertilizer and application rates recommended through GesCoN

over the entire growing season. ..................................................................................... 17

Table 2. Irrigation strategies effects on fresh-market tomato yield and yield components of

each harvest. ................................................................................................................... 41

Table 3. Irrigation strategies effects on fresh-market tomato cumulative yield and yield

components until each harvest. ...................................................................................... 42

Table 4. Irrigation strategies effects on fresh-market tomato average fruit numbers per plant

and mean fruit weight at each harvest. ........................................................................... 43

Table 5. Irrigation strategies effects on fresh-market tomato average cumulative fruit

numbers per plant and mean fruit weight until each harvest.......................................... 44

Table 6. Fruit total soluble solids (TSS) and dry matter in the 2nd and 3rd harvest. .............. 46

Table 7. Tomatoes plant dry biomass of leaves, stems, fruits, and total in 4th harvest. ........ 47

Table 8. Total irrigation water usage per area and irrigation water use efficiency (iWUE). 47

x

Acknowledgments

This research was supported in part by the United States Department of Agriculture’s (USDA)

National Institute of Food and Agriculture (NIFA) Federal Appropriations under Project

#PEN04653; Accession #1016510, Project #PEN04723 and Accession #1020664; the

Pennsylvania Vegetable Grower Association (PVGA); the State Horticultural Association of

Pennsylvania (SHAP), and the USDA Northeast SARE Research and Extension Grant Project

(Grant #19-378-33243). Any opinions, findings, and conclusions, or recommendations expressed

in this thesis are those of the author and do not necessarily reflect the views of funding agencies.

I would like to express my gratitude to Dr. Long He and Dr. Francesco Di Gioia for their

continuous support on my master study and research. They have patiently guided my research

project and professional writing throughout the time. Without their guidance and support, I could

not complete what I have presented in this thesis in the time given. Besides, I would also like to

give my great appreciation to Dr. Daeun Choi and Dr. Paul Heinemann. They have given me

many valuable comments on my research project and helped me to improve my scientific writing

as well.

1

1. Introduction

In the United States, agriculture is a major consumer of ground and surface water, accounting for

approximately 80% of the nation’s consumptive water use, and this percentage can be higher in

the western states characterized by a dryer climate (USDA-ERS, 2019). As the global population

continues to increase, food-crop production is expected to increase dramatically while water

resources are increasingly limited (Howell, 2001; Di Gioia, 2018). Therefore, it is very important

to use water efficiently, especially for crops such as vegetables, characterized by shallow roots

and relatively high water content, and thus very sensitive to water stress. Water can be in excess

or deficit, and thus irrigation water management can substantially impact vegetable yield and

quality (Shock, 2007; Poh, 2011). Conventionally, farm managers determine when and how

much to irrigate vegetable crops based on their experiences and often varies by their time

availability, which may not be optimal leading to inefficient water usage and crop yield and

quality reduction either by over-irrigating or under-irrigating. Precision irrigation is defined as a

modern irrigation management strategy that allows growers to avoid plant water stress at critical

growth stages by applying only the necessary amount of water directly to the crop, varying rate

and duration as needed based on site-specific conditions (Casadesus, 2012). By applying

precision irrigation on agricultural crops, farmers are expected to benefit from lower cost of

irrigation water and manpower, and improvement of crop yield and quality. Adoption of

precision irrigation for crop production systems requires the development of integrated sensing,

decision-making strategies, and control systems, eventually to precisely control the timing, rate

and distribution of water as needed (Smith and Baillie, 2009).

2

The application of irrigation can be related to soil, plant, or weather (Romero et al., 2012).

Different sensor systems and technologies have been investigated and tested for precision

irrigation, including evapotranspiration (ET)-based, plant-based, and soil moisture-based systems

(Pardossi and Incrocci, 2011). ET-based irrigation requires a complete set of weather parameters

from a nearby weather station or a Class A pan evaporimeter to estimate the ET rate (FAO, 1998;

Di Gioia et al. 2009). For plant-based irrigation, canopy temperature is usually used as an

indicator to schedule irrigation based on plant infrared thermal response to water status (Conaty

et al., 2012). Sap flow, as a plant parameter, is also used to schedule irrigation (Fernandez et al.,

2008). Among these methods, soil moisture sensor-based precision irrigation has been widely

tested and used in vegetable fields and protected culture systems. Soil volumetric water content

(VWC) and soil matric potential (MP) are two indicators for available water in the soil which can

be used to implement soil moisture-based irrigation systems (Osroosh et al., 2016). VWC is the

ratio of water volume and soil volume with a unit of percentage or m3/m3. MP describes the force

with which soil matrix holds water (Bianchi et al., 2017). MP measures the required energy for

water movement in the soil which is always a negative number with a unit of kPa. In this study,

the soil moisture-based irrigation method was used throughout the experiments. Based on

preliminary study, the soil MP sensors can fit the IoT systems better and showed sensitivity to

variations of soil moisture for vegetable irrigation. Thus, MP sensors were selected over VWC

sensors in this study.

Wired or wireless sensor network is one of the key technologies for precision and automated

irrigation systems (Kim et al., 2008). Vellidis et al. (2008) developed and evaluated a real-time

smart sensor array which measured soil moisture and temperature for scheduling cotton

irrigation. In similar research, a microcontroller was used to provide real-time feedback control

3

for a drip-irrigation system, toggling system control valves to apply water under the appropriate

conditions (Prathyusha and Suman, 2012). Different embedded control technologies have been

applied for automated irrigation systems, such as Xbee-PRO technology (Ramya and

Palaniappan, 2012), GSM Bluetooth-based remote-control systems (Gautam and Reddy, 2012),

and Dual Tone Multiple Frequency (DTMF) signaling (Dubey et al., 2011). Coates and Delwiche

(2009) developed a mesh network system for wireless valve controllers and sensors to limit

power consumption in addition to controlling water usage. Applications for mobile phones and

wireless personal digital assistants (PDAs and “tablets”) have been developed to enable access to

remote sensor data and control over physical irrigation systems from a distance (Ahmed and

Ladhake, 2011; Sumeetha and Sharmila, 2012).

Internet of Things (IoT), which was coined as a term in 1999 by Kevin Ashton, is a combination

of networked sensors and machines for capturing, transmitting, managing, and analyzing data.

Firstly, the data from sensors are uploaded wirelessly to a server. Then data are available on the

internet for analysis and computing. Finally, the server sends commands wirelessly to the

actuators for executing tasks. A project called SWAMP tested the effect of their IoT-based

irrigation system at four pilot locations in Brazil, Italy and Spain (Kamienski et al., 2019). Goap

et al. (2018) developed an IoT-based smart irrigation management system with machine learning

algorithm to improve irrigation control.

Various wireless technologies in IoT systems have been investigated for crop irrigation

management, such as Wi-Fi, cellular network (GPRS, LTE), ZigBee, and LoRaWAN. Zhao et al.

(2017) compared the performance of Wi-Fi, ZigBee, GPRS, and LoRaWAN for irrigation

systems. The results indicated that Wi-Fi and ZigBee had low coverage and only worked for the

vegetable fields near to the gateway. GPRS is good for long-distance communication, but it has

4

high-power consumption, and high cost of maintenance and deployment. LoRaWAN technology

has a maximum range of 10 km with low power consumption and low-cost, which could allow

for affordable precision irrigation systems for small farms. This technology was originally

published in 2015 by LoRa Alliance, a non-profit association supporting the LoRaWAN protocol

(LoRa Alliance, 2015); however, it is still not widely used in precision irrigation systems. To the

best of knowledge, the LoRaWAN based system has been proposed for vegetables, but has not

been tested under field conditions.

Therefore, the primary goal of the present study was to develop an effective IoT-based precision

irrigation system using LoRaWAN technology for vegetable production. Different irrigation

treatments were established to evaluate the performance of the soil moisture sensor-based

irrigation strategies, and the functionality and robustness of the IoT system.

Specific objectives were:

1. to develop a LoRaWAN technology-based IoT wireless sensing network system for

precision irrigation for vegetable irrigation,

2. to conduct the functionality evaluation on the irrigation system in terms of data

communication, irrigation execution and power consumption, and

3. to evaluate the efficacy of the developed IoT-based precision irrigation system at field

scale using fresh-market tomato as a test crop and comparing the MP sensor-based, ETc-based,

decision support system-based irrigation scheduling.

It is hypothesized that for vegetable fields often located at distance from the farm center, and

thus far away from an internet connection-point such as a gateway, LoRaWAN could be a good

5

choice for the wireless technology in IoT-based irrigation. By using a LoRaWAN IoT-based

precision irrigation system, vegetable farmers will rationalize and optimize the irrigation

management saving water and improving crop yield, quality, and irrigation water use efficiency

(iWUE).

6

2. Literature Review

2.1 Introduction

Precision irrigation is a modern irrigation management strategy which applies only the necessary

amount of water to the crop, with varying rate and duration as needed based on site-specific

conditions. Precision irrigation has been proven to increase the water use efficiency compared to

conventional irrigation management approaches based primarily on operator’s experience or pre-

set irrigation scheduling (Zhang et al., 2002). Irrigation water use efficiency (iWUE) is defined

as the dry mass of plant tissue or yield produced per unit of water volume used for irrigation.

Water can be applied precisely based on the sensor data and scheduling strategies in the

precision irrigation system. In precision irrigation, parameters can be related to soil, plant, or

weather. Common irrigation scheduling strategies include soil moisture-based, plant-based, and

evapotranspiration (ET)-based irrigation. Various control strategies have been used for automatic

irrigation systems. The simplest automatic irrigation system is based on timer with a

programmed setting. Another widely used method is on-off control of solenoid valves based on

sensor data. There are also some advanced control theories, which may have better accuracy in

irrigation, including proportional–integral–derivative (PID) control, fuzzy control, neural

network, and model predictive control (MPC) (Romero et al., 2012).

IoT-based irrigation is a new irrigation solution and has been attracting attention in recent years.

A wireless sensor network (WSN) is usually a core component in an IoT-based irrigation system,

which connects sensors and server wirelessly. Users can access the field data through the Internet

in real-time, and the data can be analyzed and processed on the server. Several types of IoT

wireless technologies can be used for agricultural application, including ZigBee®, ONE-NET,

7

Sigfox, WirelessHART, ISA100.11a, 6LoSWPan, Bluetooth Low Energy (BLE),

LoRa/LoRaWAN, DASH7, and low-power Wi-Fi (Tzounis et al., 2017). Typically, the energy

efficiency, network features, scalability, and robustness should be taken into consideration in

selecting the IoT system technology.

In the following sections, commonly used precision irrigation strategies, automatic control

strategies, and application of IoT systems in irrigation are reviewed with highlighting the

features of the LoRaWAN technology.

2.2 Precision irrigation strategies

Soil water status or soil moisture, measured through either soil volumetric water content (VWC)

or soil matric potential (MP), can be used directly for irrigation scheduling. Soil VWC, which is

the ratio between water and soil volume, has been used widely for precision irrigation

applications. Zotarelli et al. (2009) applied a three-year field experiment to test the effects of

irrigation scheduling on fresh-market tomatoes. Three irrigation treatments were included: SUR

(surface drip irrigation, at 0 cm soil depth, 10% VWC), SDI (subsurface drip irrigation, at 15 cm

soil depth, 10% VWC), and TIME (conventional control, drip irrigation at 0 cm soil depth). The

results showed that SUR and SDI reduced 15%-51% and 7%-29% of irrigation water use

compared to TIME. SUR and SDI had 11%-80% higher yield than TIME. The study concluded

that VWC sensor-based irrigation consistently increased tomato yield and irrigation water use

efficiency (iWUE).

Soil MP sensors have also been used in precision irrigation. Seidel et al. (2017) applied three

different irrigation strategies including soil water balance calculations, soil MP measurements,

and crop growth model simulations in a white cabbage field test for two years. The results

8

showed that soil MP measurements had better performance at a threshold of -25 kPa measured at

30 cm soil depth. Simulation-based strategy required a robust model calibration to be acceptable.

Soil water balance calculations caused over-irrigation because of overestimated crop

coefficients.

Soil MP threshold for optimal irrigation scheduling varies for different crops, soil types, and

changes with sensor installation depth. Researchers commonly tested different thresholds for

certain plants and soil types to get the optimal threshold. A recent review focusing on soil MP

irrigation management strategies for several agronomic and horticultural crops suggested that

higher thresholds (closer to field capacity: -10 to -30 kPa) are usually set for crops that are more

sensitive to water stress such as rice, apples, strawberries, lettuce, broccoli, eggplant, onion and

tomatoes, while lower thresholds (-40 to -60 kPa) are suitable for crops that are more tolerant to

water stress such as most cereals (Bianchi et al., 2017).

Instead of testing different thresholds to get the optimal threshold, some researchers used

simulation models to find optimal soil MP thresholds. Cancela et al. (2015) testing a precision

irrigation system for grapes, implemented a series of actions to define a suitable soil MP

threshold, including calibrating of crop coefficients, establishing the soil-water retention curves,

and measuring plant water status in the field. For example, the SIMDualKc model was calibrated

to estimate the soil water balance. By combining measured soil and plant water status data, the

authors defined that in correspondence of a soil MP threshold of -100 kPa, the grape vines begin

to be stressed.

Many plant-based irrigations are based on the crop water stress index (CWSI) calculated by the

canopy temperature and air temperature. Irrigation is applied once the CWSI reaches a pre-

9

defined level. CWSI is a value between 0 and 1, and higher CWSI means the higher chance of

crops under water stress. O’Shaughnessy et al. (2012) used the CWSI to express plant water

stress and build an irrigation strategy. The authors compared two methods using four irrigation

volume treatments (80%, 55%, 30%, 0% of full replenishment to field capacity in the top 1.5 m

soil) in each method in a precision irrigation system. In method 1, manual irrigations were

applied according to the VWC. In method 2, the crop water stress index with time threshold

(CWSI-TT) was used to define the irrigation scheduling. Results showed in the treatments with

higher irrigation amount, method 2 was better than method 1 considering the yield. But there was

no obvious difference when the irrigation amount was small. During the experiment, over

irrigation happened in the early season and it was a common shortcoming of thermal-based

indices, which needed to be solved.

ET-based irrigation is based on the accumulated water deficit, which is calculated by the ET of

plant-soil system using different equations (FAO, 1998). ET-based irrigation requires a complete

set of weather parameters from a nearby weather station to evaluate daily ET rate. ET-based

irrigation has been widely implemented for various crops. Crop evapotranspiration (ETc) shows

the estimated ET of a certain kind of crop. It is calculated by reference evapotranspiration (ET0)

multiplying crop coefficient (Kc). ET0 can be calculated by different equations. Crop coefficients

are different for different plants and growth stages of plants. Unlike soil moisture or plant-based

irrigation, ET-based irrigation does not need sensors in the soil or near the plants. However,

since it does not measure the water status directly in the soil or plant, it could generate some

cumulated errors over period. Giuliani et. al (2016) carried out a two-year study on tomatoes

using ETc-based irrigation with the treatments of minimal irrigation (only irrigation at

transplanting and fertigation), deficit irrigation (DI, irrigation by 60% of ETc), regulated deficit

10

irrigation (RDI, irrigation by 60%-80% of ETc), full irrigation (FI, irrigation by 100% ETc), and

farmer irrigation (FaI, irrigation by subjective farmer method). ET0 was calculated by Penman-

Monteith’s equation. The results showed in two years, RDI had the highest iWUE, followed by

FI, DI, and FaI. However, RDI had significantly less yield than FaI. FI, DI, and RDI had

significantly higher soluble solids than FaI. Thus, RDI was a better irrigation method compared

with FaI considering the water saving and fruit quality.

2.3 Automatic irrigation control strategies

Precision irrigation can be applied both manually and automatically, whatever the kind of sensor

used. In automatic irrigation, different automatic control strategies are used. The most

fundamental and cheapest automatic irrigation systems employ timers to initiate and end an

irrigation event based on pre-set intervals of time. Farm operators irrigate by a certain time

interval or irrigate at certain time every day. Sudarmaji et al. (2019) developed a time-based drip

and sprinkler irrigation system for horticulture cultivation. Onion and cabbage flower were

selected as experiment crops. Soil VWC sensors were used to measure the soil moisture in real-

time and estimated the performance of irrigation strategies. Results showed that it was effective

to apply water on adjusted time (7 am, 11 am, 5 pm) with 15 min irrigation at each time in a day.

On-off threshold, which is a common and traditional method in automatic irrigation, can be used

along with various sensors. Researchers set a threshold to initiate the irrigation and sometimes

another threshold to end the irrigation. When to end the irrigation can also be decided by a

certain volume of water or a certain duration.

Dynamic threshold is a development of static on-off threshold, where the threshold changes

under different conditions. Osroosh et al. (2015) developed a precision irrigation scheduling with

11

an adaptive algorithm relying on a theoretical crop water stress index (CWSI) for apple trees.

The threshold was dynamically determined by the CWSI trend. The dynamic non-stressed

threshold avoided erroneous stress signals caused by wind, shoot growth or other unexpected

factors. It solved the problem of over-irrigation during early growth period of crop caused by

inaccurate canopy temperature measurement. It was concluded that CWSI can be more efficient

and improve iWUE when combined with a well-developed control algorithm.

Other advanced control theories, such as PID control, fuzzy control, neural network, genetic

algorithms, and model predictive control are also used for automatic irrigation (Abiove et al.,

2020). These strategies usually combined several factors from soil, plant, and weather variables

to optimize the irrigation. However, they require more time and fund expense of farmers and

researchers considering the installation and calibration of the sensors.

2.4 Application of IoT in irrigation

IoT-based precision irrigation is gaining interest from researchers and farm operators in recent

years. A wireless sensor network (WSN) is core component in the IoT-based precision irrigation.

Several kinds of wireless communication technologies are used in irrigation, which have

different data rate, power, and transmission range, including Wi-Fi, Bluetooth, LoRaWAN,

cellular networks, and among others.

Nawandar and Satpute (2019) developed a low cost IoT-based irrigation system with Wi-Fi

network. It had functions of user interaction, one-time setup for irrigation schedule estimation,

neural network-based decision making, and remote data monitoring. Air temperature, humidity,

and soil VWC were monitored by sensors in this system. The data was updated to a server by

Wi-Fi. The server processed three inputs with a neural network and made the decision of

12

irrigation. They tested the system in a sample spinach field. Results showed their system

conserved 30% of irrigation water volume compared with normal drip irrigation.

Gutierrez et al. (2014) developed an automated irrigation system with General Packet Radio

Service (GPRS) network. Soil VWC and soil temperature were measured by sensors. The data

was updated to the gateway through a GPRS module. The gateway unit could process sensor

data, trigger valves, and transmit data to a web application. An algorithm was written into the

gateway to calculate the threshold of temperature and VWC. The system was powered by solar

panels. Researchers could monitor the data and program the irrigation strategy through a website.

The system was tested in a sage crop field for over four months. Results showed the system

conserved 90% of irrigation water volume compared with applying fixed amount of water three

times a week.

Sacaleanu and Kiss (2018) built an IoT system with LoRaWAN technology to monitor walnut

orchards. The test field was a six-year-old walnut orchard whose area was 600,000 m2. Nodes

ware installed 150 m, 300 m, and 450 m away from the base station. Sensors at the node

measured air temperature, light intensity, air humidity, soil temperature, and soil VWC at 20 cm

and 70 cm. A Lithium-ion 2300 mAh or 6600 mAh rechargeable battery supplied the power for

each node. The data transmitted from the nodes to the base station by LoRaWAN and the base

station was connected to a laptop where the data were shown in a software. Results showed that

the node had a 3.13 mA average current consumption. With a 6600 mAh rechargeable battery,

the node can be used for 87 d. Due to dense vegetation, the nodes only reached a maximum

range of 500 m without signal loss though LoRaWAN is designed for long range transmission.

The experiment showed their LoRaWAN system could work properly with low power

consumption and stable connectivity in the walnut orchard.

13

Tzounis et al. (2017) summarized various wireless technologies used in agriculture and their

specifications. Their work shows that compared with traditional wireless network GPRS and Wi-

Fi, LoRaWAN consumes less power. LoRaWAN also has a maximum transmission range of

over 10 km, like cellular networks. The crop field can be far from the gateway and users cost less

power through LoRaWAN. LoRaWAN was originally developed in 2015, thus, it has not been

widely used in precision irrigation system for vegetables.

14

3. Methodology

3.1 Experimental setup

To achieve the proposed goal and objectives, an open field irrigation management study was

conducted at the Horticultural Research Farm of the Penn State Russell E. Larson Agricultural

Research Center (Furnace, PA) during 5/21/2020 – 9/23/2020. The soil type in the experiment

field was silty clay loam with 13.5% of sand, 47.8% of silt, and 38.8% of clay. The soil pH was

6.5. Contents of phosphate, potash, magnesium, and calcium were 156, 266, 521, and 3203 lb/A,

respectively. Total nitrogen and carbon soil content were 0.13% and 1.13%, respectively. The

content of ammonium nitrogen was 2.34 mg/kg.

Fresh-market tomatoes (Solanum lvcopersicum L.) cv. Red Deuce F1 (HM Clause, Davis, CA)

were used as the test crop. Seedlings were planted at the 4th true-leaf stage on May 21st, 2020 on

raised beds mulched with black polyethylene film served by a single drip tape per bed. Beds

were 0.91 m wide and 2.13 m away from each other, and plants were planted at in-row distance

of 0.46 m, establishing a density of 1.03 plants m-2. The 16 mm in diameter drip tape had

emitters spaced 0.30 m and had a flow rate of 1 L/h per emitter at 55 kPa (T-Tape, Rivulis

Irrigation Ltd. San Diego, CA). Plants were trellised using the stake and Florida weave method

(Di Gioia et al., 2016) and the crop was managed according to local practices using an integrated

pest management approach. A view of the open field experiment is shown in Figure 1.

15

Figure 1. View of the open field irrigation experiment.

Four treatments using different irrigation scheduling were designed and tested. Between 0 and 34

days after transplanting (DAT), the whole irrigation system and IoT system were tested, and the

irrigation was the same for four treatments during this period. The treatments started on 36 DAT.

Treatment 1 (standard control, denoted ET) was based on crop evapotranspiration (ETc) and

consisted of irrigating the crop with a volume equivalent to 100% of the ETc estimated on daily

basis using a dual crop coefficient model according to FAO Paper 56 (FAO, 1998). Irrigation

was applied after reaching 12 mm of cumulative water deficit on average. Treatment 2 and 3

were based on soil matric potential (MP) sensors (Watermark 200SS-5, Irrometer company, Inc.,

Riverside, CA) at different setpoints of -60 kPa (denoted MP60) and -40 kPa (denoted MP40),

respectively. The thresholds were suggested by the introduction of the MP sensors (Irrometer,

2021). Once the average sensor readings at 20 cm depth reached the thresholds, the IoT system

would send a notification and valves were manually controlled through the IoT system interface.

The irrigation durations for the MP60 and MP40 were determined by monitoring the sensor

readings in the pre-tests, and some adjustments were made during the experiment according to

the response of the sensors at 20 and 40 cm soil depth, which were on average 114 and 117 min

between 36 and 66 DAT, 177 and 151 min between 67 and 83 DAT, and 285 and 241 min

16

between 84 and 125 DAT, respectively. After the irrigation, sensors at 20 and 40 cm would show

readings higher than -20 kPa and did not change much in several hours. Treatment 4 (denoted

GesCoN) was based on recommendations provided by a decision support system (DSS) named

GesCoN. When irrigation was needed, the system would send the suggestion of irrigation

volume and time. GesCoN is a model based DSS developed by researchers at the University of

Foggia (Italy) to estimate crop N and water requirements and manage fertigation in open field

grown vegetables (Elia and Conversa, 2015) and calibrated on tomatoes (Conversa et al. 2015).

Using the real-time, historic, and forecasted local weather data, the basic information on soil

(soil texture and mineral content), the specific crop (species, variety, planting date, planting

density, estimated production), on the area of the field and on the irrigation system (drip tape

flow rate and distance between drippers) the GesCoN integrates water balance, plant growth, and

nitrogen uptake sub-models to estimate crop dry matter production, crop yield,

evapotranspiration, soil moisture, drainage flow, soil nitrogen dynamics, and nitrate leaching and

provides recommendations on irrigation and N fertigation. The DSS GesCoN has been integrated

with the web-application Ecofert (www.ecofert.it) and an Android App (Ecofert). The DSS can

be connected to weather stations using the RESTful API method to automatically retrieve real-

time on-site or location specific climate data (Gallardo et al. 2020).

All treatments received N fertilizer via fertigation at the same time, in the same form and amount

according to the recommendations provided by GesCoN as reported in Table 1.

17

Table 1. Fertigation dates, fertilizer and application rates recommended through GesCoN over the entire growing season.

Date GesCoN suggested fertilizer

N% Applied N (kg ha-1)

6/26/2020 Urea 46 18 6/30/2020 Urea 46 30 7/15/2020 Urea 46 30 7/21/2020 Urea 46 19 Calcium nitrate 15.5 8 7/31/2020 Urea 46 14.5 Calcium nitrate 15.5 9.5 Total N applied 129

The layout of the field is shown in Figure 2. The overall field (planting area) was 27.74 m ×

42.67 m in dimension. The field was constituted by 12 raised beds 42.7 m long. Three beds

constitute a block, and each block was divided into 4 sections, each hosting an irrigation strategy

treatment. Overall, the field was divided into 16 experimental units of the same size. Each

experimental unit was constituted by three beds about 9.14 m long. Twenty plants were placed in

each bed, for a total of 60 plants per experimental unit. Treatments were arranged in a

randomized complete block design (RCBD) with four replicates for each. Four treatments were

randomly distributed in each block.

Figure 2. Experimental setup and the locations of the sensors and dataloggers.

18

An irrigation system was setup in the field, connecting irrigation pipelines, solenoid valves,

pressure sensors, soil MP sensors and data loggers (DL). Six data loggers were used to connect

the sensors and valves, including one for valves (DL 1), four for soil MP sensors (DL 2~5), and

one for pressure sensors (DL 6). DL 1 connected to four valves. DL 2, 3, 4, and 5 connected to

four soil MP sensors of treatment ET, MP60, MP40, and GesCoN, respectively. DL 6 connected

to four pressure sensors. The detail of sensors, datalogger, and valves are introduced in the

following sections.

3.2 Irrigation system setup

Figure 3 shows the structure of the irrigation system, including pipelines, valves, pressure

sensors, flow meters, and other accessories. The water was supplied from the main water line of

the farm. The main irrigation line after a screen filter was divided into four pipelines, one for

each treatment. Four latching DC solenoid valves (PGV Series 2.54 cm, Hunter Inc., San

Marcos, CA) were installed at the beginning of each line, followed by the pressure sensors (0.64

cm 5V 0-1.2 MPa) installed one per line behind the solenoid valves to measure the water

pressure. The pressure indicated the status of the valves. The pressure regulators (241 kPa)

behind the pressure sensors limited the water pressure reaching the following components. The

flow meters measured the cumulative flow volume during the experiment. The fertigation system

constituted by a by-pass with a Venturi injector was used to apply fertilizers when recommended

by the DSS GesCoN. When fertigation started, the injector was opened first, then the switch in

the pipeline was closed. When fertigation ended, the switch in the pipeline was opened first, then

the injector was closed. The pressure gauges were used to measure the real-time water pressure

and calibrate pressure sensors. The last components were pressure regulators at 103 kPa, which

19

is the suggested pressure for the irrigation drip-tape lines used. Drip tapes were placed

underneath the mulch for each section. Then the drip lines on the same treatment were connected

to the pipes in the field pathway. Finally, these pathway pipes were connected to the

corresponding pipelines (2.54 cm) for each treatment.

Figure 3. The schematic of the overall irrigation system setup with one treatment irrigation pipeline as an example.

3.3 Sensor system setup

As indicated earlier, in total six data loggers were used in the system. The major components of

the data loggers included a base control board (Vinduino LLC, Temecula, CA) and a LoRaWAN

wireless communication unit with antenna (LM130-H1, GlobalSat WorldCom Corp., New Taipei

City, Taiwan). Each data logger was powered by 3.7 V lithium-ion polymer (LiPo) battery, and a

solar panel (70 mm × 70 mm) was connected to charge the battery. DL 1 and 6 used 10000 mAh

batteries and Data loggers 2~5 used 5000 mAh batteries. In each treatment, four soil MP sensors

20

(Watermark 200SS-5, Irrometer company, Inc., Riverside, CA) were used. Figure 4 shows the

connection of the soil MP sensors to the data loggers.

(A) (B)

Figure 4. Connection of soil MP sensors with the data logger. (A) schematic illustration. (B) connection with the datalogger

In each treatment, the sensors were installed in two experimental units and at two depths (20 cm,

40 cm). During the installation, a soil sampler was used to dig a cylinder hole at 40 cm depth. Then,

two soil MP sensors, which were soaked in water for 24 hours before installation, were placed in

the hole. The original soil was back filled, and water was added to ensure the close contact between

the sensors and soil. The average soil MP of two experiment units at 20 cm was used for starting

the irrigation in the treatments MP60 and MP40. The sensors at 40 cm were used as a reference to

check whether the water went down to the depth when the irrigation ended. The sensors installed

in treatments ET and GesCoN were used only for monitoring the soil moisture levels.

In each treatment pipeline, a pressure sensor was installed behind the solenoid valve to indicate

the valve status (on or off). The sensors were calibrated with the pressure gauges installed on the

same line. By adjusting the opening of the ball valve, eleven pairs of pressure sensor readings and

21

pressure gauge readings from 0 to 207 kPa were used to calibrate the pressure sensor. Figure 5

shows the connection of the pressure sensors to the data loggers.

(A) (B)

Figure 5. Connection of pressure sensors with the data logger. (A) schematic illustration. (B) connection with the datalogger

3.4 Valve control

DC latching solenoid valves were used to control irrigation by connecting and disconnecting the

water supply. Figure 6 shows the connection of valves and the data logger. The Vinduino board

originally supports only one valve. For using it to control four valves, the sensor ports on the

board were used to extend the capability. Sensor ports can send high/low signal at 3.3/0 V, but

the current is not enough for executing the valves. To use the battery voltage (Vbat, 3.7 ~ 4.2 V)

as power of the valves, four 2-way relays were added for the valve control. Two sensor ports

send high/low signals to the IN1 and IN2 port of the 2-way relay to change electric potential at

OUT1 and OUT2. High or low of IN corresponds to Vbat or 0V of OUT. OUT1 and OUT2

connect to the positive and negative electrode of the valve, respectively. When IN1/IN2 is

high/low, the valve will open. When IN1/IN2 is low/high, the valve will close. When IN1/IN2 is

the same or no input, the valve will not respond.

22

(A) (B)

Figure 6. Connection of valves with the data logger. (A) schematic illustration. (B) connection with the datalogger

3.5 IoT system and data collection

An Internet of things (IoT) system was established to connect the sensors, valves, and data loggers

in the field. A LoRaWAN gateway (Sentrius™ RG191, Laird) was placed in an office which was

150 m away from the field. Figure 7 illustrates the procedure of the IoT system development. The

gateway and data loggers were configurated in a free IoT server named The Things Network. Then

an IoT platform AllThingsTalk (AllThingsTalk NV, Mechelen, Belgium) was used to store and

display the sensor data and control the valves.

Figure 7. Structure of the experimental IoT system.

23

The first step was to configure the gateway and the data logger. The LoRaWAN gateway was first

connected to a computer with Wi-Fi for matching the parameters, including using “The Things

Network” as server and “915MHz” as band frequency. In the server interface, a gateway was

created with the recorded gateway ID. Then an “Application” was built in the gateway to represent

the proposed tomato field irrigation system. An Application Extended Unique Identifier (EUI) of

the application was generated automatically by the server. Six devices were created under the

“Application” to connect the six data loggers, respectively. They shared the same Application EUI

while had unique App Keys. Algorithms were developed for these data loggers with different

functions, including soil MP recording, valve control, and water pressure recording. The

Application EUI and App Key were used for the data loggers to be connected to the gateway for

uploading data (sensor data) and downloading data (control signal). Once the data loggers were

powered, they were connected to the server after a few seconds.

The next step was to transmit the data to the AllThingsTalk IoT platform and process the data. In

the web-based interface, integration “AllThingsTalk” was added to the created application. Then

the IoT platform was linked to the server The Things Network. Sensor data was uploaded and

stored in the IoT platform for monitoring and irrigation control. In the platform, six “devices” were

added corresponding to the six data loggers. Then the “assets” were created in the “devices” to

represent the corresponding sensors or valves in each data logger. The data communication

frequency was applied with a pre-set time interval, which was 1 minute. The sensor data (uplink)

and valve control (downlink) were communicated as binary payload. Each value, such as the

battery voltage, sensor reading, and valve status was parsed from a byte between 0 and 255. These

were converted to actual values by editing the payload format in the “devices” interface. The valves

can be separately controlled to turn on or off through the IoT platform.

24

For DL 1, the battery voltage, valve status, and valve control switch were shown in the

AllThingsTalk display. And for DL 2~6, the real-time sensor readings and battery voltages were

displayed. The historic data of sensors and batteries were stored and can be exported in a comma-

separated values (CSV) file. The status of all assets of six devices could be displayed to a pinboard

(Figure 8). The status of the solenoid valves was shown by a circle indicator with black color for

close and green color for open. The valves could be controlled by inputting a number between 0

and 15 in the textbox. The soil MP sensor readings were compared with the pre-set threshold, and

a notification (Figure 9) can be sent by email or mobile application when the threshold is reached.

However, full automation was not applied for the irrigation, instead a manual control through the

IoT platform was used to start and end irrigations. The major reason was that there was signal loss

occasionally on the farm which may lead to delayed response of valves with automatic setting.

Figure 8. Sensor data display and irrigation valve control on the IoT platform.

25

Figure 9. The notification of the IoT platform on its mobile application.

3.6 Harvest, yield components, biometric assessment, and fruit quality

3.6.1 Yield

At harvest, ten continuous representative tomato plants were selected from the middle row of

each experimental unit. Tomatoes ranging from mature green to ripe red in color were harvested

on 8/7, 8/19, 9/1, 9/11, and 9/23/2020 in correspondence of 78, 90, 103, 113, and 125 days after

transplanting (DAT), respectively. The experiment ended with the last harvest on 125 DAT.

Harvested tomatoes were sorted into marketable and unmarketable fruit, and marketable fruit

were graded into several size categories: extra-large (d > 6.99 cm), large (6.35 cm < d ≤ 6.99

cm), medium (5.72 cm < d ≤ 6.35 cm), and small (5.40 cm < d ≤ 5.72 cm), according to the U.S.

Standards for Grades of Fresh Tomatoes (USDA, 1997). Fruit number and fresh weight were

recorded for each category.

3.6.2 Fruit quality

Corresponding to the 2nd and 3rd harvest (90 and 103 DAT), subsamples of representative fruit

for each experimental unit were used for fruit quality determinations. Five representative

26

tomatoes per experimental unit were selected from the harvested tomatoes. Fruit were split into

quarters longitudinally, and one quarter of the fruit was used to measure the total soluble solids

(TSS) content and the remaining portion was dried in a forced-air oven at 65 ˚C until constant

weight to determine the fruit dry matter content. Fruit TSS were measured using a portable

refractometer (Brix-Stix BX 100 Hs; Techniquip Corporation, Livermore, CA) and values were

expressed in ˚Brix at 20 ˚C. Fruit dry matter content was calculated as a percentage of the sample

fruit fresh weight.

3.6.3 Biometric assessment

Corresponding to the 4th harvest (113 DAT), a biometric assessment was conducted using two

representative plants per experiment unit. For each sampled plant, leaves, stems, and fruits were

separated and placed in labeled paper bags and dried in a forced-air oven at 65 ˚C until constant

weight to measure leaf, stem, fruit, and total plant dry biomass. In the case of the fruit, the total

fruit fresh weight of each plant was measured, and a representative subsample was dried in the

oven to calculate the dry matter content used to calculate the total fruit dry biomass. The total

plant dry biomass was the sum of the dry weight of fruits, stems, and leaves.

3.6.4 Irrigation water use efficiency

Water usage was recorded after the last harvest by reading the flow meter. Water usage was

measured by treatment and assumed to be the same for the four replications. Irrigation water use

efficiency (iWUE) was calculated dividing the marketable fruit fresh weight by the total water

usage.

27

3.7 Weather

Weather data during the experiment (5/12/2020 – 9/23/2020) were acquired from PSU Rock

Spring weather station (40.710 °N, 77.952 °W, 376.0 m). Hourly solar radiation, rainfall, and air

temperature (average, maximum, minimum) data at height of 2 m was recorded. Cumulative

solar radiation, rainfall, and air temperature were calculated.

3.8 Statistic analysis

The data of the of the IoT system between 6/25/2020 (35 DAT) and 9/23/2020 (125 DAT) was

downloaded from the IoT platform AllThingsTalk. Data loss rate was calculated by the numbers

of signals received divided by the expected numbers of signals received. Each data logger’s and

the average data loss rate was calculated. Each data logger’s battery voltage, the soil MP sensor

readings, weather data including solar radiation, rainfall, and air temperature during the

experiment were processed using Microsoft Excel and presented in figures.

Analysis of variance (ANOVA) was applied to fruit fresh weight, fruit numbers, fruit total

soluble solids (TSS), fruit dry matter, mean fruit weight, plant dry biomass, and iWUE for four

different treatments and four different blocks. These statistical analyses were performed using

PROC GLM procedure in SAS 9.4 (SAS Institute Inc., Cary, NC, USA., 2016). All means were

compared by the least significant difference (LSD) test at P ≤ 0.05.

28

4. Results and Discussion

4.1 Weather data

Figures 10-12 show the daily rainfall, air temperature and solar radiation during the experiment

period. The mean daily air temperature was on average 20.3 ℃ during the experiment. Daily

minimum and maximum air temperatures ranged from -1.8 to 21.8 ℃ and 16.3 to 34.2 ℃,

respectively. The average daily solar radiation was 26.8 MJ·m-2 and cumulative precipitation was

300 mm. The maximum daily solar radiation occurred at 23 DAT and gradually decreased

afterwards. During 0 - 34 DAT, the cumulative precipitation was 167 mm and the mean daily air

temperature on average was 18.9 ℃. During this period, irrigation was applied for only a few

times. The amount of irrigation water was the same for the four treatments and the irrigation

volume was not measured. During 35 – 95 DAT, the cumulative precipitation was 79 mm and the

mean daily air temperature on average was 22.6 ℃. Irrigation was frequently applied during this

period. Irrigation was usually applied once every two or three days. During 96 – 125 DAT, the

cumulative precipitation was 54 mm and the mean daily air temperature dropped continuously.

Irrigation was applied less frequently. Frost happened on 117 DAT, thus, on 125 DAT all the

tomatoes in the field were harvested.

29

Figure 10. Daily and cumulative rainfall during the experiment.

Figure 11. Mean, minimum, and maximum daily air temperature, and growing degree days during the experiment.

30

Figure 12. Daily and cumulative solar radiation during the experiment.

4.2 Feasibility of the IoT system

4.2.1 Data loss in communication

The system generally worked well during 35 - 125 DAT. There were a few occasions of data loss

in the IoT system. Six data loggers had similar data loss rate and the average data loss rate was

5.51 % (Figure 13). The gateway was placed inside a building about 150 m away from the field,

which may have led to some signal loss. This can be improved if the gateway is installed outside

with less occlusion. Meanwhile, the disconnection of any component such as the gateway, the

server The Things Network, and the IoT platform AllThingsTalk can contribute to the signal

loss. However, since data were uploaded at high frequency every minute, and there was no long

period of data loss, the sensing system was not affected significantly. A few continuing data loss

31

for more than 10 minutes was observed. If the irrigation was applied during these periods, it is

possible that the valves would not respond on time.

Figure 13. Signal loss rate of six data loggers during the experiment.

4.2.2 Battery

The batteries in the six data loggers were charged by solar panels and maintained sufficient power

throughout the three-month experiment period. Figure 14 shows the voltage of these batteries over

time. On 103 and 114 DAT there was voltage drop for all data loggers due to the continuing cloudy

and rainy weather. A disconnection of the solar panel wires occurred for the data logger 3 between

60 and 70 DAT, resulting in an obvious voltage drop compared with other time period. Overall,

the battery voltages during the experiment were always higher than 4 V, which showed that the

5000 mAh LiPo battery with solar panel could support a data logger continuously in the open field.

32

Figure 14. Battery voltage of six data loggers during the experiment

4.2.3 Valve control

During the experiment, the irrigation was applied successfully by controlling the status of the

solenoid valves with the implemented IoT system. When the soil moisture threshold was

reached, the platform sent a notification through the mobile application and an email to the end-

user. Then the operator turned the corresponding valves on, and the irrigation started. When the

valves were opened, the status of the valves were showing “Green” in the IoT interface, and the

pressure of the water line was indicated with a positive reading. When the valves were closed

after certain hours of irrigation, the valve status was showing “Black”, and the reading of the

pressure sensors returned to zero. It was also observed that, the valve status changed to “Green”

while the pressure sensor reading did not change accordingly. In this situation, the valve did not

actually open with pressure water going through. In other words, the pin in the solenoid did not

respond even if the signal was received by the controller. By sending the signal several times, the

33

valves did finally open or close. This can be attributed to the clog in the solenoids or small

voltage for powering the valves (4 V). In the future, a higher voltage (9 V) can be used to

improve the response of the valve. Meanwhile, remote control with manual operation was used

in the experiment, for a fully automatic irrigation system, the start and end of an irrigation will

be programed into the controller according to the soil moisture thresholds.

4.3 Soil moisture monitoring

Figures 15-18 shows the daily average soil MP sensor readings, rainfall, and irrigation of four

treatments between 35 and 125 DAT. Generally, the sensor readings presented the changes of the

soil moisture level in the test field. When irrigation or rainfall happened, the sensor readings were

increasing. The soil at 40 cm were generally dryer than the soil at 20 cm. It indicated that the

irrigation did not drain the soil at 40 cm depth completely. Sometimes the sensors at 40 cm had no

response after the irrigation, suggesting the irrigation volume was not enough and the water could

not reach the depth of 40 cm.

For treatment ET (Figure 15), most sensor readings were higher than -100 kPa. The sensor reading

did not respond consistently because the irrigation was based on ETc, and the sensor readings were

not considered. The readings of MP 1-1 were usually lower than MP 1-2 (treatment 1-ET, the

second sensor installation place), which might be caused by uneven soil texture and hydraulic

properties. The place where MP 1-2 was installed tended to get wetter than MP 1-1.

34

Figure 15. Daily average soil MP sensor readings, rainfall, and irrigation of treatment ET during the experiment

For the treatment MP60 (Figure 16), the sensor readings were all above -140 kPa throughout the

experiment. When the average sensor reading at 20 cm reached -60 kPa, the irrigation was

applied. Thus, the daily average readings of sensors at 20 cm were usually above -60 kPa during

the experiment. However, between 35 and 55 DAT, MP 2-1 20 cm continuously got dryer, which

was abnormal since there was irrigation. A new sensor was installed to replace that one at a near

position, and the readings of the new sensors were consistent with the other sensors. Before the

replacement, the irrigation was based on MP 2-2 20 cm only. There were two possible reasons for

this abnormality. One possible explanation is that the position where the sensor was installed might

have a dryer condition. A second explanation could be that the sensor was not in close contact with

35

the soil. The volume of each irrigation was adjusted according to the sensor response during the

experiment. The sensor readings of MP 2-2 40 cm kept low between 65 and 72 DAT but returned

to higher level after irrigation on 72 DAT. This indicated that the irrigation volume was not enough

during this period. Thus, the water could not reach the 40 cm depth. After 85 DAT, the irrigation

depth was increased to over 10 mm each time. At this point, the response of the four sensors

became more consistent.

Figure 16. Daily average soil MP sensor readings, rainfall, and irrigation of treatment MP60 during the experiment.

For the treatment MP40 (Figure 17), all the sensor readings were above -120 kPa during the

experiment. When the average sensor reading at 20 cm reached -40 kPa, the irrigation was applied.

Thus, the daily average sensor readings at 20 cm were usually above -40 kPa. The sensor response

36

kept consistent between 35 and 125 DAT, and all the sensors showed a high reading compared

with other treatments. However, the total irrigation volume was much less than other treatments.

In the field, treatment MP40 was dry during the experiment and tomato plants were more stressed

because of lack of water. This outcome indicated that the sensor readings did not show the real

condition of treatment MP40. A possible reason could be variations of soil texture within the

experimental field. The sensors of each treatment were installed at only two positions, and the two

positions were not far from each other because the wire of the sensors was only 5 m long and they

needed to connect to the same data logger. Meanwhile, the position where the sensors were

installed was wetter than the average condition of treatment MP40. Thus, the sensors readings of

treatment MP40 showed wetter readings than the actual soil moisture level, making the irrigation

not adequate.

37

Figure 17. Daily average soil MP sensor readings, rainfall, and irrigation of treatment MP40 during the experiment.

For treatment GesCoN (Figure 18), most sensor readings were above -140 kPa. Two sensors at

MP 4-2 continuously dropped between 70 and 80 DAT and between 100 and 110 DAT. The drip

tapes of the area where the sensors were installed was found disconnected during that period, and

the area did not get any irrigation. However, this issue affected only one of the guard rows from

which tomatoes were not harvested. Therefore, the issue did not affect the results. However, the

issue could be detected by monitoring the sensors, which suggests that integrating the use of soil

moisture sensors with GesCoN could provide some benefits.

38

Figure 18. Daily average soil MP sensor readings, rainfall, and irrigation of treatment GesCoN during the experiment.

Figure 19 shows the sensor readings around an irrigation event of treatment MP60 at 77 DAT. The

irrigation was based on the average of two sensor readings at 20 cm. When the average readings

reached -60 kPa at 18:50, the notification was sent, and the valve was opened remotely. The

increased water pressure readings showed that the water was on. The sensor readings increased in

30 minutes and stopped increasing in about 2 hours. Sensors at 40 cm and at the first installation

position responded first. Commonly, water reached 20 cm first, then 40 cm. But the sensors were

not close to the drip tapes, and the irrigation water spread in both vertical and horizontal direction

to reach the sensors. For this reason, sometimes the irrigation water reached the 40 cm sensors

first. At 22:20, the irrigation ended, and the water pressure readings went back to 0. Before the

39

irrigation, the sensors readings at 20 cm were near -60 kPa, and the sensor readings at 40 cm were

at -100 and -140 kPa. During the irrigation, all the sensor readings increased to above -20 kPa, and

sensors at 40 cm were still dryer than sensors at 20 cm. After the irrigation, the readings of MP 2-

1 40 cm decreased a little, and then became higher. The possible reason is the water flowed through

the area and that the soil was not evenly moist just after irrigation. Sensors at 40 cm would become

dryer faster than sensors at 20 cm.

Figure 19. Soil MP sensors and pressure sensor readings of treatment MP60 during an irrigation event on 77 DAT.

4.4 Fruit yield components

Tomato fruit yield components were recorded at each harvest for all four treatments. Table 2 shows

the yield of marketable fruit by size categories. In general, the marketable yield was the highest in

the GesCoN, followed by the MP60 and the ET, with the MP40 as the lowest. The 3rd harvest was

the most abundant in terms of fruit fresh weight, and differences between treatments became more

apparent with this harvest. Large and medium tomatoes represented a limited fraction compared

40

to extra-large tomatoes in the first four harvests. In the last harvest, there were more large and

medium tomatoes because all tomatoes were harvested. MP60 was not significantly different from

ET and GesCoN at most categories. But MP60 had more marketable yield than ET at the 4th harvest,

and more unmarketable weight than GesCoN at the 3rd harvest. Comparing MP40 with ET, MP40

had less extra-large and marketable fruit fresh weight than ET in the 2nd and 3rd harvest, and more

unmarketable weight than ET in the 1st harvest. MP40 provided the lowest yield.

Table 3 shows the cumulative fruit fresh yield determined cumulating the yield after each harvest.

MP60 was not significantly different from ET and GesCoN for most yield components. However,

MP60 had more large fruit fresh weight than ET until the last harvest, more unmarketable weight

than GesCoN until the 3rd, 4th, and 5th harvest, and more marketable and total weight than ET until

the 4th harvest. Comparing MP40 with ET, MP40 had lower extra-large fruit fresh weight than ET

until the 3rd harvest, lower marketable yield than ET until the 2nd, 3rd, and 4th harvest, and lower

total yield than ET until the 3rd and 4th harvest but more total yield in the 1st harvest since there

was a high proportion of unmarketable fruit. MP60 and GesCoN had a total marketable yield 15.2%

and 22.1% higher than ET, respectively. MP40 had a marketable yield 12.5% lower than ET. MP60

was not significantly different from ET and GesCoN. However, MP40 had significantly lower

marketable yield than MP60 and GesCoN but was not significantly different from ET. MP60 and

MP40 are sensor-based irrigation using different soil moisture levels of thresholds. MP40 was

expected to be wetter than MP60, thus, it was expected to suffer less water stress and potentially

have more yield. However, as mentioned in the soil moisture monitoring section, MP40 used less

water than other treatments, presumably because of improper sensor positioning or positioning of

the sensors in an area that was not representative of the entire field, which led to lower yield.

41

Nevertheless, the results of MP60 shows the sensor-based irrigation scheduling with the IoT

system is effective and suitable for precision irrigation management at field scale.

Table 2. Irrigation strategies effects on fresh-market tomato yield and yield components of each harvest.

Harvest DAT

Treatment Fruit Fresh Weight (Mg ha-1) XL L M Cull TM T

78

ET 3.68 0.70 0.23 2.03 b 4.62 6.65 b

MP60 4.13 0.71 0.09 1.43 b 4.92 6.34 b

MP40 4.21 0.74 0.26 3.48 a 5.21 8.69 a

GesCoN 4.70 0.78 0.15 1.48 b 5.62 7.09 b P-value 0.69 0.92 0.30 0.001 0.68 0.03

90

ET 12.23 a 0.15 0.06 7.00 12.44 a 19.44

MP60 13.23 a 0.32 0.00 4.90 13.55 a 18.44

MP40 8.49 b 0.27 0.16 8.23 8.92 b 17.15

GesCoN 12.22 a 0.17 0.00 6.25 12.39 a 18.64 P-value 0.002 0.72 0.07 0.21 0.002 0.58

103

ET 18.08 b 0.68 bc 0.11 ab 12.29 a 18.88 b 31.16 a

MP60 20.07 ab 1.33 ab 0.24 a 12.26 a 21.64 ab 33.90 a

MP40 11.75 c 0.43 c 0.00 b 10.02 ab 12.18 c 22.19 b

GesCoN 23.43 a 1.49 a 0.20 a 7.09 b 25.12 a 32.21 a P-value 0.001 0.03 0.05 0.03 0.001 0.0002

113

ET 6.38 bc 0.45 0.24 3.07 7.06 b 10.13

MP60 9.79 ab 1.19 0.13 3.62 11.11 a 14.74

MP40 5.21 c 0.64 0.05 3.89 5.90 b 9.79

GesCoN 9.84 a 0.74 0.24 3.49 10.82 a 14.31 P-value 0.03 0.30 0.51 0.92 0.02 0.11

125

ET 5.98 2.54 2.70 1.35 11.22 12.57

MP60 5.50 2.91 2.80 1.46 11.22 12.67

MP40 8.49 3.36 3.28 1.88 15.13 17.01

GesCoN 6.53 2.78 2.94 1.69 12.25 13.94

P-value 0.52 0.67 0.65 0.22 0.43 0.34 XL = Extra-Large, L = Large, M = Medium, Cull = Unmarketable, TM= Total marketable, T = Total Treatments are grouped by LSD t-test if they are significantly different.

42

Table 3. Irrigation strategies effects on fresh-market tomato cumulative yield and yield components until each harvest.

Harvest Day Treatment Fruit Fresh Weight (Mg ha-1) XL L M Cull TMY TY

78

ET 3.68 0.70 0.23 2.03 b 4.62 6.65 b

MP60 4.13 0.71 0.09 1.43 b 4.92 6.34 b

MP40 4.21 0.74 0.26 3.48 a 5.21 8.69 a

GesCoN 4.70 0.78 0.15 1.48 b 5.62 7.09 b P-value 0.69 0.92 0.30 0.001 0.68 0.03

until 90

ET 15.92 a 0.85 0.29 9.03 ab 17.06 a 26.09

MP60 17.35 a 1.03 0.09 6.32 b 18.47 a 24.79

MP40 12.70 b 1.01 0.42 11.71 a 14.13 b 25.84

GesCoN 16.92 a 0.94 0.15 7.73 b 18.01 a 25.74 P-value 0.03 0.71 0.08 0.03 0.03 0.90

until 103

ET 34.00 b 1.53 0.40 21.32 a 35.93 b 57.25 a

MP60 37.42 ab 2.36 0.32 18.58 a 40.11 ab 58.68 a

MP40 24.45 c 1.44 0.42 21.72 a 26.31 c 48.03 b

GesCoN 40.35 a 2.44 0.34 14.82 b 43.13 a 57.94 a P-value 0.002 0.11 0.89 0.003 0.001 0.01

until 113

ET 40.37 b 1.98 0.64 24.39 ab 43.00 b 67.38 b

MP60 47.21 ab 3.55 0.45 22.20 b 51.22 a 73.42 a

MP40 29.66 c 2.07 0.47 25.61 a 32.21 c 57.82 c

GesCoN 50.19 a 3.18 0.58 18.31 c 53.95 a 72.26 a P-value 0.0008 0.16 0.87 0.001 0.0004 0.0005

until 125

ET 46.35 bc 4.52 b 3.34 25.73 ab 54.21 bc 79.95 ab

MP60 52.71 ab 6.46 a 3.26 23.66 b 62.43 ab 86.09 a

MP40 38.16 c 5.43 ab 3.75 27.49 a 47.34 c 74.83 b

GesCoN 56.72 a 5.95 a 3.52 20.00 c 66.19 a 86.20 a

P-value 0.01 0.04 0.66 0.002 0.01 0.06 XL = Extra-Large, L = Large, M = Medium, Cull = Unmarketable, TM= Total marketable, T = Total Treatments are grouped by LSD t-test if they are significantly different.

4.5 Fruit numbers

Average fruit numbers per plant were recorded by fruit size categories. Table 4 shows the fruit

numbers of each harvest. In general, the number of marketable tomatoes was the highest in

GesCoN, followed by MP60, ET, with MP40 as the lowest. Comparing MP60 with ET and

GesCoN, for most categories there were no significant differences. But MP60 had more extra-

43

large, marketable, and total fruit numbers than ET in the 4th harvest. Comparing MP40 with ET,

for most fruit size categories there were no significant differences. But MP40 had a lower number

of extra-large, marketable, and total tomato than ET in the 3rd harvest.

Table 4. Irrigation strategies effects on fresh-market tomato average fruit numbers per plant and mean fruit weight at each harvest.

Harvest Day

Treatment Fruit (number/plant)

Mean fruit weight (g)

XL L M Cull TM T TM T

78

ET 1.18 0.38 0.18 0.78 b 1.73 2.50 b 254.98 253.35

MP60 1.35 0.38 0.08 0.55 b 1.80 2.35 b 257.85 254.83 MP40 1.47 0.40 0.23 1.65 a 2.10 3.75 a 234.50 222.48

GesCoN 1.50 0.43 0.13 0.75 b 2.05 2.80 b 259.93 240.08 P-value 0.65 0.84 0.35 <.0001 0.53 0.01 0.32 0.07

90

ET 3.30 a 0.10 0.08 1.83 3.48 5.30 340.35 346.70

MP60 3.45 a 0.20 0.00 1.33 3.65 4.98 349.95 351.28 MP40 2.50 b 0.18 0.15 2.47 2.83 5.30 301.75 302.90

GesCoN 3.20 a 0.10 0.00 1.58 3.30 4.88 352.30 357.28 P-value 0.01 0.72 0.13 0.10 0.12 0.68 0.39 0.12

103

ET 5.78 b 0.45 bc 0.10 ab 3.88 a 6.33 b 10.20 a 281.88 289.98

MP60 6.78 ab 0.88 ab 0.23 a 4.10 a 7.88 ab 11.98 a 261.73 270.00 MP40 3.93 c 0.30 c 0.00 b 3.18 ab 4.23 c 7.40 b 271.35 284.38

GesCoN 7.50 a 1.00 a 0.20 a 2.35 b 8.70 a 11.05 a 273.60 276.85 P-value 0.002 0.03 0.04 0.03 0.001 0.002 0.65 0.63

113

ET 1.93 b 0.28 0.18 0.98 2.38 b 3.35 b 280.15 286.60

MP60 3.15 a 0.60 0.10 1.25 3.85 a 5.10 a 273.15 273.35

MP40 1.68 b 0.33 0.05 1.30 2.05 b 3.35 b 272.75 275.48

GesCoN 3.08 a 0.43 0.23 1.13 3.73 a 4.85 a 277.28 280.55 P-value 0.03 0.46 0.45 0.75 0.01 0.04 0.87 0.75

125

ET 2.28 1.68 2.60 0.78 6.55 7.33 163.60 164.20

MP60 2.15 1.85 2.70 0.90 6.70 7.60 160.80 158.93

MP40 3.48 2.08 3.05 0.98 8.60 9.58 170.05 174.05

GesCoN 2.48 1.78 2.90 0.85 7.15 8.00 165.20 167.30

P-value 0.45 0.85 0.83 0.74 0.50 0.49 0.82 0.51 XL = Extra-Large, L = Large, M = Medium, Cull = Unmarketable, TM= Total marketable, T = Total Treatments are grouped by LSD t-test if they are significantly different.

44

Table 5. Irrigation strategies effects on fresh-market tomato average cumulative fruit numbers per plant and mean fruit weight until each harvest.

Harvest Day

Treatment Fruit Numbers (number/plant)

Mean fruit weight (g)

XL L M Cull TM T TM T

78

ET 1.18 0.38 0.18 0.78 b 1.73 2.50 b 254.98 253.35

MP60 1.35 0.38 0.08 0.55 b 1.80 2.35 b 257.85 254.83

MP40 1.47 0.40 0.23 1.65 a 2.10 3.75 a 234.50 222.48

GesCoN 1.50 0.43 0.13 0.75 b 2.05 2.80 b 259.93 240.08 P-value 0.65 0.84 0.35 <.0001 0.53 0.01 0.32 0.07

until 90

ET 4.48 0.48 0.25 2.60 b 5.20 7.80 314.15 317.83 a

MP60 4.80 0.58 0.08 1.88 b 5.45 7.33 319.85 319.78 a

MP40 3.98 0.58 0.38 4.13 a 4.93 9.05 270.28 267.65 b

GesCoN 4.70 0.53 0.13 2.33 b 5.35 7.68 319.50 315.95 a P-value 0.13 0.73 0.13 0.004 0.53 0.07 0.10 0.02

until 103

ET 10.25 b 0.93 0.35 6.48 ab 11.53 b 18.00 294.98 302.13

MP60 11.58 ab 1.45 0.30 5.98 b 13.33 ab 19.30 285.38 288.18

MP40 7.90 c 0.88 0.38 7.30 a 9.15 c 16.45 270.68 275.13

GesCoN 12.20 a 1.53 0.33 4.68 c 14.05 a 18.73 290.60 292.83 P-value 0.001 0.11 0.96 0.002 0.001 0.07 0.50 0.22

until 113

ET 12.18 b 1.20 0.53 7.45 b 13.90 b 21.35 b 292.88 299.63

MP60 14.73 a 2.05 0.40 7.23 b 17.18 a 24.40 a 282.98 285.30

MP40 9.58 c 1.20 0.43 8.60 a 11.20 c 19.80 b 271.40 275.75

GesCoN 15.28 a 1.95 0.55 5.80 c 17.78 a 23.58 a 288.03 290.88 P-value 0.001 0.18 0.88 0.001 0.001 0.003 0.52 0.28

until 125

ET 14.45 bc 2.88 3.13 8.23 b 20.45 bc 28.68 250.43 a 263.85

MP60 16.88 ab 3.90 3.10 8.13 b 23.88 ab 32.00 247.25 a 254.18

MP40 13.05 c 3.28 3.48 9.58 a 19.80 c 29.38 227.98 b 241.93

GesCoN 17.75 a 3.73 3.45 6.65 c 24.93 a 31.58 251.00 a 258.00

P-value 0.03 0.07 0.75 0.001 0.04 0.19 0.04 0.08 XL = Extra-Large, L = Large, M = Medium, Cull = Unmarketable, TM= Total marketable, T = Total Treatments are grouped by LSD t-test if they are significantly different.

Table 5 shows the average cumulative fruit numbers until each harvest by categories. Comparing

MP60 with ET and GesCoN, there was no significant difference for most categories. But MP60

had more extra-large, marketable, and total fruit numbers than ET until the 4th harvest. Comparing

MP40 with ET, MP40 had a lower number of extra-large fruit compared to ET until the 3rd and 4th

harvest, and a higher number of cull fruit until the 2nd, 4th, and 5th harvest, and a lower number of

45

marketable fruit until the 3rd and 4th harvest. MP60 and GesCoN had 16.8% and 21.9% more total

marketable fruits than ET, respectively. MP40 had 3.2% less total marketable fruits than ET. the

results of t-test groups of marketable fruits numbers are the same as the marketable yield. In

general, the variation in fruit numbers is smaller than yield. According to the results of MP60, the

sensor-based irrigation is suitable for practical irrigation applications.

4.6 Fruit Quality

4.6.1 Mean fruit weight

Mean fruit weight in each harvest and cumulative mean fruit weight until each harvest is shown in

Table 4 and 5. For each harvest, there was no significant difference among the treatments. The

treatments also did not have a significant difference until each harvest for marketable and total

mean fruit weight. But MP40 had a lower total mean fruit weight than other treatments until the

2nd harvest, and lower marketable mean fruit weight than other treatments until the last harvest.

MP60 and MP40 had 1.3% and 9.0% lower marketable mean fruit weight than ET. GesCoN had

0.2% higher marketable mean fruit weight than ET. There was no significant difference among

ET, MP60, and GesCoN. But MP40 had a significantly lower mean fruit weight. Thus, lack of

water decreased the mean fruit weight. However, better irrigation scheduling may not increase the

mean fruit weight.

4.6.2 Total soluble solids (TSS) and dry matter

Table 6 shows the fruit TSS and dry matter measured in correspondence to the 2nd and 3rd harvest.

In the 2nd harvest, MP40 fruit had significantly higher dry matter content than ET and MP60 fruit

because of the water deficit observed in MP40. There was no significant difference in the dry

46

matter among the other three treatments. In the 3rd harvest, there was no significant difference

among the treatments. There was no significant difference of the TSS among the treatments in

both the 2nd and 3rd harvest.

Compared with yield and fruit numbers, irrigation scheduling has less effects on fruit quality. But

lack of water can still affect the fruit quality.

Table 6. Fruit total soluble solids (TSS) and dry matter in the 2nd and 3rd harvest.

Treatment

TSS (°Brix)

Dry matter

TSS (°Brix)

Dry matter

(g/100 g FW) (g/100 g FW)

2nd harvest 3rd harvest

ET 4.9 3.6 b 4.7 2.6 MP60 5.1 3.6 b 4.7 2.5 MP40 5.1 4.5 a 4.7 2.7

GesCoN 4.9 4.2 ab 4.4 2.6 p-value 0.36 0.04 0.17 0.42

4.7 Tomatoes plant dry biomass

Tomatoes plant dry biomass was measured in correspondence to the 4th harvest. Table 7 shows the

plant dry biomass of leaves, stems, fruit, and total. For fruit and total, there is no significant

difference among the four treatments. For leaves and stems, MP60 was not significantly different

from ET and GesCoN. But MP40 had lower dry biomass than MP60 and GesCoN. Water deficit

decreased the plant dry biomass. As the results of MP60, the sensor-based irrigation scheduling

with IoT system is available considering the plant dry biomass.

47

Table 7. Tomatoes plant dry biomass of leaves, stems, fruits, and total in 4th harvest.

Treatment Plant Dry Biomass (g/plant)

Leaves Stems Fruits Total

ET 189.75 ab 82.88 ab 184.44 457.06 MP60 194.50 a 87.25 a 160.23 441.98 MP40 166.94 b 75.38 b 151.99 394.30

GesCoN 204.56 a 89.44 a 198.96 492.96

P-value 0.02 0.05 0.19 0.06

4.8 Irrigation water use efficiency (iWUE)

Table 8 shows the total water usage per area and iWUE of four treatments. Treatment MP40 used

much less irrigation water compared with other treatments, leading to the low yield. ET, MP60,

and GesCoN had similar irrigation volume. MP60, MP40, and GesCoN had 19.2%, 25.7%, and

27.7% higher iWUE than ET. MP60 was not different from the other treatments. However, MP40

and GesCoN had significantly higher iWUE than ET. Considering the lack of yield in MP40, its

high iWUE is not an advantage.

Table 8. Total irrigation water usage per area and irrigation water use efficiency (iWUE).

Treatment Volume (m3 ha-1) iWUE (kg m-3) ET 2440 22.22 b

MP60 2357 26.49 ab MP40 1695 27.94 a

GesCoN 2339 28.38 a

4.9 Future Improvement

In this experiment, the irrigation was controlled manually through the IoT platform. The signal

loss of the system and no response of valves were problems to apply automatic irrigation control.

48

In the future, a battery with larger voltage will be used and automatic irrigation rules will be

written in the IoT platform or in the data logger to achieve fully automatic irrigation.

In general, sensor-based irrigation (MP60) and Decision support system (DSS)-based irrigation

(GesCoN) had better performance than ET-based irrigation. However, they both had advantages

and disadvantages. To use sensor-based irrigation, adequate number of sensors need to be

installed in the field dispersedly. If part of sensors does not represent the average condition of the

field, like the condition in MP40, the number of sensors can mitigate the effects of these sensors.

Sensors directly measure the moisture in the field. Thus, once the sensors have representative

readings, the sensor-based irrigation is effective. To use the Decision support system, a model

needs to be built and calibrated for specific plants. If the model is not accurate, the irrigation

cannot have ideal performance. It also needs a weather station or public local weather report near

the field to get accurate weather data. It does not need sensors to apply irrigation. In the future,

the two method can be combined. The sensor readings can be used as a double check for the

DSS-based irrigation. Besides, the sensor readings can be used as a parameter in the DSS model

to improve accuracy.

49

5 Conclusion

An IoT-based precision irrigation system with LoRaWAN technology was developed and

evaluated in the study. In general, the system worked well for the irrigation application in an

open tomato field. The batteries for the data loggers were charged sufficiently by the attached

solar panels. A series of data were recorded and displayed for the soil matric potential sensors

and pressure sensors. Irrigation was applied successfully by controlling the solenoid valves based

on the soil moisture level. Even though no significant negative effect was observed in the sensor

data monitoring and valve control, the IoT system had a data loss rate of 5.5% in average, which

may be attributed to the indoor gateway placement and disconnection of each component in the

network. High frequency of data communication (every minute) mitigates this problem.

Meanwhile, it was observed a couple of times that the valves did not respond with the first sent

signal, which may be caused by the low voltage power used for the valves. The fruit yield, fruit

numbers, quality, plant dry biomass, and irrigation water use efficiency were analyzed and

compared for all four treatments. MP40 used much less irrigation volume than other treatments

because of improper installation position. Other three treatments used similar irrigation volumes.

MP40 has less yield, fruit number, and worse fruit quality. For the other three treatments,

considering the yield, fruit numbers, and iWUE, GesCoN was the best, followed by MP60 and

ET. However, there was no significant difference between MP60 and other two treatments. For

plant dry biomass, mean fruit numbers, total soluble solids, and dry matter, the three treatments

are relatively similar. According to the results of MP60, the developed IoT-based system using

LoRaWAN technology can be potentially used for precision and automatic irrigation application

for practical vegetable production.

50

Bibliography

Abioye, E. A., Abidin, M. S. Z., Mahmud, M. S. A., Buyamin, S., Ishak, M. H. I., Rahman, M.

K. I. A., … Ramli, M. S. A. (2020). A review on monitoring and advanced control

strategies for precision irrigation. Computers and Electronics in Agriculture, 173(April),

105441.

Ahmed, V., Ladhake, S. A. (2011). Innovative cost effective approach for cell phone based

remote controlled embedded system for irrigation. Proceedings - 2011 International

Conference on Communication Systems and Network Technologies, CSNT 2011, 419–

422.

Bianchi, A., Masseroni, D., Thalheimer, M., de Medici, L. O., Facchi, A. (2017). Field irrigation

management through soil water potential measurements: A review. Italian Journal of

Agrometeorology, 22(2), 25–38.

Cancela, J. J., Fandiño, M., Rey, B. J., Martínez, E. M. (2015). Automatic irrigation system

based on dual crop coefficient, soil and plant water status for Vitis vinifera (cv Godello

and cv Mencía). Agricultural Water Management, 151, 52–63.

Casadesús, J., Mata, M., Marsal, J., Girona, J. (2012). A general algorithm for automated

scheduling of drip irrigation in tree crops. Computers and Electronics in Agriculture, 83,

11–20.

Coates, R. W., Delwiche, M. J. (2009). Wireless mesh network for irrigation control and sensing.

Transactions of the ASABE, 52(3), 971-981.

Conaty, W. C., Burke, J. J., Mahan, J. R., Neilsen, J. E., Sutton, B. G. (2012). Determining the

optimum plant temperature of cotton physiology and yield to improve plant-based

irrigation scheduling. Crop Science, 52(4), 1828–1836.

Conversa, G., Bonasia, A., Di Gioia, F., Elia, A. (2015). A decision support system (GesCoN)

for managing fertigation in vegetable crops. Part II—model calibration and validation

under different environmental growing conditions on field grown tomato. Frontiers in

Plant Science, 6(JULY), 1–14.

51

Di Gioia, F., Ozores-Hampton, M., Hong, J., Kokalis-Burelle, N., Albano, J., Zhao, X., …

Thomas, J. (2016). The effects of anaerobic soil disinfestation on weed and nematode

control, fruit yield, and quality of florida fresh-market tomato. HortScience, 51(6), 703–

711.

Di Gioia, F., Rosskopf, E. N., Leonardi, C., Giuffrida, F. (2018). Effects of application timing of

saline irrigation water on broccoli production and quality. Agricultural Water

Management, 203, 97-104

Dubey, V., Dubey, N., Chouhan, S. S. (2011). Wireless sensor network based remote irrigation

control system and automation using DTMF code. Proc. - 2011 Int. Conf. on

Communication Systems and Network Technologies, CSNT 2011, 34–37.

Elia, A., Conversa, G. (2015). A decision support system (GesCoN) for managing fertigation in

open field vegetable crops. Part I—methodological approach and description of the

software. Frontiers in Plant Science, 6(MAY), 1–18.

FAO. (1998). Crop evapotranspiration - Guidelines for computing crop water requirements -

FAO Irrigation and drainage paper 56.

Fernández, J. E., Diaz-Espejo, A., Cuevas, M. V., Montaño, J. C., Romero, R., & Muriel, J. L.

(2006). A device for scheduling irrigation in fruit tree orchards from sap flow readings. In

V Int. Symp. on Irrigation of Horticultural Crops, 792, 283-290.

Francesco Di Gioia, Eric Simonne, Diane Jarry, Michael Dukes, Robert Hochmuth, D. S. (2009).

Real-time drip-irrigation scheduling of watermelon grown with plasticulture. In Proc. of

Florida State Horticultural Society, 122, 212–217.

Gallardo, M., Elia, A., Thompson, R. B. (2020). Decision support systems and models for aiding

irrigation and nutrient management of vegetable crops. Agricultural Water Management,

240(February), 106-209.

Gautam, I., Reddy, S. R. (2012). Innovative GSM Bluetooth based remote controlled embedded

system for irrigation. International Journal of Computer Applications, 47(13), 1–7.

52

Giuliani, M. M., Gatta, G., Nardella, E., Tarantino, E. (2016). Water saving strategies assessment

on processing tomato cultivated in Mediterranean region. Italian Journal of Agronomy,

11(1), 69–76.

Goap, A., Sharma, D., Shukla, A. K., Rama Krishna, C. (2018). An IoT based smart irrigation

management system using Machine learning and open source technologies. Computers

and Electronics in Agriculture, 155(September), 41–49.

Gutierrez, J., Villa-Medina, J. F., Nieto-Garibay, A., Porta-Gandara, M. A. (2014). Automated

irrigation system using a wireless sensor network and GPRS module. IEEE Transactions

on Instrumentation and Measurement, 63(1), 166–176.

Howell, T. A. (2001). Enhancing water use efficiency in irrigated agriculture. Agronomy

Journal, 93(2), 281–289.

Irrometer. (2021). Soil water basics. Available at https://www.irrometer.com/basics.html.

Accessed on May 20, 2021.

Kamienski, C., Soininen, J.-P., Taumberger, M., Dantas, R., Toscano, A., Salmon Cinotti, T., …

Torre Neto, A. (2019). Smart water management platform: IoT-based precision irrigation

for agriculture. Sensors, 19(2), 276.

Kim, Y., Evans, R. G., Iversen, W. M. (2008). Remote sensing and control of an irrigation

system using a distributed wireless sensor network. IEEE Transactions on

Instrumentation and Measurement, 57(7), 1379–1387.

LoRa Alliance. (2015). LoRaWAN R1.0 open standard released for the IoT. Available at

https://www.businesswire.com/news/home/20150616006550/en/LoRaWAN-R1.0-Open-

Standard-Released-IoT. Accessed on April 1, 2021.

Nawandar, N. K., Satpute, V. R. (2019). IoT based low cost and intelligent module for smart

irrigation system. Computers and Electronics in Agriculture, 162(May), 979–990.

O’Shaughnessy, S. A., Evett, S. R., Colaizzi, P. D., Howell, T. A. (2012). A crop water stress

index and time threshold for automatic irrigation scheduling of grain sorghum.

Agricultural Water Management, 107, 122–132.

53

Osroosh, Y., Peters, R. T., Campbell, C. S., Zhang, Q. (2016). Comparison of irrigation

automation algorithms for drip-irrigated apple trees. Computers and Electronics in

Agriculture, 128, 87–99.

Osroosh, Y., Troy Peters, R., Campbell, C. S., Zhang, Q. (2015). Automatic irrigation scheduling

of apple trees using theoretical crop water stress index with an innovative dynamic

threshold. Computers and Electronics in Agriculture, 118, 193–203.

Pardossi, A., Incrocci, L. (2011). Traditional and new approaches to irrigation scheduling in

vegetable crops. HortTechnology, 21(3), 309–313.

Poh, B. L., Gazula, A., Simonne, E. H., Di Gioia, F., Hochmuth, R. C., Alligood, M. R. (2011).

Use of reduced irrigation operating pressure in irrigation scheduling. I. Effect of

operating pressure, irrigation rate, and nitrogen rate on drip-irrigated fresh-market tomato

nutritional status and yields: Implications on irrigation and fertilization ma.

HortTechnology, 21(1), 14-21.

Prathyusha, K., & Suman, M. C. (2012). Design of embedded systems for the automation of drip

irrigation. International Journal of Application or Innovation in Engineering &

Management, 1(2), 254-258.

Ramya, V., Palaniappan, B., George, B. (2012). Embedded system for automatic irrigation of

cardamom field using Xbee-PRO technology. International Journal of Computer

Applications, 53(14), 36–43.

Romero, R., Muriel, J. L., García, I., Muñoz de la Peña, D. (2012). Research on automatic

irrigation control: State of the art and recent results. Agricultural Water Management,

114, 59–66.

Sacaleanu, D. I., Kiss, I. K. (2018). Monitoring walnut orchards with LoRa technology. In 2018

IEEE 24th Int. Symp. for Design and Technology in Electronic Packaging (SIITME) (pp.

76–79). IEEE.

Seidel, S. J., Werisch, S., Schütze, N., Laber, H. (2017). Impact of irrigation on plant growth and

development of white cabbage. Agricultural Water Management, 187, 99–111.

54

Shock, C.C., Pereira, A.B., Hanson, B.R. and Cahn, M.D. (2007). Vegetable irrigation. Irrigation

of agricultural crops, 30, 535-606.

Smith, R. J., Baillie, J. N. (2009). Defining precision irrigation: A new approach to irrigation

management. Irrigation and Drainage Conf., 1–6.

Sudarmaji, A., Sahirman, S., Saparso, Ramadhani, Y. (2019). Time based automatic system of

drip and sprinkler irrigation for horticulture cultivation on coastal area. IOP Conf. Series:

Earth and Environmental Science, 250(1), 012-074.

Sumeetha, S., & Sharmila, D. (2012). Embedded based remote control application using mobile

phone in irrigation. International Journal of Power Control Signal and Computation,

3(1), 76-79.

Tzounis, A., Katsoulas, N., Bartzanas, T., Kittas, C. (2017). Internet of Things in agriculture,

recent advances and future challenges. Biosystems Engineering, 164, 31–48.

USDA. (1991). United States standards for grades of fresh tomatoes. United States Department

of Agriculture, 1991(January), 1–13.

USDA-ERS. (2019). Irrigation and water use. Available at

https://www.ers.usda.gov/topics/farm-practices-management/irrigation-water-use/.

Accessed on April 1, 2021.

Vellidis, G., Tucker, M., Perry, C., Kvien, C., Bednarz, C. (2008). A real-time wireless smart

sensor array for scheduling irrigation. Computers and Electronics in Agriculture, 61(1),

44–50.

Yunseop Kim, Evans, R. G., Iversen, W. M. (2008). Remote sensing and control of an irrigation

system using a distributed wireless sensor network. IEEE Transactions on

Instrumentation and Measurement, 57(7), 1379–1387.

Zhang, N., Wang, M., Wang, N. (2002). Precision agriculture - A worldwide overview.

Computers and Electronics in Agriculture, 36(2–3), 113–132.

55

Zhao, W., Lin, S., Han, J., Xu, R., Hou, L. (2017). Design and implementation of smart irrigation

system based on LoRa. In 2017 IEEE Globecom Workshops (GC Wkshps) (Vol. 2018-

Janua, pp. 1–6).

Zotarelli, L., Scholberg, J. M., Dukes, M. D., Muñoz-Carpena, R., Icerman, J. (2009). Tomato

yield, biomass accumulation, root distribution and irrigation water use efficiency on a

sandy soil, as affected by nitrogen rate and irrigation scheduling. Agricultural Water

Management, 96(1), 23–34.