Synthesis Paper for “Piloting Automated Sensor Applications for Improving Rice Water Use Efficiency Project (P161216)” Contents Overview ....................................................................................................................................................... 2 Summary of Sensors used in Agriculture ...................................................................................................... 3 Water-level Sensors .................................................................................................................................. 6 Basic Concepts on Feasibility Study .............................................................................................................. 9 Basics of Internet of Things (IoT) ................................................................................................................ 11 Fundamentals of IoT Network Connectivity: Design Space Approach ................................................... 12 Wireless Connectivity Technologies ....................................................................................................... 14 Wireless Personal Area Networks (WPAN) ......................................................................................... 16 Wireless Local Area Networks (WLAN) ............................................................................................... 17 Wireless Wide Area Networks (WWAN) ............................................................................................. 17 Mobile Connectivity Expansion............................................................................................................... 22 Synthesis of lessons learned from the sensor application ......................................................................... 22 Challenges of Implementing IoT for Agriculture in Developing Countries ............................................. 22 Practical Challenges from Local Start-ups in Developing Countries ....................................................... 25 TechAguru ........................................................................................................................................... 25 MimosaTEK ......................................................................................................................................... 26 Lessons for Future Work and Scale-up ....................................................................................................... 26 1 Overview Today’s global agriculture faces unprecedented challenges driven by rapid population growth, dietary change, resources constraints, climate change, to name a few. One of the ramifications of these challenges is that the world will need to produce more with less. Indeed, the Food and Agriculture Organisation (FAO) estimates that the global food production will need to rise by 70% to meet with projected demand by 20501. These challenges, in turn, imply the urgent need for the efficient management and the optimized use of natural resources such as water and soil, and other inputs like fertilizers, pesticides. However, managing these inputs efficiently – let alone optimizing their usage – is difficult without consistent and precise monitoring. Undoubtedly, for smallholder farmers, who account for 4/5 of agricultural production from developing regions2, the production gains could be monumental as many farmers in developing regions still rely on personal experiences, rather than data for their farming decisions. Internet of Things (IoT) holds potential to help feed the world’s growing population. IoT refers to networks of objects – or things – that communicate with other things based on the wireless connectivity. Simply put, IoT not only enables the inter-communication between the objects, such as sensors and actuators, but also provides decision analytics based on the precise data accumulated by these devices. In fact, this concept of improving agricultural decisions based on the information and data gathered by sensors is not new. Sensors and wireless sensor networks (WSNs) have long contributed to agriculture, particularly by promoting precision agriculture in crop production and management3. Precision agriculture (PA) is defined as a holistic strategy in which farmers can vary input use and cultivation methods to match varying soil and crop conditions across a field4, enabling farmers "to do the right thing, at the right time, in the right place, in the right way.5" In this sense, sensors are intrinsic part of PA as they facilitate and enable its effects. Indeed, the use of sensors and wireless sensor networks made possible the precise monitoring and management even in the areas in which controlling variables was considered challenging, such as arable farming and livestock management. The integration of sensor data and device with various enterprise applications makes IoT solutions offer much more than simply connecting sensors and devices 6 . Unlike the simple machine-to-machine (M2M), which relies on direct point-to-point communication, IoT solutions enable multiple devices to access data through the cloud system where data is stored. This ability to access data from multiple end-devices, in turn, provides additional benefits of signaling, or even automating, desired actions to be made based on the information collected. Furthermore, with the Internet, you can access to the connected devices and take actions whenever and wherever you are because the data stored in the cloud system is readily accessible through the Internet. This is where the potential of IoT is realized as sensors can now be more easily connected to other devices, particularly mobile phones, through the Internet. Thus, the sensors-based IoT can help farmers by providing the practical knowledge of optimal input levels required for their field conditions as well as by making an overall farming process more efficient and effective. While the up-to-date IoT application in agriculture has been visible mostly in developed countries, it also holds prospects of increasing agricultural productivity in 2 developing countries thanks to the increasing presence of two enabling elements. First, the costs of sensors, or embedded devices, have rapidly declined over the past decade – as much as 100 times lower7 – contributing to the potential growth in IoT application in agriculture in emerging economies. Furthermore, the global mobile broadband coverage has rapidly increased in the past 15 years: the proportion of the global population covered by 2G mobile-cellular network grew from 58% in 2001 to 95% in 20158, by 3G network from 45% in 2011 to 69% in 20159. Thus, the combination of innovative technologies and high mobile broadband penetration rate serves as the basis for developing countries to enjoy the benefits of IoT. Against this background, this synthesis report on the use of wireless sensors and IoT system in agriculture is prepared as part of the ongoing “Piloting Automated Sensor Applications for Improving Rice Water Use Efficiency project” in the Agriculture Global Practice. While not exhaustive, the report aims to provide a review on the practical application of wireless sensors in agriculture and identify the environment in which the effects of the wireless sensor networks and IoT system can be best utilized. While each innovative application of sensors across the whole agricultural value chain is worth noting, the report aims to survey wireless sensors and sensor networks in crop farming in open fields, with a focus on water-level sensors. This focus on water- level sensors is to increase the linkage with the feasibility study of deploying sensors and wireless networks to promote the use of alternate wetting and drying (AWD) irrigation method currently in progress in Vietnam and India, the main pillar of the project. Also, the report surveys the basic components that comprise of the IoT network system, particularly on connectivity. Understanding wireless connectivity solutions will shed light on how the global expansion of the mobile broadband networks and mobile phone usage can spur the benefits of IoT to the farmers in developing regions. Summary of Sensors used in Agriculture The application of sensors in agriculture is widely ranged, from collection of weather and crop information to differentiated usage of fertilizer and water. As wide as the range of the sensors applied to agriculture, the technical design of these sensors also varies according to their purpose of use and application context. Some of the most common types of the sensors used in agriculture include capacitance, ultrasonic, infrared, and tensiometric sensors. The non- exhaustive list of sensor types used in agriculture is provided with a brief description and sample usage in Table 1. Table 1. List of Sensor Types Used in Agriculture Technical Types Sensor Description Sample Usage Dynamic measurements to estimate Measure distance to a target by illuminating fruit-tree leaf area; Combined with Lidar sensors that target with a laser light GPS have been applied for 3D map generation in vine plantations 3 Fourier Transform Obtain an infrared spectrum of absorption or Assessment of structural differences Infrared emission of a solid, liquid or gas of celluloses of various origins Spectroscopy (FTIR) Work as a capacitor coupled with a capacitance Soft water level- Hydrological behavior in agriculture to frequency converter and measures water based sensors catchments level at an adjustable time step acquisition Use capacitance to measure the dielectric Capacitance probes Soil moisture permittivity of a surrounding medium Reflectometers Measure the transit time of waves along a (Frequency Domain probe in the soil based on the properties of Soil moisture content Reflectometry, FDR) electromagnetic waves10 Determine the distance to a target by emitting Ultrasonic ranging an ultrasonic pulse and measuring the elapsed Analyzing apple tree canopies sensors time between the emission and return of the pulse as an echo reflected from the target11 Produce an electrical signal proportional to the Optoelectronic amount of light incident on its active area and Weed detection in wide row crops to sensors respond to light to recognize patterns, images, detect accuracy and feasibility motion, intensity, and color12 Measurements of variables in the soil pH soil-based Measure soil pH level oriented toward crop sensors productivity Infrared Detect infrared by receiving heat energy Measurement of the leaf thermocouple radiated from the objects and measure surface temperature of forests or agricultural (IRt/c) temperatures of materials without contact13 plants Optical sensors observe visible lights and infrared rays (near infrared, intermediate infrared, thermal infrared); Microwave sensors Optical and receive microwaves, which is longer wavelength Characterizing olive grove canopies microwave sensors than visible light and infrared rays, and observation is not affected by day, night or weather14 Source: Pajares, G., Advances in Sensors Applied to Agriculture and Forestry, MDPI, 2011 The application of the sensors in precision agriculture can be further categorized by the agricultural application purposes, such as irrigation, fertilization, and pest control, or by the target inputs, such as temperature, humidity, moisture, that the sensors are used to measure. In addition, the types of sensors can be also broken down into the technical functions that they serve – environmental, chemical, mechanical, acoustic, ultrasonic, electric, optical, computer vision systems, biological, Micro-Electro-Mechanical (MEMS), radio-frequency identification (RFID)15 – or the measurement platforms to which sensors are attached or mounted, including remote platforms such as unmanned aerial vehicles (UAVs) like drones, mobile proximal ones such as agricultural machines and robots, and in-situ proximal, fixed sensors16. 4 The existing literature17,18 categorizes the WSN application in agriculture into three main areas, soil, ambient, and plant. A few types of sensors are applied and tested in measuring soil properties: electromagnetic, optical, mechanical, electrochemical, airflow, and acoustic19. A brief explanation for each sensor type is summarized below: • Electromagnetic Sensors use electric circuits to measure the capability for soil particles to conduct or accumulate electrical charge by making soil part of an electromagnetic circuit. Electromagnetic soil properties are mostly affected by soil texture, salinity, organic matters, and moisture content; • Optical sensors use light reflectance to characterize soil, simulating the human eye as well as measuring near-infrared, mid-infrared, or polarized light reflectance. They are usually mounted to vehicles or airborne platforms for remote sensing; • Mechanical Sensors can be used to estimate soil mechanical resistance such as compacting by using a mechanism that penetrates or cuts through the soil and recording the force measured by strain gauges or load cells; • Electrochemical Sensors can measure soil nutrient levels and pH, an essential type of information needed for precision agriculture, whereby electrodes detect the activity of specific ions in the soil; • Airflow sensors measure soil air permeability on the go. The pressure required to squeeze a given volume of air into the soil at fixed depth could be compared to several soil properties; • Acoustic sensors can determine soil texture by measuring the change in noise level based on the interaction of a tool with soil particles Table 2. Summary of In-field Sensor Applications in Agriculture Sensor Usage Sensor Application Types of Sensors Soil moisture Rain/water flow Soil Water level Soil temperature Electromagnetic Conductivity Optical Salinity Mechanical Electrochemical Humidity Airflow Ambient temperature Acoustic Atmospheric pressure Ambient Wind speed Wind direction Rain fall 5 Solar radiation Moisture Temperature Hydrogen Plant Wetness CO2 Photosynthesis Source: Wireless sensor networks for agriculture: The state-of-the-art in practice and future challenges. Tamoghna Ojha, Sudip Misra a, Narendra Singh Raghuwanshi; Soil and Crop Sensing. Institute of Agriculture and Natural Resources. http://cropwatch.unl.edu/ssm/sensing Water-level Sensors This report primarily focuses to survey the liquid- and water-level sensors used in agriculture to contribute to the successful design and implementation of the feasibility study, which is discussed later in the report. For water level monitoring and measurement, commonly used sensors can be categorized by four main measuring features: pressure, supersonic waves, heat, and image20. The type of water-level sensors is further divided by the sensor’s contact or non - contact characteristic with water. The contact type sensors include pressure type, capacitance type, shaft encoders (also known as rotary encoder), and bubblers (hydrostatic pressure type), and the non-contact type includes ultrasonic, radar, and mobile canal control sensors21. Table 3. Water-level Sensors by Measuring Features Measuring Description Pros Cons Sensor Types Features Measure quantify of water Pressure and directly translate to Need to be calibrated and sensors, water level through force replaced frequently due to hydrostatic Pressure per unit area represented Easy to use potential breakdown of pressure by product of mass and sensors by continuous sensors, staff acceleration of gravity of water pressure gages water Measure the time of travel Free from water Inaccurate returning Ultrasonic, of supersonic wave pulse pressure; Lifespan values when temperature radar, Supersonic from emitter to receiver is temporary as it fluctuates between high electromagnetic reflected by the water does not contact and low, or when water sensors surface water directly fluctuates rapidly Measure the amount of Need to be well insulated temperature drop of a and the temperature of solid, which is proportional the solid should be to its contact area with Lifespan is controlled near the Heat fluid, and translate the temporary measurement area; heat temperature change to the type sensors are used for depth information of the restricted applications solid in water such as the measurement 6 of cooling water level in nuclear plants Costly compared to other types; Need a large and secured space to install Measured data Provide the surrounding data, and capacity to can be confirmed information around the transmit large data size; Mobile canal Image by the sensor as well as the water Measurement accuracy is control surrounding level affected by lighting information conditions and minimum light is required during night time Source: Yu and Hahn, Remote Detection and Monitoring of a Water Level Using Narrow Band Channel, 2010 & Nakka, Sensors for Agriculture and Water Use Efficiency, http://www.slideshare.net/saibhaskar/sensors- for-agriculture-and-water-use-efficiency, 2016 There is an extensive range of experiments, pilot projects and initiatives that have been conducted in developing in-situ sensors in paddy fields. The available academic literature confirms the effectiveness of sensor deployment in enhancing rice cultivation by allowing more precise measurement of critical factors in paddy fields such as water level, water salinity and soil moisture. While some projects have specific project objective, others have deployed a more comprehensive network system to improve rice crop monitoring. However, the current state of research remains largely experimental or pilot-level, which calls for scaling up of proved sensor technologies. Also, many of these projects deployed sensors in controlled settings. Below table 4 is the list of available academic literature on sensor application in rice fields. Table 4. Selected List of Academic Research for Water-level Sensors Use in Rice Fields Projects Objective Sensor Description Developed a wireless sensor network To develop an automatic salinity comprised of low-cost electrical Smart on-line water salinity sensor system in response to conductivity (EC) measurement nodes measurement network to increasing salinity and excessive and an autonomous power supply based manage and protect rice usage of irrigation water caused on energy harvesting, that is capable of fields (SMART-PADDY) by using coastal groundwater for rice transmitting readings of water salinity in European Commission22 farming paddy fields in real-time to a central server 7 For on-farm water monitoring the ultrasonic sensors was used with RBC Pilot initiative of To obtain real time information and Flume for water inflows and outflows; WALAMTARI by ClimaAdapt establish decision support system For measuring the water in the fields project23 to improve water use efficiency and ultrasonic sensors fitted to Bowmen water productivity in rice fields water tube are used; Other parameters measured are temperature and relative humidity24 To design and develop a paddy field moisture content and water depth Integrated soil moisture and sensor, calibrate the sensor with water depth sensor for Wireless, frequency domain soil various soil samples, and test the paddy fields (Xiao et al, moisture and water depth sensors for performance of WFDSS in terms of 2013)25 paddy field measurement range, stability, system error, energy consumption and transmission distance research various areas related to Paddy Field Sensor, Initiative sensor application in paddy fields Water level and temperature sensors of Japanese Government26 such as measurement, (sensor type not specified) communication, installation and cost analysis Development and Electro-mechanical sensor to measure To resolve the soil acidity of the Deployment of Wireless water level, in which the mechanical part paddy fields in Kuttanad, India, and Sensor Network in Paddy floats on water and the electrical part thus to improve rice production in Fields of Kuttanad27 (Simon produces signals based on the portion of the region and Jacob, 2012) the floating device Monitoring the Paddy Crop Humidity, pH sensor, temperature, and Field Using ZigBee Network28 To test fully automated rice field pressure sensors (Sriharsha, Rao, Pravin and monitoring (sensor type not specified) Rajasekhar, 2012) Sensors for (1) prediction of paddy growth stages, (2) water level control A Fully Automated Water corresponding to growth stages, (3) Management System for To develop a system for an estimation of mean water level, (4) Large Rice Paddies29 automatic water management for prediction of water consumption, and (5) (Sekozawa) large-scale paddy fields optimal water allocation for minimizing of damage (sensor type not specified) To develop a system that Application of Sensor Sensor nodes for Barometric Pressure, guarantees a low-cost, high Networks for Monitoring of Ambient Light, Relative Humidity, performance and flexible Rice Plants: A Case Study30 Temperature distributed monitoring system with (Kumar et al.) an increased functionality in rice (sensor type not specified) fields A floating sensing system to To develop a sensing system that Floating soil sensor (FloSSy) as evaluate soil and crop can float, acquire and process electromagnetic induction technique variability within flooded detailed geo-referenced soil does not require physical contact with paddy rice fields31 (Islam et information within flooded rice the soil al., 2011) paddy fields 8 Basic Concepts on Feasibility Study The feasibility study is concurrently conducted in two pilot sites, Vietnam and India, to deploy low-cost water-level sensors to promote the use of alternate wetting and drying (AWD) irrigation method in rice fields to increase irrigation efficiency. AWD is a proven technique to optimize irrigation water use in rice, which entails frequent drying of the paddy fields and irrigating once the water level reaches a prescribed limit, depending on the stage of plant growth as depicted in the Figure 1. To provide details on how AWD irrigation method works, it requires farmers to re-flood the field to a depth of 5cm a few days after the disappearance of the ponded water when the water level drops to 15cm below the soil surface. From one week before to a week after flowering, the field should be kept flooded, topping up to a depth of 5cm as needed. After flowering, during grain filling and ripening, the water level can be allowed to drop again to 15cm below the soil surface before re-irrigation32. The AWD can be implemented by using a ‘field water tube’ in Figure 2. It is estimated that AWD can save water usage up to 30% and reduce greenhouse gas (GHG) emissions, particularly methane by up to 48% while maintaining yields33. Figure 1. Graphics for AWD Schedule Figure 2. Images of AWD Field Water Tubes 9 The successful deployment of AWD method requires accurate estimation of water-level in the ground so farmers use that to make irrigation decision making. However, manual measurement of water, often using measuring tapes, is challenging, inaccurate, and labor- intensive. Against this background, the feasibility study aims to develop sensor-based measurement to increase the potential utilization of AWD, and therefore contribute to significant water-saving efforts. The proposed model will use (a) water-level sensors to measure standing water-level in rice crop and subsequently to determine irrigation schedule; (b) mobile-based pump control system and a decision support system to implement irrigation scheduling and monitor the pump operation remotely; and (c) monitoring system for the volume of water extracted along with energy consumption. At the time of this report, the information on the sensors and IoT system design of the feasibility study taking place in India became available. To fully capture the benefits enabled by the optimization of water use, such as saving ground water, lowering carbon footprint and methane emission, and reducing irrigation costs, various sensors are considered for the pilot study: soil moisture sensors, standing water level sensors, rain gauge, water flow meter, ground water measurement sensors, and temperature & humidity sensors. The application purpose of each sensor category is explained below with brief descriptions of different technical types considered for the feasibility study as shown in Table 5. • Soil moisture sensors measure the moisture level of soil during AWD • Standing water-level sensors measure the water level in the field during ponding period and alternate wetting and drying period • Rain gauge measures the quantity of rainfall • Water flow meter measures the volume of water pumped out • Ground-water measurement sensors measure the distance of water presence in borewell from the ground level • Temperature and humidity sensors measure the temperature and humidity of ambience/atmosphere in the field Table 5. Summary of Sensors used and considered in Feasibility Study in India Sensor Types Considered Major Comparable Characteristics Soil Moisture Capacitive Corrosion free Resistive Corrosive Standing Water-level Ultrasonic water-level sensor Contactless; More expansive than pressure- based sensors Hydrostatic Pressure sensors Can be clogged by muddy water in the fields and may not give accurate values Rain Gauge Tipping bucket Simple, reliable, and cost-efficient Water Flow Meter Electromechanical Has a magnetic drive and dry dial for visual readings in the field 10 Water flow sensor Suitable for small diameter medium duty pipelines Clamp-on type water meter Portable and easy to install but expensive; More suitable for one time measurement; Requires calibration every time when the pipe dimensions change Ground water Barometric Sensor Challenges faced when Installing the sensors measurement directly inside pumping tube wells; Acoustic contactless ground water-level sensors will be considered in the future Temperature & Electromechanical Gives ambient temperature, humidity, and Humidity atmospheric pressure directly; Can be integrated to a controller or data logging device directly; gives decent accuracy; needs installation inside a weather shield for outdoor applications Note: Yellow cells indicate the type of sensor selected for the pilot study Basics of Internet of Things (IoT) Thanks to the recent decline in the cost of sensors based on the advances in computing capability and the rapid expansion of the global mobile broadband penetration over the years, IoT can benefit the farmers in developing regions alike. For example, sensors can be deployed – on the ground or on unmanned aerial vehicles (UAV) like drones – to collect data on target inputs such as soil moisture, fertilizer level, and crop health. The data are logged and stored to a gateway, which integrates all connected sensors and devices, to a server or cloud system wirelessly. Then, farmers access the information gathered by the sensors with mobile apps or analytical software in any end-devices such as mobile phones, tablets, and computers (See Figure 3 for the flow of 11 the process). Depending on the context, farmers can choose to manually control connected devices or to automate processes for any required actions. Figure 3. Overview of IoT Architecture (source: https://www.embitel.com/blog/embedded-blog/how- iot-works-an-overview-of-the-technology-architecture-2) When you take a deeper look into how IoT works, as in the above scenario, sensors and embedded devices explain only one part of the equation. Indeed, Internet Society (ISOC) recognizes ubiquitous connectivity, widespread adoption of IP-based networking, computing economics, miniaturization, advanced in data analytics, and the rise of cloud computing as the technologies that enable the wide adoption of IoT in real life 34 . Similarly, ITU identifies that connectivity, object identification, real-time information, smart devices, and advances in miniaturization as key technologies that need to be further developed and deployed at affordable costs. Therefore, comprehending the underlying technical concepts behind IoT is useful to think about the ways in which the global development can benefit from IoT. Specifically, this part of the report focuses on the IoT network connectivity to envisage another enabling ingredient, the expansion of the global mobile broadband coverage, and to introduce opportunities and challenges ahead. Fundamentals of IoT Network Connectivity: Design Space Approach Wireless communication and wireless networks are important element of IoT as they are the “glue” that holds Internet of Things systems together35. Without connectivity of IoT networks, data cannot be collected nor transmitted to end users—or farmers in the context of IoT application in agriculture. As ITU states, the basic idea behind the IoT is the connection of anything, at any time, from any place, as indicated in Figure 4. Recalling the definition of precision agriculture mentioned in the introduction, the goal behind IoT shares much similarity with the “right thing, right time, right place, right way” principles of precision agriculture, 12 Figure 4. Basic Idea behind the IoT: any place, any time, and any thing In the IoT network design space Figure 5), there are three axes in the graph: battery life, data rate (“duty cycle”), and range between device and gateway. Battery life indicates power consumption of device. Depending on the settings, you may opt to use powered devices or the devices that have years of battery lifetime. The second axis is the device’s data rate, which is the speed at which data is transferred within a computer or between computing devices measured in bytes per second 36 . Simply said, it is the number of bytes per day that you would like to communicate between your devices and the servers on the Internet. Your IoT system may use as little as a few bytes every day or as many as gigabytes (GB) per day depending on the applications. The below Table 6 is help you understand the approximate volume of data required for each selected application example 37 . Additionally, it is important to note that these axes are not independent, which makes the IoT connectivity design space even more complicated38. Table 6. Data volume required for selected applications (source: https://www.computerhope.com/issues/chspace.htm) Data volume Comparable Scale Example Bit Binary: 1 or 0 - Byte 8 bits 1 character, e.g. “a” Kilobyte (KB) 1,024 bytes 2 or 3 paragraphs of text Megabyte (MB) 1,024 Kilobytes 4 books (200 pages) Gigabyte (GB) 1,024 Megabytes 341 digital pictures (with 3MB average file size) 13 Figure 5. IoT Network Design Space (Source: https://6s062.github.io/6MOB/2017/materials/lec5-IOTx-WirelessNetworkConnectivity.pdf) Wireless Connectivity Technologies With the IoT network connectivity design in mind, there are many wireless connectivity technologies used for different purposes and applications. An easy way to categorize different wireless networks is by the network range of the connectivity technologies. A network’s range is typically classified into four classes: personal area networks (PANs), local area networks (LANs), neighborhood area networks (NANs) and wide area networks (WANs)39. Figure 6. Different ranges and applications for PANs, LANs, NANs and WANs 14 PANs are usually wireless and cover a range of about 10m. Wireless PAN (WPAN) devices usually have low radio transmission power and run on small batteries —for example, a smartphone connected over Bluetooth to a smart watch or fitness device40. Common WPANs include ZigBee, Bluetooth, and RFID. LANs are either wired or wireless, and Wireless LANs (WLANs) usually cover a range up to 100m. A home Wi-Fi network providing internet access to computers, smartphones, and IoT-enabled home appliances is a common example of WLAN. NANs are usually wireless and can reach more than 25km. Lastly, WANs cover a very large area, as big as the entire globe41. The internet is considered a WAN and comprises a complex mix of wired and wireless connections42. In addition, WWANs include 2G, 3G, 4G/LTE, and low-power wide-area networks (LPWANs) like LoRA 43 . The summary of the characteristics of different wireless telemetric options is available in the Annex 1. Figure 7. Comparing IoT Connectivity Technologies (Source: ITU and Cisco) One thing to note here is that many of the wireless connectivity technologies the report introduces are operating in the industrial, scientific and medical (ISM) band where spectrum use is free or “unlicensed.” However, the licensed frequency band, such as cellular communication, can be also used as one method to connect devices in IoT systems. As we have covered the basic concepts of wireless connectivity and the considerations for designing IoT networks, below are the predominant wireless connectivity technologies in more detail (non-exhaustive). 15 Wireless Personal Area Networks (WPAN) ZigBee ZigBee is an open global standard and is the only standards-based wireless technology 44 designed specifically to be used in machine-to-machine (M2M) networks 45 . It is designed to provide a whole connectivity solution for device interoperability and cloud connectivity46. The technology is inexpensive to run and doesn’t require a lot of power, making it an ideal solution for many industrial applications. Zigbee’s module cost ranges US$1-10 and its data transmission range typically falls between 10-100 meters, power output and environmental characteristics47. Moreover, the technology has a low latency, and a low duty cycle, allowing products to maximize battery life. Specifically, Zigbee has data throughput48 rates of 250kbps, 40kbps, and 20kbps49. Devices using Zigbee are usually low data rate emitting sensors and actuators and vastly minimizing air time and therefore power consumption based on the low-power performance of the underlying 802.15.4 link layer protocol 50 . The technology is also based on mesh networks, which allow nodes to be connected together through multiple pathways. One of Zigbee’s major benefits is that it provides a complete solution that enables true device interoperability between different manufacturers51. The Zigbee protocol suite incorporates the Zigbee cluster library, which refers to a standard library of devices types, data models and behaviors built by original equipment manufacturers (OEMs) operating in different domains and proven in actual deployments for many years52. Zigbee’s ability to connect multiple devices together simultaneously and seamlessly makes it ideal for a connected home environment such as smart locks, lights, robots and thermostats to talk to one another 53 . Service providers like Comcast, Deutsche Telekom and Alibaba endorsed Zigbee as the protocol of choice when introducing their home automation services to consumers, and lighting manufacturers like Philips and Osram have a whole line of Zigbee-based wireless LED products54. Bluetooth Bluetooth is a wireless technology primarily used for short-range communication. The Bluetooth became successful in mobile phones, and is frequently used in small devices, such as fitness accessories or speaker systems, that connect users to smartphones and tablets. The technology operates in the 2.4GHz ISM band based on point-to-point communication design, also known as start network topology55. The Bluetooth technology is fairly low-power and devices typically use small rechargeable batteries or two alkaline batteries 56 . It supports data throughput up to 2Mbps, with up to eight connected devices. If a Wi-Fi connection is available, as the case in a mobile device, Bluetooth can operate over Wi-Fi, known as Bluetooth High Speed, and provide transfer rates of up to 25Mbps. Bluetooth Low Energy (BLE), or Bluetooth Smart, is a more recent addition to the Bluetooth specification that is more suitable for IoT connectivity application. While BLE uses less power than the classic Bluetooth with the trade-off of lower data throughput57, it enables lengthy operation using coin-cell batteries58. BLE also uses 2.4GHz ISM band but is not compatible with Classic Bluetooth 59 . The current standard specification, Bluetooth 5, offers improved the data transmission range of 200m (quadruples the previous specification Bluetooth 4.2), doubles the speed up to 2Mbps, and provide an eightfold increase in data broadcasting capacity by increasing the advertising data length60. The increases in range and data rates make BLE increasingly attractive in the growing segments such as industrial data loggers or smart energy meters61. Bluetooth and BLE has the advantage of built-in compatibility with mobile devices. Its mobile device connectivity offers an excellent avenue for data display and retrieval, Internet connectivity, and initial 16 provisioning and configuration 62 . Another strength of the Classic Bluetooth standard is that includes application profiles such as for audio/video remote control and heart-rate monitors 63 . The recorded profiles help to understand how applications exchange information to achieve specific tasks and include comprehensive Bluetooth Special Interest Group (SIG)-defined certification program to ensure interoperability in the market64. BLE also offers similar functionality by offering generic attribute (GATT) profiles. GATT provides a structured list that defines the services, characteristics, and attributes of a given node, enabling BLE to quickly adapt to new applications65. Wireless Local Area Networks (WLAN) Wi-Fi Wi-Fi is probably the most familiar and common wireless technology. Based on IEEE 802.11 standard, it was developed as a wireless replacement for the popular wired IEEE 802.3 Ethernet standard and natively integrated with the TCP/IP-based protocol for Internet connectivity66. Furthermore, Wi-Fi networks rely on a star topology, where the center node is called an access point (AP) and the other nodes are called stations 67 . In Wi-Fi networks, therefore, AP functions as the Internet gateway. In large buildings, more than one AP is often deployed in different locations to increase network coverage, and some Wi-Fi products have two antennas for network diversity68. Therefore, Wi-Fi would be a suitable choice for IoT devices located in buildings or homes where Wi-Fi networks are already present. The most visible advantage of using Wi-Fi is interoperability as Wi-Fi is already integrated into all new laptops, tablets, smartphones and TVs. Taking advantage of its ubiquity, Wi-Fi is also widely used in IoT applications that can leverage installed Wi-Fi infrastructures without custom gateways 69 . Despite its widespread presence, Wi-Fi is complex in its network design and has relatively higher power consumption. Since most IoT products do not need the maximum data rates that Wi-Fi offers, which has been a major barrier for IoT developers to choose Wi-Fi as IoT connectivity solution. However, to reduce the average power consumption, the latest Wi-Fi silicon devices apply advanced sleep protocols and fast on/off times, enabling Wi-Fi integration into emerging IoT applications and battery-operated devices. Wireless Wide Area Networks (WWAN) Cellular Networks Unlike many of the IoT connectivity technologies, including above-explained ZigBee and Bluetooth, cellular networks operate on licensed spectrum. The licensed spectrum means that users need to buy a license from the local regulator to operate a radio transmitter in a designated frequency channel70. The clear benefit of cellular networks is their widespread coverage across a country and a region. The ubiquity of mobile cellular coverage is also true in developing countries: GSMA estimated 3G coverage surpassed almost two thirds of population in developing economies in 201471. However, as witnessed by the short-range nature of other IoT connectivity solutions, many of the IoT systems do not require the long-range communication and connection, restraining the adoption of cellular networks as the optimal connectivity solution. Furthermore, cellular networks are high power-consuming, thus not ideal for many of IoT applications that need long battery life. Unlike 2G networks, in which the radios didn’t have to be on all the time waiting for messages, also known as store and forward architecture, 3G and 4G adopt always-on approach: the radio is always on to get the signal for data reception for the connected device72. Nevertheless, cellular carriers have been working to standardize several low-power wide area (LPWA)— LTE-M, Narrowband (NB)-IoT, and EC-GSM—to meet requirements for massive IoT deployment 17 characterized by low-cost devices with low energy consumption and good coverage73. Below is the brief description 74 for some of the sought-after cellular network connectivity solutions tailored to IoT applications: • EC (Extended Coverage)-GSM o Compatible with existing GSM network, the wireless protocol by which 80% of smart phone are operated globally o Could leverage 2G base • LTE-M o Compatible with the existing LTE network o Capping the maximum system bandwidth to 1.4MHz o Targeting LPWAN applications like smart metering • NB-IoT o Doesn’t operate in the LTE bands, thus requiring the new infrastructure to deploy NB-IoT o Eliminates the need for a gateway, potentially making the less expensive option Cellular networks could provide other benefits when used as a bridging option 75 , for example, as an aggregation and routing solution. This capillary network approach allows end devices to utilize varying access solutions from either the short range or LPWA domain and access the cellular networks via a gateway choice76. This utilization of existing and new cellular networks for the IoT connectivity solution introduces the potential uptake of IoT in developing countries alike driven by the rapid expansion of mobile connectivity across the globe. Low-Power Wide-Area Networks (LPWAN) Sigfox Sigfox is a proprietary LPWAN technology aimed at IoT applications such as remote sensing that transmit small amounts of data relatively infrequently77. It sets up antennas on towers, like a cell phone company, and receives data transmissions from devices like parking sensors or water meters78. Compared to cellular, Sigfox offers significant cost and battery-life advantages to such applications. It has the lowest cost radio modules under $5, compared to around $10 for LoRa and $12 for NB-IoT79. The Sigfox network provides a very long-range connectivity, up to 30 km in rural areas between the base station and the node using a star network topology. The ability to connect such a long range comes with a very low data transmission rate, up to 100bps using 12 byte packets, and no more than 140 messages per node per day80. Moreover, Sigfox can be used for uplink only, the function of which is limited to sending data from sensors to a gateway81, and has the problem of signal interference82. LoRaWAN LoRaWAN is an open standard network protocol assembled by LoRa Alliance led by IBM, Actility, Semtech, and Microchip. It uses underlying chip, LoRa, to implement a full LoRaWAN stack provided by Semtech83. LoRa’s functionality is similar to SigFox as it is primarily for uplink-only applications with many end-points. Because LoRaWAN spreads out information on different frequency channels and data rates using coded messages, there is less likely interference issue, thereby increasing the capacity of the gateway84. 18 Figure 8. Technologies addressing different IoT segments by coverage and performance requirements (Source: Ericsson) Additionally, understanding different communications models by which IoT devices connect with one another can further help narrow down connectivity options. According to the Internet Architecture Boards (IAB) and the Internet Society (ISOC), there are four common communication models used by IoT devices: device-to-device, device-to-cloud, device-to-gateway, and back-end data sharing85. The device-to-device communication model indicates the IoT devices that directly connect and communicate between on another, rather than through an intermediary application server86. This type of networks allows devices that adhere to a particular communication protocol to communicate and exchange messages to achieve their function87. These devices often use low- power radio connectivity technologies such as Bluetooth, Z-Wave, and Zigbee. Another IoT communication type is a device-to-cloud model, whereby the IoT device connects directly to an Internet cloud service like an application service provider to exchange data and control message traffic88. A common example of this communication model would be popular consumer IoT devices such as the Net Labs Learning Thermostat, in which the device transmits data to a cloud database where the data can be used to analyzed home energy consumption 89 . This cloud connection enables the user to obtain remote access to their thermostat via a smartphone or web interface, and it also supports software updates to the thermostat90. Another model common to IoT networks is the device-to-gateway model, or more typically, the device-to-application-layer gateway (ALG) model91. Simply, ALG service indicates that there 19 is application software operating on a local gateway device, which acts as an intermediary between the device and the cloud service and provides security and other functionality such as data or protocol translation. A common example of this communication model is when the local gateway is a smartphone running an app to communicate with a device and relay data to a cloud service 92 . For example, Fitbit, a popular personal fitness tracker, relies on smartphone app software to serve as an intermediary gateway to connect the fitness device to the cloud, since it does not have the native ability to connect directly to a cloud service93. The other common form of this device-to-gateway model is the emergence of “hub” devices in home automation applications, that serve as a local gateway between individual IoT devices and a cloud service94. The last communication model is the back-end data-sharing model. It is a communication architecture that grants third parties to access the data collected from IoT devices, enabling the data from multiple sources to be aggregated and analyzed95. This model is an extension of the earlier mentioned single device-to-cloud communication model, which can lead to data silos where “IoT devices upload data only to a single application service provider96.” However, ISOC notes that the back-end data-sharing does not fully eliminate data silos as it cannot overcome closed system designs, putting emphasis on the interoperability need in the underlying IoT system designs. Table 7. Examples of IoT Connectivity Models (i) Device-to-Device Communication (ii) Device-to-Cloud Communication 20 (iii) Device-to-Gateway Model (iv) Back-End Data Sharing Model Therefore, choosing a suitable IoT network connectivity technology depends on IoT applications97, based on the IoT network design space and the type of communication models in consideration. In precision agriculture applications, for example, the farm size in which the sensors are located, the distance between the sensors and between the nodes and the gateway, the desired frequency of and data volume required for data transmission, should be understood in advance before selecting an appropriate IoT connectivity solution. The feasibility study showcased earlier in the report used public band wireless radio frequencies for the connectivity between the gateway and sensors while it used cellular-based GSM gateway between the gateway and the cloud. 21 Mobile Connectivity Expansion Combined with the cost decline of the sensors and embedded devices, the increase in the mobile broadband coverage and access to mobile phone is another critical factor behind spurring the application of IoT in agriculture. In particular, the increasing adoption of smartphone connections, about 3.8 billion worldwide at the end of 2016 according to GSMA estimate 98, will further open doors for IoT adoption in agriculture. As one survey indicates, almost half of the respondents already use mobile devices less than their laptop or desktop computers as a main monitoring end-device for precision agriculture work99. Mobile solutions are already highly integrated into an array of agricultural applications in development context as well, resolving the last-mile service delivery to individuals. Mobile phones are used for data collection such as mobile surveys, for transactions purpose such as mobile money payments, and for information exchange such as access to market information and weather data100. The increase in the use of mobile phones and smartphones in developing regions can also function as a convenient tool to monitor the data collected by the sensors and control the devices connected to any IoT systems. Synthesis of lessons learned from the sensor application Challenges of Implementing IoT for Agriculture in Developing Countries Despite the potential role that IoT can serve in agriculture, there are challenges that must be considered before considering the use of wireless sensor networks. The most predominant issue is the lack of connectivity in many of the areas where farmers in developing countries are located. GSMA estimated that 10% of the global population lacked access to basic service and text services, and about a third lacked access to 3G or 4G mobile broadband Internet. The majority of these unconnected populations live in the rural regions of Asia and Sub-Saharan Africa, which account for 3.4 billion of the 4.8 billion people not yet connected to the internet 101. Even though there are IoT connectivity solutions that operate over unlicensed spectrum, the basic cellular connectivity is expected an important role of connecting people to access data collected by the sensors via mobile phones. When deployed in the context of developing countries, sensors and IoT connectivity can convey bigger trade-offs between cost, accuracy and precision, frequency and duration, data retrieval, and power requirements. As implied in IoT network design, major challenges of the agricultural monitoring systems are power management and network lifetime 102 . Because batteries lose power over time and require recharging or replacement, power consumption must be limited to the bare minimum required for effective performance. Managing the sleeping mode of sensor nodes to enable quick re-activation and appropriate sleep timing are very important. However, these problems can be resolved by using solar panels in agriculture fields to recharge the battery and upgrading the power class and antennas to increase the radio range. Existing agricultural monitoring systems are mostly suitable for a small area. 22 The cost of deploying and commercializing IoT system can be still high for farmers in developing countries. Depending on the sensor devices and IoT connectivity solutions of choice, the cost structure and implication can vary. The basic cost elements of building an IoT system include development costs, manufacturing costs, and operational costs103 as provided in Table 8 and Table 9. Table 8. Cost Elements of Building IoT System (Source: Link-Labs) Elements of cost Details Cost Hardware design and Engineering outsourcing: $10-30,000; prototyping Prototype Spin: $100-200/each Industrial design and Engineering outsourcing: $20 – 100,000; mechanical engineering Mold development: $5-20,000 Firmware development Engineering outsourcing: $10-100,000+ Development cost Software/application Engineering outsourcing: $10-100,000+ development Design for manufacturing N/A Testing N/A Certification $3,000-$100,000+ Contract manufacturing, lead Manufacturing times/EOL, unit testing, boxing Varies costs and shipping, warehousing, distribution Demos and pilots, installations, support and troubleshooting, Operational costs Varies returns and repairs, connectivity costs Table 9. IoT connectivity cost (Source: Link-Labs) Connectivity Option Module cost Connectivity cost Infrastructure LTE-M $10-15 $3-5/mo for 1MB NB-IoT $7-12 <$1/mo for 100kb (adequate for water meter but not for asset tracker) Sigfox $5-10 <$1/mo LoRaWAN Public $9-12 $1-2/mo LoRaWAN Private $9-12 $0.25/mo $500 If the IoT system in consideration relies on cellular connectivity, the cost can further increase due to the infrastructure cost and data package plans. This is not just the case for smartphones application as an extra cost would incur even for the sensor networks opting to use basic phone- based mobile messaging services. There are two types of messaging services that are widely used and inherently built within mobile phones: short message service (SMS) and Unstructured Supplementary Service Data (USSD)104. SMS messages usually have a limit of 160 characters in 23 Latin-based alphabets and enable people to communicate via text between phones, or also from a computer to phones via an SMS gateway that provides bulk messaging services. USSD messaging allows phones to communicate directly with their mobile network operator’s computers. Users can access by dialing a number that starts with an * and ends with a # (e.g. *1234#). USSD allows callers to navigate and input information through text-based menu interface. It is often used for messaging that requires multiple responses, as multiple interactions can take place during one USSD session, which is less costly than sending multiple SMS messages to obtain the same information. However, USSD services are more complicated than SMS and require an agreement with a mobile network operator (MNO) or an aggregator, which are third- party service providers that negotiate agreements directly with the MNOs. Other challenges are non-technical as some of them require innovative business models, adequate environmental factors, and human skills development. Karim et al. identifies major design challenges in agricultural monitoring systems include cost, reliability, readiness of information system, crop models, the gap between designers and users, and farmer capacity building105: • Cost – Cost of sensors is the most important factor. A single framework of machine-to- machine (M2M) communication network can be used in multiple applications, which requires densely deployed sensor nodes • Reliability – In most developing countries, sensor nodes are scarce and climates can be extreme. Hence, sensor nodes must be protected from outdoor conditions including moisture and heat. Availability and reliability of telecommunications/wireless infrastructure in rural areas of developing countries is another challenge. • Developing a capable and intelligent information system for developing regions is very important. Otherwise, collecting only raw sensor data would be useless. • Crop models – simulate the growth of crops in different environmental constraints and play an important role in designing decision support systems. Two existing crop models used in agricultural projects, DSSAT and APSIM, consider only the crop growth and yields and do not consider the impact of pests and diseases and the timing of farming operations. Moreover, these models do not consider the recent development of environmental monitoring such as monitoring soil moisture. In such cases, data gathered from sensor- based MTC devices help to build, design and test more effective crop models. In addition, as it is very important to be roughly correct rather than precisely wrong, more environmental factors should be considered in the decision support systems/crop models that the farmers of developing regions require. • The failures of information systems projects in developing countries are mainly due to the gap between the design and actual requirements, that is, the distances between the designers and users. Any participatory design does not guarantee any success until it alleviates the gap between the designers and users. • Another major challenge in technology based agriculture systems in developing countries is the capability of users (farmers) to understand, use and own the system. 24 Practical Challenges from Local Start-ups in Developing Countries Due to the nature of context-specific design, it is equally important to identify practical challenges that emerge in the process of implementing sensor-enabled IoT system in agriculture in developing countries. The informal interviews were conducted as part of synthesizing relevant information to the project. TechAguru TechAguru is an agriculture technology start-up company based in Philippines and Japan. It has conducted research on how to help farmers adopt AWD irrigation technology since January 2015 and prototyped water level sensors to be used with the AWD method106. The final prototype of the sensor was tested with an infrared distance sensor, a SD card module to store data, and a GSM module to send the distance information to a cloud service. These prototyped sensors are not made commercially viable. The challenges that TechAguru faced in developing prototype sensors to be placed in the tube were multifold (also available in TechAguru’s blog 107 ). Firstly, TechAguru faced technical challenges – sensor technicality and communication infrastructure – in developing sensor prototypes for AWD tubes. It began with using ultrasonic sensor but had to develop a new prototype with infra-red sensor as ultrasonic sensors generated inaccurate data and needed to be waterproofed and protected from dirt and soils. While there are other sensor types better suitable for the AWD context, TechAguru narrowed down the menu options to aforementioned two types considering the cost implications for farmers. It was also noted that the cost of developing a prototype and manufacturing at a bulk amount will differ. Another technical challenge that TechAguru experienced was the insufficient level of infrastructure in the region. The prototypes were GSM-based so they had to be connected to cell-tower and the Internet. However, not all fields were covered with GSM and this led to a connectivity problem. Moreover, TechAguru mentioned that WiFi connection is not desirable as it is designed to send bulk amount of data while the sensor used for the AWD tube does not transmit as much data. Another set of challenges was associated with the specific farming context. TechAguru mentioned that the field must be leveled for the water level sensors to be useful. Unless the sensors are deployed in properly leveled fields, they will generate inaccurate data and the system would be inefficient. Also, having multiple plots requires more water level sensors, therefore rice field plots need to be consolidated and leveled to make the use of water level sensors more efficiently. Consolidating multiple plots may require association or cooperatives of farmers to collaborate with the use of tractors with leveling attachments108. Moreover, farmers experienced stealing water from neighboring irrigated rice fields as some farmers closed the irrigation gates of other farms so they can get more water in their farms. Considering this issue, TechAguru suggested the introduction of water level sensors and irrigation control may not be feasible in some irrigated rice paddies. 25 Last set of challenges was related to the farmers’ conservative mindset. TechAguru said many farmers were hesitant, if not reluctant, to adopting the AWD simply because they were taking change as a risk. MimosaTEK MimosaTEK, a Vietnamese IoT for agriculture start-up, shared its experiences of helping farmers to have more accurate sense of their crop water demand through the application of IoT solutions. It has two main business approaches: B2B and B2C. B2B customers include corporations whose farm sizes range from 300ha to 1,000ha. Its B2C (business-to-customer) focuses on bringing smallholder farmers to use MimosaTEK’s solutions through 20 pilot projects across Vietnam. While its experiences are not specifically focused on rice paddy fields nor sensors for AWD tubes, MimosaTEK introduced and shared its experiences with the current business chain and model. MimosaTEK has more than one B2C business approach. It first let farmers try products for free so that the farmers could test out the effectiveness of the system. Also, MimosaTEK provides a one-year leasing contract to long-pepper farmers through a cooperative. Under this contract, each farmer pays $100 deposit for one controller and one soil moisture node and the deposit will be paid back to the farmers if the products are in good conditions upon return. Besides the deposit, each farmer has a monthly payment of around $10, 20% subsidized level from the original fee. It mentioned that this cooperative has incentives to use the IoT solutions because they can grow pepper in a sustainable manner, which will increase their crop value in export market, in addition to more visible advantages of precise irrigation and traceability enabled by sensor devices. In terms of technology and communication infrastructure, MimosaTEK did not have as many challenges as TechAguru given that Vietnam has a relatively high broadband penetration rate and that many farmers use smartphones. Lessons for Future Work and Scale-up Sensors undoubtedly play a central role of collecting and transmitting data, thereby delivering valuable information to farmers. However, there is still little information on the cost comparison of sensors and characteristics of network connectivity solutions tailored to development, particularly in agricultural applications. Also, the research shows that the successful deployment and utilization of IoT solutions requires a comprehensive understanding of inter-linked components that comprise of an overall IoT system. Smart farming ecosystem enabled by IoT incorporates multifaceted elements to consider such as enabling technologies, business models, barriers for adoption, knowledge- sharing109, as shown in Figure 6. In this sense, the application of IoT developments, particularly to promote precision agriculture as part of smart farming, requires a whole farm management perspective110. 26 Table 10. Smart Farm and Food Safety Stakeholders Yet, to fully capture the agricultural opportunities introduced by IoT in development context, it is equally important to understand existing barriers and think to effectively overcome them. One of the technical barriers to scaling up IoT in agriculture is based on the decentralized and fragmented state of IoT technological development, leading to the lack of interoperability and standardization. This lack of standardized approach is applicable to both hardware and software sides as many of the IoT platforms are decoupled from one another as they are developed based off from open source. In addition, proprietary architectures, platforms and standards could lead to the risks of side-effects such as vendor lock-in and incompatibility with other systems 111 . Another set of technical challenges associated with IoT deployment in developing countries is the lack of enabling environment such as the cost affordability of sensors and network communications infrastructure. There are other technical issues such as the need for more advanced and flexible data analytics services and data security that are currently discussed in developed economies112. While these issues convey valid concerns for the future direction of IoT application in agriculture, their presence is not yet as visible in developing countries context. The non-technical barriers are also present when applying IoT in agriculture in development context. These challenges include the sustainability of business models, digital literacy for farmers to fully reap the benefits of IoT, and government regulations. 27 Annex 1. Comparison of Different Telemetric Options Telemetry Operating Specification Suitable for Cons Range Module Cost Method frequency Global GSM (global Throughput and delay are system for mobile variable, and they 2G 35km $1-$15 communications) Wherever there is depend on the number of bands cellular connection; other users Periodic monitoring Cellular 3G Licensed spectrum applications than real- - Up to 100km $35-$50 time; Relieves the range limitation 4G/LTE Licensed spectrum - Up to 100km $80-$120 Wi-Fi Wherever wireless (802.11n, 802.11a/g, 2.4 GHz / 5GHz Internet connection is 300m $10+ 802.11b, 802.11ac) available Wireless - Low power Sub-1GHz ISM Wi-Fi consumption and bands Up to 1000m N/A (802.11ah113) long-range data (differs in countries) transmission 2.4GHz mesh local area network (LAN) protocol; Originally designed for Standards-based wireless building automation technology intended to Low data rates (requires and control – suitable Zigbee meet the needs of device- 2.4 GHz / 900MHz low specification 10-100m $1-$15 for wireless to-device hardware) thermostats and communications; Uses lighting systems small, low-power radios for data transmission 28 Endnotes 1 "FAO's Director-General on how to Feed the World in 2050." Population and Development Review 35.4 (2009): 837- 9. EconLit. Web. 2 International Food Policy Research Institute. "SMALLHOLDER FARMING " http://www.ifpri.org/topic/smallholder- farming. Web. 3 P.28. Srinivasan, Ancha. Handbook of Precision agriculturePrinciples and Applications . Food Products Press, 2006. Print. 4 P.3. Ibid. 5 P3. Ibid. 6 https://www.automationworld.com/article/topics/cloud-computing/know-difference-between-iot-and-m2m 7 Lesser, Adam. "Declining sensor costs open up new consumer applications." Gigaom. Jan 25 2015. Web. . 8 International Telecommunication Union (ITU). "ICT Facts & Figures." International Telecommunication Union (ITU). May 2015. Web. . 9 Ibid. 10 A method of measuring soil moisture by time-domain reflectometry, Ledieu et al., Journal of Hydrology, 1986 11 "Distance Sensors: Ultrasonic ranging sensors." Campbell Scientific. Web. . 12 "Optoelectronic Sensors in Medical Applications." Sept 1, 2003. Web. . 13 http://www.omega.com/temperature/Z/pdf/Z076-080.pdf 14 "Type of Sensor." Web. . 15 P19. ALLIANCE FOR INTERNET OF THINGS INNOVATION. AIOTI WG06 – Smart Farming and Food Safety. Smart Farming and Food Safety Internet of Things Applications: Challenges for Large Scale Implementations., 2015. Print. 16 P.2. Bogue, Robert, and Robert Bogue. "Sensors Key to Advances in Precision Agriculture." Sensor Review 37.1 (2017): 1-6. Print. 17 Wireless sensor networks for agriculture: The state-of-the-art in practice and future challenges. Tamoghna Ojha, Sudip Misra a, Narendra Singh Raghuwanshi 18 A Review of Wireless Sensors and Networks’ Applications in Agriculture , Aqueel-ur-Rehman et al. 19 "Soil and Crop Sensing." University of Nebraska-Lincoln. Web. . 20 Yu and Han, Remote Detection and Monitoring of a Water Level Using Narrow Band Channel, Soongsil University, 2010. https://pdfs.semanticscholar.org/b29c/c338a845e82da929cdbba5f83896ff13d950.pdf 21 Nakka, Sensors for Agriculture and Water Use Efficiency, http://www.slideshare.net/saibhaskar/sensors-for-agriculture-and- water-use-efficiency, 2016 22 "Smart on-line water salinity measurement network to manage and protect rice fields." Web. . 23 "SENSORS FOR WATER MONITORING IN PADDY FIELDS FOR IMPROVED ON FARM WATER MANAGEMENT." ClimaAdapt. Web. . 24 "SENSORS FOR WATER MONITORING IN PADDY FIELDS FOR IMPROVED ON FARM WATER MANAGEMENT " Web. . 25 Xiao, Deqin, et al. "Integrated Soil Moisture and Water Depth Sensor for Paddy Fields." Computers and Electronics in Agriculture 98 (2013): 214-21. Print. 26 http://archive.apan.net/meetings/apan39/Sessions/7/Hiroshi_Shimamura_APAN_Meeting_15.03.04.pdf 27 Poulose Jacob, K., and Simon Santhosh. "Development and Deployment of Wireless Sensor Network in Paddy Fields of Kuttanad." (2012). Print. 28 Monitoring the Paddy Crop Field Using ZigBee Network. K Sriharsha, et al. International Journal of Computer and Electronics Research [Volume 1, Issue 4, December 2012] 29 Sekozawa, Teruji. "A Fully Automated Water Management System for Large Rice Paddies". Proceedings of the 14th WSEAS International Conference on Systems: part of the 14th WSEAS CSCC multiconference. World Scientific and Engineering Academy and Society (WSEAS), Stevens Point, WI, USA, I. 2010. 325-330. Print. 30 Kumar, Suman, et al. "Application of Sensor Networks for Monitoring of Rice Plants: A Case Study." Science (2009): 1-7. Print. 31 Islam, Mohammad Monirul, et al. "A Floating Sensing System to Evaluate Soil and Crop Variability within Flooded Paddy Rice Fields." Precision agriculture 12.6 (2011): 850. Print 32 "Saving Water with Alternate Wetting Drying (AWD)." Rice Knowledge Bank. Web. . 33 "Overview of AWD." IRRI. March 2016. Web. . 34 Internet Society. The Internet of Things (IoT): An Overview ., 2015 29 35 Hari Balakrishnan.  NETWORK CONNECTIVITY FOR IoT., 2017. 36 https://www.pcmag.com/encyclopedia/term/40833/data-rate 37 https://www.computerhope.com/issues/chspace.htm 38 Hari Balakrishnan.  NETWORK CONNECTIVITY FOR IoT ., 2017 39 P 5. Texas Instrument. Wireless Connectivity for the Internet of Things: One Size does Not Fit Al l., 2017. 40 Ibid. 41 Ibid. 42 Ibid. 43 Ibid. 44 ZigBee. http://www.zigbee.org/zigbee-for-developers/ 45 LinkLabs. https://www.link-labs.com/blog/types-of-wireless-technology 46 P.9. Texas Instrument. Wireless Connectivity for the Internet of Things: One Size does Not Fit Al l., 2017. 47 ZigBee. https://web.archive.org/web/20130627172453/http://www.zigbee.org/Specifications/ZigBee/FAQ.aspx 48 In data transmission, network throughput is the amount of data moved successfully from one place to another in a given time period, and typically measured in bits per second (bps). See: http://searchnetworking.techtarget.com/definition/throughput 49 Zigbee. http://www.zigbee.org/non-menu-pages/faq/ 50 P.10. Texas Instrument. Wireless Connectivity for the Internet of Things: One Size does Not Fit Al l., 2017. 51 Ibid. 52 Ibid. 53 LinkLab, https://www.link-labs.com/blog/types-of-wireless-technology 54 P.10. Texas Instrument. Wireless Connectivity for the Internet of Things: One Size does Not Fit Al l., 2017. 55 Ibid. 56 P 8. Ibid. 57 Ibid. 58 Ibid. 59 Ibid. 60 Ibid. 61 Ibid. 62 Ibid. 63 Ibid. 64 Ibid. 65 P 9. Ibid. 66 P 7. Ibid. 67 P.5. Ibid. 68 P. 7. Ibid. 69 Ibid. 70 P 2. Ibid. 71 GSMA. https://www.gsmaintelligence.com/research/2014/12/mobile-broadband-reach-expanding-globally/453/ 72 LinkLabs. https://www.link-labs.com/blog/cellular-iot 73 P 4. Ericsson. Cellular Networks for Massive IoT: Enabling Low Power Wide Area Applications. , 2016. 74 https://www.iotforall.com/cellular-iot-explained-nb-iot-vs-lte-m/ 75 P 3. Ericsson. Cellular Networks for Massive IoT: Enabling Low Power Wide Area Applications. , 2016. 76 Ibid. 77 P 13. Ibid. 78 Link-Labs. A Comprehensive Look at Low Power Wide Area Networks. 79 Link-Labs. https://www.link-labs.com/blog/nb-iot-vs-lora-vs-sigfox 80 P. 14. Ericsson. Cellular Networks for Massive IoT: Enabling Low Power Wide Area Applications. , 2016. 81 https://www.leverege.com/blogpost/iot-connectivity-comparison-lora-sigfox-rpma-lpwan-technologies 82 https://www.leverege.com/blogpost/iot-connectivity-comparison-lora-sigfox-rpma-lpwan-technologies 83 https://www.link-labs.com/blog/nb-iot-vs-lora-vs-sigfox. 84 https://www.leverege.com/blogpost/iot-connectivity-comparison-lora-sigfox-rpma-lpwan-technologies 85 http://www.thewhir.com/web-hosting-news/the-four-internet-of-things-connectivity-models-explained 86 P 18. Internet Society. The Internet of Things (IoT): An Overview ., 2015 87 Ibid. 88 P.19. Ibid. 89 Ibid. 90 Ibid. 91 P.20. Ibid. 30 92 Ibid. 93 Ibid. 94 Ibid. 95 P.21. Ibid. 96 P.22. Ibid. 97 Hari Balakrishnan.  NETWORK CONNECTIVITY FOR IoT ., 2017 98 P.13. GSMA. The Mobile Economy., 2017 99 https://efarmer.mobi/blog/agriculture-technology/ 100 USAID and Feed the Future. Digitizing the Agricultural Value Chain. 101 https://www.gsmaintelligence.com/research/?file=53525bcdac7cd801eccef740e001fd92&download 102 P.77. Karim, Lutful, et al. "Sensor-Based M2M Agriculture Monitoring Systems for Developing Countries: State and Challenges." Network Protocols and Algorithms 5.3 (2013): 68-86. Print. 103 https://www.link-labs.com/thank-you-webinar-costs-in-iot-lte-m-nb-iot-sigfox-lora?submissionGuid=0cd436a2-f931-43b8- 9187-5134bac88d36 104 USAID et al. Integrating Mobiles into Development Projects., 2014. Print 105 P.79. Karim, Lutful, et al. "Sensor-Based M2M Agriculture Monitoring Systems for Developing Countries: State and Challenges." Network Protocols and Algorithms 5.3 (2013): 68-86. Print. 106 Auxillos, John. "Water Level Sensors for Rice Field Irrigation." TechAguru. Jul 14 2016. Web. . 107 Ibid. 108 Ibid. 109 ALLIANCE FOR INTERNET OF THINGS INNOVATION. AIOTI WG06 – Smart Farming and Food Safety. Smart Farming and Food Safety Internet of Things Applications: Challenges for Large Scale Implementations., 2015. Print. 110 P.139. Sundmaeker, Harald, et al. "Internet of Food and Farm 2020." Digitising the Industry-Internet of Things Connecting Physical, Digital and Virtual Worlds.River Publishers, Gistrup/Delft (2016): 129-51. Print. 111 P.142. Ibid. 112 P.143. Ibid. 113 802.11ah, also known as “the low power WiFi”, is 900 MHz WiFi developed as an IEEE protocol for the wireless Internet network. However, there is no global standard for 900 MHz as 80% of the world uses 2.4 GHz Wifi. It is expected WiFi AH products would appear in the next 12 months to 1.5 years. Link-lab mentions that this technology is similar to LPWAN (LoRA). 31