Non Technical Summary
Internet of Things (IoT) technology and the digitalization of agriculture stands poised to transform production via the smart farming framework (e.g., Gubbi , 2013; Abbasi et al., 2014; Wolfert et al., 2014;Ferrandez-Pastor, 2016; Nukala et al., 2016). Deficiencies that are currently limiting progress are: (1) lack of appropriate low-cost sensing technology, (2) lack of sensor connectivity to the internet, especially in rural areas, and (3) cloud-based software that can use data from multiple data streams to generate new information for decision support. This project aims to address these deficiencies in serval crucial areas of Ag production and sustainability.Accurate weather data is essential for almost every agriculture endeavor, especially in the smart farming framework where decisions are based on current and forecasted conditions. Most states have an agriculture weather network like Colorado's CoAgMet, where a relatively small number of stations report regional weather conditions across the state. However, these stations tell us little about site specific environmental conditions on individual farms, fields, beef feedlots, or dairy barns. We need a much higher density of weather information (both spatially and temporary) to feed into a cloud-based IoT data analytic framework. Agricultural producers will likely need multiple weather stations positioned at key areas across their operations. The open source weather station design proposed here (Objective 1) will meet that need. Furthermore, the station could be used by researchers, citizen scientists, hobbyists, and schools.Agriculture air quality is a rapidly growing concern at multiple scales. Ammonia emissions from beef feedlots in Colorado are affecting Rocky Mountain National park (e.g., Benedict et al., 2013), and greenhouse gas emissions, especially during a period of intensification, are a serious global concern (Burney et al., 2010; Campbell et al., 2014, Snyder et al., 2014 ). Dust and odors from feedlots and dairies are perennial issues (e.g., Guo et al., 2011). On farm air quality is a safety and health concern. Recently a Wisconsin farmer and 16 cows died of hydrogen sulfide or methane asphyxiation from a manure tank - these accidents are more common than one might think (Beaver and Field, 2007). We need low-cost air monitoring equipment that can be deployed at multiple locations across the farm to measure emissions and insure worker safety. The air quality monitoring module proposed in this project (Objective 2) will address that need, especially when paired with the proposed weather station. Soil water content affects almost every ecological, agricultural, and hydrological process at the land-atmosphere interface. Water is a special concern in Colorado; where municipalities, oil and gas, and agriculture all compete for limited water resources. Improved irrigation management is paramount for the sustainability of many cropping systems. Unfortunately, commercial sensors that provide automated, continuous soil water measurements and send data to the internet are expensive. Commercial sensors and dataloggers cannot be economically deployed in large enough numbers for practical irrigation management. We think the new low-cost soil moisture sensors and electronics developed in this project (Objective 3) will be used by farmers, crop consultants, researchers, and citizen scientists for irrigation management and a wide range of other applications - this is an active area of research (e.g., Bitella et al., 2014). The sensors could also be a boon for citizen science networks like CSU's CoCoRaHS precipitation monitoring program (Reges et al., 2016) - who would like to add soil moisture monitoring to the network (Cosh et al., 2016). Data from the sensors could be used by researchers when studying a wide range of soil and plant processes that are influenced by soil water content.While new sensors and low-cost data collection are essential aspects of IoAgT, it is the integration of this information into real-time virtual "dashboards" and cyberdata products (data Analytics) that truly harnesses the power of IoT smart farming. IoT dashboards and web based software products (data aggregators like ThingSpeak) allow users to see real-time data, look for trends, or combine variables for data analytics. Furthermore, the stored data allows for historical analysis and algorithm development. Thus, while sensor development is essential, you also need the math and statistics to make optimal use of the data the new sensors provide.

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
111	7299	2070	50%
404	7299	2020	50%

Knowledge Area
404 - Instrumentation and Control Systems; 111 - Conservation and Efficient Use of Water;

Subject Of Investigation
7299 - Research equipment and methods, general/other;

Field Of Science
2070 - Meteorology and climatology; 2020 - Engineering;

Goals / Objectives
The goal of this project is to develop new measurement technologies that increase efficiency, save water, and protect the environment when using smart farming practices. Here we use the term "smart farming" to describe a broad range of new technologies that provide data to guide on-farm decisions and planning. This could include sensors for almost any variable of interest; everything from soil moisture sensors for irrigation control to electronic monitors for livestock health. Driving this revolution is the Internet of Things (IoT), an ever-growing network of sensors and devices with internet and machine-to-machine (M2M) connectivity. Data collected by these technologies is immediately uploaded to the internet "cloud" (i.e., a network of servers) where results can be viewed and combined with broader information in a data analytics and decision support framework. Internet-of-Ag-Things (IoAgT) extends this technology to the farm. For example, soil moisture data could be combined with weather forecasts and satellite imagery of crop stress to better manage the timing and amount of irrigation - perhaps via robotic control. Unfortunately, the sensing technology and software needed to implement smart farming is often too expensive or simply does not exist. For smart farming to work, we need a lot of "cheap" data from a network of low-cost sensors distributed across the farm or livestock operation - all linked to the internet. This project will use new developments in low-cost electronics, novel fabrication technologies (e.g., 3D Printing), IoT connectivity, and cloud computing to develop next generation sensing technologies for growing crops and protecting the environment. Several areas will be targeted for development, including: (1) weather measurement, (2) air quality monitoring, and (3) soil moisture sensing. In all cases, the goal is to design instrumentation that can be deployed by people without specialized engineering or electronics training.Objectives:Design, build, and test a mini weather station capable of research grade results using low-cost sensors, 3D-printed parts and open-source electronics. Compare the performance of the mini weather station to research grade sensors under field conditions (e.g., co-deploy at CoAgMet sites).Design an air quality module for the weather station that can detect volatile organic compounds, ammonia,carbon dioxide and dust in agricultural settings.Data will be used to monitor air quality, trigger warnings, and estimate emissions.Design and test low-cost soil moisture sensors that use a capacitance-based approach (measures dielectric permittivity) to detect changes in volumetric soil water content.The sensor will cost less than $15 and readily interface with low cost dataloggers and microcontrollers.Build on-line IoAgT virtual dashboards that can display and store data collected by the sensors described in objectives 1-3. These software dashboards will use existing IoT services like Thingspeak and Microsoft Azure to provide feedback to the user via web browser, tablet, or smart phone.We will also demonstrate how to use the all the sensors in IoT frameworks and include data analytics for calculating Reference ET, air emissions, and other agricultural weather-based metrics. For all instruments, we will post the complete build instructions and software on the web for open source distribution.

Project Methods
Mini Weather StationObjective 1 is to design a mini weather station for under $300 that can deliver continuous, real-time, research-grade data to an IoT website via a cell phone of Wi-Fi interface. The key parameters and sensor selections are described in Table 1.Table 1. Weather station variables and sensors Parameter Sensor(s) NotesAir Temperature and RHIntegrated CircuitSensiron SHT31, Bosch BME280Solar RadiationPhotodiodeAdvanced Photonics, PDB-C139, 3D printed mountWind SpeedCup AnemometerInspeed VortexWind DirectionVane3D printed design (Ham)PrecipitationAcoustic disdrometerDisdrometrics B.V.Soil Temperature*One-wire Temp SensorMaxim DS18B20Soil Water Content*Capacitance based transducerVegetronix VH400 or custom I2Csensor* OptionalThe mini weather station will be connected to the internet via WiFi or cell phone using one of several different technologies. WiFi connectivity will be provided via Adafruit's Huzzah ESP8266 microcontroller ($9) or Particle's Photon device ($19). For locations without WiFi, a Particle Electron microcontroller ($55) with 3G cellular connectivity will be used. These devices come with a SIM card and a $3/month low-bandwidth 3G data service. A cell phone contract with providers like Verizon or AT&T is NOT required. Our team has been experimenting with these platforms for various projects with good success. The loggers will be housed in waterproof boxes with solar panels and batteries for power. Data from all sensors will be sampled every 30-60 s and 30 min or hourly averages will be uploaded to IoT platforms.Prototype versions of the mini weather station will be co-deployed next to the CoAgMet weather station at the ARDEC experimental farm near Fort Collins..Air Quality Monitor (Objective 2)We will develop an agricultural air quality module that can be co-deployed with the mini weather station (Objective 1). The initial design will be based somewhat on EPA's new SPoD fenceline monitor (EPA, 2016). While SPoD was designed for fence line monitoring of VOCs at oil refineries, we cooperate closely with EPA's Eben Thoma, the director the SPoD project. Thus, we are confident we can adapt the design for air quality monitoring at cattle feedlots and dairies. The SPoD uses many 3D printed parts, so our lab is well equipped to proceed with fabrication. Initially the system will focus on using a $450 PID detector (Baseline MOCON Inc.) to measure VOCs and ammonia. While the sensor may seem expensive, most real-time NH3 instruments cost between $40 and 50 thousand dollars. The PID has a low response to NH3 directly; however, recent work shows very strong correlations between NH3 and other VOCs emitted from livestock manure and urine patches - and the PID has a strong response to these VOCs.. Thus, it may be possible to make a low fidelity estimate of NH3 using a PID sensor once it has been calibrated for a specific location or type of CAFO. In addition to the PID, the air quality module will also have options for a CO2 monitor (K33, CO2Meter Inc.) and a low-end dust sensor (Amphenol/Telaire SM- PWM-01A). Carbon dioxide can be used to determine if the sensor is being impacted by a plume coming directly from the livestock. Because Ag sites are often very dusty, air for the PID and CO2 sensors will be drawn into the system using a small pump (0.6 LPM) via a section of teflon tubing with a filter at the inlet. The system will also have the ability to purge the instruments with NH3-free and CO2-free air in order to "zero" the analyzers.Prototype versions of the air quality monitor will be deployed at a cattle feedlot or dairy in combination with research grade instruments. We (Jay Ham, Azer Yalin) have funded livestock project underway so the new sensors will be tested in tandem with ongoing research. The NH3 data will be compared to results from a Picarro cavity ring down analyzer and the CO2 compared a Li-Cor 840 analyzer.Soil Moisture Sensors (Objective 3)The I2C soil moisture sensor by wemakethings.com provides a good starting point for an open source design (https://www.tindie.com/products/miceuz/i2c-soil-moisture-sensor/). This $13 sensor uses a capacitance technique that detects the dialectic permittivity of the soil. This is the same principle used by +$100 commercial probes such as the Decagaon ECH2O sensors. The probe measures the charge time of a capacitor that is governed by the dialectic properties of the soil, a parameter that is a strong function of water content (or air filled porosity). We have made a 3D printed housing for the sensor that protects the electronics and allows long-term installation in the soil . Our group has been testing the weatherproof version of the sensor under turfgrass and results are encouraging. The sensor also measures soil temperature to within 0.1 C with a thermistor.For the proposed project, we will test a stock version of the waterproofed sensor and a customized version that includes several design modifications, including: 1) changing the sensors frequency to something closer to 50-70 MHz in order to make the readings less sensitive to salinity and soil type, 2) changing the geometry of the traces used for the capacitor to improve the response, 3) explore replacing the electronics with dedicated capacitance measurement IC like the Texas Instrument FDC004 ( 4) building a custom mold for potting/waterproofing the electronics to improve longevity/durability, and 5) adding a more durable finish on the PCB to make it less sensitive to damage and weathering.All three sensor designs: 1) Stock I2C sensors, 2) customized I2C sensor, and the 3) Vegetronix VH400; will be calibrated over a range of water contents using soils with textures ranging from a fine sand to a silty clay loam. Bulk soil will be brought to a target moisture content by adding water to sieved and oven dry soil. The well mixed sample will be packed to a known bulk density in a brass cylinder that includes the sensor. Sensor output will be collected for several hours before the sensor is removed and the core is oven dried and weighed. The resulting data will show the sensor response to different water contents for each soil type. Calibrations will be repeated at two different temperatures to check for a thermal effect. The main goal of the lab calibration is: 1) determine the functional form of the water content vs. sensor output calibration curve, 2) examine how calibration is affected by soil type and bulk density, 3) examine variation among sensors, and 4) provide a test bed to refine the customized sensors described earlier (e.g., evaluate different trace configurations in the PCB). We will also attempt to evaluate the effect of soil salinity on the calibrations. We recognize that sensors will likely require a site-specific soil calibration.Data from all sensors (Objectives 1, 2 and 3) will be routed to IoT data aggregators like ThingSpeak, Blynk, and Microsoft Azure. These services all one to display real-time data, store data for historical analysis, and combine data from multiple sources to evaluate multifactorial processes. Example calculated parameters include: 1) Reference Crop ET from weather and forecast data using the ASCE formula, 2) emissions of NH3 from feedlots using an inverse model like FIDES (Loubet et al. 2001) and 3) the weekly change in soil moisture content at an irrigated corn site. Furthermore, these services can be programed to send users text messages or emails when a parameter falls outside a defined limit. For example, the user could receive a notification if soil water content drops below a set value or air NH3 concentrations exceed a defined limit. We plan to test of these notification feature for NH3 at feedlots and dairies, and the soil water notification at corn research plots managed by CSU and USDA.