MeteoMex: open infrastructure for networked environmental monitoring and agriculture 4.0

Microcontroller board, circuit, and PCB design

For connecting sensors to wireless networks, Arduino-compatible, single-board microcomputers are used. The low-end Wemos D1 mini board is based on the ESP-8266EX chip (80 MHz, 4 Mb flash memory) and provides WiFi. The board measures 34.2 mm × 25.6 mm and weighs 3 g. With a micro-USB connection, the Wemos D1 can be powered and programed (https://www.wemos.cc/). The board features 11 digital input/output pins and one analog input (3.2 V max.). The price of a standard Wemos D1 mini board is approximately 5 USD.

For mounting sensors and other electronic components, printed circuit boards (PCB) were designed that can be stacked on the Wemos boards-so-called shields.

The open-source software Fritzing (https://fritzing.org) was used for designing the circuits and printed circuit boards (PCB) (Monk, 2015). For fabrication, the PCB layouts were exported to the Gerber file format (https://www.ucamco.com/en/gerber). PCB figures in this article were produced with a Gerber file viewer (https://www.pcbgogo.com/GerberViewer.html). ALLPCB, China (https://www.allpcb.com/) produced testing lots of 100 pieces each.

The editable Fritzing files and example programs are available from the MeteoMex GitHub repository (https://github.com/robert-winkler/MeteoMex).

The configuration of the different shields is explained in the following subsections.

MeteoMex aeria

The MeteoMex aeria (Fig. 1) shield uses a BME280 chip (Bosch, Stuttgart, Germany, https://www.bosch-sensortec.com/products/environmental-sensors/humidity-sensors-bme280/) to monitor the ambient air parameters temperature, relative humidity, and barometric pressure.

An additional CCS811 sensor (https://www.sciosense.com/products/environmental-sensors/ccs811-gas-sensor-solution/), permits the detection of total volatile organic compounds (VOC). Both chips use the I²C bus for sensor data transfer. If the board is powered by batteries, programming a DeepSleep mode is recommendable for reducing the energy consumption in idle mode. In DeepSleep mode, all functions except the real-time clock are switched off, reducing the power consumption of the ESP8266 chip from 70 mA to 20 μA (https://www.instructables.com/ESP8266-Pro-Tips/). For waking up the Wemos D1 mini, the jumper J1 between RST and D0 (GPIO16) must be closed. To enable the DeepSleep function, therefore, a wire bridge has to be soldered into the jumper J1.

MeteoMex terra

The MeteoMex terra (Fig. 2) shield is designed to connect an analog sensor, such as a conductive or capacitive soil moisture sensor, and a digital DS18B20 temperature sensor (Maxim Integrated). For operating the DS18B20 sensor, a pull-up resistor of 10 kΩ (R1) is required.

The power supply of the analog and digital sensor is connected to the pins D1 and D3, respectively. However, some analog sensors require a connection to the 5V pin for reliable operation. The maximal permitted input voltage for the A0 analog pin is 3.2 V. Thus, a voltage divider needs to be used when using sensors with higher signal levels.

This board as well provides a J1 option.

MeteoMex WasteWater prototype

The MeteoMex WasteWater (Fig. 3) configuration is a slight variation of the terra circuit. An analog Arduino turbidity sensor (TresD Print Tech) is used for measuring total suspended solids (TSS). For determining tank filling levels, a Jsn-sr04t waterproof ultrasonic sensor (Ranmex) is connected. Both sensors operate on 5 V. For this custom design, no PCB was printed, but the circuit was built on a Perfboard.

All boards were designed for the use of through-hole electronic components for facilitating manual assembly and soldering.

Programming and Internet-of-Things infrastructure

An overview of the IoT infrastructure is shown in Fig. 4. The programming was done on Windows 10 and standard Linux distributions (Fedora and Ubuntu). The IoT platform is running on a Virtual-Private-Server, hosted by IONOS (https://www.ionos.de). The operating system of the IoT server is Ubuntu 18.04 LTS (https://ubuntu.com).

Housing and power supply

MeteoMex kits provide no housing by default, which saves costs and reduces unnecessary plastics waste. The circuits operate at low voltages (5 V), so the devices are safe for humans and animals. For indoor air monitoring, the devices can be simply connected to a USB power supply (Fig. 5A). For outdoor conditions, the protection against dust, insects, and water might be necessary (Figs. 5B–5E). However, re-used plastic beakers, for example, from dairy products, usually fulfill this purpose.

Alternatively, to USB port power, the devices can be operated with 3 × A.A. rechargeable batteries or with solar panels and a power bank. A MeteoMex aeria device with BME280 sensor works about 3 months with 3 × AA rechargeable batteries, with hourly measurement and using a DeepSleep/WakeUp routine. However, additional electronic components and batteries have a negative impact on the system’s environmental footprint.

The building of the device and housing shown in Fig. 5E is described in an Instructable (https://www.instructables.com/member/RobertWinkler/instructables/).

Internet-of-Things integration of MeteoMex units

According to the legend, Internet-of-Things (IoT) started in the early 1980s, when a Coke-vending machine at the Carnegie Mellon University was connected to the internet (https://www.cs.cmu.edu/~coke/). In these pioneering times, integrating custom hardware into a computer network was a challenge for technical specialists. Nowadays, different IoT standards and protocols facilitate the data telemetry and processing (Ponnusamy & Rajagopalan, 2018).

The IoT platform ThingsBoard supports various telemetry protocols. For the MeteoMex project, HTML was chosen for its simplicity in programing and testing. The correct set-up of a device can be verified by sending data from a computer console. For example, the following command sends the value 99 of the variable humidity for a device with the token uVCGuzjBqV:

curl -v -X POST -d“{\“humidity\”:99}”

http://www.meteomex.com:8080/api/v1/uVCGuzjBqV/telemetry

--header“Content-Type:application/json”

Wemos boards also could be connected to other IoT platforms. The integration into Blynk (https://blynk.io) and Thinger.io (https://thinger.io) was successful. But apart from technical questions, the licensing scheme of IoT platforms is relevant. ThingsBoard (https://thingsboard.io) is licensed under the Apache 2.0 license (https://www.apache.org/licenses/LICENSE-2.0). That is, besides being open-source, ThingsBoard offers a free Community Edition and allows its employment in commercial applications. ThingsBoard is cross-platform compatible and can be installed on Windows, Mac, and Linux (https://thingsboard.io/docs/user-guide/install/installation-options/). The system used in this study was installed on Ubuntu 18.04 LTS (https://ubuntu.com). The computational capacity of the Virtual-Private-Server, hosted by IONOS (https://www.ionos.de), can be adjusted to the IoT server load. However, the installation of a ThingsBoard Community Edition IoT server is possible at no software cost.

The open-source PostgreSQL (https://www.postgresql.org) database server for storing the IoT data has high performance and robustness. Database maintenance and data manipulation are possible with system tools and external programs. Exported data can be further analyzed, for example, with statistics and data mining software, such as R/Rattle (https://rattle.togaware.com) (Williams, 2011).

Transferring data with radio frequency (R.F.) and Bluetooth was tried as well. However, the direct connection of the devices to WiFi networks turned out the technically easiest solution. WiFi networks are ubiquitously available, of fair security, and no additional adaptors are necessary to send collected data to the internet.

The ThingsBoard dashboards are visualized on a standard web browser, making it compatible with standard personal computers and mobile devices. The web platform also permits the setting-up of data processing pipelines, the IoT Rule Engine, and the definition of alarm levels and actions.

At the time of writing this manuscript, the IoT server was running for more than 180 days without interruption, demonstrating technical robustness. Although the IoT server was located in Germany, and the sensor units in Mexico, no telemetry data transfer problems were noticeable. Local power cuts or internet failures only affect devices in a particular zone. Since the wireless network settings are saved to the ESP8266 flash memory, the Wemos boards reconnect when rebooting.

Thus, the overall infrastructure takes into account the main aspects of an IoT system for agriculture (Elijah et al., 2018), such as cost, simple and robust technology, localization, scalability, and interoperability. Open licenses of hardware and software, and the use of common standards (WiFi network, HTML telemetry, SQL database) assure a long-term, cost-efficient, and provider-independent service.

Different application examples are presented in the next part.

Example 1: weather station for climate and volatile organic compounds

Climatic conditions and air pollution directly affect human health and welfare. Therefore, public monitoring data inform the citizens about possibly hazardous levels of contamination.

Figure 6 shows the recording of climate and total volatile organic compounds (tVOC) of a MeteoMex aeria device installed outdoor at the National Institute of Respiratory Diseases, INER, Mexico City, Mexico (http://iner.salud.gob.mx/). The apparatus is installed next to an aerobiology station for the continuous monitoring of pollen and microbial spores.

The readings show a daily tVOC peak at about 8 a.m., which could be caused, for example, by the morning traffic. Sensitive persons should avoid physical activities in periods of increased air contamination. Widely distributed public sampling stations could provide more localized data for warning people about possible health risks. The information also could motivate to reduce contaminating activities in affected areas.

The climate parameters temperature, humidity, and pressure, which are simultaneously recorded, also could be used to estimate the daily global solar radiation using a neuronal network model (Jimenez et al., 2016). I²C capable sensors for measuring photosynthetically active radiation (PAR) or Lux sensors could provide additional information about radiation and light conditions.

Example 2: greenhouse monitoring (air and soil)

For the optimal growth of plants, adequate air and soil conditions are essential. The conductivity and the dielectric properties of soil depend on its composition, structure, moisture, and salinity (Sreenivas, Venkataratnam & Rao, 1995; Malicki & Walczak, 1999; Wang & Schmugge, 1980). Those can be easily measured, and it was shown that soil electrical conductivity measurements correlate with Local yield (Grisso et al., 2005).

The MeteoMex terra device uses either conductive or capacitive probes to estimate the soil moisture. The capacitive probes are protected against corrosion and, therefore, preferable. Since the analogous measurement only provides an integer value between 0 and 1,024, the signal needs to be calibrated. The most straightforward procedure is measuring the dry sensor’s output and the sensor when completely submerged in water. A more realistic calibration is possible by adding water to dry soil until saturation. The soil type is crucial for the water storage capacity and the conductivity/ dielectric properties. Another strategy is the acquisition of raw readings and subsequent data interpretation.

Figure 7 shows on the left side, the air data, and on the right side, the soil data, which are measured hourly in a domestic greenhouse with automated irrigation. The capacitive soil moisture sensor was calibrated to 0–100% with purified water. The actual sensor readings range between 110% and 124%, which reflects the dielectric properties of soil.

The average soil temperature of about 22 °C is ~8 K lower than the air temperature, demonstrating the heat capacity and temperature buffer properties of the soil. Besides, the temperature variations during the day are ~15 K for air and ~7 K in the soil.

The low average barometric pressure of 80,360 Pa is consistent with the theoretical value of 80,572 Pa, which was calculated for an ambient temperature of 30 °C and an altitude of 1,990 m above sea level (https://www.mide.com/air-pressure-at-altitude-calculator).

As the temperature peak at about 12 p.m. with the following decline indicates, the measuring point is only exposed to direct sunlight for a short period of day. Temperature and solar radiation profiles have a direct effect on plant growth. Therefore, micro-climate engineering, for example, by planting shading trees, could become a common practice in future agriculture (Trilnick, Gordon & Zilberman, 2018).

The well-being of farm animals also depends on climate conditions. Temperature and humidity measurements in the barn demonstrated that the heat stress of dairy cattle in Canada was underestimated when using data from meteorological stations. Therefore, reading environmental data in the barn is recommended to determine the actual conditions the cows are exposed (Shock et al., 2016).

The economical sensors and the open IoT platform enable numerous applications in agriculture 4.0, smart farming, and urban greening (Madushanki et al., 2019).

Example 3: high-performance computing room monitoring (air conditioning)

Delicate equipment such as scientific instruments and high-precision machines require controlled ambient conditions. Overheating or condensation could result in serious device damages. Figure 8 shows the monitoring of a high-performance computing server room at the National Laboratory of Genomics for Biodiversity, Cinvestav Irapuato, Mexico (https://langebio.cinvestav.mx). The charts demonstrate a tight temperature control with an average temperature of 20.45 °C and less than 1.5 K variation.

The constant monitoring of ambient conditions also helps to detect potentials for energy savings, such as changing the settings of an air conditioning system at night, or during periods of inactivity.

Example 4: domestic wastewater plant

Using treated wastewater for irrigation saves sweet water reserves in dry areas. Figure 9 shows the block chart of a domestic wastewater treatment plant. The activated sludge process for reducing organic soluble solids by microbes requires aeration. This aeration step and the pumping for filtration and irrigation require electric energy. The food-energy-water nexus describes the strong interconnection between these resources (D’Odorico et al., 2018).

The presented facility uses photovoltaic energy. For saving electricity, the aeration is not operated continuously, but only during the daytime, and in half-hour intervals. The quality of the treated water needs to monitored to avoid either over-purification (wasting energy) or insufficient purification (clogging filters). In addition, the irrigation scheme has to be adjusted to the generation of treated water. Unused treated water passes through an overflow-pipe to the public sewer and is lost for irrigation.

The MeteoMex WasteWater prototype (Figs. 3 and 10) has two sensors: An analog Arduino turbidity sensor for measuring total suspended solids (TSS), and an ultrasonic distance sensor for estimating the tank filling. The Wemos board and the sensors are located inside the wastewater treatment plant, which is built from ferroconcrete. Thus, a direct WiFi connection to the wireless network is not possible. Thus, a Wemos D1 mini Pro board with an external antenna was used.

The data of both sensors are noisy (Fig. 11). Air bubbles and particles can create spikes in the turbidity measurements. Scattering and sometimes erratic readings of the ultrasonic distance sensor are also common and need to be addressed in the data processing. Two methods to filter noisy signal data are presented here.

Sustainability and socio-economic impact

As for any technology, the possible negative aspects of its adoption need to be discussed. IoT devices consume energy for themselves and for the data transfer and processing infrastructure. Further, their production consumes resources, and at the end of their life-time, they generate electronic waste (https://whatis5g.info/). Additional environmental issues arise when the devices are powered with batteries.

On the other side, the IoT technology can significantly contribute to saving natural resources (see examples above). The energy consumption of the presented domestic wastewater treatment plant (Example 4, “Example 4: Domestic Wastewater Plant”) is several kWh/day, compared to only some Wh/day of the MeteoMex device. The energy demand of the IoT equipment corresponds to about 0.1% and enables the continuous monitoring and optimization of the plant. Even little improvements in the wastewater treatment and irrigation process that reduce the energy consumption by a few percent justify the environmental and economic cost of the IoT integration.

In other cases, the IoT devices could be mobile and only connected for project-specific tasks, for example, for determining the day-night temperature profile of a production facility.

Sustainability was a central design goal of the complete platform. The used boards are highly integrated and provide the necessary computation and networking functions with a minimum of material (3 g for a complete Wemos D1 mini board) and energy. WiFi technology and 5 V USB power supplies are globally available, and no special adapters are necessary.

Importantly, the low cost of the hardware components and the permissive licenses for all parts of the infrastructure—circuit and PCB board design, database, and IoT platform—make the adoption of an IoT system in marginal production sites feasible. Further, weather enthusiasts and environmental activists could form networks for regional and global data collection. The program code can be re-used by the community, for example, to integrate more sensors, and for education.