How can a crop be irrigated intelligently without evidence that enough water is available? Will there be sufficient rain to ensure the expected water supply? Does the amount of water used for this purpose comply with the permit issued by the water board? Answering these and other questions as a service provided to farmers and managers of other businesses is the aim of a project that is being conducted by Espectro in Campinas (São Paulo State, Brazil) and is supported by FAPESP’s Innovative Research in Small Business Program (PIPE) da FAPESP.
The project, which began in December 2022, is designed to enable producers to see how much water is available in their sources for six months ahead, so that they can make plans, invest in water reserves, and even plant a different crop, among other possible decisions.
In an era of climate change and altered rainfall patterns, this predictability will help producers maintain quality even in the event of a water crisis such as the severe drought that ravaged São Paulo State in 2021, leading municipal water boards to seal off irrigation pumps and redirect water to consumers.
For PIPE Phase 2, Espectro is conducting a study of rainfall regimes in selected drainage basins using weather forecasting data from official agencies and satellite images. By analyzing changes to the storage units involved, it will be able to create an artificial intelligence algorithm that will show farmers how their water sources and irrigation systems behave.
The project is called PalmaFlex UmiSolo-Total, and is a modular extension to the startup’s IoT (internet of things) platform (PalmaFlex UmiSolo), which in 2019 began real-time monitoring of available groundwater as a basis for irrigation recommendations.
In the first module, sensors installed underground at various depths collect data as often as the customer wishes. The data goes to the cloud, where it is converted into stratified soil moisture measurements. The platform correlates this with other data and presents the results in graphs, tables and alerts, simplifying the lives of farmers, agronomists and other people who work in the countryside.
Complete data communication solution
The system was the brainchild of electrical engineers Adilson Chinatto and Cynthia Junqueira, Espectro’s owners. They wanted to develop a product of their own inspired by their experience in signal transmission and telemetry so as to take the firm to a level beyond consulting, research and development, in which they have been working for 20 years.
They set out to create a complete data communication solution that would be modular and easily scalable and would have a wide array of applications in agriculture and industry, with a common core. Its mission was to meet farmers’ needs in monitoring and data communications.
The result was PalmaFlex, as the platform is called. Its logo is an Opuntia or prickly pear cactus (palma in Portuguese), a valuable forage source with an abundance of water as well as essential nutrients. The common core consists of a radio transceiver capable of inexpensively establishing long-distance links with various devices, especially sensors.
In the soil moisture module, a battery- and solar-powered device receives data from the sensors and sends it to the cloud. Calculations are performed, and the results can be accessed on the platform’s website via cell phone, tablet or computer.
The hardware, software and communication protocols were all developed by Espectro. “In urban areas, data networks are available both via cell phone and for IoT specifically, including LoRa, Sigfox and ZigBee, but these networks don’t cover rural areas, which aren’t profitable for the providers,” Chinatto said.
This gap led the startup to opt for vertical solutions. It began by developing a LoRA-based concentrator or gateway using the same techniques as large-scale metropolitan networks but in a small device capable of covering 3,000 hectares (the average area of medium-size farms). Users install the device themselves, connecting it via Wi-Fi or cable, with an antenna on the roof of their home or some nearby structure.
The data collector and sensors are installed out in the fields, transmitting data autonomously at preset intervals. “In an agricultural setting, the frequency doesn’t have to be very high. Every 5 minutes is more than enough. In industrial settings, we can shorten the interval between data bursts,” Chinatto said.
The cost of this coverage is less than BRL 1.00 (now about USD 0.18) per hectare. “That’s very little compared with other existing systems,” Junqueira said. Moreover, it reduces the amount of checkups and note-taking, while economizing inputs, water and power in a range of applications, assuring gains for the customer, whether agricultural or industrial.
More modules
The evolution of the system was driven by demand from customers, who asked for soil monitoring to be complemented by a weather station to monitor wind speed, rainfall and solar radiation, among other variables. This was achieved by adding aerial sensors that provide weather-related data as well as indicating daily evapotranspiration and crop spraying windows.
The monitoring was later extended to other types of farm infrastructure, such as electric currents in motors and pumps. “When irrigation pumps are turned on at night because the power is cheaper then, farmers often realize one phase isn’t working but maintenance crews aren’t available to fix it because they only do daytime shifts,” Junqueira said.
PalmaFlex detects such events by design, sending a phase loss alert to enable the user to take emergency action, fix the problem, or offset the loss on the following day. In artesian and semi-artesian wells, water flow and static and dynamic water levels can be monitored, with preventive measures including alerts in the event of imminent pump failure, for example, so that less periodic maintenance is required.
The platform is fully customizable and can be configured according to the farmer’s requirements. To customize irrigation monitoring, for example, the user provides details of methods, equipment, soil texture, and crop (type, sowing or planting date etc). The platform correlates this information with data from the field sensors in order to calculate when and how much to irrigate. The platform’s business intelligence software displays the results in tables and graphs, and the user can export them in spreadsheets. The information is kept in the database for a year.
The platform currently offers four modules: Agriculture, Animal Comfort & Water Quality, Industry, and Water Storage Forecast. The latter, partly developed under the PIPE project, is already operational and being marketed on the platform. “The module developed so far monitors the level of water or vinasse, the liquid residue left over from ethanol and sugar production. We recently installed ten systems in a sugar plant in São Paulo State to monitor ten vinasse tanks,” Junqueira said.
In this module, the sensor does not touch the liquid but measures the level and calculates the volume on the basis of the dimensions of the tank informed by the user. By measuring the volume of vinasse, the system enables the operator to prevent spillage, which would pollute the soil and incur a fine. In addition, the platform enhances the rationality of vinasse use for fertigation (application of fertilizer or nutrients via the irrigation network), reducing costs and assuring operation in compliance with environmental laws and standards.
Because the module developed under the aegis of the PIPE project combines everything the platform already does, including new types of monitoring and forecasting, Espectro considers it an ecosystem and has given it the brand name PalmaFlex Total.
Novel sensors
Besides the Water Storage Forecast module, a second aspect of the ongoing PIPE project is the development of simpler and more precise sensors. The water storage level sensor is an example. Its application in vinasse tanks has already been mentioned, and several have been installed in strategic water tanks or reservoirs to collect data for use by the AI algorithms the firm is developing.
A new soil moisture sensor nearing completion and undergoing laboratory tests will collect data at various depths using nothing more than a rod. It can work with the existing four-port data collector or transmit data by radio autonomously using the PalmaFlex protocol.
The key advantage of this system, according to the firm’s owners, is the integration of local and remote sensing modules with weather forecasting and other predictive algorithms for calculating and estimating the future water balance and water supply throughout the catchment area in which an irrigation system is located. The firm has experience with the estimation of local water balances using soil and weather sensors thanks to the agricultural modules it has sold to date.
“We estimate the depth of the roots, which absorb water from the soil, and with this parameter plus sensor readings we provide an alert dashboard showing the amount of water available and the day-to-day changes in water storage,” Chinatto said.
Similarly, the firm already supplies weather forecasts and related information as a crop spraying guide (e.g. temperature, relative humidity, wind speed, and delta T, a key parameter for agronomists to judge whether the conditions are right for spraying).
The challenge posed by the PIPE project entails extending these techniques until they provide regionalized rather than localized information. This requires working in a multiplatform environment in which sensors, data merging, modelling, machine learning and AI are integrated.
Highly qualified team
Complex missions call for first-class brains in order to succeed. “Finding the right people was hard, but we managed to bring two PhD candidates on board from the School of Electrical and Computer Engineering [FEEC] at UNICAMP [the State University of Campinas]: Marcos Ricardo Covre and Douglas Henrique Siqueira Abreu,” Junqueira said.
The researchers are working on the development of algorithms, testing, and validation. Covre is analyzing the available theoretical and practical studies. For example, he keeps in touch with EMBRAPA Cerrados, a unit of the Brazilian Agricultural Research Corporation (EMBRAPA) that conducts R&D on sustainability technologies for this savanna-like biome and has studied the dynamics of water bodies used for irrigation in various regions. Abreu focuses on cross-referencing the available databases in order to refine the information provided to irrigators.
In addition, Covre is working on mathematical modeling of water balances and the flow of water into and out of natural reservoirs. “Part of the challenge is discovering the shape of the reservoir in order to measure the volume of water it contains and make predictions for a six-month horizon,” he said.
Discovering the shape and depth without going to the site is Abreu’s challenge. He is working with satellite imagery and other kinds of data to surmount it. A combination of statistical analysis, rainfall tables and AI is expected to lead to predictions of possible future lake areas. “We’re aiming at a generalist algorithm that can work with any lake,” he said.
Although there are more models to work on, more data to collect and more lakes to measure so as to assure the generality of the model and obtain a good product involving AI, the current model gets about 80% of the variations right, according to Chinatto.
Besides Covre and Abreu, who are funded by FAPESP, the project also involves Fernando Ikedo, an engineer and marketing person with a scholarship from the São Paulo unit of SEBRAE, the Brazilian Small Business Support Service. During the first year of the project, it also involved José Matheus de Britto, an agronomist who helped design and formulate test specifications for the new soil sensor.
Without these outstanding professionals and grants, it would not have been possible to develop PalmaFlex Total, the firm’s owners stressed. “Without FAPESP we wouldn’t have had access to professionals with the baggage they brought to the firm. They guaranteed the success of the project,” Chinatto said.
The goal is to launch PalmaFlex Total on the market by end-2025, including information relating to state water board permits to use water for irrigation and to water availability.
* Image by Fernando Augusto from Pixabay