Artificial intelligence analyzes leaf temperature to identify water needs

A low-cost autonomous plant water stress sensing device was developed by Embrapa Tropical Agroindustry. The technology is based on leaf energy balance and can contribute to making more accurate and assertive decisions in irrigation management. A partnership between Embrapa, the Federal University of Ceará (UFC), the Atlântico Institute’s Technological Innovation and Scientific Exploration Laboratory (LITEC) and the Ceará-based company 3V3 Tecnologia will develop a commercial version in the coming years.

The Embrapa researcher Cláudio Carvalho says that the technology uses artificial intelligence (AI) tools to control the information collected through sensing. While the effects of water deficiency on the energy balance of leaf tissues are known, Carvalho states that the use of AI to identify patterns and control irrigation is unprecedented.

The technology is based on leaf energy balance and can contribute to more accurate decision-making in irrigation management

For Otto Sousa, a computer engineer and master's student at UFC's Postgraduate Program on Teleinformatics Engineering who is responsible for monitoring and developing the technology's computing abilities, the device offers the chance to create equipment for crop irrigation management at more affordable costs for medium-scale farmers and smallholders. 

Existing devices are very expensive, since the Brazilian industry imports almost everything that involves electronic equipment. Hence we can establish the maxim that, if we must import something, we will build something efficient, but with the cheapest electronic components possible″, Sousa adds.

Water is a fundamental resource

In agriculture, water is a crucial resource that significantly influences plant health and yield. Conditions such as lack of water in the soil, unfavorable climate conditions or even unsuitable agricultural practices can generate water stress.

Investment in sensing technologies provides an automated solution for water resource management, reducing costs and environmental impacts associated with water waste and, most importantly, avoiding damage resulting from water deficit.

The sensing system

The sensing system consists of three devices: leaf temperature sensor, aspiration psychrometer and pyranometer. The leaf temperature sensor is composed of glass encapsulated thermistors that are attached to the surface of the leaves and connected to a data collection system. The collector system uses the Steinhart-Hart equation to calculate leaf temperature in relation to air temperature and humidity. In those devices, temperature readings are taken every minute and the data is sent immediately after the collection to the data server used. Transmission is done via LoRa, a low-power radio frequency protocol.

The aspiration psychrometer, in turn, collects air temperature and humidity data, adds a timestamp to each reading and sends this set of information to the data server as well. Finally, the pyranometer checks the level of solar radiation on the plants. As it is a sensor that provides information as analog signals, the researchers developed an analog-to-digital conversion circuit that receives the information from the pyranometer and digitizes it. The converted signals ??are sent via Wi-Fi to the server.

With the information on temperature, air humidity and incidence of solar radiation, the system assesses and determines the plants' water needs. If water needs are identified, the system automatically activates the irrigation devices.

The experiment

To validate the technology in the field, the study conducted the experimental cultivation of BRS Gorutuba corn plants provided by  pela Embrapa Maize and Sorghum and Embrapa Coastal Tablelands.

Forty corn seedlings were grown in pots containing sifted soil taken from the Pacajus Experimental Station, in the countryside of Ceará state, and coconut shells. Automatic drip irrigation provided 100mL of water per minute. 

The arrangement was installed inside a greenhouse that is open on the side and covered by a commercial transparent plastic film that prevented 100% of the incidence of rain, but reduced the incident solar radiation by around 5% to 10%, which was monitored every minute using the pyranometer.

The experiment started with the sowing of four corn seeds per pot. During the germination period, all pots were hydrated daily without reaching the soil's water saturation point, and liquid fertilizer was applied. Post germination, the most robust plants were selected.

The next step was the placement of the temperature sensors, psychrometers and pyranometers; and the data started to be monitored. The results obtained in the experiment demonstrated that the device built to monitor leaf temperatures enabled the inference of the crop’s states between good hydration and water deficit. 

Next step: reaching the market

According to professor Atslands Rego da Rocha, from the Department of Teleinformatics Engineering (DETI) at the Federal University of Ceará (UFC), the prospects are that the device and its AI foundation will reach the field within two years. "We are building version 2.0 of the hardware and increasing the database for modeling using AI tools. The prospects point to an interesting solution in a few months. However, full trade will take some time", he explains. 

There is a partnership that is still informal but already ongoing, between Embrapa Tropical Agroindustry, UFC, Atlântico Institute’s Technological Innovation and Scientific Exploration Laboratory (LITEC) and the local company 3V3 Tecnologia, specializing in developing technological solutions for irrigated agriculture.

Award-winning study

During the 44^th Congress of the Brazilian Computing Society held in July in Brasília, DF, the study was presented at the Workshop on Applied Computing for Environment and Natural Resource Management (WCAMA). The presentation earned an award in the ″Best Paper″ category.

The study was carried out by Otto Sousa, Guilherme Alves, and Atslands Rocha, from the Department of Teleinformatics Engineering at the Federal University of Ceará, and co-supervised by researcher Cláudio Carvalho, from Embrapa Tropical Agroindustry.

For more information, access the study  ″Aplicação de sensores de baixo custo no suporte a tomada de decisão em irrigação de precisão″ [Application of low-cost sensors to support decision making in precision irrigation].