Research explores new AI methods for monitoring water stress in regenerative cotton
Plants and agricultural systems in the Mediterranean region face significant challenges from climate change and extreme weather events like drought. Although cotton is generally considered a drought-resistant crop, continuous water stress can affect the yield and fibre quality. It’s important therefore to be able to rapidly monitor the plant’s water status, to enable irrigation to be scheduled at the right time and avoid damage and losses.
New non-invasive methods of monitoring have been explored within the Apulia Regenerative Cotton Project. This Living Lab 1.0 is supported by the Circular Bioeconomy Alliance in collaboration with the Sustainable Markets Initiative’s Fashion Task Force, coordinated by the European Forest Institute (EFI) together with the Council for Agricultural Research and Economics of Italy (CREA) and PRETATERRA. The experimental agroforestry regenerative cotton site in Rutigliano, southern Italy is testing and scientifically assessing new ways to implement sustainable cotton production in Italy.
Traditional methods for assessing cotton’s water status include measurements of soil moisture, water potential in the leaves and in the stem (the most stable and reliable indicator). However, the measurement of stem water potential involves enclosing a leaf in a foil bag and using a pressure chamber. This is a labour-intensive and time-consuming process using specialised equipment, which limits its practicality for large-scale or frequent monitoring.
In 2023, remote sensing and machine learning (AI) techniques were tested to see if they could help. Remote sensing technology measures the radiation reflected or emitted from objects, without direct physical contact. Satellites, aircraft and unmanned aerial vehicles are used to collect the data. Researchers working on the Apulia project used data from Sentinel-2 – two satellites from the European Space Agency’s Copernicus programme which provide high-resolution multispectral images. That data was then analysed using machine learning to identify patterns in the data, allowing predictions to be made.
Previous studies have investigated the use of remote sensing data to see the water status for cotton, for example using drones. However, this is the first time remote sensing data from satellites has been integrated with machine learning techniques, especially in the Mediterranean area. Different machine learning algorithms were tested and compared, to see which performed best at estimating the cotton plants’ water status, with the ‘random forest’ (RF) model coming out on top.
This approach demonstrates how high-frequency, non-invasive monitoring of cotton’s water status could help support smart irrigation strategies, improving water use efficiency in Mediterranean cotton production.
More information
Garofalo, S.P.; Modugno, A.F.; De Carolis, G.; Sanitate, N.; Negash Tesemma, M.; Scarascia-Mugnozza, G.; Tekle Tegegne, Y.; Campi, P. Explainable Artificial Intelligence to Predict the Water Status of Cotton (Gossypium hirsutum L., 1763) from Sentinel-2 Images in the Mediterranean Area. Plants 2024, 13, 3325. https://doi.org/10.3390/plants13233325
