New remote sensing method to identify how Xylella affects trees

Researchers from the Institute of Sustainable Agriculture (IAS), of the Higher Council for Scientific Research (CSIC), in Córdoba, have demonstrated the existence of specific spectral indicators that allow differentiating the stress in trees associated with the bacterium Xylella fastidiosa from other causes of stress , such as those derived from the lack of water. The finding, which has been published in the journal Nature Communications, has been the result of the use of remote sensing techniques monitoring areas affected by this plant disease.

Xylella fastidiosa is the highest risk pathogen internationally, being able to infect more than 550 plant species. Coming from America, its identification in Europe, devastating the olive grove of southern Italy and later in Spain, poses a threat to agriculture at an international level due to its rapid expansion.

“In this work we use image spectroscopy techniques using hyperspectral sensors on board manned aircraft to scan more than a million trees in areas infected by this bacterium and different levels of water stress in healthy trees. We demonstrate the existence of specific spectral indicators that make it possible to differentiate physiological changes associated with these diseases from those caused by water stress ”, says Pablo J. Zarco-Tejada, lead author of the article.

Together with him, six other researchers from the IAS have worked (JA Navas-Cortes, BB Landa, V. Gonzalez-Dugo, A. Hornero, M. Román-Écija and MP Velasco-Amo), and they have had the collaboration of the University of Melbourne (Australia), Cornell University (United States), the Joint Research Center (JRC) of Ispra (Italy), the University of Swansea (United Kingdom) and the Institute per la Protezione Sostenibilie delle Piante, of the CNR (Italy ).

Detection of diseases through remote sensing techniques is a critical step to monitor infected areas in early stages that allow their eradication and possible treatment. Previous studies demonstrate the use of remote sensing images for this purpose, but the results obtained when different types of stress are mixed (biotic vs. abiotic) make it difficult to use them in large-scale plant health programs.

“In this study we show that hyperspectral remote sensing and machine learning algorithms, fed by physical models of radiative transfer, allow us to differentiate stress caused by pathogens from that caused by causes associated with abiotic origin. We show that there are spectral indicators characteristic of each disease, and that these patterns are specific for each species and pathogen ”, explains the researcher.

“And, fundamentally, we proved that these spectral indicators are modulated by the level of water stress. This specificity and characterization of the modulation allow the use of image spectroscopy to monitor large areas and detect differences between types of stress that occur simultaneously naturally, obtaining results that exceed 90% accuracy in the detection of these diseases «, the investigator ends.

More information: