Multiscale remote sensing of plant pathogens: Detecting and monitoring myrtle rust
Rene Heim: Macquarie University
<div>Australia’s natural and managed landscapes, dominated by the plant family Myrtaceae, are under threat from a devastating, invasive disease called myrtle rust (<em>Austropuccinia psidii)</em>. Already, the lemon myrtle (<em>Backhouse citriodora</em>) essential oil industry has suffered yield losses up to 70 percent per year. Due to recurring infections, natural populations of several highly susceptible native plant species seem likely to go extinct. Currently, detecting and monitoring disease outbreaks is only possible by eye, aided by forecasting infection periods, including co-incident timing of spore release and host growth flush, using weather data. These methods can be resource intensive and sometimes unreliable. Fungicides can only be applied once clear disease symptoms are visible, which is wasteful, ineffective and resulting in high costs. However, optical remote sensing techniques are well known for objective and reliable automated diagnosis of plant diseases. Combined with advanced data analysis techniques, sustainable and targeted pest management systems can be developed. This study aims at establishing a proof-of-concept for monitoring and detection of myrtle must. At the ICPP 2018 in Boston, we would like to present results of our first sub-project where we could discriminate healthy and infected lemon myrtle trees with an accuracy of 95% based on their leaf spectral characteristics. In particular, we would like to present (i) the results of our classification of hyperspectral data collected at the leaf level; (ii) a self-designed myrtle rust-specific vegetation index and introduce and an approach to validate this index with multispectral aerial imagery, and iii) provide an outlook on upcoming projects, especially on data we collected for other plant species to extend our spectral library of myrtle rust specific signatures.</div>
View Presentation |
