About

Citation

Heim, RHJ; Wright, IJ; Allen, AP; Geedicke, I and Oldeland, J. (2018) Developing a spectral disease index for myrtle rust (Austropuccinia psidii), Plant Pathology, XX(xx), pp. XXX. DOI: 10.1111/ppa.12996.

Pre-print version

not available

Abstract

Since 2010 Australian ecosystems and managed landscapes have been severely threatened by the invasive fungal pathogen, Austropuccinia psidii. Highly susceptible native species are already being locally extirpated. For the lemon myrtle industry, yield losses of up to 70 percent per year have been reported and have led to increased fungicide application and costs. Yet, detecting and monitoring disease outbreaks is currently only possible by human assessors, which is slow, labour intensive and depends on an assessor’s experience. Over the last 25 years, spectral vegetation indices have been designed to assess variation in biochemical or biophysical traits of vegetation. Such indices provide an automatic and objective alternative to visual disease assessments by humans. However, diagnosis of individual diseases based on classical spectral vegetation indices is currently not possible since these indices lack disease specificity. To meet the needs of modern plant-disease detection, a new type of spectral indices must be developed. These indices should detect disease severity and incidence before symptoms become visible and discriminate between different pathogens, host-species and abiotic stresses. Here we develop a novel, spectral disease index, the lemon myrtle-myrtle rust index (LMMR). We show that the LMMR predicts infections caused by A. psidii on lemon myrtle (Backhousia citriodora) plantations with an overall accuracy of 90% for our dataset. We compare the LMMR classification accuracy to three classical spectral vegetation indices (PRI, MCARI, NBNDVI) used for stress detection, and show that the accuracies of these indices are all significantly lower (58%, 67%, 60%, respectively). If the LMMR can be validated on independent datasets from similar and different host-species, it could enable land managers to reduce costs by swiftly locating disease hotspots in their crop fields and applying fungicides in a targeted fashion. After validation, the LMMR could be transferred to hyperspectral data from different kinds of sensors, be applied on ground, aerial and satellite scales. We provide the analysis code in R to enable other researchers developing their own unique index.

Licences

Data: CC-0 attribution requested in reuse
Manuscript: CC-BY-4.0
Code: MIT year: 2017, copyright holder: René Heim