Early detection of Phytophthora root rot in Eucalyptus using hyperspectral reflectance and machine learning

Root rot caused by Phytophthora is a growing problem in forests. It is hard to catch because it starts in the roots. By the time a tree looks sick, the disease is well established, and current tests are slow and damage the tree.

The team infected Eucalyptus benthamii trees from 19 commercially planted families with Phytophthora alticola, then measured the light reflected from their leaves with a field instrument that splits light into many narrow bands of colour, far more than the human eye can tell apart.

Infected trees reflected light differently, and the changes were linked to shifts in leaf pigments and to water stress. The researchers trained three machine-learning models on the measurements. The best, a type of neural network, identified infected trees with 97% accuracy using a small set of key wavelengths.

Why it matters: a quick check of the leaves, without harming the tree, could let foresters find infected trees early. The study, in Computers and Electronics in Agriculture, points towards sensors that could one day check plantation health from the air.