Phenotyping is critical in tree breeding, but traditional methods are often labour-intensive and not easily scalable. Resistance to biotic and abiotic stress is a key focus in tree breeding programmes. While heritable traits derived from spectral remote sensing have been identified in trees, their application to tree phenotyping remains unexplored. This study investigates in situ high-throughput hyperspectral and thermal imaging for assessing Dothistroma needle blight (DNB) resistance in Pinus radiata D.Don.

Using UAV-based hyperspectral and thermal imaging during a severe DNB outbreak in a clonal trial in New Zealand, we computed narrow-band hyperspectral indices (NBHIs), canopy temperature indices, radiative transfer inverted plant traits, and solar-induced fluorescence. Visual severity scores and remote sensing indices were modelled using spatially explicit mixed-effect linear models integrating pedigree and genomic data in a single-step genomic evaluation. Multi-trait models and sampling simulations were used to evaluate the potential of remote sensing indices to supplement or replace traditional phenotyping.

Remote sensing indices exhibited narrow-sense heritability values comparable to severity scores (up to 0.37) and high absolute correlation coefficients with severity scores (up to 0.79). Carotenoid and chlorophyll-related NBHIs were the most informative, reflecting the physiological impacts of DNB. Combining partial visual scoring with NBHIs maintained high estimated breeding value (EBV) accuracy (0.68) at 50% scoring and moderate accuracy (0.59) at 20% scoring. EBV correlation with full scoring was above 0.8 even at 20% scoring. Using solely the most heritable NBHI, achieved 0.71 breeding value accuracy and 0.79 absolute EBV correlation with severity scores, suggesting NBHIs can replace visual scoring with minimal precision loss.

By utilising UAV-based hyperspectral and thermal imaging to capture single-tree phenotypes related to disease in a forestry trial and pairing the data to genomic evaluation, this study establishes that remote sensing data offers an efficient, scalable alternative to traditional phenotyping. Our approach constitutes a major step towards characterising specific physiological responses, facilitating the discovery of the genetic architecture of physiological traits, and significantly enhancing genetic improvement.