Tree pathogens and abiotic stresses are increasing threats to tree diversity with global change. Accurately diagnosing the cause of tree stress symptoms is critical in evaluating forest health and mitigating risks. We used a greenhouse experiment of two northern US deciduous oaks, pin oak (Quercus ellipsoidalis) and bur oak (Q. macrocarpa), to treat saplings with either 1) drought (4% soil VWC), 2) stem inoculations of Bretziella fagacearum, the vascular wilt pathogen that causes oak wilt, or 3) leaf inoculations of Tubakia iowensis, the pathogen responsible for bur oak blight, which causes leaf and petiole necrosis. We measured leaf-level hyperspectral reflectance over 8 weeks and monitored leaf physical symptoms and physiological function (water potentials, photosynthetic rates, stomatal conductance). We expected similarities between drought and oak wilt symptoms and expected more severe symptoms of oak wilt in Q. ellipsoidalis than in Q. macrocarpa, given known differences in their infection responses. We asked whether physiological symptoms were different among treatments, and whether and when we could correctly classify treatments from hyperspectral profiles or predict physiological symptoms (PLS-DA and PLS-R models, respectively).
Results/Conclusions
Disparities in leaf- to plant-level effects of stress influence detection. Physical symptoms appeared most rapidly in leaf-specific bur oak blight (7 days), then oak wilt (Q. ellipsoidalis 15 days, Q. macrocarpa 60 days). The whole plant effects of drought, and later in the experiment, leaf-vascular disruption of bur oak blight, caused treated plants to have significantly lower water potentials, stomatal conductance, and photosynthetic rates than oak wilt or control plants. We found significant treatment differences in spectral signatures (coefficients of PLS-R models) predicting physiological traits or stress progression. Classification model (PLS-DA) accuracy was greatest when constructed from spectra collected before the appearance of most symptoms for Q. ellipsoidalis (18-25 days, Kappa = 0.22 +/- 0.12, accuracy = 0.48 +/- 0.08) and assignment sensitivity was higher for control and drought treated plants than oak wilt infected plants. Classification was weaker for Q. macrocarpa treatments, where oak wilt symptoms also appeared much later (35 days, Kappa = 0.13 +/- 0.09), and sensitivity was highest for bur oak blighted plants. Leaf-level spectral measurements are not yet standalone diagnostic tools, however, clear spectral changes emerged among treatments. Differences in whole plant (drought), leaf specific (bur oak blight), or plant-controlled segmented responses (oak wilt) underscore the need to integrate whole plant spectral reflectances to bridge the gap between leaves and canopies in stress diagnosis.