Spectral differentiation of oak wilt from foliar fungal disease and drought is correlated with physiological changes.

Hyperspectral reflectance tools have been used to detect multiple pathogens in agricultural settings and single sources of infection or broad declines in forest stands. However, differentiation of any one disease from other sources of tree stress is integral for stand and landscape-level applications in mixed species systems. We tested the ability of spectral models to differentiate oak wilt, a fatal disease in oaks caused by Bretziella fagacearum ``Bretz'', from among other mechanisms of decline. We subjected greenhouse-grown oak seedlings (Quercus ellipsoidalis ``E.J. Hill'' and Quercus macrocarpa ``Michx.'') to chronic drought or inoculation with the oak wilt fungus or bur oak blight fungus (Tubakia iowensis ``T.C. Harr. & D. McNew''). We measured leaf and canopy spectroscopic reflectance (400-2400 nm) and instantaneous photosynthetic and stomatal conductance rates, then used partial least-squares discriminant analysis to predict treatment from hyperspectral data. We detected oak wilt before symptom appearance, and classified the disease with high accuracy in symptomatic leaves. Classification accuracy from spectra increased with declines in photosynthetic function in oak wilt-inoculated plants. Wavelengths diagnostic of oak wilt were only found in non-visible spectral regions and are associated with water status, non-structural carbohydrates and photosynthetic mechanisms. We show that hyperspectral models can differentiate oak wilt from other causes of tree decline and that detection is correlated with biological mechanisms of oak wilt infection and disease progression. We also show that within the canopy, symptom heterogeneity can reduce detection, but that symptomatic leaves and tree canopies are suitable for highly accurate diagnosis. Remote application of hyperspectral tools can be used for specific detection of disease across a multi-species forest stand exhibiting multiple stress symptoms.

Using foliar spectral properties to assess the effects of drought on plant water potential.

Drought frequency is predicted to increase in future environments. Leaf water potential (ΨLW) is commonly used to evaluate plant water status, but traditional measurements can be logistically difficult and require destructive sampling. We used reflectance spectroscopy to characterize variation in ΨLW of Quercus oleoides Schltdl. & Cham. under differential water availability and tested the ability to predict pre-dawn ΨLW (PDΨLW) using spectral data collected hours after pressure chamber measurements on dark-acclimated leaves. ΨLW was measured with a Scholander pressure chamber. Leaf reflectance was collected at one or both of two time points: immediately (ΨLW) and ~5 h after pressure chamber measurements (PDΨLW). Predictive models were constructed using partial least-squares regression. Model performance was evaluated using coefficient of determination (R2), root-mean-square error (RMSE), bias, and the percent RMSE of the data range (%RMSE). ΨLW and PDΨLW were well predicted using spectroscopic models and successfully estimated a wide variation in ΨLW (light- or dark-acclimated leaves) as well as PDΨLW (dark-acclimated leaves only). Mean ΨLWR2, RMSE and bias values were 0.65, 0.51 MPa and 0.09, respectively, with a %RMSE between 8% and 20%, while mean PDΨLWR2, RMSE and bias values were 0.60, 0.44 MPa and 0.01, respectively, with a %RMSE between 9% and 20%. Estimates of PDΨLW produced similar statistical outcomes when analyzing treatment effects on PDΨLW as those found using reference pressure chamber measurements. These findings highlight a promising approach to evaluate plant responses to environmental change by providing rapid measurements that can be used to estimate plant water status as well as demonstrating that spectroscopic measurements can be used as a surrogate for standard, reference measurements in a statistical framework.