Tree mortality due to global change—including range expansion of invasive pests and pathogens— is one of the greatest threats to forest ecosystems. As such, there is increasing interest in predicting and monitoring tree decline. Several studies have identified ecophysiological indicators of tree decline. Despite their accuracy, measuring these indicators is time-consuming and often limits the type and number of measurements that can be taken. However, advances in spectroscopic technology enable acquisition of biologically meaningful, high resolution data from leaf to landscape scales. Here, we use leaf-level visible and near-infrared hyperspectral reflectance to estimate several ecophysiological indicators of tree wilt and decline. We performed two experiments: 1) a field experiment in which red oak (Quercus ellipsoidalis and Quercus rubra) saplings inoculated with oak wilt (Bretziella fagacearum) experienced crown wilt leading to death and 2) an outdoor potted experiment in which Q. ellipsoidalis saplings experienced drought stress. We measured leaf spectral reflectance (400 – 2400 nm) and several ecophysiological variables related to leaf water or photosynthetic status. We developed partial least square regression (PLSR) models to predict ecophysiological indicators of tree decline based on spectra. We applied these models to an independent set of trees not used to build the models.
Results/Conclusions
Models developed from spectral reflectance strongly predicted ecophysiological indicators of tree wilt and decline, including relative and volumetric water content, turgor loss recovery capacity, osmotic potential, maximal quantum efficiency (Fv/Fm), nonphotochemical quenching (qN) and, to a lesser degree, water and pressure potentials. Models were able to accurately predict plant physiological status of non-stressed, oak wilt diseased, and droughted saplings even when trained with independent field experimental data. Regardless of treatment, most ecophysiological indicators showed relationships between observed and predicted values close to the 1:1 line. These results suggest that leaf spectral reflectance captures ecophysiological processes occurring within plants. As such, spectral models are able to accurately estimate tree crown decline regardless of the cause. Our findings show that it is possible to obtain rapid estimates of several ecophysiological variables and may allow us to address previously intractable questions due to limitations imposed by classic ecophysiological approaches. In particular, it may now be possible to address questions requiring a multi-trait approach for large sample sizes and questions that address canopy to landscape-level ecophysiology.