COS 268-4 - Can we map tree diseases from space? The case of oak wilt disease.

Abstract: Protecting the future of forests relies on our ability to observe changes in forest health conditions. Thus, developing tools for remotely sensing diseases in a timely fashion is critical for managing threats at a large scale. Here we propose a novel workflow for mapping oak wilt —a disease caused by a pathogen fungus (Bretziella fagacearum)— by targeting the temporal pathology through symptoms using land surface phenology (LSP) from spaceborne observations. We hypothesize that phenological changes in pigments of oak trees due to oak wilt disease can be tracked using LSP metrics derived from the Chlorophyll/Carotenoid Index (CCI). We used dense time-series observations from Sentinel-2 to create Analysis Ready Data across Minnesota and Wisconsin and to derive three LSP metrics: the value of CCI at the start and end of the growing season, and the coefficient of variation of the CCI during the growing season. We also integrate high-resolution airborne imagery from different locations and years to select pixels (n = 3914) from the most common oak tree conditions on land: healthy, symptomatic to disease, and dead trees. These pixels were used to train an iterative Partial Least Square Discriminant (PLSD) model and derive the probability of an oak tree (ie., pixel) of being in such conditions and its uncertainty. We assessed these models spatially and temporarily, revealing that it is feasible to discriminate diseased trees among oak tree conditions with overall accuracies that ranged between 80 to 82%. Within conditions, our predicting models suggest that spatial variations in these three LSP can predict healthy (Area Under the Curve (AUC) = 0.98 ± 0.02), symptomatic (AUC = 0.88 ± 0.05), and dead (AUC = 0.94 ± 0.04) oak trees with low rates of false positive. Such performance is also observed at different years, suggesting that LSP metrics from CCI are robust descriptors of oak tree conditions; thus, indicators of disease. Our results highlight the capabilities of LSP metrics from dense spaceborne observations to map diseases and their importance for monitoring changes in biodiversity at large scales.