Characterizing spectral-temporal patterns of defoliator and bark beetle disturbances using Landsat time series

Defoliators and bark beetles are natural disturbance agents in many forest ecosystems around the world. Mapping the spatial and temporal patterns of insect disturbance dynamics can help in understanding their impacts on forest ecosystem resilience and functioning, and in developing adaptive management strategies. In recent years, much progress has been made in landscape-level analyses of insect-induced disturbances using remotely sensed data. However, many studies have focused on single insect agents or aggregated different insect agents into a single group. In this study, we characterized the temporal-spectral patterns associated with bark beetle and defoliator disturbances using Landsat time series between 1990 and 2013, with the objective to test if the two insect disturbances can be separated with Landsat data. We analyzed a recent outbreak of mountain pine beetle (Dendroctonus ponderosae Hopkins) and western spruce budworm (Choristoneura freemani Razowski) in British Columbia, Canada. To characterize the disturbance and recovery trends associated with insect disturbances we used the LandTrendr segmentation algorithm. We fitted LandTrendr spectral trajectories to annual normalized burn ratio (NBR) and Tasseled Cap (TC) time series, from which we then extracted a set of disturbance metrics. With these disturbance metrics, two random forest models were trained to a) distinguish insect disturbances from harvest and fire disturbances; and to b) attribute the insect disturbances to the most likely agent, i.e. mountain pine beetle or western spruce budworm. Insect disturbances were successfully mapped with an overall accuracy of 76.8%, and agents were successfully attributed with overall accuracies ranging from 75.3% to 88.0%, depending on whether only pure host-stands or mixed stands with both insect hosts were considered. In the case of mixed host stands, nearly 45% of the western spruce budworm disturbances were falsely attributed to mountain pine beetle. Spectral metrics describing disturbance magnitude were more important for distinguishing the two insect agents than the disturbance duration. Spectral changes associated with western spruce budworm disturbances had generally lower magnitudes than mountain pine beetle disturbances. Moreover, disturbances by western spruce budworm were more strongly associated with changes in TC greenness, whereas disturbances by mountain pine beetle were more strongly associated with changes in TC brightness and wetness. The results reflect the ephemeral nature of defoliators versus the tree mortality impacts of bark beetles in our study area. This study demonstrates the potential of Landsat time series for mapping bark beetle and defoliator disturbances at the agent level and highlights the need for distinguishing between the two insect agents to adequately capture their impacts on ecosystem processes.