Change Detection in Noisy Time-Series Data: Forest Health Applications

The U.S. Forest Service is developing general and scalable methods for the detection and monitoring of forest health concerns in near real-time from time-series analysis of MODIS imagery. The system will be used to identify ‘hotspots’ of insect and disease activity, provide flight prioritization information to increase the efficiency of the national Aerial Detection Survey (ADS) program, and possibly enable the production of more consistent tabular and spatial forest health summary data across the nation. We use a new change detection method to generate an evolving disturbance map at an 8-day timestep and 250-meter resolution for the years 2004-2007 over much of the western U.S. Aerial Detection Survey data collected over those years is then used to evaluate the potential for detection and monitoring of spruce budworm attacks in the U.S. Pacific Northwest and mountain pine beetle outbreaks and aspen dieback in the central Rockies. Initial results suggest that the methodology can be very useful for hotspot detection and flight prioritization, and has the potential to contribute toward the creation of more consistent forest health summary data for the continental U.S.