Relative impact of previous disturbance history on the likelihood of additional disturbance in the Northern United States Forest Service USFS Region

Abstract:

The Landsat archive is increasingly being used to detect trends in the occurrence of forest disturbance. Beyond information about the amount of area affected, forest managers need to know if and how disturbance regimes change. The National Forest System (NFS) has developed a comprehensive plan for carbon monitoring that requires a detailed temporal mapping of forest disturbances across 75 million hectares. A long-term annual time series that shows the timing, extent, and type of disturbance beginning in 1990 and ending in 2011 has been prepared for several USFS Regions, including the Northern Region. Our mapping starts with an automated detection of annual disturbances using a time series of historical Landsat imagery. Automated detections are meticulously inspected, corrected and labeled using various USFS ancillary datasets. The resulting maps of verified disturbance show the timing and types are fires, harvests, insect activity, disease, and abiotic (wind, drought, avalanche) damage. Also, the magnitude of each change event is modeled in terms of the proportion of canopy cover lost. The sequence of disturbances for every pixel since 1990 has been consistently mapped and is available across the entirety of NFS. Our datasets contain sufficient information to describe the frequency of stand replacement, as well as how often disturbance results in only a partial loss of canopy. This information provides empirical insight into how an initial disturbance may predispose a stand to further disturbance, and it also show a climatic signal in the occurrence of processes such as fire and insect epidemics. Thus, we have the information to model the likelihood of occurrence of certain disturbances after a given event (i.e. if we have a fire in the past what does that do to the likelihood of occurrence of insects in the future). Here, we explore if previous disturbance history is a reliable predictor of additional disturbance in the future and we present results of applying logistic regression to obtain predicted probabilities of occurrence of additional disturbance types. We describe responses in additional disturbance and prominent trends for each major forest type.