WWETAC Projects

Project Title:  A multi-scale approach for remotely mapping pine-beetle attacks over time and associated fire hazard

Principal Investigator:   Russell Parsons, USDA Forest Service RMRS Fire lab, Missoula, MT

rparsons[at]fs.fed.us

Collaborators:   W. M. Jolly, USDA Forest Service RMRS Fire lab; Erin Landguth, University of Montana Computation Ecology Lab; Zachary Holden, USDA Forest Service Region 1, Missoula, MT

Key Issues/Problem Addressed:              

Forests throughout the western United States and Canada are currently undergoing rapid changes as a result of extensive outbreaks of mountain pine beetle (MPB, Dendroctonus ponderosae Hopkins).  Because the changes in forest composition and structure associated with these attacks have significant implications for fire management, as well as hydrology and other ecosystem processes over time, there is a pressing need for early detection and mapping systems to provide reliable information regarding the nature and extent of beetle attacks.

Setting and Approach:  

Current literature addressing the impacts of beetle attacks on fire behavior offers conflicting results. Much of the current misunderstanding stems from the use of fire behavior models that are not adequate for evaluating fire hazard in beetle-kill conditions. An emerging new class of dynamic, physics-based fire behavior models, however, show great promise in increasing our understanding of implications of beetle attacks for fire behavior because they are capable of addressing fuel heterogeneity as well as changes in fuel structure over time. Despite the promise these models show for increasing our understanding of interactions between insect-induced change and fire behavior, making this information useful for landscape and fire planning requires translating CFD model outputs to field-based and remotely sensed structure and fuels attributes.

We will develop a prototype methodology for detecting and mapping beetle attacks over time. Importantly, we will develop methods for identifying the occurrence of beetle outbreak, and the magnitude of change (canopy cover, canopy base height) associated with beetle attack. Our approach will incorporate multi-scale imagery: rapid detection of beetle attacks with MODIS data, trajectory-based change detection analysis carried out on a Landsat TM image time series, and detailed change assessments using NAIP high resolution imagery. Image data will be supported by detailed field data consisting of 210 intensified Forest Inventory and Analysis (FIA) plots that were installed in 2007, and then re-measured in 2009 following a severe mountain pine beetle infestation that caused high tree mortality in many plots. In Phase 2, fuels maps developed with the remote sensing analysis in Phase 1 will be used to provide detailed fuels inputs to these advanced models to facilitate exploration of MPB/fire interactions.

Progress to Date:

We have processed 425 Landsat TM5 images (PR4028) over the Helena National Forest and have downloaded and processed all available (2000‐2011) MODIS NDVI and EVI images. We developed and tested a new method for detecting changes in image time series, using the Bayesian Change Point algorithm (BCP). We have applied the algorithm to both Landsat and MODIS time series data. Our preliminary results suggest that detection of onset date with Landsat performs poorly, but that MODIS data are capable of detecting onset date with reasonable (~60% accuracy). We have applied the algorithm across the study area, producing maps of estimated date of onset of attack. We are now in the process of comparing the BCP algorithm with another recently developed change detection algorithm called Breaks For Additive Seasonal Trend (BFAST). The algorithm that performs best will be then be refined and used to develop final maps of onset of MPB‐induced mortality across the study area. These outputs will then be used to assign potential fire behavior attributes for each pixel across the study domain.

Impacts/Applications:  

The mapping effort, in conjunction with the intensively-sampled FIA field data, will provide a new method to estimate the magnitude and trajectory of fuel structure changes in beetle-killed stands. Additionally, developing methods to produce fire behavior modifier surfaces that can be applied across the landscape will significantly advance our ability to apply computationally intensive CFD models to address landscape scale problems.

WWETAC ID:      FY11NG98