ABSTRACT FINAL ID: B31K-07;

TITLE: A Statistical Methodology for Detecting and Monitoring Change in Forest Ecosystems Using Remotely Sensed Imagery

SESSION TYPE: Oral

SESSION TITLE: B31K. Identifying and Quantifying Change in Ecological Systems I

AUTHORS: Richard T. Mills^1,2, Jitendra Kumar¹, Forrest M. Hoffman¹, William W. Hargrove³, Joseph Spruce⁴

INSTITUTIONS:
¹Oak Ridge National Laboratory, Oak Ridge, TN, United States.
²Electrical Engineering and Computer Science, University of Tennessee, Knoxville, TN, United States.
³Eastern Forest Environmental Threat Assessment Center (EFETAC), USDA Forest Service, Asheville, NC, United States.
⁴NASA Stennis Space Center, Bay St. Louis, MS, United States.

ABSTRACT BODY: Variations in vegetation phenology, the annual temporal pattern of leaf growth and senescence, can be a strong indicator of ecological change or disturbance. However, phenology is also strongly influenced by seasonal, interannual, and long-term trends in climate, making identification of changes in forest ecosystems a challenge. Forest ecosystems are vulnerable to extreme weather events, insect and disease attacks, wildfire, harvesting, and other land use change. Normalized difference vegetation index (NDVI), a remotely sensed measure of greenness, provides a proxy for phenology. NDVI for the conterminous United States (CONUS) derived from the Moderate Resolution Spectroradiometer (MODIS) at 250 m resolution was used in this study to develop phenological signatures of ecological regimes called phenoregions. By applying a quantitative data mining technique to the NDVI measurements for every eight days over the entire MODIS record, annual maps of phenoregions were developed. This geospatiotemporal cluster analysis technique employs high performance computing resources, enabling analysis of such very large data sets. This technique produces a prescribed number of prototypical phenological states to which every location belongs in any year. Analysis of the shifts among phenological states yields information about responses to interannual climate variability and, more importantly, changes in ecosystem health due to disturbances. Moreover, a large change in the phenological states occupied by a single location over time indicates a significant disturbance or ecological shift. This methodology has been applied for identification of various forest disturbance events, including wildfire, tree mortality due to Mountain Pine Beetle, and other insect infestation and diseases, as well as extreme events like storms and hurricanes in the U.S. Presented will be results from analysis of phenological state dynamics, along with disturbance and validation data.

KEYWORDS:
[0430] BIOGEOSCIENCES / Computational methods and data processing,
[1914] INFORMATICS / Data mining,
[0480] BIOGEOSCIENCES / Remote sensing,
[0468] BIOGEOSCIENCES / Natural hazards.

SPONSOR NAME: Richard Mills

CONTACT: Richard Mills <rmills at ornl dot gov>