Detecting Forest Disturbance in the Pacific Northwest From MODIS Time Series Using Temporal Segmentation

Damien Sulla-Menashe, Boston University, sullamenashe@gmail.com (Presenter)
Zhiqiang Yang, Oregon State University, zhiqiang.yang@oregonstate.edu
Justin Braaten, Oregon State University, justin.braaten@oregonstate.edu
Olga Krankina, Oregon State University, olga.krankina@oregonstate.edu
Robert E Kennedy, Oregon State University, robert.kennedy@oregonstate.edu
Mark Friedl, Boston University, friedl@bu.edu

Changes to the land surface of the Earth are occurring at unprecedented rates with significant implications for the surface energy balance and regional to global scale cycles of carbon and water. Data from the Moderate Resolution Imaging Spectroradiometer (MODIS) aboard the Aqua and Terra satellite platforms provide over 11 years of consistent, synoptic observations of the biosphere. New methods have recently emerged to analyze time series of remote sensing images, thereby providing ecologically important information about disturbance and succession over large regions. In particular, the Landtrendr algorithm was developed to characterize long-term trends, including punctual and gradual disturbance events and subsequent vegetation regrowth, in dense time series of Landsat imagery. While this approach has shown to be useful and robust in a wide range of ecosystems, its application is limited to areas with sufficient Landsat archive depth and relatively cloud-free periods. Additionally, the approach requires significant effort in atmospheric correction and normalization steps, increasing the cost for large-area application. Here we present an adaptation of the Landtrendr algorithm to an 11-year time series of MODIS Normalized BRDF-Adjusted Reflectance (NBAR) data to detect forest disturbance in the Northwest Forest Plan (NWFP) area of Washington, Oregon, and California. The NWFP area represents a dynamic zone of forest management with an active disturbance regime that includes insect defoliation, wildfires, and logging. This work aims to explore how the size and severity of disturbance events influence detection and characterization of such events using MODIS data. We sampled disturbance events across gradients of size and severity that occurred during the MODIS era (2000-present) using a high-quality database of forest disturbance information derived from Landsat. One-third of these disturbance records were used to calibrate the model using MODIS NBAR time series, and the remaining two-thirds were kept aside for assessment of model performance. The results demonstrate how the Landtrendr approach can be expanded for use at continental and global scales with data of moderate spatial resolution such as MODIS.

Presentation Type:  Poster

Session:  Global Change Impact & Vulnerability   (Tue 11:30 AM)

Associated Project(s): 

• Friedl, Mark: Using MODIS to Monitor Dynamics in Land Cover and Phenology at Seasonal to Decadal Time Scales ...details
• Krankina, Olga: Contribution to studies of LCLUC in Northern Eurasia ...details

Poster Location ID: 288