Protocol for Landsat-based monitoring of landscape change in North Coast and Cascades Network parks: Version 2.1

Copass, Catharine; Antonova, Natasha

doi:10.36967/2294109

Landscape change refers to a large suite of ecological and geomorphological processes occurring across broad spatial scales. As part of its Vital Signs Monitoring, the North Coast and Cascades Network (NCCN) of the National Park Service (NPS) developed a protocol for monitoring landscape change by detecting and mapping landscape disturbances originating from both natural and anthropogenic causes. This protocol is presently implemented for four NCCN parks: Mount Rainier (MORA), North Cascades (NOCA), and Olympic (OLYM) National Parks; and Lewis and Clark National Historical Park (LEWI). Protocol implementation involves using images from the Landsat family of satellites to detect landscape change using a set of algorithms called LandTrendr (Landsat-based Detection of Trends in Disturbance and Recovery) developed by Oregon State University’s Environmental Monitoring, Analysis, and Process Recognition (eMapR) Lab. LandTrendr tracks the spectral trajectory of Landsat image pixels through time. The main processing steps in generating landscape change data using LandTrendr include uploading the study area outline to Google Earth Engine and processing the Landsat imagery using LandTrendr scripts by smoothing the spectral index signature of each pixel into coherent segments describing periods of stability or change. The primary outputs from LandTrendr are mapped pixels where change occurred, attributed with the year when the change began (year of onset), the duration of spectral change (measured in years), and the magnitude of the spectral change. Change pixels are identified within the NPS boundaries of the four parks, plus a 16 km (10 mi) buffer modified using watershed boundaries surrounding the larger parks (MORA, NOCA and OLYM), and watershed-based areas surrounding the various units that comprise LEWI. Pixels are then grouped into polygons, or patches, based on their year of onset and adjacency. The minimum patch size for inclusion is nine pixels, or 0.8 ha (2 acres). Patches are then manually labeled by an ecologist who selects from a list of target change types: Avalanches, Clearings, Development, Fires, Mass Movements, Progressive Defoliation, Riparian, and Tree Toppling. Due to time limitations, manual labeling of patches is limited for the larger parks (MORA, NOCA and OLYM) to areas within NPS boundaries plus USDA Forest Service designated Wilderness Areas and British Columbia Provincial Parks and Ecological Reserves. For the remaining patches in the study areas surrounding the larger parks, change type labels are assigned using a machine-learning classification algorithm, Random Forests (RF), after which an ecologist validates a subset of the results to determine the accuracy of the RF classification. For LEWI, all patches within the study area are labeled manually. Data summaries are generated every four years for the entire LEWI study area and the manually labeled areas within and surrounding MORA, NOCA and OLYM. Detection of trends for the target natural and anthropogenic change types is planned every 12 years. With the development of the Landscape Change Monitoring Protocol, NCCN is one of the first natural resource management agencies in the country to use the dense Landsat time-series data for monitoring.

Protocol for Landsat-based monitoring of landscape change in North Coast and Cascades Network parks: Version 2.1

Series

Other versions

Collections