Estimating vegetation height from WorldView-02 and ArcticDEM data for broad ecological applications

Arjan JH Meddens, University of Idaho, ameddens@uidaho.edu
Lee A. Vierling, University of Idaho, leev@uidaho.edu (Presenter)
Jan Eitel, University of Idaho, jeitel@uidaho.edu
Jyoti Jennewein, University of Idaho, jjennewein@uidaho.edu
Natalie Boelman, Lamont-Doherty Earth Observatory, Columbia Univ., nboelman@ldeo.columbia.edu
Joanne C. White, Canadian Forest Service, joanne.white@canada.ca
Michael A. Wulder, Canadian Forest Service, mike.wulder@canada.ca

Boreal and arctic regions are warming at an unprecedented rate, and at a rate higher than in other regions across the globe. Ecological processes are highly responsive to temperature and therefore substantial changes in these northern ecosystems are expected. Recently, NASA initiated the Arctic-Boreal Vulnerability Experiment (ABoVE), which is a large-scale field campaign that aims to gain a better understanding of how the arctic responds to environmental change. High-resolution data products that quantify vegetation structure and function will improve efforts to assess these environmental change impacts. Our objective was to develop and test an approach that allows for mapping vegetation height at a 5m grid cell resolution across the ABoVE domain. To accomplish this, we selected three study areas across a north-south gradient in Alaska, representing an area of approximately 130 km2. We developed a RandomForest modeling approach for predicting vegetation height using the ArcticDEM (a digital surface model produced across the Arctic by the Polar Geospatial Center) and high-resolution multispectral satellite data (WorldView-2) in conjunction with aerial lidar data for calibration and validation. Vegetation height was successfully predicted across the three study areas and evaluated using an independent dataset, with R2 ranging from 0.58 to 0.76 and RMSEs ranging from 2.2 to 2.5 m. This predicted vegetation height dataset also led to the development of a digital terrain model using the ArcticDEM digital surface model by removing canopy heights from the surface heights. Our results show potential to establish a high resolution pan-arctic vegetation height map, which will provide useful information to a broad range of ongoing and future ecological research in high northern latitudes.

Associated Project(s): 

• Vierling-01
• Boelman-01
• Eitel-01

Poster Location ID: 31

Session Assigned: Vegetation Dynamics and Distribution