Mapping vegetation community types in a highly-disturbed landscape: integrating hierarchical object-based image analysis with digital surface models

Authors: Rachel Snavely*, The University of Texas at Austin
Topics: Remote Sensing, Biogeography
Keywords: Object-based image analysis, high resolution aerial imagery, digital surface models, vegetation mapping, Channel Islands
Session Type: Illustrated Paper
Day: 4/13/2018
Start / End Time: 10:00 AM / 11:40 AM
Room: Canal St. Corridor, Sheraton, 3rd Floor
Presentation File: No File Uploaded

Increased demand for more accurate vegetation cover data and greater availability of high spatial resolution imagery have furthered object-based image analysis (OBIA) for faster and more reliable mapping of vegetation communities. Definitions of boundaries with spectrally heterogeneous vegetation types may be inconsistently or poorly defined, or challenges pertaining to varying percent cover or minimum mapping units (MMU) arise, rendering community-level vegetation mapping even more complex. Advances in OBIA technology, however, facilitate novel remote sensing methods to address these challenges. A combination of OBIA, high-resolution imagery, and a LiDAR-derived normalized digital surface model (nDSM) was used to map vegetation communities on San Clemente Island, the southernmost of California's Channel Islands. Its long history of ecological disturbance, ranging from exotic herbivore grazing to tactical military use, has created an extremely heterogeneous vegetative landscape. A multi-level segmentation routine was developed: the individual shrub object or patch level, and a larger community level. Community level objects were tested for MMU and percent cover rules per vegetation cover type, and boundaries between same-class adjacent objects were generalized. Next, classification sensitivity was compared for whether an nDSM was incorporated. Accuracies for the combined areas of interest (AOIs) were equal (63%) for both classification methods, but individual AOI results reveal that nDSM inclusion resulted in higher mapping accuracies (84% and 70% compared to 72% and 53%), and nDSM-aided individual vegetation classes were generally greater. These results demonstrate the effectiveness of this integrated OBIA approach, emphasizing advantages and limitations of conducting multi-scale analyses when characterizing a highly-disturbed landscape.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login