Repository logo
 

Estimating pre-fire forest structure with stereo imagery and post-fire lidar

Date

2016

Authors

Filippelli, Steven, author
Lefsky, Michael, advisor
Rocca, Monique, committee member
Sibold, Jason, committee member

Journal Title

Journal ISSN

Volume Title

Abstract

Lidar has become an established tool for mapping forest structure attributes including those used as inputs for fire behavior and effects modelling. However, lidar is rarely available to document pre-fire conditions due to its sparse availability. In contrast, aerial imagery is regularly collected in many regions, and advances in stereo image matching have enabled the creation of dense photogrammetric point clouds similar to those from lidar. As part of a study of the physical and ecological impacts of the 2012 High Park Fire, we generated a photogrammetric point cloud from pre-fire aerial imagery collected in 2008 and calculated forest height using a digital terrain model generated from a 2013 post-fire lidar collection. A suite of canopy height and density metrics were created from both the pre-fire photogrammetry and the post-fire lidar point clouds. These metrics were compared to each other and to forest structure attributes measured in the field. For unburned areas, we found strong relationships between corresponding lidar and photogrammetry height and density metrics with biases that were consistent with known differences in each sensor’s method of sampling the canopy. Regressions models of field-measured forest structure attributes incorporating both lidar and photo metrics demonstrated that a single equation could estimate some forest structure attributes without significant intercept or slope bias due to the source of the metrics (i.e. photo or lidar). Models of aboveground biomass on unburned plots had similar root mean square errors for lidar (29.3%), photogrammetry (31.0%), and combined data sources (RMSE = 29.1% and source intercept bias = 34.64 Mg ha-1 and slope bias = -0.28). Similar results were obtained for Lorey's height, basal area, and canopy bulk density. Models of structure in burned areas derived from post-fire lidar had lower performance than photogrammetry due to the fire's consumption of canopy materials which generally reduced the explanatory power of lidar density metrics. Pre-fire forest structure information could aid assessments of contributing factors such as canopy fuels and fire effects such as loss of biomass. The wide spatial and temporal coverage of aerial photos and growing coverage of lidar could enable many other applications of combining photogrammetry with lidar, including assessments of changes in forest carbon storage.

Description

Rights Access

Subject

Citation

Associated Publications