Evaluating terrain elevation estimates from stereogrammetric high resolution spaceborne imagery with airborne lidar

Abstract:

Circumboreal forests contain vast amounts of carbon and are currently subjected to rapid climate warming that may alter their productivity and status as a carbon sink. Currently tools are available to monitor these forests with digital surface models of vegetation and terrain elevations (DSMs) from commercial high-resolution spaceborne imagery (HRSI). These data provide detailed forest structural information at the site scale. However, digital terrain models (DTMs) at a similar resolution are needed to extract the most precise vertical structure estimates, such as canopy height. Typically, light detection and ranging (LiDAR) data and/or synthetic aperture radar (SAR) data are used as a DTM for bare ground with stereo DSMs representing canopy top to extract this information, but to our knowledge, no openly available HRSI DTM is available for the circumboreal forest. To address this, we explored approaches to extract DTMs by filtering stereo optical point clouds in the sparse canopy of boreal forests to provide broad-scale stereo derived DTMs where other data are not currently openly available. Our approach consisted of two stereogrammetric methods, with three different point cloud search radii, at six different canopy cover intervals. We found 2 m resolution DTMs were robust for estimating bare ground where canopy cover is less than 40% using two airborne platforms, Goddard’s LiDAR Hyperspectral and Thermal Imager (G-LiHT) small footprint LiDAR and Land, Vegetation, and Ice Sensor (LVIS) wide-swath imaging laser altimeter as reference. Our results indicate filtering HRSI stereo optical point clouds provide valuable information about the ground surface that is required to characterize vegetation structure in climate-vulnerable northern high-latitude forests.