Siberian boreal forest structure estimates from concurrent multi-angle WorldView acquisitions

Abstract:

Estimating forest structure from space is important for monitoring the distribution and abundance of forest carbon stocks. Very-High Spatial Resolution (VHSR, 1 m or less) optical data could be used to estimate forest structure in remote and difficult to access forests of the world, but little information exists about the utility of multi-sensor cross-track stereo pairs for this purpose. We estimated Siberian boreal forest structure in Tura Krasnoyarsk, Russia from a tasked dense 2014 summer time-series of WorldView-1 and 2 in multi-angle combinations of Ames Stereo Pipeline (ASP) runs to generate point clouds from parallax that are used to produce digital surface models (DSMs). We evaluated single pair point cloud DSMs and accumulated point cloud DSMs with different viewing geometries from ASP to estimate root mean square errors (RMSEs). Our results suggest that a dense multi-angle time series from the WorldView constellation is a useful tool for estimating forest canopy height and dense multi-temporal observations can reduce height RMSEs if they have the appropriate viewing geometry.