Bayesian state-space model estimation of canopy height with next-generation space-borne LiDAR

A large fraction of the Earth's surface vegetation is found in forests, which provide rich natural resources and diverse habitats as well as play a key role in regulating water and atmospheric processes. As a significant component of atmospheric carbon exchange, above-ground biomass in forests remains a large source of uncertainty in current climate models. To address the need for forest structure monitoring on a global scale we propose an algorithm for estimating ground elevation, canopy height, and canopy cover from the forthcoming NASA earth-observing satellite, ICESat-2, planned for launch in 2017. The payload, a next-generation micro-pulse photon-counting LiDAR sensor known as the Advanced Topographic Laser Altimeter System (ATLAS), is primarily designed for monitoring ice-masses. Thus several significant challenges need to be overcome to generate canopy measurements. This work presents a Bayesian state-space generative statistical model to estimate canopy height and cover accurately from the low signal-to-noise data expected from ATLAS, and demonstrates the effectiveness with data simulations. The method performs comparably to existing methods with several advantages including robustness to noise, extensibility, and ease of data fusion.