2018/01/06 by Daniel Taylor‐Rodríguez, Andrew O. Finley, Taylor-Rodriguez, Daniel +13 · 1 citation
Environmental Science · #Remote Sensing and LiDAR Applications #Soil Geostatistics and Mapping #Atmospheric and Environmental Gas Dynamics
paper · pdf · doi:10.48550/arxiv.1801.02078
Gathering information about forest variables is an expensive and arduous\nactivity. As such, directly collecting the data required to produce\nhigh-resolution maps over large spatial domains is infeasible. Next generation\ncollection initiatives of remotely sensed Light Detection and Ranging (LiDAR)\ndata are specifically aimed at producing complete-coverage maps over large\nspatial domains. Given that LiDAR data and forest characteristics are often\nstrongly correlated, it is possible to make use of the former to model,\npredict, and map forest variables over regions of interest. This entails\ndealing with the high-dimensional (\∼102) spatially dependent LiDAR\noutcomes over a large number of locations (~105-106). With this in mind, we\ndevelop the Spatial Factor Nearest Neighbor Gaussian Process (SF-NNGP) model,\nand embed it in a two-stage approach that connects the spatial structure found\nin LiDAR signals with forest variables. We provide a simulation experiment that\ndemonstrates inferential and predictive performance of the SF-NNGP, and use the\ntwo-stage modeling strategy to generate complete-coverage maps of forest\nvariables with associated uncertainty over a large region of boreal forests in\ninterior Alaska.\n