High-resolution mapping of aboveground shrub biomass in Arctic tundra using airborne lidar and imagery

Abstract

Abstract Accurate monitoring of climate-driven expansion of low-stature shrubs in Arctic tundra requires high-resolution maps of shrub biomass that can accurately quantify the current baseline over relevant spatial and temporal extents. In this study, our goal was to use airborne lidar and imagery to build accurate high-resolution shrub biomass maps for an important research landscape in the American Arctic. In a leave-one-out cross-validation analysis, optimized lidar-derived canopy volume was a good single predictor of harvested shrub biomass (R2 = 0.62; RMSD = 219 g m− 2; slope = 1.08). However, model accuracy was improved by incorporating additional lidar-derived canopy metrics and airborne spectral metrics in a Random Forest regression approach (pseudo R2 = 0.71; RMSD = 197 g m− 2; slope = 1.02). The best Random Forest model was used to map shrub biomass at 0.80 m resolution across three lidar collection footprints (~ 12.5 km2 total) near Toolik Field Station on Alaska's North Slope. We characterized model uncertainty by creating corresponding maps of the coefficient of variation in Random Forest shrub biomass estimates. We also explore potential benefits of incorporating lidar-derived topographic metrics, and consider tradeoffs inherent in employing different data sources for high-resolution vegetation mapping efforts. This study yielded maps that provide valuable, high-resolution spatial estimates of aboveground shrub biomass and canopy volume in a rapidly changing tundra ecosystem.