Mapping Vegetation and Seasonal Thaw Depth in Central Alaska Using Airborne Hyperspectral and LiDAR Data

Field collection of permafrost vegetation and seasonal thaw depth measurements are a time consuming and labor intensive procedure. To address this, in this study we assessed the capacity of airborne hyperspectral and LiDAR data for upscaling in-situ point data to map vegetation and summer thaw depth at an experimental site near Fairbanks, Alaska, USA. The site represents roughly three quarters of the most common vegetation in the boreal and taiga permafrost zones. An overall accuracy of 93% was achieved for mapping five different vegetation types, and a correlation coefficient (r) of 0.66 was produced for thaw depth modeling using a machine learning algorithm when two remote sensing data sources were combined. Object-based vegetation and thaw depth maps were produced. This study demonstrates that fine spatial resolution hyperspectral data could be valuable for permafrost characterization, while the contribution from LiDAR was marginal at the testing site.