Mapping Fractional Cover and Change for Tundra, Boreal, and Ecotone Plant Functional Types using Modeled Landsat Reflectance

Abstract:

Increasing shrub expansion and densification has been widely documented in Arctic tundra ecosystems (Tape et al. 2006, Myers-Smith et al. 2011, Elmendorf et al. 2012, Myers-Smith et al. 2015). In contrast, lichen cover is thought to be decreasing in both boreal and Arctic zones due to competition with vascular plants (like shrubs), warmer and drier climate, disturbance and herbivory (Fraser et al. 2014, Moffat et al. 2016). Change in lichen community composition is a critical measure of ecosystem function that may impact herbivores important for other faunal and human subsistence (especially caribou), and myriad other ecosystem services. We have compiled and harmonized an extensive “calibration and validation” (cal-val) data set for modeling vegetation cover over large areas, including detailed in situ plots compiled by resource management agencies and others in the domain, sUAS image mosaics classified to plant functional type, and lidar-derived shrub metrics. These include 30 m resolution topography and canopy height metrics from the entire AAC LVIS data acquisitions. The LVIS metrics provide extensive cal-val data for shrub height and structure across domain-wide ecological and climate gradients. Both seasonal (e.g. early summer, mid-summer) and percentile (e.g. 10/25/50/75/90% quality screened non-snow) Landsat reflectance composites have been used as predictors for shrub and lichen mapping (Macander et al. 2017, 2018, 2020), but sparser historical Landsat time-series and variations in seasonal collection strategies over time make historical reflectance composites more prone to noise and bias, and less reliable for plant functional type mapping. Here we assess the utility of modeled Landsat reflectance derived from the Continuous Change Detection and Classification model (CCDC, Zhu and Woodcock 2014) for mapping of plant functional type fractional cover across multiple epochs, 1985–2020.