Transferability of Lidar-Derived Aboveground Biomass Models in Piñon-Juniper Woodlands of the Western United States

Piñon-juniper woodlands (PJ) are a spatially-expansive and temporally-dynamic dryland ecosystem that encapsulates a wide range of functional diversity and climatic variability. Fluctuations in PJ's extent and abundance over space and time dictate the woodlands' carbon storage capacity, influencing land-climate interactions. Understanding the role that PJ plays in the global carbon cycle requires that accurate maps can be made that quantify the living, aboveground biomass (AGB) throughout its vast range. Such maps could also provide critical insight into PJ's role as a habitat for wildlife, its fuel structure for fire behavior analysis, and inform land management practices. However, to consistently and accurately map PJ AGB throughout its entire range, a deeper understanding is needed of the extent to which ecological differences dictate model transferability - the degree to which AGB models built in one PJ woodland can be used to predict AGB in another ecologically-dissimilar and spatially-distant PJ woodland. Airborne light detection and ranging (lidar) remote sensing, with its capacity to generate precise, three-dimensional models of vegetation structure, has emerged as a leading approach to mapping AGB. This study aims to determine the degree to which lidar-derived AGB predictions can be applied across the entire spatial range of PJ, from southeastern New Mexico to northwestern Nevada. An extensive database of field plots and random forest modeling is utilized to assess AGB model transferability between ecologically similar clusters to determine how differences in vegetation structure, species composition, and climate affect model transferability. The results provide important implications for understanding the strengths and limitations of broad-scale remote sensing studies of vegetation structure and function in dryland woodland ecosystems.