Increasing wildfire activity at high northern latitudes has the potential to mobilize large amounts of terrestrial mercury (Hg). However, understanding implications for Hg cycling and ecosystems is hindered by sparse research on peatland wildfire Hg emissions. In this study, we used measurements of soil organic carbon (SOC) and Hg, burn depth, and environmental indices derived from satellite remote sensing to develop machine learning models for predicting Hg emissions from major wildfires in the permafrost peatland of the Yukon-Kuskokwim Delta (YKD) in southwestern Alaska. Wildfire Hg emissions during summer 2015─estimated as the product of Hg:SOC (0.38 ± 0.17 ng Hg g C1-), predicted SOC stores (mean [5th-95th] = 9.1 [5.3-11.2] kg C m-2), and burn depth (11.3 [8.2-13.9] cm)─were 556 [164-1138] kg Hg or approximately 6% of Hg emissions from wildfire activity >60°N. Modeling estimates suggest that wildfire nearly doubled summertime Hg deposition within 10 km, despite advection of more than 75% of total emissions beyond Alaska. YKD areal emissions combined with remote sensing estimates of burned area suggest that wildfire Hg emissions from northern peatlands (25.4 [14.9-33.6] Mg y-1) are an important component of the northern Hg budget. Additional research is needed to refine these estimates and understand the implications for Arctic and global Hg cycling.
Keywords: Hg; deposition; emissions; machine learning; peat; transport; uncertainty.