Spatio-temporal analysis of remotely-sensed forest mortality associated with road de-icing salts

Abstract

Forest mortality along highways has long been a concern in areas where de-icing compounds are applied during winter. This study combined the spatial advantage of high-resolution remote sensing imagery and the temporal advantage of long-term archival imagery to quantify forest mortality and to detect the subtle and chronic effects of road de-icing salts for a large mountain watershed in the Sierra Nevada Mountains, USA. IKONOS-derived mortality was used in a fine-scale spatial analysis to assess road proximity and roadside topography effects on forest mortality and to compare two potential mechanisms of de-icing salt damage, i.e. aerial deposition and soil uptake. These mechanisms were modeled using spatial proxy variables that were constructed from LiDAR topographical data. The analysis revealed a clear trend of increasing mortality with increasing potential for aerial deposition of de-icing salt onto tree crowns, mainly occurring within 10m from roads. The effect of soil uptake of salt was weaker than that of aerial deposition but had a broader potential effect zone that extended to at least 100m from roads. Landsat TM-derived mortality from 1989 to 2010 provided a long-term time series that indicated both immediate and lagged effects of salt application on forest mortality. Immediate effects of de-icing salt were only distinct in wet years when salt application and spray generation by passing traffic and snow plowing were likely high and other damaging factors, such as bark beetles or drought mortality, were likely weak. A strong and consistent one-year lag in the effect of salt application on incidence of mortality suggested that longer-term impacts of de-icing salt on forest health likely involved more complex pathways than simply aerial deposition. Our multi-scale remote sensing approach provided convincing evidence that de-icing salt was a significant factor for roadside forest mortality and allows for efficient future monitoring at the large-watershed scale.