High-resolution snow depth prediction using Random Forest algorithm with
topographic parameters and an ecosystem map: a case study in the Greiner
Watershed, Nunavut
Abstract
Increased surface temperatures (0.7℃ per decade) in the Arctic affects
polar ecosystems by reducing the extent and duration of annual snow
cover. Monitoring of these important ecosystems needs detailed
information on snow cover properties (depth and density) at resolutions
(< 100 m) that influence ecological habitats and permafrost
thaw. As arctic snow is strongly influenced by vegetation, an ecotype
map at 10 m resolution was added to a method with the Random Forest (RF)
algorithm previously developed for alpine environments and applied here
over an arctic landscape for the first time. The topographic parameters
used in the RF algorithm were Topographic Position Index (TPI) and
up-wind slope index (Sx), which were estimated from the freely available
Arctic DEM at 2 m resolution. Ecotypes with taller vegetation with
moister soils were found to have deeper snow because of the trapping
effect. Using feature importance with RF, snow depth distributions were
predicted from topographic and ecosystem parameters with a root mean
square error = 8 cm (23%) (R² = 0.79) at 10 m resolution for an arctic
watershed (1 500 km²) in western Nunavut, Canada.