Glacial surface feature identification using multispectral airborne laser scanning data

Capturing the Earth's surface in 3D with lidar has found widespread application and driven fundamental advances in many fields of the Earth sciences. In addition to spatial data, most lidar sensors record return signal amplitude data (often called intensity) from single or multi wavelength laser returns; this provides the opportunity to detect and classify surface features beyond basic 3D mapping because the amplitude of the return signal is proportional to the scattering characteristics of the surface at the wavelength of the incident energy. While laser scanning has been mainly used for detailed mapping and analysis of glacial topography (e.g., roughness, elevation, etc.) and detection of spatiotemporal change (e.g., elevation, volume, and mass), a few studies have used lidar intensity for glacier surface feature identification, demonstrating the potential of lidar intensity data. Glacial surfaces and their surroundings are composed mainly of snow, firn, ice, and various transition facies. It is essential to distinguish these facies to better understand the accumulation and removal of snow and ice on glaciers. However, since fresh, white snow cannot be easily distinguished from white ice, these two surface facies have commonly been merged into one class when remotely sensed.

A novel approach is presented for separating glacial snow and ice and mapping their spatial distribution on Canada Glacier, outlying frozen lakes (Lake Hoare and Lake Fryxell), and surrounding areas in the McMurdo Dry Valleys, Antarctica. Multi-wavelength lidar data, collected by an Optech Titan sensor in the austral summer of 2014/2015, are used to differentiate snow, firn, white and blue ice, and bare rock. The intensity data were corrected for range and incidence angle effects, thereby converting raw intensity to values proportional to surface reflectance. A set of rules describing different surface facies based on corrected laser intensity was defined. In addition to laser intensity, surface parameters such as roughness and slope were used in a Support Vector Machine (SVM) classification in an effort to enhance the identification of glacial surface facies.