Extraction of Physical Snow Surface Characteristics from LiDAR Return Intensity

Abstract:

Light Detection and Ranging (LiDAR) has recently gained a strong foothold in the snow science community, as its ability to produce high-spatial resolution maps of snow depth emerges as an invaluable tool for snow hydrological and avalanche forecasters alike. However, an often-overlooked data point produced by many LiDAR units is the return intensity, hence the magnitude of backscattered radiation, and thus the ability to utilize LiDAR as both a spatial and spectral analysis tool. At the near-infrared wavelength of 1064 nm, where many LiDAR units operate, reflectance is known to be sensitive to snow microstructure - namely specific surface area (SSA) - in a manner that has been well-quantified by empirical studies as well as radiative transfer and ray tracing models. However, few, if any, of these models account for the polarization state of incoming irradiance. This is thought to be a salient point, as many commercial LiDAR units emit a beam of predominately linearly polarized light, which may indeed influence how energy is scattered off a snow surface. To better understand the relevance of incident beam polarization state on detected backscatter, as well as how this LiDAR-specific reflectance varies with snow microstructure, experiments from within a controlled cold laboratory environment were conducted. Snow samples were generated featuring a range of different physical parameters; grain size, grain shape, density, SSA, and penetration resistance were recorded for each sample, prior to being scanned by a terrestrial LiDAR unit at (28) different incidence angles. In contrast to studies focusing on randomly polarized light sources, it was determined that LiDAR reflectance did not correlate well with SSA, and thus LiDAR may not be a suitable tool for extraction of effective grain diameter. However, reflectance did exhibit a strong logarithmic relationship with snow density for certain grain shapes, generating optimism for future field-based avalanche and hydrology applications. Further, reflectance deteriorated with increasing incidence angle in a consistent manner regardless of varying snow microstructure. Therefore, normalization to compare direct vs. oblique hits across differing snow samples proved reasonable.