The semi-analytical snow retrieval algorithm and its application to MODIS data

Grain size is a key parameter of a snowpack, affecting its thermodynamic state and influencing the spectral snow albedo. Differently from visible wavelengths, where the sensitivity to grain size is very low, in the near-infrared band there is a strong sensitivity of the reflectance to the grain size. This sensitivity provides the basis for the retrieval of grain size. In this paper we introduce a new snow retrieval algorithm that makes use of near-infrared measurements in which snow is modeled as a semi-infinite, weakly absorbing medium. It is assumed that the dense packing effects can be neglected and the radiative transport in snow can be studied using the standard radiative transfer equation extensively used, e.g., in cloud optics. The shape of grains is accounted for in the framework of fractal snow grain model. The performance of the algorithm is evaluated using ground-based measurements of snow albedo and results from a different retrieval algorithm. The technique is applied to study the changes of snow properties before and just after snow fall as seen by two MODIS sensors on TERRA and AQUA satellites. These satellites fly approximately 3 h and half apart (10:30 a.m. and 1:30 p.m. equator crossing time). The values of grain size retrieved from MODIS are also compared with values of grain size collected on ground. However, the area observed by MODIS including the locations of ground measurements was completely covered by clouds on the date of the measurements and the comparison could be performed only for the two previous days. A sensitivity analysis of the retrieval error due to atmospheric correction is also performed. Results show that the error on grain size retrieval induced by atmospheric correction ranges between ± 5% and ± 40%, depending on the grain size.