The behaviour of snow and snow-free surface reflectance in boreal forests: Implications to the performance of snow covered area monitoring

Abstract

Optical methods for snow mapping typically exploit wavelengths in the visible and near-infrared range, where reflectances from snow and snow-free ground may significantly vary, especially during the melting season. In this study, the variability of ground reflectance in the boreal forest zone was investigated in order to assess the feasibility of satellite sensors to the mapping of snow covered area (SCA). This aims at the improvement of the existing snow mapping algorithms, such as the reflectance model-based snow monitoring method SCAmod of the Finnish Environment Institute (SYKE). We acquired and identified some statistical features for reflectance spectra of seasonally snow covered and snow-free terrain by using a field spectroradiometer (ASD Field Spec Pro JR). Extensive measurement campaigns were carried out in 2007–2008 in northern Finland, resulting to hundreds of spectral samples between 350 and 2500 nm. The main emphasis was put on the determination of the melting snow reflectance under different weather conditions and stages of snow metamorphosis as well as over different terrain types. The gained reflectance spectra provide useful information for optical snow mapping studies in general. In this investigation, the primary function of the spectrometer data was the accuracy assessment and optimal band selection when applying the SCAmod-method to different space-borne instruments (MODIS, MERIS and AVHRR). The correspondence of small scale field observations with scene reflectance was also addressed. This was performed by comparing field spectrometer data with mast-based observations. Based on the inversibility of the SCAmod reflectance model, we addressed the standard deviation (standard error) of SCA estimation contributed by wet snow and snow-free ground reflectance fluctuations. An average error in the determination of the fractional Snow Covered Area (SCA) of about 5–7 %-units was obtained (range from 0% = snow-free ground to 100% = full snow cover), the maximum error of 10–12 %-units occurring at full snow cover conditions (for the moment when first open patches are emerging). These investigations show that the variability in the reflectances of snow and snow-free ground is a significant error source in snow mapping. However, providing the SCA with descriptive error statistics is very beneficial for further use as the error statistics are needed for data assimilation approaches, e.g. in using SCA-values as input to hydrological models.