The behaviour of snow and snow-free surface reflectance in the boreal forest area

The Northern Hemisphere seasonal snow cover influences highly the interactive Earth's surface and atmosphere system. For this reason, snow cover is a sensitive climate change indicator both regionally and on global scale. As being the largest element of the cryosphere, snow pack is also an important temporary fresh water storage. Accordingly, terrestrial snow cover is a rapidly varying constituent of the hydrological cycle in particular during the spring and autumn transition months. Spatial variability and long-term trends in snow cover distribution and related climate patterns have been analyzed based on observation data and climate change prediction models. The models project significant changes in the spatial and temporal distribution of snow in the boreal forest zone. These changes can have widespread impacts on ecosystems and human activities, e.g. flooding, water resources management, agriculture, transportation, hydropower production, and recreational activities. A topical and vital task is the preparation of national and international strategies for adaptation to climate change. One recognized strategy is to improve climate variable observation systems.

Accurate information on snow spatial and temporal distribution is essential for climate research, numerical weather prediction and hydrological applications. Regardless, that point measurement records can have a long history, they may not be representative for the large areas or spatially altering snow patterns of Eurasia. Earth observation provides a spatially and temporally effective means to obtain information on the snow cover extent in addition to the traditional in-situ weather station and snow-gauging network. Both optical and passive microwave satellite data are used in snow remote sensing. Yet, the importance of both point and remote sensing observations is evident. Field spectroscopy is advancing remote sensing through feasibility studies, image analysis and vicarious calibration of satellite products. As with imaging spectroscopy different variables of the Earth's surface can be mapped with unprecedented accuracy, the field spectral data exploitation has gained relevant importance during the past decades. At present, the arising challenge is to ensure the high quality of the spectral data by establishing and defining common measurement protocols and terminology.

Several methods for snow mapping exploit wavelengths in the optical and near-infrared range, where snow and snow-free ground reflectance are sensitive to variation specifically in the melting season. The objective of this work is to study the variability of surface reflectance in the boreal forest area in order to enable the improvement of existing snow mapping algorithms, such as the reflectance model-based SCAmod snow monitoring method originally developed at the Finnish Environment Institute (SYKE). For this purpose, we acquired and identified generally applicable average values for the reflectance of seasonally snow covered terrain and their variability by using a ground-based field spectroradiometer. The primary function of the field spectral measurements is the collection of an elaborate spectral library during several years to be utilized in the accuracy assessment of the SCAmod-method and other methods alike. To perform the actual accuracy assessment of the SCAmod, we calculated the contributions of both wet snow and snow-free ground reflectance fluctuations to the standard deviation (standard error) of the Snow Covered Area (SCA) estimate.

The results show that the temporal and spatial variability in the average reflectance is an important error source in snow mapping in the boreal forest zone. The estimates for the variance in snow covered and snow-free reflectance spectra directly provide information for the future SCAmod algorithm modification. The results give a good overview particularly on the variability of wet snow reflectance. Besides, optimal band alternatives can be found for current and future satellite instruments by determining the wavelength areas where the variability is only minor or has only a slight effect on the snow covered area algorithm performance.