Mapping Fractional Snow in Boreal Forests
Metsämäki, Sari; Heinilä, Kirsikka
Finnish Environment Institute, FINLAND

Seasonal snow accumulation and melting concerns extensive areas in the Northern Hemisphere. Interannual changes in the extent and duration of the seasonal snowpack is an important climate change indicator and these are therefore intensively studied. Since boreal forests cover vast areas in the Northern Hemisphere, snow assessments particularly in boreal forest zone is of great interest. This is emphasized by the fact that remote sensing techniques often suffer from inaccuracies over forested areas. Mapping of Fractional Snow Cover (FSC) is very useful in mapping of year-to-year fluctuations in the extent of the snow pack. The transitional zone exhibiting patchy snow cover can be wide, implying that mapping of FSC would be much more useful information than binary (‘snow/not snow’) classification. Indeed, accurate mapping of FSC also gives indication of start of the melting season and is therefore useful e.g. to hydrological applications. Here FSC refers particularly to undercanopy snow, e.g. if terrain is fully snow-covered then FSC=100% even if forest canopy is snow free.

The optical remote sensing method SCAmod by the Finnish Environment Institute (SYKE) is particularly designed for FSC-retrievals in boreal forests. The masking effect of forest canopy is accounted for by apparent forest transmissivity introduced by the radiative transfer-based reflectance model (Metsämäki et al., 2005, 2012). The SCAmod method is employed in the European Scape Agency's DUE-project GlobSnow, where it is applied in the production of 15 years' Climate Data Record on Snow Extent, using mostly Envisat/AATSR-data.

It has been earlier verified that SCAmod performs well for boreal forest belt of the Northern Hemisphere. The in situ data for validations, however, was limited to Finland. This absolute validation showed RMSE < 0.15 in the range [0-1]. In Metsämäki et al. (2012) the accuracy assessment was carried out through relative validation i.e. by comparison the FSC-retrievals with those from independent high-resolution data-based FSC. Although these may exhibit biased values, the fact that results from absolute and relative validations are very similar, implies a good performance of SCAmod in general.

Several databases for the Northern Hemisphere snow extent have been established. Here we demonstrate the difference of SCAmod-based SE and one of commonly used SE-products, NASA MODIS MOD10_C1 for the Northern Eurasia. The purpose is to highlight the differences of these two, without particular considerations on the absolute accuracies. While doing this, we can identify the locations of discrepancies and focus the future investigations to those particular areas. Since GlobSnow SE products have a limited geographical coverage due to the narrow swath width of the AATSR, we employ MODIS daily global reflectance data to provided FSC with SCAmod. This is well justified as MODIS and AATSR operate at similar spectral bands and is has also been verified by Metsämäki et al. (2012) that SCAmod provides similar FSC using either of these data. Our on-going studies also focus to the future employment of Sentinel-3 SLSTR in FSC mapping with SCAmod.

The forest transmissivity employed by SCAmod is in a key role as to the success of the FSC-retrievals. Therefore much effort has been put to determinate generally applicaple values for local transmissivities at pixel-level. The current approach for Northern Hemisphere transmissivity relies on MODIS reflectance observations acquired at full snow cover and Global land cover (ESA GlobCover) data. Also GlobAlbedo data and GlobSnow SWE (Snow Water Equivalent) are utilized in identifying the densest forests in terms of transmissivity. The feasibility of this approach has been verified e.g. by comparing the gained transmissivities with national crown coverage map of Finland as well as with Lidar-based local crown coverage maps. The forest transmissivity is applicaple not only in under-canopy FSC retrievals but also in mapping the snow-on-canopy. This application is based on the predicted maximum reflectance proportional to the local transmissivity; should it be exceeded, it is likely that forest canopy is at least partly snow-covered. This approach is demonstrated here.


Metsämäki, S., Anttila, S., Huttunen, M., & Vepsäläinen, J. (2005). A feasible method for fractional snow cover mapping in boreal zone based on a reflectance model. Remote Sensing of Environment, 95, 77-95.

Metsämäki, S., Mattila, O.-P., Pulliainen, J., Niemi, K., Luojus, K., Bottcher, K. (2012). An optical reflectance model-based method for fractional snow cover mapping applicable to continental scale. Remote Sensing of Environment, 123, 508-521.