|
Snow is the most important freshwater resource in northern Xinjiang, which is a typical inland arid ecosystem in western
China. Snow mapping can provide useful information for water resource management in this arid ecosystem. An
applicable approach for snow mapping in Northern Xinjiang Basin using MODIS data was proposed in this paper. The
approach of linear spectral mixture analysis (LSMA) was used to calculate snow cover fractions within a pixel, which
was used to establish a regression function with NDSI at a 250-meter grid resolution. Field campaigns were conducted to
examine whether NDSI can be used to extend the utility of the snow mapping approach to obtain sub-pixel estimates of
snow cover. In addition, snow depths at 80 sampling sites were collected in the study region. The correlation between
image reflectivity and snow depth as well as the comparison between measured snow spectra and image spectra were
analyzed. An algorithm was developed on the basis of the correlation for snow depth mapping in the region. Validation
for another dataset with 50 sampling sites showed an RMSE of 1.63, indicating that the algorithm was able to provide an
estimation of snow depth at an accuracy of 1.63cm. The results indicated that snow cover area can reach 81% and
average snow depth was 13.8 cm in north Xinjiang in January 2005. Generally speaking, the snow cover and depth had a
trend of gradually decreasing from north to south and from the surroundings to the center. Temporally, the cover reached
a maximum in early January, and the depth reached a maximum was ten days later. Snow duration was so different in
different regions with the Aletai region having the longest and the Bole having the shortest. In the period of snow
melting, snow depth decreased earlier, afterward snow cover dwindled. Our study showed that the spatial and temporal
variation of snow cover was very critical for water resource management in the arid inland region and MODIS satellite
data provide an alternative for snow mapping through dedicated development of mapping algorithms suitable for local
application.
|
