4.4 Remote sensing image uncertainty
This study utilizes the MODIS day-by-day cloud-free snow cover dataset
to mitigate the impact of cloud cover to some extent. However, the
spatial resolution of the MODIS data is 500m, which may not be
sufficient to detect small-scale areas of snow accumulation in the two
watersheds examined in this study. Furthermore, the dataset employs the
NDSI normalized snow index with a threshold of 0.4 to extract the
snowpack, which has demonstrated reliable accuracy in North America
(Hall et al. 1995,Hao et al. 2008a). However, numerous studies have
demonstrated that employing an NDSI threshold of 0.4 in China leads to a
substantial underestimation of ground snowpack (Hao et al. 2008b,Zhang
et al. 2019). Meanwhile, the western Sichuan Plateau is situated in the
eastern part of the Tibetan Plateau, characterized by extensive forest
coverage. Zhao et al. (year) demonstrated that the accuracy of NDSI in
extracting the snowpack in forested areas was inadequate (Zhao et al.
2016). Combined with the aforementioned reasons, the utilization of
MODIS day-by-day cloud-free snowpack area data in this study may result
in an underestimation of the snowpack area. Consequently, this
underestimation may lead to inaccuracies in the classification of
snowpack types in certain regions. Therefore, future research conducted
in the area should prioritize addressing the concerns of data accuracy
and its applicability