11:15 AM - 11:30 AM
[STT39-03] Estimation of snow cover in Hokkaido using ALOS-2/PALSAR-2 backscatter coefficient differences
Keywords:snow depth, land cover, topographic dependency, snow properties
The difference in backscatter intensity between snow-free and snow-covered seasons using satellite-mounted synthetic aperture radar has been pointed out as an effective method for estimating the spatio-temporal distribution of snow cover. However, the accuracy of the estimation and error factors have not yet been fully investigated. In this study, we attempted to estimate the snow cover condition during the dry season in two areas in Hokkaido, one near Mikasa City in central Hokkaido and the other near Rikubetsu Town in eastern Hokkaido. The two regions have significant differences in snow depth and snow properties. The Mikasa area belongs to the heavy snowfall area in Hokkaido, where the maximum snow depth is over 1 m with compacted snow. On the other hand, the Rikubetsu area, while having a snow depth of only about 50 cm, is the coldest region in Japan, and the development of the depth hoar layer within the snow cover layer is remarkable.
For the analysis, ALOS-2/PALSAR-2 imaging data for February (maximum snow depth period) and the snow-free period in summer since 2017 were used. The backscatter coefficient (dB value) for each period was calculated by SARscape. The difference between the snow-free period and snow cover period of the previous year was taken to provide an index of snow cover conditions. For the comparison of snow depth, the data measured by AMeDAS within both areas and the data from the field survey of snow depth and density were used. In addition, in considering the spatial distribution of differences, we examined the relationship between the difference in vegetation conditions during the snow-free season and before the start of snow cover. We used the Normalized Difference Vegetation Index (NDVI) for the analyses based on visible and near-infrared images from the Planet Scope satellite, separately for natural vegetation and cultivated land. In addition, natural vegetation was examined with topographic factors such as elevation, slope, and slope orientation using the GSI's digital elevation model (10 m).
The backscatter intensity attenuated during the snow cover period for natural vegetation. The difference between the snow cover period and the snow-free period (negative value for the snow cover period - snow-free period) showed an elevation dependence, suggesting a relationship with snow depth, as shown in previous studies. In particular, in the Rikubetsu area, an increasing trend of differential values was also observed on the northwestern slopes in mountainous areas with natural vegetation, which may correspond with the topographical distribution of snow cover. On the other hand, differences in differential values were observed in the cultivated land in both regions according to the plot of cultivated land. The difference in the backscatter coefficient was directly related to the backscatter coefficient in the snow-free season. The relationship suggests that the difference in the backscatter coefficient due to crop growth during the snow-free season may be an error factor in estimating the snow cover conditions of cultivated lands. As another issue, it was found that in some years, there was no difference in backscatter coefficient difference, or the difference was smaller in the Mikasa region, despite the difference in snow depth between the two regions being more than twice as great. The reason for this is unknown. However, based on local snow cover survey results, the difference may be due to differences in snow properties (compacted snow vs. depth hoar) or ice sheets in the snow layer (due to the influence of Rain on Snow and winter thaw in the Mikasa area).
For the analysis, ALOS-2/PALSAR-2 imaging data for February (maximum snow depth period) and the snow-free period in summer since 2017 were used. The backscatter coefficient (dB value) for each period was calculated by SARscape. The difference between the snow-free period and snow cover period of the previous year was taken to provide an index of snow cover conditions. For the comparison of snow depth, the data measured by AMeDAS within both areas and the data from the field survey of snow depth and density were used. In addition, in considering the spatial distribution of differences, we examined the relationship between the difference in vegetation conditions during the snow-free season and before the start of snow cover. We used the Normalized Difference Vegetation Index (NDVI) for the analyses based on visible and near-infrared images from the Planet Scope satellite, separately for natural vegetation and cultivated land. In addition, natural vegetation was examined with topographic factors such as elevation, slope, and slope orientation using the GSI's digital elevation model (10 m).
The backscatter intensity attenuated during the snow cover period for natural vegetation. The difference between the snow cover period and the snow-free period (negative value for the snow cover period - snow-free period) showed an elevation dependence, suggesting a relationship with snow depth, as shown in previous studies. In particular, in the Rikubetsu area, an increasing trend of differential values was also observed on the northwestern slopes in mountainous areas with natural vegetation, which may correspond with the topographical distribution of snow cover. On the other hand, differences in differential values were observed in the cultivated land in both regions according to the plot of cultivated land. The difference in the backscatter coefficient was directly related to the backscatter coefficient in the snow-free season. The relationship suggests that the difference in the backscatter coefficient due to crop growth during the snow-free season may be an error factor in estimating the snow cover conditions of cultivated lands. As another issue, it was found that in some years, there was no difference in backscatter coefficient difference, or the difference was smaller in the Mikasa region, despite the difference in snow depth between the two regions being more than twice as great. The reason for this is unknown. However, based on local snow cover survey results, the difference may be due to differences in snow properties (compacted snow vs. depth hoar) or ice sheets in the snow layer (due to the influence of Rain on Snow and winter thaw in the Mikasa area).