Figure 7. LOS-changes of ALOS2 interferograms overlaid on shaded relief map. Details of each image are described in Table 1; imaging was performed by ascending, right-looking orbit. Warm and cold colors indicate LOS changes away from and toward the satellite, respectively. Black dashed line indicates the boundary between the burned and unburned area confirmed with Landsat optical images.
Figures 7a—7h show ALOS2 interferograms, each of which covers nearly one-year after October 2015 with some overlaps in its temporal coverages. Figure 7a, derived at the earliest period after the fire, indicates the maximum one-year subsidence to be as much as 10 cm or more.
If the amplitude and timing of seasonal subsidence/uplift cycle are invariable over time, a one-year interferogram will tell us only the irreversible displacements regardless of the acquisition times of master/slave images, which corresponds to the “pure ice” model in Liu et al (2015). Figure 7 sequentially shows the periods from October 2015 to June 2019 and indicates that the yearly subsidence rate slowed down. However, the variations of the one-year LOS changes in Figures 7 suggest that the actual deformation processes were more complex.
Figure 8a shows the cumulative LOS changes from October 2015 to June 2019 derived from SBAS-type time-series analysis, and that the maximum LOS extension reached as much as 25 cm; the 2\(\sigma\) errors for Figure 8a were ±1.5cm. Considering that the LOS changes during the first year after the 2014 fire were not included, the total LOS changes were presumably much greater than 25 cm, which meant that the subsidence was greater than 30 cm on account of the 36° incidence angle. As mentioned earlier, however, the higher-elevation areas such as the ridge did not undergo significant deformation, which probably would have been the case even during the first year after the fire. In addition to the high elevation areas, we realized clear contrasts in the LOS changes between the east- and the west-facing slopes near the northwestern area and the central north-south trending ridge; this spatial heterogeneity could also be recognized in Sentinel-1 (Figure 4). Their possible mechanisms comparing the burn severity (Figure 8b) and local landform (Figure 8c) are discussed in section 5.2.
We show the estimated time-series data at four representative sites (Figures 9a-9d), whose locations are indicated in Figure 8a. The sites (a) and (b) underwent nearly the same cumulative LOS changes by roughly 20 cm but were located at different slopes that are 4.3 km apart. On the other hand, the cumulative LOS changes at the site (d) were relatively small (approximately 10 cm). The site (c) located in the ridge did not show either significant seasonal or long-term deformation.
Time series data in Figures 9a and 9b clearly indicate that the largest subsidence took place from 2015 and 2016. We believe, however, that the most significant subsidence probably occurred only during the thaw season in 2016, as we have observed earlier, that no deformation occurred from December to March. Thus, the actual subsidence rate from October 2015 to July 2016 should have been more complicated than that expected from the linear trend in Figures 9a and 9b. The error bars in Figures 9a-9d indicated an estimated standard deviation with 2σ and attained ±1.5cm in the last epoch.