Surface height adjustments in pyroclastic-flow deposits observed at Unzen volcano by JERS-1 SAR interferometry

Pyroclastic flows from the 1990–1995 eruption of Unzen, a dacitic volcano in the southwest of Japan, descended the mountain along a variety of routes causing widespread damage and loss of life. Interferograms constructed from JERS-1 L-band Synthetic Aperture Radar (SAR) images show a number of features related to these pyroclastic flows and their secondary effects. The most useful interferogram in this respect is based on images acquired on 22 July 1993 and 1 December 1993 and shows the descent paths for pyroclastic flows occurring in four valley systems within this time window as zones of decorrelation caused by the repeated resurfacing. The 22 July 1993 SAR image was, through considerable good fortune, acquired only 2.6 days after a major pyroclastic flow had descended into the Mizunashi valley so that, in the absence of rainfall, the fresh 2-m-thick deposits were dry when first imaged. The largest differential surface height changes observed in the interferogram represent height decreases in the vertical of ∼12 cm and, significantly, lie within a small region of the Mizunashi valley which was resurfaced by the pyroclastic flow of 19 July 1993 but not subsequently. Within this small region, radar coherence is higher (maximum correlation value of ∼0.75) in a center-valley site where ash but relatively few large boulders are present. In a qualitative sense, the new ash surfaces exhibit higher levels of radar coherence than the older (pre-19 July) deposits. In other Unzen valleys visited by pyroclastic flows, smaller differential surface height decreases (∼4 cm) are observed where the surface deposits were emplaced by events taking place between 1–3 months before the acquisition date of the 22 July 1993 image. The ‘extra’ ∼8 cm of surface height decrease observed in the case of the freshly laid Mizunashi deposits must result from a deflationary mechanism (or mechanisms) operating in a spatially uniform manner in order for radar coherence to be maintained. A number of possible causative mechanisms including hydroconsolidation initiated by the seepage of rainwater into a newly deposited ash sheet, deflation on gas release, basal erosion, settlement and substrate relaxation are discussed here. Given the relatively low emplacement temperatures of the Unzen ash (300–500°C), it is unlikely that viscous deformation within the body of the fresh ash deposits is responsible for the subsidence. Comparison of the 22 July 1993–1 December 1993 interferogram with a LANDSAT 4 Thematic Mapper image of Unzen, acquired on 10 January 1994, highlights a ribbon-like zone of radar decorrelation running along the course of the Mizunashi valley which marks the path of debris flows that occurred within the time interval between the two SAR images used in the construction of the interferogram. Overall, these results demonstrate the utility of interferometric SAR-derived surface height and coherence measurements for monitoring the pyroclastic- and debris-flow regimes of active volcanoes.