Journal of Geophysical Research: Solid Earth

Reconstruction of eruption column dynamics on the basis of grain size of tephra fall deposits: 2. Application to the Pinatubo 1991 eruption


  • Takehiro Koyaguchi,

  • Marekazu Ohno


The granulometric methods to reconstruct eruption column dynamics developed in paper 1 are applied to the tephra fall deposits of the climactic Plinian phase of the 1991 eruption at Pinatubo. The tephra fall deposits are composed of two units: layer C1, which corresponds to the first half of the climactic phase, and layer C2 the second half. The granulometric estimates of the expansion rate of the umbrella cloud for layers C1 and C2 are 7×1010 and 3×1010 m3/s, respectively, which agree with the observations of satellite images. These estimates indicate that the magma discharge rate decreased from 9×108 to 3×108 kg/s during the eruption. The grain-size distribution at the top of the eruption column is characterized by depletion of coarse clasts, suggesting a distinct deceleration in the gas thrust region down to several tens of meters per second, particularly during the second half of the climactic phase. The total amount of layers C1 and C2 is estimated to be 3×1012 kg, and the very fine particles, which did not deposit in the accessible on-land area, may occupy up to 60% of the total ejecta. The effective duration of the eruption estimated by the present methods (a few hours or less) is substantially shorter than a previous estimate from the real-time observation based on infrasonic data (∼10 hours). This discrepancy may be attributed to the overestimation of the magma discharge rate due to the entrainment of the ambient air at the cloud top and/or to the underestimation of the total amount of the ejecta due to the effect of the contemporaneous Plinian activity and generation of pyroclastic flows. It is also suggested that the intensity of eruption was fluctuating with time so that the instantaneous observations such as cloud height in satellite images do not always show a quantitative agreement with the time-averaged features predicted by the present methods.