Over the last few decades, the mechanism of flow and deposition of debris avalanches and pyroclastic flows has been the subject of great interest and much debate. The objective of this thesis is to discuss the mechanisms of these gravity currents, based on the detailed field surveys, observation of the currents, fluidization experiment, computer simulations, and critical evaluation of previous models. A plug flow model is used to explain the field evidence of Iwasegawa and Kaida debris avalanche deposits. A density-modified grain flow model is used to explain the characteristics and the results of the laboratory experiment and simulations of Komagatake 1929 and Unzen 1991-93 pyroclastic flow deposits. The Iwasegawa debris avalanche deposit, ca. 0.1 km^3 in volume, is distributed over the southeastern foot of Tashirodake Volcano, northern Honshu. At least 12 types of fragile debris avalanche block (DB) were identified in this avalanche deposit. Poorly-sorted debris avalanche matrix (DM) fills spaces between them. Normal grading of large clasts and reverse grading of wood fragments were observed. Preferred orientation of 311 pieces of wood fragments within this deposit coincided with the local flow direction. At the bottom part of the deposit, (1) erosion and incorporation of the basement material; (2) higher proportion of the basement materials than the upper part; and (3) reverse grading of the clasts were observed. The Kaida debris avalanche deposit, >0.3 km^3 in volume, is distributed at the eastern foot of Ontake Volcano, central Japan. At least 5 types of fragile DBs were included in this avalanche deposit. Many deformation structures were observed in the DBs such as elongation, folding, conjugate reverse faults, and numerous minor faults. Lithic components within the DM tended to have a higher ratio of the material which was captured from the adjacent basement formation during flowage.
The plug flow model is proposed to explain the flow and deposition mechanisms of these debris avalanches as follows. In the inner plug, the fragile DBs were transported without major rapturing due to high yield strength and viscosity of the DM. As the debris avalanches flowed down the steep slope at high speed, basement materials were eroded and incorporated into the laminar boundary layer under a strong shear stress field. The DM was produced by the shear stress between two DBs, and DBs and basement formations. Collision between DB and basement formations resulted in fracturing. Wood fragments included within the DM in the laminar boundary layer were rotated and aligned parallel to the flow direction. The density grading of clasts and wood fragments were formed by the density differences relative to the DM. The elongated DBs were formed by shear stress within the laminar boundary layer, and reverse grading of clasts was formed by the dispersive pressure on collision of each clast at the bottom of the flows. When the shear stress in the laminar boundary layer became smaller than the yield strengths of the flows due to deceleration, the debris avalanches froze and stopped thixotropically.
The Komagatake 1929 pyroclastic flow deposit, ca. 0.14 km^3 in volume, is distributed around Komagatake Volcano, southwestern Hokkaido. Yield strengths were estimated by the measurement of widths and heights of 53 lobes and 37 natural levees of pyroclastic flow deposits at the northwestern slope using methods proposed by Moore et al. (1978). The order of estimated yield strengths were 10^2-10^4 Pa. One flow unit of the pyroclastic flow deposit at the northern slope can be subdivided into three layers: such as (1) fines-depleted layer (DL), (2) fines-rich layer (FL), and (3) pumice-concentrated layer (PL), in ascending order, where the deposit contains many carbonized wood fragments. Fluidization experiments using the material of the pyroclastic flow deposits and the glass beads of various sizes revealed that the formation mechanism of the DL and FL due to concentration of the fine particles toward the top through the segregation pipes. In this experiment, particles larger than 1 mm were not affected by the upward flow at all. The Unzen 1991-93 pyroclastic flow deposits, ca. 0.2 km^3 in total volume, have been produced by more than 8600 collapses of lava domes extruded at Jigokuato crater. Volume of each pyroclastic flow was estimated using the seismographic record of the tremor produced by the pyroclastic flow. Relationship between the volume and mobility of pyroclastic flows was investigated to predict the disaster area. Observation showed that the pyroclastic flow deposit is composed of a few-meters thick, poorly-sorted, dense basal avalanche deposit and very thin, better-sorted, ash-cloud surge deposit. Video images revealed that the pyroclastic flows descended along the channels, with developing vortices of overlying dilute ash-cloud, but ingestion of air or jetting from the flow head which were proposed by Wilson and Walker (1982) were not observed. Excavation by machines of the toe of the June 8, 1991, pyroclastic flow deposit revealed that the deposit, ca. 4 m thick, was partly fines-depleted and accompanied a <5 cm-thick surge deposit and a 15 cm-thick layer 2a (Sparks, et al., 1973) at the bottom. The measured average velocities of five pyroclastic flows from May 24 to May 29, 1991, were up to 42 m/s. Numerical simulations of pyroclastic flows using energy line/cone (Malin and Sheridan, 1982) and Bingham flow (McEwen and Malin, 1989) models were made to simulate the velocity changes and arrival times of the flows, and to predict the disaster areas. In the simulation using the Bingham flow model, obtained yield strengths and viscosities were 850-2000 Pa and 50-90 Pas, respectively.
The density-modified grain flow model is proposed to explain the flow and deposition mechanisms of these pyroclastic flows. At the measured velocities, the maximum grain size that can be transported by the turbulence is at most 0.1-1.0 cm in diameter. Therefore, grains finer than this size can form the dilute ash-cloud surge/fall, but grains larger than this size must be concentrated on the bottom and form the poorly-sorted, dense basal avalanche. Theoretical curve of fluidization (Sparks, 1976) shows that the maximum grain size can be fluidized in the basal avalanche is at most 1.0 cm, when no ingestion of air from the flow head occurs. Therefore the transport mechanism of the clasts larger than this size is not fluidization as proposed by Wilson (1980, 1984), but "grain interaction" and "matrix strength". Such a kind of flow is called as "the density-modified grain flow (Lowe, 1976)". The roundness of the clasts in the both pyroclastic flow deposits indicates the severe collision of clasts. Both pyroclastic flows have yield strengths large enough to support the large clasts (10^2-10^4 Pa). The surge deposit at the bottom of the pyroclastic flow deposit may have been formed by ash-cloud surge which was detached from the basal avalanche at the bottom of the steep slopes due to hydraulic jump and proceeded ahead of the tip of the basal avalanche.