TECHNICAL REPORT

MOBILITY OF A LARGE-VOLUME PYROCLASTIC FLOW--EMPLACEMENT OF THE CAMPANIAN IGNIMBRITE, ITALY

Richard V. Fisher, Dept. Geol. Sci., Univ. Calif., Santa Barbara, CA 93106

Giovanni Orsi, Dipt. Geo. Vulcan., Univ. Napoli, Largo S. Marcellino 10, I-80138, Napoli, Italy

Michael Ort, Dept. Geology, N. Ariz. Univ., PO Box 6030, Flagstaff, AZ 86011-6030

Grant Heiken, ESS-1, MS 575, Los Alamos Natl. Lab., Los Alamos, NM 87545

(published in the Jour. Volcan. Geotherm. Res. vol. 56, p. 205-220)

ABSTRACT

The trachytic Campanian Ignimbrite, which originally covered an area of 30,000 km2 around Naples, Italy, is the product of a highly energetic, gas-rich eruption. The deposit lies in valleys and isolated watersheds. In its medial and distal extent to the south, north and east of Naples, the Campanian pyroclastic current encountered mountains exceeding 1000 meters. Anisotropy of magnetic susceptibility (AMS) measurements indicate that the Campanian pyroclastic current (the transport system) traveled radially outward from the Phlegrean Fields area, but the ignimbrite-forming flow (the deposition system that developed from the base of the transport system) moved downslope from mountainsides to valleys, including slopes facing the eruption source, and flowed down drainage systems from intermontane basins. The Campanian pyroclastic current flowed ~35 km over the water of the Bay of Naples to deposit >43 m of ignimbrite on the south shore and also overtopped a 685-1000 m high ridge of the Sorrento Peninsula to deposit more on the other side. The distribution of the ignimbrite and the measured flow directions suggest that the Campanian pyroclastic current moved across the landscape as an expanded (and therefore turbulent) decompressing flow rather than as a high density, nonturbulent sheet-like current moving over mountains by momentum acquired by eruption column collapse. Strong expansion is corroborated by shard morphology that indicates derivation from highly inflated pumice and suggests vesicles must have been at least 80% by volume of the original magma.

GENERAL BACKGROUND STATEMENT The runout length of pyroclastic flows and their ability to surmount topographic barriers are topics of continuing research germane to the distribution of ignimbrite sheets. Pumice-rich pyroclastic flows are known to have crossed topographic barriers of considerable height (Yokoyama, 1974; Miller and Smith, 1977; Koch and McLean, 1975; Rose et al., 1979). The 22,000 yr B.P. Ito pyroclastic flow (Japan) traveled 70 km over topographic barriers as high as 600 m (Aramaki and Ui, 1966; Yokoyama, 1974). The 18,000 B.P. Taupo Ignimbrite, only ~30 km3 in volume, is spread out over a ~20,000 km2 area and mantles mountains as high as 1500 m above the inferred vent as far as 45 km from the source (Wilson, 1985; Wilson and Walker, 1985). Pyroclastic flows from Aniakchak and Fisher calderas in the Aleutian Islands traveled as far as 50 km over mountainous barriers between 250 and 500 m high (Miller and Smith, 1977).

Currently, there are two general models that describe the way that pyroclastic currents move across the landscape -- (1) as expanded flows (EFs) thicker than the height of the mountains they traverse, or (2) as dense flows (DFs) moving as a nonturbulent ground-hugging sheet across the landscape (Sparks, 1976). The purpose of the present paper is to test these ideas by analyzing the stratigraphy and flow directions, as determined by anisotropy of magnetic susceptibility (AMS) measurements (see below), of the Campanian Tuff.

The EF model was introduced first and applied to the Ito pyroclastic flow, Japan (Aramaki and Ui, 1966; Yokoyama, 1974). The DF model was applied by Miller and Smith (1977) to pyroclastic currents that moved outward from Fisher and Aniakchak calderas, Alaska. This latter model was given quantitative credence by Sparks et al. (1978) who concluded that a pyroclastic flow develops around the base of a collapsing eruption column, deflates, and then moves outward across the landscape under its own momentum. In their model, the momentum that a pyroclastic flow acquires is proportional to the height from which the eruption column collapses.

The two models require fundamentally different hydrodynamic behaviors. For EFs, the pyroclastic current must remain turbulent to maintain its expansion to a thickness greater than the topography that it overtops. Also, an expanded current can travel across water because expansion reduces its bulk density so that only its basal part interacts with the water (Sigurdsson et al., 1991). Dense pyroclastic flows probably cannot easily travel above water. Moreover, should they enter and travel beneath water, viscous boundary effects, mixing with water, and other conditions would inhibit flow, and it is unlikely that they could re-emerge.

One critical observation applicable to the problem discussed herein is that pyroclastic currents such as nuees ardentes are known to separate gravitationally into a lower part containing most of the solid fragmental mass. This natural density stratification in initially turbulent pyroclastic currents (Valentine, 1987) and other sediment gravity flows (Fisher, 1983, 1984) commonly results in flow transformations from turbulent to nonturbulent behavior in basal zones where concentration values become high. Having different densities and turbulent behaviors, the different parts of the current can decouple and travel different paths, depositing material independently (Fisher, 1990). In mountainous terrain, flow transformations, decoupling and divergent flow directions are especially amplified. Study of these effects can contribute to a better understanding of pyroclastic flow emplacement processes.

Sparks et al. (1978) postulated that DFs quickly originate following fall-back of a turbulent, collapsing eruption column and then move outward as a non-turbulent flow. Their calculations, using flow velocities ranging from 10 to 200 m/s, a drag coefficient of 0.01, and terminal velocity measurements of pyroclastic particles by Walker (1971), showed that grains >1 mm could not be carried in suspension.

The conclusion that momentum was the main cause of transport of pyroclastic flows influenced the "energy line" concept of Sheridan (1979). He explained that the slope of the energy line as proposed by Hsu (1975) for avalanche runout, traces the potential flow head from the top of the gas-thrust region of an eruption column to the distal toe of a flow along the line of transport. McEwen and Malin (1989), however, argued that the energy-line model predicts velocities that are too high, resulting in flow paths that are insufficiently responsive to topography. They suggest that velocity-dependent resistance factors such as Bingham or turbulent models are needed for accurate velocity predictions. We further suggest that in mountainous regions, flow transformations, density stratification, decoupling and blocking which are discussed below need to be considered as resistance factors that affect the forward progress of flows. Fisher (1990) shows that mountainous terrain itself can be considered as a roughness element that significantly effects runout distance.

Valentine (1987) concluded that pyroclastic flows may become density stratified and do not necessarily completely collapse to a non-turbulent condition of flow. According to his model, density stratification does not necessarily form a surface above which is mostly gas and below which is a dense flow, but rather there is a continuous gradation from one to the other. At flow velocities of 100 m/s or more and 300 m/s or more, particles as large as 1 cm and 10 cm repectively can be turbulently supported -- considerably larger than sizes calculated by Sparks et al. (1978). The differences in supportable clast sizes stem from the choices of substrate roughness and boundary layer thickness. Valentine (1987) proposed a rougher terrain than Sparks et al. (1978), with roughness elements (such as tree stumps) up to 1 m. Sparks et al. (1978) assumed a flat terrain with a roughness of 1 cm and considered the whole flow as a boundary layer.

The EF model contends that a pyroclastic current may initially be of medium- to low-density, but unlike the DF model, it remains expanded as it travels over the landscape leaving behind a depositional carpet deposited from its basal part, a model proposed by Fisher (1966) and extended by Branney and Kokelaar (1992). Flow of DFs is based upon a plug flow model whereby deposition is thought to occur by en masse freezing of the debris, similar to nonvolcanic debris flows, rather than layer by layer accretion (Sparks, 1976).

DESCRIPTION OF THE CAMPANIAN IGNIMBRITE

The Campanian Ignimbrite is a trachytic tuff that crops out regionally around Naples, Italy, and occurs in isolated watersheds and valleys. In its medial and distal extent, to the south, north and east of Naples, the Campanian flow encountered mountains 1000 meters and more in height. In the proximal area near the Phlegrean Fields, identification of rocks that belong to the Campanian Ignimbrite eruption is controversial. Rosi et al. (1983) and Rosi and Sbrana (1987) hold that its source is in the Phlegrean Fields, possibly in the Bay of Pozzuoli. Di Girolamo et al. (1984) and Scandone et al. (1991), however, propose a source north of Naples beneath the Volturno Plain. Our data from anisotropy of magnetic susceptibility (AMS) (discussed below) suggests a source within the Phlegrean Fields area, therefore we use the center of Pozzuoli Bay as a reference point.

Recently obtained high precision dates using single-crystal, laser-fusion 40Ar/39Ar methods from samples of the Campanian Ignimbrite near Avellino indicates an age of 35.5 +-0.8 ka (n=18), and one from near Maddaloni indicates an age of 36.0 +- 0.6 ka (n=21)(Deino et al. 1992). But other than these two dates from distal Campanian Ignimbrite localities, the age of the Campanian Ignimbrite is poorly constrained. Paterne et al. (1988) dated five trachytic marine tephra layers at 38.7, 36, 33.5, 26.9, and 24.1 Ka that they call the Campanian Ignimbrite series. While we do not dispute that there are five marine tephra layers that may be genetically related to one another and to the ignimbrite on land, we prefer to include them under the name "Campanian eruptive events." There appears to be only one major ignimbrite -- the Campanian Ignimbrite -- associated with the eruptive events. According to Paterne et al. (1988), the layer dated at 33.5 Ka corresponds to the eruption that produced the deposit called the Campanian Ignimbrite. C14 dates (Capaldi et al., 1985; Alessio et al., 1971, 1973, 1974) and K/Ar dates (Curtiss, 1966; Cassignol and Gillot, 1982) give a scatter of dates between 25 and 42 Ka for the Campanian eruptive events.

The Campanian Ignimbrite is a gray, poorly to moderately welded, trachytic ignimbrite sheet. It consists of pumice and lithic fragments in a devitrified matrix that contains sanidine, lesser plagioclase rimmed by sanidine, two clinopyroxenes, biotite, and magnetite. Pumice clasts ranging from trachyte to alkali trachyte generally increase in size upward, and lithic fragments tend to increase in size and abundance downward in individual sections. Pumice fragments are invariably rounded. Regionally, pumice and lithic fragments decrease in size away from the Bay of Naples region (Barberi et al., 1978).

The Campanian Ignimbrite is thickest beneath the plain crossed by the lower reaches of the Volturno river and in the many valleys that drain surrounding fault-block limestone mountains. It crops out in valleys on both sides of 1000+ m high mountain ridges to the north, east and south of Naples. East of Naples it lies within the rugged terrain of the Appenine Mountains; northward across the Volturno Plain the ignimbrite crops out on the sides and within the crater of Roccamonfina Volcano. To the south the pyroclastic current crossed the Bay of Naples (presently as deep as 200 m) and encountered the Sorrento Peninsula (650 to >1000 m elevation), leaving thick deposits on the north and south sides of the peninsula. To the west is the Tyrrhenian Sea, where sampling was not done. The ignimbrite is exposed as far as 60 to 70 km from Pozzuoli Bay, deposits at the most distal exposures being up to 5 m thick. We therefore consider that the original pyroclastic current extended much farther, possibly 100 km. The extent of the Campanian Ignimbrite, though dispersed in isolated outcrops because of ponding during flow and later extensive erosion, suggests that the original pyroclastic current flowed across some 30,000 km2 of area, and laid down a deposit estimated to be about 500 km3 in volume (bulk volume, not DRE). This was crudely estimated by circumscribing a circle of deposits with a radius of 100 km, 100 m thick at the center that become zero at the perimeter of the circle. It was the largest eruption of the last 200,000 years in the Mediterranean region (Barberi et al., 1978).

DISTAL FACIES OF THE CAMPANIAN IGNIMBRITE

Glass shards of the Campanian Ignimbrite are unique; long, straight to slightly curved plates that were derived from a highly vesiculated magma that consisted of large (centimeters) vesicles stretched into pita-bread-like pockets. Some shards are semicircular remnants of large spherical vesicles. The magma just before eruption is estimated to have exceeded 80% vesicles by volume. Pumice clasts have more poorly-developed vesicles, ranging in volume from 37% to 62%. This mixture of ash and pumice implies that the vesiculated magma was a bimodal mixture of highly-inflated vesicles and patches with poorly developed vesicles and lower overall vesicularity. Most of the deposit consists of the distinctive plate-like shards.

The stratigraphic sequence of the Campanian Ignimbrite in medial and distal areas can be divided into two parts based upon the idealized ignimbrite defined by Sparks et al. (1973); a very thin (1 - 10 cm), discontinuous, fines-poor layer called layer 1, above which lies the bulk of the ignimbrite called layer 2. Layer 2 is divided into layer 2a, the finer grained basal zone, and 2b, which forms the bulk of the deposit. Layer 3 is not preserved in the Campanian Ignimbrite sequence. The Campanian Ignimbrite is easily correlatable from place to place because it commonly is a nonwelded tuff with a characteristic blue-gray color, contains identifiable sanidine, pyroxene and some biotite, and it rests on soils developed upon older volcanic rocks or limestone. In many areas the tuff has been diagenetically altered to a yellowish color.

The highest elevation reported for the Campanian Ignimbrite is 1000 m above sea level on top of a dome in the crater of Roccamonfina Volcano (Giannetti, 1979), 60 km north of the center of the Bay of Pozzuoli. At one place 55 km due east of the Bay of Pozzuoli, the Campanian Ignimbrite lies at 970 m above sea level.

Layer 1.
Layer 1 is composed predominantly of phenocrysts of clinopyroxene, sanidine and plagioclase, and lithic clasts consisting of altered, poorly sorted brown tuffs, trachytic lavas and hornfelsed lavas. The layer commonly rests on limestone soil or limestone talus. In some areas, notably Roccamonfina Volcano, it rests upon older volcanic rocks. In some outcrops to the east and southeast of Pozzuoli Bay, the ignimbrite lies upon a co-genetic Plinian fallout pumice layer as thick as 100 cm. Layer 1 is mixed with material from the underlying terrains -- volcanic lapilli where it lies above Plinian fallout, and limestone lapilli where it lies above limestone basement.

Layer 1 has been identified throughout the distal region, from Roccamonfina Volcano 40 - 70 km north of Pozzuoli Bay, to Salerno 65 km to the south. In all exposures showing the base of the Campanian Ignimbrite, layer 1 is a lenticular to continuous layer as thick as 10 cm. In its thickest occurrences it is cross bedded showing flow directions downslope irrespective of source direction. Layer 1 consists mostly of sand-size lithic fragments and crystals, is devoid of medium to very fine ash, and in some places contains lithic lapilli to 9.5 cm. In places, it occurs in pockets or in lenticular mounds above a flat surface that was sheared and eroded during emplacement of the overlying layer 2.

Layer 2.
At some localities, as far north as Mondragone and as far south as Salerno, layer 1 is overlain by 10 to 20 cm of fine-grained and generally pale gray to yellowish ignimbrite (layer 2a). This is followed transitionally upward by coarser grained and darker gray layer 2b ignimbrite, the uppermost part commonly being a yellow color caused by diagenetic alteration.

Textural discontinuities at some localities give a crude bedding that is internal within otherwise homogeneous-appearing ignimbrite. The crude bedding features range from distinct textural breaks to extremely subtle color changes that are discontinuous across exposures of 10 to 20 meters in length. Because the bedding are not continuous, we suggest that the bedding was formed from overlapping of lobes of the pyroclastic current that formed by drainage of flows off mountain slopes from different directions. We favor this possibility because the best examples of these textural discontinuities are in the lee of high (>1000 m) mountain ranges. A similar phenomenon is reported from blast surge deposits of the May 18, 1980 eruption of Mount St. Helens (Fisher, 1990). Such crude bedding features may also be caused by unsteady flow (Branney and Kokelaar, 1992).

ANISOTROPY OF MAGNETIC SUSCEPTIBILITY MEASUREMENTS (AMS) AND FLOW DIRECTIONS

Determination of flow directions in ignimbrites is commonly difficult because the fabric in these deposits is visually nearly isotropic. In some cases flow directions have been successfully determined from textural indicators. These include use of imbricated logs (Froggatt et al., 1981), as well as orientation of glass shards, crystals, pumice, and lithic fragments (Potter and Oberthal, 1987; Elston and Smith, 1970; Ui et al., 1989; Suzuki and Ui, 1982, 1983, 1988; Buesch, 1991; Seaman and Williams, 1992). Imbrication of pumice and other features indicate not only a flow azimuth (bidirectional) but the flow direction as well (Kamata and Mimura, 1983; Suzuki and Ui, 1988). These methods, however, are time consuming and the features are commonly absent in specific outcrops.

Determination of anisotropy of magnetic susceptibility (AMS) is another method for determining lineations and foliations in ignimbrites. This method detects the alignment of magnetic minerals in the ignimbrite. The shape anisotropies cause the ignimbrite to be variably susceptible to the acquisition of an induced magnetic moment, forming a susceptibility ellipsoid with axes K1, K2, and K3. The K1 axis, the axis of greatest susceptibility, is generally interpreted as the axis in which the flow lineation lies. The K3 axis is commonly perpendicular to the plane of foliation.

Several studies have shown that the K1 susceptibility axis lies in the direction that the flow moved (e.g. Ellwood, 1982; Incoronato et al., 1983; Knight et al., 1986; Wolff et al., 1989; MacDonald and Palmer, 1990; Palmer, et al., 1991; Hillhouse and Wells, 1991). These studies have confirmed that AMS can determine flow directions that agree with physical flow textures such as imbricated clasts. In addition, Ellwood (1982) demonstrated that magnetic lineations in ignimbrites of known origins are oriented radially away from those sources, as would be expected. The method is relatively fast and therefore can include many sites, and relatively weak lineations in ignimbrites can be determined with a high degree of precision.

Paleomagnetic analyses of the Campanian Ignimbrite were made on a Kappabridge KLY-2 susceptibility bridge. Results from 25 sites were plotted and Bingham statistics for each site were determined using the Stereonet program. Results are presented as eigenvectors, including their azimuth and plunge, and eigenvalues, in which values closer to unity indicate better clustering of the AMS data. Eigenvectors may give a sense of the direction of flow. Previous studies (Knight et al., 1986; Wolff et al., 1989; MacDonald and Palmer, 1990) have noted an imbrication of AMS directions, such that the vectors dip up-flow. At each sampling site where the base is exposed, the substrate angles were measured in order to determine imbrication directions. Where the base is unexposed, substrate angles were estimated.

Sample localities were generally chosen to investigate the interaction of topography with the Campanian pyroclastic current. These include (1) the base of slopes facing the probable source direction of the current as well as in the lee of slopes, (2) in areas of enclosed drainage with single canyon outlets, and (3) on valley sides along drainage systems.

The AMS trends are nearly always parallel to local stream valleys or slope directions, and commonly at high angles to the radial direction from any reasonable source area. In addition, most AMS plunge directions point up slopes, implying that they reflect imbrication and provide information on flow directions. Some of the measured flow directions point toward the source area. This is counter intuitive, because it is commonly assumed that pyroclastic currents move away from their source, therefore flow directions measured in the ignimbrite should also be away from source. How does a pyroclastic current bypass an area and then return from the opposite direction, particularly if this occurs on successive slopes at increasing distance from the source? The explanation of this apparent paradox has profound implications for determining flow mechanisms of large volume ignimbrites that occur on both sides of mountain ranges and within intermontane basins.

Six of the sites used to measure flow directions of the Campanian Ignimbrite were sampled in the 300 Ka Roccamonfina Volcano area because the presence of this edifice and its crater would have affected the directions of movement of the pyroclastic current and provide an enclosed basin, which in this case has a single valley outlet. AMS eigenvectors for 4 samples have directions indicating radial flow off the volcano. The direction of flow indicated by another sample is east-west, parallel to the valley that drained the volcano, rather than northward from the source of the pyroclastic flow. This suggests that much of the material flowed radially off the outer slopes of the volcano irrespective of the direction from the original source area. Within the crater, one locality shows that the flow moved off the dome into the caldera. This is the same dome on which Campanian Ignimbrite was reported at an elevation of 1000 m.

Three sites in the Volturno Plain were measured where there would have been no topographic interference with the movement of the pyroclastic current. The eigenvectors from these sites suggest radial flow outward from an area in the vicinity of the Phlegrean Fields. These results do not locate a source precisely, but they strongly indicate a source south of the Volturno Plain near or within the Bay of Pozzuoli.

A major stumbling block in interpreting AMS directions is that the sites of magnetism in ignimbrites and the causes of imbrication (plunge) of eigenvectors is not known. Therefore the length of movement over which the magnetic orientation is set within the ignimbrite cannot be estimated. It may occur over a distance of centimeters in the very last stages of movement and therefore be imprinted by minor creep on steep slopes. However, many sites were sampled in open valley drainage systems where the underlying surface is nearly horizontal. Samples OF37, OF38, OF39, and OF40 (Fig. 4), for example, exhibit strong alignment of the AMS directions that correspond to the valley trend. This suggests that the AMS eigenvector is a measure of the direction that the ignimbrite was moving at the time of deposition, and not due to downvalley creep. Most other AMS sites also show a strong correlation between topographic orientations and flow directions. At Acqua Fidia, 970 m above sea level, the flow direction is down and parallel to the slope, suggesting that the flow may have drained from an area higher up the mountain.

DISCUSSION

The distribution of the Campanian Ignimbrite together with AMS measurements indicates that the Campanian pyroclastic flow surmounted ridges higher than 1000 m to enter some intermontane basins that have restricted external drainage through single narrow valleys. AMS data indicate that the Campanian flow moved down drainage systems originating in the intermontane basins. It did not move up the valleys into the basins. For example, Tocco Caudio, an abandoned village situated upon 15 m of Campanian Ignimbrite 55 km from source, lies in the lee of limestone ridges with elevations exceeding 1000 m. Another example is the watershed in which Cusano village is located 66 km northeast of the Bay of Pozzuoli. To gain access to the watershed, the pyroclastic flow had to (1) surmount ridges higher than 1000 m, or else (2) flow up a narrow, winding, deeply incised canyon before spreading out within the watershed. There is no Campanian Ignimbrite within the narrow confines of the canyon, but AMS data from the ignimbrite within the watershed shows down-valley current directions. We hold that the above relationships are best explained by the EF model.

An explanation of the apparent contradiction that AMS data do not show a correlation between measured flow directions and the direction to the source is aided by application of the idea of a transport system that differs from the deposition system as proposed for blast surge deposits at Mount St. Helens (Fisher, 1990). At Mount St. Helens the outward flowing pyroclastic current was the transport system that carried fragments to accumulation sites, and rapidly accumulating material at the base of the transport system formed the deposition system that had the ability to flow before coming to a complete rest. Geological evidence (Fisher et al., 1987; Brantley and Waitt, 1988; Fisher 1990; Druitt, 1992) indicates that the deposits from the expanded cloud decoupled from the transport system, moved gravitationally in the directions of existing slopes, and drained down some valleys as secondary pyroclastic flows. Photographs of the initial 1980 eruption at Mount St. Helens (Lipman and Mullineaux, 1981) indicate that the pyroclastic current (blast surge) extended as a turbulent cloud several hundred meters higher than the topography.

As indicated at Mount St. Helens, the transport of debris in an expanded flow with consequent deposition can continue across and in the lee of high ridges with possible generation of basal high-density sediment gravity flows draining into every watershed. Blocking of the flows occurs only in their basal parts where density values rise to such high levels that they cannot surmount the mountains (Valentine, 1987; Fisher, 1990).

In the case of the plug flow model (Sparks, 1976), a dense flow moving across the landscape under its own momentum may deposit all of its materials as a continuous sheet before any post-flow downhill movement takes place. Following emplacement, deposits of the sheet would then begin to creep or to flow downhill, imprinting the AMS flow fabric within the last few minutes of movement. Simultaneous emplacement of the entire sheet, however, would require that there is no momentum loss from the front to the back of the pyroclastic flow, and that the front of the flow comes to rest at the same time as the back of the flow. The space problem inherent in the emplacement of dense pyroclastic flows and the problems associated with en masse deposition have been pointed out by Branney and Kokelaar (1992).

Even allowing that a DF could move as a modified plug flow over steep mountainous terrain, and that deposition did not occur by instantaneous freezing throughout its extent, it is still difficult to explain how a high density ignimbrite sheet could be emplaced without the development of severe damming effects on distribution. Consider, for example, a DF moving up a steep mountain front with an average inclination of 45 degrees or more and with 1000 m of relief. In such steep terrain, materials being laid down by a dense flow could develop downward counter-flow against the forward moving current. AMS data in the Campanian Ignimbrite do not indicate opposing flow directions on steep slopes. Flow decoupling (Fisher, 1990, 1995; Buesch, 1992) could occur, but opposite moving dense flows would tend to create considerable drag and perhaps complex mixing at their interface, greatly reduce momentum on source-facing slopes, and restrict distribution of the deposits. This condition would be compounded if more than one ridge were encountered along the path of flow as is the case in the Campanian Ignimbrite distribution area. Furthermore, if it were possible for sheet-like dense flows to move over mountainous ridges into watersheds and then drain out through stream valleys, they could be blocked by another part of the same dense current moving up the valley.

Downslope gravity-driven turbidity currents, subaqueous analogues of pyroclastic flows and surges, exhibit behavior similar to that described herein for pyroclastic currents. "Upslope flow" is a term applied to observations that turbidity currents with enough momentum and/or flow thickness, can deposit sand and silt hundreds of meters up slopes and across bathymetric barriers (Shipley, 1978; Damuth, 1979; Damuth and Embley, 1979; Moore et al., 1982; Cita et al., 1984; Underwood, 1987; Dolan et al., 1989). Pickering et al. (1992) measured paleocurrent directions from a modern deep-marine basin, giving evidence that turbidity currents can be deflected or reflected at submarine barriers. From laboratory experiments, Muck and Underwood (1990) conclude that the most important variable influencing upslope movement of a turbidity current is the flow thickness.

The idea of reflection of deflected turbidity currents was used by van Andel and Komar (1969) to explain ponded sediments in marine basins. There have been some studies of flow reversals in turbidity current deposits (Ricci Lucchi and Valmori, 1980; Pickering and Hiscott, 1985), and they indicate that reverse-flow deposits can immediately follow those deposited by the primary flow. Such sequences are believed by researchers to be flow reversals caused by reflection rather than by later independent currents from the opposite direction. Reflection has been experimentally reproduced in flumes (Pantin and Leeder, 1987).

A key circumstantial relation suggesting that the Campanian pyroclastic transport system was an expanded flow is the fact that the Campanian Ignimbrite is over 43 m thick on the north side of the Sorrento Peninsula, some 35 km from Pozzuoli Bay across the Bay of Naples. Moreover, Campanian Ignimbrite up to 15 m thick occurs within the watershed draining into the sea at Maori on the south side of the Sorrento Peninsula. It is virtually impossible for the Campanian pyroclastic flow to have flowed around the Sorrento Peninsula to enter the Maori watershed from the south, therefore it must have overtopped the peninsula and flowed down the valleys. The lowest pass on the crest of the watershed is at 685 m altitude. At the time of the emplacement of the Campanian Ignimbrite, the Gulf of Napoli existed. Sea level may have been close to its present level between about 35,000 and 25,000 years ago and then receded as the last full glacial episode began (Kennett, 1982, p. 269). If sea level during the eruption of the Campanian Ignimbrite were nearly the same as today, the pyroclastic flow would have had to flow 25 to 35 km across open water to reach the Sorrento Peninsula. If the current were a dense pyroclastic flow, it would have probably entered the water, as has occurred in some historic pyroclastic currents (e.g., 1815 Tambora, Sigurdsson and Carey, 1989; 1902 Mt. Pele, Lacroix, 1904; 1976 Augustine Volcano, Fisher and Schmincke, 1984). At Mont Pelee the dense part of the nuee ardente entered the water while the dilute cloud above it flowed across the water, capsizing some of the boats and setting others on fire.

At Krakatau in 1883 a pyroclastic flow entered the sea, greatly altered the bathymetry around the volcano and generated tsunamis responsible for the deaths of many of the 36,000+ people (Self and Rampino, 1981; Sigurdsson et al., 1991). Sigurdsson et al. (1991) concluded that the initial phase of the Krakatau pyroclastic current was less dense than water, but after partial runout the flow base became denser than sea water by gravitational segregation, a process postulated by Fisher and Heiken (1982) for the 1902 nuee ardente at Mont Pelee. The upper part of the Krakatau pyroclastic current remained less dense than seawater and traveled >50 km over open water to the Sumatra coast where >2000 people lost their lives by burns. According to Sigurdsson et al. (1991), at 20 km from source, the current was still expanded to 800 m in height.

Because of profound mixing conditions that would occur between a submerged pyroclastic current and ambient water, it is not possible for the entire Campanian pyroclastic current to have entered and moved under water for 25-35 km to re-emerge as a hot, gas-inflated sediment gravity flow, and then move over the 650-1000 m Sorrento Peninsula to deposit more ignimbrite on the other side. It is more likely that a significant portion of it moved above the waters of the Bay of Naples as an expanded, turbulent suspension, similar to that postulated by Sigurdsson et al. (1991) for the 1883 Krakatau pyroclastic flow. The expanded transport system surmounted the Sorrento Peninsula and drained down its south and north slopes. At present there is no information available to us that indicates if Campanian Ignimbrite lies at the bottom of the Bay of Naples.

To conclude, our study indicates that the Campanian Ignimbrite was emplaced by an expanded turbulent pyroclastic current. The transport system is assumed to have been thicker than the highest topography and moved outward at 360 degrees from the possible source in the Bay of Pozzuoli, and the depositional system drained off ridges and down valleys in directions dictated by slope direction.

ACKNOWLEDGMENTS

We thank Peter Kokelaar, Nancy Riggs, Richard Waitt and an anonymous reviewer for their critical reviews, and express gratitude for research funds provided by the National Geographic Society to RVF in 1988 and the National Science U.S. - Italy Cooperative Research; National Science Foundation, Grant # INT - 9003232, to RVF and GH; and a grant from CNR, Italy to GO.

REFERENCES

Alessio, M., Bella, F., Improta, S., Belluomini, G., Cortesi, C., and Turi, B., 1971. University of Rome Carbon-14 Dates IX. Radiocarbon, 13(2): 395-411.

Alessio, M., Bella, F., Improta, S., Belluomini, G., Cortesi, C., and Turi, B., 1973. University of Rome Carbon-14 Dates X. Radiocarbon, 15(1): 165-178.

Alessio, M., Bella, F., Improta, S., Belluomini, G., Cortesi, C., and Turi, B., 1974. University of Rome Carbon-14 Dates XII. Radiocarbon, 16(3): 358-367.

Aramaki, S. and Ui, T., 1966. The Aira and Ata pyroclastic flows and related caldera depressions in southern Kyushu, Japan. Bull. Volcanol., 29: 29-47.

Barberi, F., Innocenti, F., Lirer, L. Munno, R. Pescatore, T. and Santacroce, R., 1978. The Campanian ignimbrite: A major prehistoric eruption in the Naples area (Italy), Bull. Volcanol., 41: 10-31.

Branney, M.J. and Kokelaar, P., 1992. reappraisal of ignimbrite emplacement: progressive aggradation and changes from particulate to non-particulate flow during emplacement of high-grade ignimbrite. Bull. Volcanol. 54:504-520.

Brantley, S.R. and Waitt, R.B., 1988. Interrelations among pyroclastic surge, pyroclastic flow, and lahars in Smith Creek valley during the first few minutes of 18 May 1980 eruption of Mount St. Helens, USA. Bull. Volcanol. 50:304-326.

Buesch, D.C., 1992, Incorporation and redistribution of locally-derived lithic fragments within a pyroclastic flow: Geological Society of America Bulletin v. 104, p. 1193-1207.

Capaldi, G. Civetta, L. and Gillot, P., 1985. Geochronology of Plio-Pleistocene Rocks from Southern Italy. Rend. Soc. Mineral. Petrol., XL(1): 25-44.

Cassignol, C. and Gillot, P., 1982. Range and effectiveness of unspiked potassium-argon dating: experimental ground work and application. In: G.S. Odin (Editor), Numerical Dating in Stratigraphy. pp. 160-179.

Cita, M.B., Beghi, C., Camerlenghi, A., Kastens, K.A. McCoy, F.W., Nosetto, A., Parisi, E., Scolari, F., and Tomadin, L., 1984. Turbidites and megaturbidites from the Herodotus abyssal plain (eastern Mediterranean) unrelated to seismic events. Marine Geology, 55: 79-101.

Crane, K., Bonatti, E., Labrecque, J., and Bell, R., 1985. Bay of Naples geophysical profiles. R/V Conrad C 2508, L-DGO, Tech. Rept.

Curtiss, G.H., 1966. The problem of contamination in obtaining accurate dates of young geologic rocks. In: O.A. Shaeffer and J. Zahringer (Editors), Potassium Argon Dating. Springer, Berlin.

Damuth, J.E., 1979. Migrating sediment waves created by turbidity currents in the northern South China Sea: Geology, 7: 520-523.

Damuth, J.E. and Embley, R.W., 1979. Upslope flow of turbidity currents on the northwest flank of the Ceara Rise: Western equatorial Atlantic. Sedimentology, 26: 826-834.

Deino, A., Curtis, G. and Rosi, M., 1992. 40Ar/39Ar dating of the Campanian Ignimbrite, Campanian Region, Ital. International Geological Congress, Kyoto, Japan, 24 August - 3 Sept., Abstracts vol. 3, p. 633.

Di Girolamo, P. Ghiara, M.R., Lirer, L., Munno, R., Rolandi, G. and Stanzione, D., 1984. Vulcanologia e petrologia dei Campi Flegrei. Boll. Soc. Geol. It., 103: 349-413.

Dolan, J.F., Beck, C., and Ogawa, Y., 1989. Upslope deposition of extremely distal turbidites: An example from the Tiburon Rise, west-central Atlantic. Geology, 17: 990-994.

Ellwood, B.B., 1982. Estimates of flow direction for calc-alkaline welded tuffs and paleomagnetic data reliability from anisotropy of magnetic susceptibility measurements: Central San Juan Mountains, southwest Colorado. Earth Planet. Sci. Letters, 59: 303-314.

Elston, W.E. and Smith, E.I., 1970. Determination of flow direction of rhyolitic ash-flow tuff from fluidal textures. Geol. Soc. Amer. Bull., 81:3393-3406.

Fisher, R.V., 1983. Flow transformations in sediment gravity flows. Geology, 11: 273-274.

Fisher, R.V., 1990. Transport and deposition of a pyroclastic surge across an area of high relief: The 18 May 1980 eruption of Mount St. Helens, Washington. Geol. Soc. America Bull., 102: 1038-1054.

Fisher, R.V., 1995. Decoupling of pyroclastic currents: hazards assessments. Special Issue, Glicken Volume, Jour. Volcanol. Geotherm. Res., 66:257-263.

Fisher, R.V. and Heiken, G., 1982. Mt. Pelee, Martinique: May 8 and 20, pyroclastic flows and surges. J. Volcanol. Geotherm. Res., 13:339-371.

Fisher, R.V. and Schmincke, H.-U., 1984. Pyroclastic rocks. Berlin, Heidelberg, New York, Tokyo, Springer-Verlag, 472 p.

Fisher, R.V., Glicken, H.X. and Hoblitt, R.H., 1987. May 18, 1980 Mount St. Helens deposits in South Coldwater Creek, Washington. J. Geophys. Res., 92: 10,267-10,283.

Froggatt, P.C., Wilson, C.J.N. and Walker, G.P.L., 1981. Orientation of logs in the Taupo Ignimbrite as indicator of flow direction and vent position. Geology, 9: 109-111.

Giannetti, B., 1979. Carta geologica della Caldera del Vulcano di Roccamonfina. Boll. Serv. Geol. d'Italia, C.

Hillhouse, J.W. and Wells, R.E., 1991. Magnetic fabric, flow directions and source area of the Lower Miocene Peach Springs Tuff in Arizona, California, and Nevada. Jour. Geophys. Res 96:12,443-12,460.

Hoblitt, R.P., Miller, C.D. and Vallance, 1981. Origin and stratigraphy of the deposit produced by the May 18 directed blast. In, Lipman, P.W. and Mullineaux, D.R., (Editors), The 1980 eruptions of Mount St. Helens, Washington. U.S. Geol. Survey Prof. Paper 1250: 401-419.

Hsu, K.J., 1975. Catastrophic debris streams (Sturzstroms) generated by rockfalls. Geol. Soc. Amer. Bull. 86:129-140.

Incoronato, A., Addison, F.T., Tarling, D.H., Nardi, G. and Pescatore, T., 1983. Magnetic fabric investigations of pyroclastic deposits from Phlegrean Fields, southern Italy. Nature, 306: 461-463.

Kamata, H., and Mimura, K., 1983. Flow directions inferred from imbrication in the Handa pyroclastic flow deposit in Japan. Bull. Volcanol., 46: 277-282.

Kennet, J.P., 1982. Marine Geology. New Jersey, Prentice-Hall, Inc., 813 pp.

Knight, M.D., Walker, G.P.L., Ellwood, B.B. and Diehl, J.F., 1986. Stratigraphy, paleomagnetism and magnetic fabric of the Toba tuffs: constraints on the source and eruptive styles. J. Geophys. Res., 91: 10,355-10,382.

Koch, A.J. and McLean, H., 1975. Pleistocene tephra and ash-flow deposits in the volcanic highlands of Guatemala. Geol. Soc. Amer. Bull., 86: 529-541.

Lacroix, A., 1904. La Montagne Pele et ses eruptions. Masson et Cie, Paris, 662 pp.

Lipman, P.W. and Mullineaux, eds., 1981. The 1980 eruptions of Mount St. Helens. U.S. Geol. Survey Prof. Paper 1250:1-844.

MacDonald, W.D., and Palmer, H.C., 1990. Flow directions in ash-flow tuffs: a comparison of geological and magnetic susceptibility measurements, Tshirege member (upper Bandelier Tuff), Valles caldera, New Mexico, USA. Bull. Volcanol., 53: 45-59.

McEwen, A.S. and Malin, M.C., 1989. Dynamics of Mount St. Helens' pyroclastic flows, rockslide-avalanche, lahars and blast. J. Volcanol. Geotherm. Res. 37:205-231.

Miller, T.P. and Smith, R.L., 1977. Spectacular mobility of ash flows around Aniakchak and Fisher calderas, Alaska. Geology, 5: 173-176.

Moore, J.C., and 13 others, 1982. Facies belts of the Middle America Trenches and forearc region, southern Mexico: Results from Leg 66 DSDP, in, Leggett, J.K., ed., Trench-Forearc geology. Geol. soc. London Spec. Publ. 10:77-94.

Muck, M.T. and Underwood, M.B., 1990. Upslope flow of turbidity currents: A comparison among field observations, theory, and laboratory models. Geology, 18: 54-57.

Palmer, H.C., MacDonald, W.D., and Hayatsu, A., 1991. Magnetic, structural and geochronologic evidence bearing on volcanic sources and Oligocene deformation of ash flow tuffs, northeast Nevada; J. Geophys. Res., 96: 2185-2202.

Pantin, H.M. and Leeder, M.R., 1987. Reverse flow in trubidity currents: the role of internal solitons. Sedimentology 34:1143-1155.

Paterne, M., Guichard, F. and Labeyrie, J., 1988. Explosive activity of the south Italian volcanoes during the past 80,000 years as determined by marine tephrochronology. J. Volcanol. Geotherm. Res., 34: 153-172.

Pickering, K.T. and Hiscott, R.N., 1985. Contained (reflected) turbidity currents from the Middle Ordovician Cloridorme Formation, Quebec, Canada: an alternative to the antidune hypothesis. Sedimentology 32:373-394.

Pickering, K.T., Underwood, M.B. and Taira, A., 1992. Open-ocean to trench turbidity-current flow in the Nankai Trough: Flow collapse and reflection. Geology 20:1099-1102.

Potter, D.B., and Oberthal, C.M., 1987. Vent sites and flow directions of the Otowi ash flows (lower Bandelier Tuff), New Mexico; Geol. Soc. Amer. Bull., 98: 66-76.

Ricci Lucchi, F. and Valmori, E., 1980. Basin-wide turbidites in a Miocene, over-supplied deep-sea plain: a geometrical analysis. Sedimentology 27:241-270.

Rose, W.I., Jr., Grant, N.K. and Easter, J., 1979. Geochemistry of the Los Chocoyos Ash, Guatemala. Geol. Soc. Amer. Spec. Paper 180: 87-99.

Rosi, M. and Sbrana, A., 1987. The Phlegrean Fields. CNR, Quaderni di La Ricerca Scientifica, 114, 175 pp.

Rosi, M., Sbrana, A. and Principe, C., 1983. The Phlegraean Fields: Structural evolution, volcanic history and eruptive mechanisms. In: M.F. Sheridan and F. Barberi (eds.), Explosive Volcanism. J. Volcanol. Geotherm. Res., 17: 89-109.

Scandone, R., Bellucci, F., Lirer, L., and Rolandi, G., 1991. The structure of the Campanian Plain and the activity of the Neapolitan volcanoes (Italy). J. Volcanol. Geotherm. Res., 48: 1-31.

Seaman, S.J. and Williams, M.L., 1992. Center-to-center analysis and flow fabric characterization in ash-flow tuffs. Bull. Volcanol. 54:319-328.

Self, S. and Rampino, M., 1981. The 1883 eruption of Krakatau. Nature 292:699-704.

Sheridan, M.F., 1979. Emplacement of pyroclastic flows: A review. Geol. Soc. Amer. Spec. Paper 180: 125-136.

Shipley, T., 1978. Sedimentation and echo characteristics in the abyssal hills of the west-central North Atlantic. Geol. Soc. America Bull., 89: 397-408.

Sigurdsson, H. and Carey, S., 1989. Plinian and co-ignimbrite tephra fall from the 1815 eruption of Tambora volcano. Bull. Volcanol. 51: 234-270.

Sigurdsson, H., Carey, S., and Mandeville, C., 1991. Krakatau. National Geographic Research and Exploration 7(3): 310-327.

Sparks, R.S.J., 1976. Grain size variations in ignimbrites and implications for the transport of pyroclastic flows. Sedimentology, 23: 147-188.

Sparks, R.S.J., Self, S. and Walker, G.P.L., 1973. Products of ignimbrite eruption. Geology, 1: 115-118.

Sparks, R.S.J., Wilson, L., and Hulme, G., 1978. Theoretical modeling of the generation, movement, and emplacement of pyroclastic flows by column collapse. J. Geophys. Res., 83: 1727-1739.

Suzuki, K., and Ui, T., 1982. Grain orientation and depositional ramps as flow direction indicators of large-scale pyroclastic flow deposits, Japan. Geology, 10: 429-432.

Suzuki, K., and Ui, T., 1983. Factors governing the flow lineation of a large-scale pyroclastic flow -- an example in the Ata pyroclastic flow deposit, Japan. Bull. Volcanol., 46: 71-81.

Suzuki, K., and Ui, T., 1988. Depositional ramps: asymmetrical distribution structure in the Ata pyroclastic flow deposit, Japan. Bull. Volcanol., 50: 26-34.

Ui, T., Suzuki-Kamata, K., Matsusue, R., Fujita, K., Metsugi, H., and Araki, M., 1989. Flow behavior of large-scale flows -- Evidence obtained from petrofabric analysis. Bull. Volcanol., 51: 115-122.

Underwood, M.B., 1987. Empirical evidence for upslope flow of turbidity currents in the Aleutian forearc. Geol. Soc. America Abstracts with Programs, 19: 874.

Valentine, G.A., 1987. Stratified flow in pyroclastic surges. Bull. Volcanol., 49: 616-630.

van Andel, T.H., and Komar, P.D., 1969. Ponded sediments of theMid-Atlantic Ridge between 22 and 23 north latitude. Geol. Soc. America Bull. 80:1163-1190.

Walker, G.P.L., Wilson, L. and Bowell, E.L.G., 1971. Explosive volcanic eruptions - I. The rate and fall of pyroclasts. Geophys. J. Roy. Astron. Soc. 22:377-383.

Wilson, C.J.N., 1985, The Taupo eruption, New Zealand II. The Taupo Ignimbrite. Philosophical Transactions, Royal Society of London A 314: 229-310.

Wilson, C.J.N. and Walker, G.P.L., 1985, The Taupo eruption, New Zealand I. General aspects. Philosophical Transactions, Royal Society of London, A 314: 199-228.

Wolff, J.A., Ellwood, B.B., and Sachs, S.D., 1989. Anisotropy of magnetic susceptibility in welded tuffs: Application to a welded-tuff dyke in the Tertiary Trans-Pecos Texas volcanic province, USA; Bull. Volcanol., 51: 299-310.

Yokoyama, S., 1974. Flow and emplacement mechanism of the Ito pyroclastic flow in southern Kyushu, Japan. Sci. Rep. Tokyo Kyoiku Daigaku, Sect. C, 12: 17-62.