In general, Hawaiian volcanoes during their most-active tholeiite shield stage can be characterized as having gentle slopes extending from the sea floor up to a summit caldera. The submarine slopes of Hawaiian volcanoes are steeper than the subaerial slopes but even they rarely exceed 14º (Mark & Moore 1987). In the alkalic stage the caldera is usually filled in and as noted earlier the slopes become steeper.
The greater proportion of explosive activity provides a good deal of ash causing Hawaiian volcanoes in their post-shield alkalic stage to resemble strato volcanoes to some degree. The most prominent large-scale structures are calderas and rift zones.
Located at the summits of both Kilauea and Mauna Loa are calderas. The areas of the calderas are 15 and 11 square kilometers, respectively, and their depths range from 140 to 170 m. The outlines of the calderas are distinctly non-circular, strong evidence that they result from the coalescence of more than one center of collapse. This is particularly evident in the shape of Moku'aweoweo, the caldera of Mauna Loa. Moku'aweoweo is strongly aligned in a direction that points towards the two Mauna Loa rift zones. The smaller collapse features that have coalesced to form Moku'aweoweo indicate the locations of zones of magma storage. This is evidence that the main magma chamber complex itself is aligned along these directions.
How do calderas actually form on basaltic volcanoes? Basaltic calderas are often shown as if they formed when a big piston-shaped cylinder of rock drops into the magma chamber, but this is not correct. A growing body of evidence is pointing to the idea that the calderas are sag structures that form when support is removed from below the summit. Walker (1988) showed that the cumulative subsidence of the Kilauea caldera has been funnel-shaped. (See below for the collapse and infilling history of Kilauea caldera since the arrival of Westerners. The bottom diagram shows the cumulative collapse to be funnel-shaped, not piston-shaped (adapted from Walker 1988). Additionally, a study of the eroded Ko'olau caldera showed that all the caldera-filling lavas have a centripetal dip (Walker 1988). The reason active Hawaiian calderas have steep walls is that surface rocks are too brittle to sag very well, and they fracture. The flat floors result from the re-surfacing by flows erupted within the calderas.
This resurfacing points to the fact that calderas are very dynamic features that are capable of collapsing and infilling many times. Indeed, ~25% of the surface of Mauna Loa is covered with lavas erupted ~600 years ago during a time when the summit caldera was full and overflowing (Lockwood & Lipman 1987)
This is not to say that distinct collapse events do not occur. Three large pyroclastic units have been mapped at Kilauea, and each can be correlated with a caldera collapse event (but not necessarily a caldera-forming event). The prevailing idea is that when magma drains suddenly from the summit region, support of the caldera floor is removed and collapse occurs. Groundwater is then able to flow inward towards the hot volcanic plumbing, and it flashes to steam, producing phreatic eruptions (as in 1924) or phreatomagmatic eruptions (as in 1790; e.g. McPhie et al. 1990).
Calderas are thus presumed to be have come and gone during the tholeiite stages of all the Hawaiian volcanoes, however, direct evidence is not present at all of them. Mauna Kea and Hualalai, for example, show no evidence of having had calderas in the past. As noted above, the "caldera" of Haleakala, although spectacular, is actually an erosional feature on the East Maui volcano, and evidence of a true volcanic caldera there has likewise not been found.
On some of the other older volcanoes the presence of old calderas manifests itself in sequences of thick flat-lying ponded flows, and areas of preferential erosion. This is particularly the case for East Moloka'i volcano. In these cases the centers of the volcanoes are now occupied by big holes. Caldera-filling lavas are usually easier to erode than flank lava flows because during the active period of the volcano's life the center of the volcano (the caldera) is the zone of greatest thermal and hydrothermal alteration and the rocks are quickly reduced to clays. An alternative idea is that the calderas happened to be in a state of collapse (rather than infilling) at the end of the tholeiite stage.
The image to the right shows the windward side of the old Ko'olau volcano (which comprises the east half of the island of O'ahu). The red line marks the crest of the pali (cliff). Numerous people have suggested that these cliffs are the scarp of the giant landslide that has removed most of the right-hand side of the Ko'olau volcano. If this were so, however, one would not expect there to be any high-standing intra-caldera rocks such as those inside the yellow marker. It seems more likely that the actual scarp is offshore from the present coastline; all the erosion between this scarp and the present pali has been accomplished by non-catastrophic processes since (but probably accentuated by) the giant landslide.
The situation seems to have been reversed on the old Kaua'i volcano. Here the ponded flows were so massive that they present more resistance to erosion compared to the flank lavas, and the outline of the old Kaua'i caldera is today marked by a high. relatively circular plateau. This also assumes that Kaua'i was only a single volcano, an idea recently challenged by Robin Holcomb of the USGS.
To the left is an air view (towards the west) of the caldera of Kaua'i volcano. Here the relatively horizontal plateau (outlined with the dashed light-blue line) consists of thick slightly more erosion-resistant lavas that ponded in the old caldera; it is topographically higher than the deeply eroded flank lavas. The yellow line marks the boundary between flank lavas and caldera lavas. Kalalau Valley and Waimea Canyon are eroding their way into the central plateau. Photo from Macdonald et al. 1983.
Radiating away from the summits of Hawaiian volcanoes are (usually two) linear rift zones. The rift zones conspicuously do not point towards adjacent volcanoes, but instead parallel the volcano-volcano boundaries. Rift zones mark preferred directions of sub-horizontal magma excursions from the magma chamber. Below is a map of the main Hawaiian islands showing rift zones in red lines and volcanic centers as red squares. Note that the rift zones tend to parallel the volcano boundaries, and avoid pointing at each other (from Fiske & Jackson 1972).
At the surface they are characterized by numerous vents, fissures, earth cracks, cinder cones, graben, pit craters, and the sources of lava flows. All of these are indications that magma preferentially intrudes into the rift zones and is also often stored there for periods of time up to a few years.
The vertical air photo on the left shows of a section of the NE rift zone of Mauna Loa. Even without the arrow it is pretty easy to figure out where the axis of the rift zone is. The red numbers give the dates of the flows (from Macdonald & Abbot 1970).
There has been much discussion about the formation and persistence of Hawaiian rift zones (e.g. Fiske & Jackson 1972; Deterich 1988). The general idea is that because Hawaiian volcanoes are close to one another relative to their size, a younger volcano is growing through the flank of an older one. The gravitational stress field caused by the pre-existing volcano tends to yield downslope-directed directions of least compressive stresses. Because dikes orient themselves so that their direction of widening is parallel to this least compressive stress, the dikes end up propagating parallel to the volcano-volcano boundary. Once a preferred direction of dike propagation is established, it is self-perpetuating as long as there is a mechanism for the flanks of a volcano to move outward to accommodate the repeated dike injections.
On the right is a schematic representation of Kilauea (purple) growing on the flank of Mauna Loa (green). Note how Kilauea has been affected by the shape (and hence the
stress orientation) of its huge neighbor, and has adopted the same rift zone orientation (from Fiske & Jackson 1972).
The most popular mechanism for this outward movement is sliding along the volcano-ocean floor interface which consists of easily-deformable sediments (e.g. Nakamura 1982). The focal mechanism for the 1975 M7.2 Kalapana earthquake indicated a slip plane that was nearly horizontal with a slight dip towards at a depth consistent with the base of the volcano (e.g. Lipman et al. 1985). Such an orientation would be expected due to the downward warping of the oceanic lithosphere under the load of the island.
Above is a schematic cross-section through Kilauea and part of Mauna Loa, viewed towards the East. This shows how the seaward flank of Kilauea (and part of Mauna Loa) is pushed southward (to the right) by the intrusion of dikes down the rift zone (away from you into the plane of the diagram). This huge bulk of volcano is probably sliding on ocean sediments that accumulated on the ocean floor during the 90 million or so years between the time that our particular part of Pacific Plate formed and when the Big Island of Hawai'i started to grow.
Rift zones probably become preferred directions of dike propagation due to stress orientations, and they evolve thermally to perpetuate themselves. This means that eruptions are rare elsewhere on the flanks of the shields. Except at the summit, the vents of Kilauea are found exclusively along the rift zones. On Mauna Loa, however, there is a class of vents called "radial vents " (Lockwood & Lipman 1987) that are found on the northern and western flanks. This is the sector on the obtuse side of the angle formed by the two rift zones, and circumferential tension caused by a bending moment set up by the rift zones and the westward push of neighboring may be leading to the formation of these vents (Walker 1990).
To the left is a map of the big island with Mauna Loa in orange. The short white lines are the "radial rifts" that do not fall into either of the rift zones (NERZ and SWRZ). Note that one of these radial rifts erupted through the flank of Mauna Kea, and that another erupted offshore (in 1877). Adapted from Lockwood & Lipman 1987.
Probably the most studied rift zone is the east rift of Kilauea. The northern flank of this rift is stable, probably because it abuts Mauna Loa. The south flank, however, is notably mobile. It has been shown to move seaward during both earthquakes and intrusive events. There is nothing in this direction to buttress the flank so the continued pressure caused by numerous dike intrusions produces this seaward displacement (Swanson et al. 1976; Lipman et al. 1985). This relative displacement between the non-mobile north flank and mobile south flank has caused a wide graben to form along the crest of the rift. Thus even though the rift axis is the locus of most eruptive activity it is in places topographically subdued. Some of the faults bounding this graben are visible near Napau crater.
Vertical air photo of Napau pit crater along the East rift zone of Kilauea. Napau has been almost filled by recent lavas (here making it look smooth relative to the surrounding forest). Note that vents, faults, fissures, and smaller pit craters are all aligned from the lower left (uprift) to upper right (downrift). The actual rift zone is wider than this photo (from Carr & Greeley 1980). Note also that differences in vegetation make flow margins traceable - the dotted white lines outline an old flow that appears to have had a source that is now engulfed into Napau crater.
Continued transport of magma down the rift zone results in the establishment of a thermally efficient conduit probably 2-3 km below the surface. Some evidence for this was provided by the first 10 km of propagation of the dike marking the onset of the Pu'u 'O'o/Kupa'ianaha eruption being aseismic (Klein et al. 1987). This indicates that there was a pre-existing conduit could be utilized by the migrating magma. This distance corresponds rather closely with the distribution of pit craters along the east rift Kilauea. Beyond the first 10 km, earthquakes marked the propagation of the dike.
Pit craters are not explosion craters or vents, but rather they are locations of localized collapse into a void. The above-mentioned conduit is the best candidate for such a void. A pit crater forms from the bottom up by stoping of a cavity.
The schematic diagram on the left shows the formation of a pit crater from the bottom up. This cross-section is cut perpendicular to the line of a rift zone, and "C" represents the main conduit at 2-3 km depth. Note that the crater is not a piston that has dropped and that the top lavas are the last to fall in (from Walker 1988). This process of a void working its way upward through solid rock (like a bubble) is called "stoping", and it commonly occurs in mines. The void and the eventual crater have a volume greater than the conduit, however, the conduit can continually carry material downrift.
Evidence for this is provided by a pit crater called the "Devil's Throat." When first noticed by Westerners, Devils Throat was a hole in the ground a few 10's of m across. A very brave man was lowered through this opening on a winch, and he soon found himself in a huge cavity, much wider than the hole he'd come through. It was evident that he was in a bell-shaped void and that the top layers of lava had not yet collapsed into it. Since then the last layers have fallen in, leaving Devil's Throat with the more typical cylindrical form of a pit crater.
On the right is a photo into Devil's Throat pit crater. Note that numerous pre-existing flows that are exposed in the walls, and the geology student for scale.
Eruptive fissures occasionally cut right across pit craters apparently without noticing the difference in topography. An eruptive fissure can extend from the floor of a pit crater, up the wall, and continue on beyond the rim.
Photo (taken in 1973 by Gordon Macdonald) showing a short curved fissure erupting across the floor of Makaopuhi pit crater. Lava is erupting (but not fountaining) from the near end whereas the far end is emitting only steam. The dotted line marks the base of the far wall of the crater. Makaopuhi is actually a double crater, and the lava cascades mark the boundary between the deeper near half and the shallower far half. The near half was previously twice as deep as the far half; eruptions prior to the one seen here almost made the two levels the same, and this eruption did eventually fill the deeper half to give the crater a single floor level.
In summary, vent, graben, and pit crater distributions yield insight to the preferred directions of magma travel within a Hawaiian volcano. These in turn can yield information about the stress directions within the edifice (e.g. McGuire & Pullen 1989; Rubin 1990). There have been some attempts to tie these stress directions to the stress field within the Pacific plate but it appears that the local stress field caused by neighboring volcanoes is much more important in determining the eventual direction of rift zone formation.