Hawaiian Calderas — MACDONALD 
331 
entered the ocean; but the volumes, and even 
the existence, of any eruptions that may have 
taken place below sea level are wholly unknown. 
A mere glance at Table 1 shows that there 
are large discrepancies between the volumes of 
several of the collapses and those of the accom- 
panying lava flows. The smaller of the discrep- 
ancies, such as those for 1840 and 1955, can 
reasonably be attributed to intrusion of magma 
as dikes in the rift zones. The volume of extru- 
sion in 1868 can only be guessed at, because 
we do not know the depth of the fill in Kilauea 
Iki crater. However, the eruption on the south- 
west rift zone had a volume of less than 200,000 
m 3 , and that in Kilauea Iki cannot be reason- 
ably assumed to have been much more than 
5,000,000 m 3 and may well have been a good 
deal less than that, so that we are left with a 
discrepancy of something more than 180,000,000 
m 3 . This also could be accounted for by the 
intrusion of an unusually large dike, averag- 
ing about 2 m thick and 2 km high above the 
top of the magma chamber, and extending 45 
km across the top of the volcano. (Both rift 
zones opened during the eruption.) The vol- 
ume of lava erupted above sea level in 1832 
also is not accurately known, but it was very 
small. Furthermore, the only known vents were 
above the level of the caldera floor, so that 
eruption from them can hardly be considered 
a simple draining of magma from beneath the 
caldera block, allowing the latter to sink. On 
the other hand, the sinking of the block may 
have helped squeeze some magma to the sur- 
face. In 1924 there was no eruption of lava 
whatever above sea level. The discrepancies 
between the subsidences of 1823, 1832, and 
1924 and the volume of known contemporane- 
ous lava flows above sea level are, respectively, 
more than 525 million, approximately 580 mil- 
lion, and 200 million m 3 . During the 1924 sub- 
sidence the east rift zone opened for a distance 
of more than 45 km, to and beyond the east 
cape of the island, and Jaggar (1934) believed 
that a submarine eruption occurred. It is cer- 
tainly a distinct possibility, and the same could 
have happened during any of the other epi- 
sodes of subsidence. An eruption in deep water 
would quite likely have escaped detection. The 
volume discrepancies can reasonably be attrib- 
uted to rift-zone intrusion and/or submarine 
eruption. 
Thus, the facts that must be taken into con- 
sideration in a theory explaining caldera forma- 
tion appear to be: Caldera collapse takes place 
generally, if not always, following a period of 
tumescence of the volcano, accompanying or di- 
rectly following an opening of one or both of 
the main rift zones, commonly accompanied by 
subaerial flank eruption and/or possible subma- 
rine eruption; and the sinking takes place on 
fractures that converge downward, which in turn 
means that sufficient outward movement of the 
circumferential mass must take place to make 
room for the increasing diameter of the sinking 
wedge at any given level. The distension of the 
summit region is very probably a part of, and is 
caused by the same mechanism as, the simul- 
taneous distension of the rift-zone regions. 
Although eruptions are nearly always accom- 
panied by a general sinking of the mountain- 
top, caldera collapse may not take place. Erup- 
tion may be confined to the caldera or the 
immediate summit region without any apparent 
opening of the rift zones, or opening of the 
rifts only close to the caldera. 
The following possible interpretation of the 
above facts, together with the other facts as 
to the general structure of the volcano pre- 
sented earlier, is suggested. Magma very rich 
in olivine crystals rises from the mantle into 
the volcano, adding to the bulk of the ultra- 
dense core; and part of the liquid portion of 
the magma rises on upward, leaving behind 
most of the load of solid crystals, to form a 
pool of fluid tholeiitic basalt in a high-level 
magma reservoir at the top of the ultra-dense 
core or, in some volcanoes, somewhat to one 
side of the top of the core, beneath the sum- 
mit of the shield and extending outward be- 
neath the near-summit portions of the rift 
zones. The volcano swells in response to the 
addition of magma. Local cupolas on the magma 
body may rise high enough to perforate the 
roof, allowing long-continued open-vent activ- 
ity such as the lava lake that existed in Hale- 
maumau crater most of the time through the 
19th and early part of the 20th centuries, or 
the eruption in Mokuaweoweo during most of 
1873 and 1874. Occasional withdrawals of 
