356 
PACIFIC SCIENCE, Vol. XV, July 1961 
the past, within 15 mi. upslope from the prob- 
able site of a diversion structure, but there is no 
known geologic reason why an eruption may not 
break through in this area. A reasoned decision 
about the necessary barrier design cannot be 
made on such data; the decision must be based 
on other considerations. 
Another, and completely unrelated, problem 
of design for which no geologic or engineering 
solution is possible rises from certain charac- 
teristics of a lava flow (Wentworth, 1954). 
Every flow of lava inevitably sends out distribu- 
tary flows from time to time and from place to 
place along its course, as one way of responding 
to frequent large fluctuations in the amount 
and rate of eruption of lava at the source vent. 
Therefore, it may be expected that more than 
one flow of lava will enter the channel of the 
diversion system during any one eruption. Inas- 
much as mobile lava becomes immobile rock as 
soon as it cools slightly, a considerable amount 
of any lava that enters the channel system will 
solidify there and form an obstruction in the 
channel. Thus, any subsequent flow of lava that 
enters the channel system at a point upgrade 
will have to override this obstruction in order 
to keep on moving downgrade. If the channel 
system has been built with enough capacity, the 
overriding flow will be contained and the sys- 
tem will continue to function; if the system has 
too small a capacity at this point, the barrier 
wall of the channel will be overrun at the ob- 
stacle and the diversion system will fail to func- 
tion. 
At the designing stage of an adequate diver- 
sion system, it is obviously impossible to antici- 
pate the point at which a future first lava flow 
will enter the system, to estimate the magnitude 
of the obstruction that it will form, or to ap- 
praise the amount of lava that may have to pass 
over the obstruction. The designer can cope 
with this situation only by overdesigning the 
entire system. He can only guess how much to 
overdesign : — twofold ? — tenfold ? 
In considering design of barriers and diver- 
sion channels, the tendency of liquids adjacent 
to a dam to cause uplift pressure and to burrow 
through should be realized. To allow for such 
tendency is standard practice in designing dams, 
because some have failed in this way. Lava bar- 
riers have also failed in this way, as was recently 
observed in some instances at Kapoho. How- 
ever, in the case of a massive barrier built of 
well-compacted rock and soil, this is thought to 
be a very remote contingency because of the 
cooling effect. Lava might retain liquidity 
through tenuous openings for a distance of 200 
or 300 ft. but would seem unlikely to do so 
through 1,000 ft. or more except in a pre-estab- 
lished tube. Such an accident is not entirely dis- 
missable, however. 
SAMPLE ESTIMATES OF DIVERSION 
CHANNEL DIMENSIONS 
We can neglect for the moment the impon- 
derable matter of overdesign and consider the 
dimensions required to convey two sample lava 
flows that may be assumed to move as simple 
flow units. 
The average natural gradient of the trough 
that leads to Hilo, which must be intercepted 
by the diversion system, is between 250 and 
300 ft/mi. The diversion channel probably 
could be laid out with an average gradient of 
no more than 200 ft/mi. Estimates of the ve- 
locity of movement of lava flows on comparable 
low gradient can be made from published de- 
scriptions of previous flows. The hot, mobile 
lava near the vent of the 1954 eruption of Ki- 
lauea (Macdonald and Eaton, 1954) moved at 
rates not less than 400 yd/hr. A channel de- 
signed to move 25,000,000 cu. yd. of hot, mobile 
lava at this velocity would need to provide space 
for a flow cross-section of 63,000 sq. yd. If a 
containing barrier on the downslope margin of 
the channel were built high enough to give an 
average depth of flow of 20 yd. in the channel, 
the width of the channel would be 3,150 yd. 
(approaching 2 mi. wide), and the maximum 
rock excavation at the upslope margin would be 
greater than 400 ft. 
A different example: the relatively cool and 
viscous lava of the 1926 flow that destroyed the 
beach village of Hoopuloa (Hawaiian Volcano 
Observatory, 1926) moved at rates not less than 
60 yd/hr. A similar relatively cool flow from a 
distant vent reaching the diversion system at a 
rate of 2,000,000 cu. yd/hr would require a 
channel cross-section of nearly 34,000 sq. yd. to 
