SECOND WORKSHOP ON MARS VALLEY NETWORKS
11
GULLEY FORMATION IN THE KEANAKAKO’I TEPHRA DEPOSIT:
IMPLICATIONS FOR
EROSION ON MARS. Robert A. Craddock1, Alan D. Howard2, Pao-Shin Chu3, Rossman P. Irwin, III1, Don
Swanson4, Stephen Tooth5, Rebecca M. E. Williams6. James Zimbelman1, 1Center for Earth and Planetary Studies,
National Air and Space Museum, Smithsonian Institution, Washington, DC 20560, [email protected]),
2
Department of Environmental Sciences, University of Virginia, Charlottesville, Virginia, 3Department of Meteorology, University of Hawaii, Honolulu, Hawaii, 4Hawaiian Volcano Observatory, U.S. Geological Survey, Hawaii
Volcanoes National Park, Hawaii, 5Institute of Geography and Earth Sciences, Aberystwyth University, Wales,
6
Planetary Science Institute, Tucson, Arizona.
Introduction: Because of their obvious association with liquid water [1] martian valley networks have
been intensely studied. As abstracts presented at this
workshop attest there have been numerous studies attempting to map and quantify various attributes related
to valley networks in an effort to better understand
when they were active, how long they were active, and
how much water was involved in their formation.
However, it is also important to consider lithology in
any watershed analysis. Lithology influences the rate
of channel incision, channel morphology, and channel
hydrology as well as the characteristics of transported
materials, but to date very little attention has been paid
to the affects we might expect from the lithology of the
martian surface. While it is not possible to say exactly
what the lithology might be at any particular location,
especially at depth, there is clear evidence that most of
the surface of Mars is basaltic in composition [e.g., 2],
most likely brecciated due to impact cratering or volcanism, and often layered [3]. Unfortunately, very few
places on Earth have similar geologic characteristics.
One notable exception, however, is the Keanakako’i
tephra deposit located on the western flank of Kilauea
volcano in the Ka’u Desert of Hawaii. We have begun
a series of analyses of gulley development in these
materials in an effort to better understand some of the
affects lithology might have in creating the martian
valley networks.
Geologic Setting: Kilauea is an active basaltic volcano that constitutes the southeastern portion of the
Island of Hawaii. The Ka’u Desert is located on the
western flank of Kilauea and is roughly bounded by
the location of Hawaii State Highway 11 to the north,
by the National Park Service’s Hilina Pali Road to the
south, by Kilauea caldera to the east, and by the Pacific
Ocean to the west; in all encompassing ~350 km2.
Typically the area only receives ~150 mm of rain
every year, but occasionally a heavy thunderstorm will
precipitate many times this amount in a single event.
Adding to the effects from the low average annual
rainfall, the Ka’u area remains a desert also because of
outgassing from the central pit caldera, Halemaumau,
and fumaroles that are concentrated along fractures
near the summit of Kilauea. On average Kilauea outgases over 4.15 x 105 tons of SO2 every year [4]. The
trade winds that blow steadily towards the west carry
these gasses into the desert, creating harsh, acidic conditions that deters almost all plant life.
The Keanakako’i Formation: The Keanakako’i
Formation essentially includes all fragmental deposits
emplaced on the rim of Kilauea by explosive eruptions.
Near the summit of Kilauea the Keanakako’i Formation is more than 10 m thick [e.g., 5] with a large exposure located in the Keanakako’i Crater, thus giving it
its name. There are many layers within the Keanakako’i Formation that contain the history of explosive
eruptions at Kilauea; however, the timing, duration,
and nature of these eruptions have been the subject of
contention, primarily because internal disconformities
within the formation have been interpreted differently.
A full discussion of these various interpretations is
beyond the scope of this abstract. However, there are
some general characteristics that most investigators
agree upon. Basically, there are two principal units
that are easily distinguishable in most exposed crosssections. The upper unit consists of ash, lapilli, and
lithic fragments and has a slight purplish color. The
lower unit consists of vitric ash and pumice and has a
slight greenish-gold color. Noting these distinct differences between the upper and lower units, Powers
[6] suggested phreatomagmatic eruptions at Kilauea
took place during two separate periods: the upper section belonging to a large eruption that was witnessed
by westerners in 1790, and the older section that was
emplaced during prehistoric times. Stone [7] and
Wentworth [8] suggested that the older section was
emplaced 300-500 years ago. However, Powers [9]
suggested that the entire formation was emplaced over
a longer period of ~1,500 years.
In the 1980’s the generally accepted interpretation
was that the Keanakako’i Formation was emplaced
entirely during the 1790 eruption [5, 10, 11]. In particular relevance to our research, Decker and Christiansen [5] note that a “careful search has revealed no
clear evidence of stream erosion, channel gravel, or
soil formation within the Keanakako’i section,” which
would have indicated that were multiple eruption episodes. Such an assertion is untenable, however, as
there is, in fact, clear evidence for at least two older
erosional surfaces [12, 13], particularly between the
SECOND WORKSHOP ON MARS VALLEY NETWORKS
upper lithic unit and the lower vitric unit. Additionally, the physical characteristics of deposits record
dramatic changes in eruption styles, and isopach maps
of the deposits show variations in how the ash was
dispersed during emplacement [12].
Including evidence from the Kulanaokuaiki tephra
[14] and the Uwekahuna ash, the current belief is that
Kilauea has experienced periodic phreatomagmatic
eruptions for at least the last 2,000-3,000 years [15]. It
is interesting to note that despite Kilauea’s reputation
for quiescent eruptions, the phreatomagmatic eruption
that took place in 1790 killed a war party of at least 80
Hawaiians, giving Kilauea the distinction as the deadliest volcano in American history [16]. There are
probably multiple reasons why Kilauea erupts explosively. Until recently, the last explosive eruption took
place in 1924 when the floor of Halemaumau quickly
withdrew to a depth of over 400 m within a period of a
few weeks [5]. In this instance an offshore eruption
caused the caldera floor to sink below the water table
[12, 13], which is located ~500 m below the surface
[17, 18]. This subsequently triggered a phreatomagmatic eruption that deposited small amounts of ash and
blocks near the summit. Kilauea suddenly began to
erupt explosively again on March 19, 2008. This current eruption is driven by a large lava lake that is situated a few tens of meters below the caldera floor in
Halemaumau. Outgassing from this lava lake is escaping through a vent in the caldera floor that periodically
collapses subsequently triggering another ash-rich explosion. To date, however, the amount of ash from this
explosive episode has been small and confined primarily to around the caldera.
Gulley Characteristics: Although the gullies that
incise the Keanakako’i tephra are several orders of
magnitude smaller than many of the martian valley
networks, it is interesting that they share many of the
same general physical characteristics. For example,
the Keanakako’i gullies often originate from a plunge
pool giving the gullies an amphitheater-shaped head
(Figure 1). Unlike the martian valley networks, however, the Keanakako’i gullies often contain a number
of these plunge pools that get progressively larger as
the contributing area increases downslope. These
plunge pools are generated by the difference in
strength between the ash and tephra layers that compose the upper lithic layer of the Keanakako’i Formation: the fine-grained, lapilli-rich ash layers are indurated and highly resistive to erosion, while the coarsergrained tephra layers are loose and friable and easily
eroded. It is this contrast in strength and erodibility
that generates these amphitheater heads and plunge
pools, not groundwater sapping as has been suggested
for martian valley networks [e.g., 19]. In fact, the wa-
12
Figure 1: Plunge pools are common characteristics of
the Keanakako’i gullies. Small pools only a few inches
deep are seen close to the gulley head (left). These pools
increase in size downslope (right) and are controlled by
the contrast in stregth between the ash layers, which are
highly resistant to erosion, and the interbedded coarsegrained tephra that is more easily eroded.
ter table is over 500 m deep in this area [17, 18], so it
is probably that these features are created entirely by
surface runoff.
Another important characteristic of the Keanakako’i gullies is that while the average channel width
increases downslope as the contributing area increases,
the local width varies a great deal (Figure 2). This
appears to be due once again to the highly friable nature of most of the tephra deposit. What is interesting
is that many valley networks often share these characteristics. For example, the valley networks seen in
Figure 2 occur on eastern rim of the Herschel impact
basin, where they have most likely incised into highly
brecciated impact ejecta.
Climatic Conditions and Discharge Rates: Because of the normally dry conditions in the Ka’u Desert and the highly permeable nature of the Keanakako’i Formation, flow in the gullies is ephemeral and
rare. As an example, on September 1, 2003 as Hurricane Jimena passed within 100 nm south of the Big
Island. The area received approximately 6 in of rain
within a few hours, but only small puddles formed on
the Keanakako’i. Until recently, the last reported flow
within the Keanakako’i gullies occurred in November
1-2, 2000 when the area received almost 30 in of rain
within a 24 hr period. However, from March 9-11,
2006, the Hawaiian Volcano Observatory measured
almost 11 total inches of rain, which was enough to
initiate surface runoff and some flow through the gullies (D. Swanson, personal communication). Our goals
were to determine the climatic conditions necessary to
generate runoff in the gullies and to determine what the
flow conditions were like in the gullies when they were
active.
SECOND WORKSHOP ON MARS VALLEY NETWORKS
13
housing development only a few miles to the west
stays relatively dry. However, the tropical weather
systems that cause the heaviest rainfall in the Ka’u area
are due primarily to stationary disturbances with strong
southerly winds that are near surface. Under this scenario, the southeastern part of Hawaii becomes windward so rainfall becomes topographically enhanced in
the elevated terrain. With the abundance of moisture
from the lower latitudes, if the atmosphere is unstable,
the condition is set for heavy rainfall that can occur
over a 48-72 hour period. As Table 1 shows, such
conditions generally occur during the winter months.
Table 1. The Ten Most Significant Rainfall Events in the
Ka’u Desert Since 1971.
Year
Month Day
Station Precipitation (in)
Hawaii NP
1979
1980
Figure 2: Aerial view of some of the gullies that have
incised the Keanakako’i Formation (top). The scale bar to
the lower left is 750 m across. Valley networks seen on
the eastern rim of the Herschell impact basin (bottom)
share some of the same characteristics. In particular,
note that the width of the valley networks in this area is
highly variable. The scale bar to the upper left is 10 km.
Climatic Conditions: Most of the information
about when the gullies were active is anecdotal. In
order to provide more quantitative information we
identified two locations where sediment transported by
the gullies superposed lava flows of known ages, including a flow that erupted from fissures near Halemaumau in September 1971 and another flow that
erupted from fissures in the Southwest Rift Zone in
December 1974. We dug trenches into fluvial deposits
at both locations and confirmed that these materials did
in fact superpose the lava flows (Figure 3). At each
location there were ten sediment couplets that alternated between very coarse and fine-grained sand, suggesting that there were ten climatic events since September 1971 that generated significant runoff with the
Keanakako’i gullies. To understand the nature of these
events we searched available climatic records for
weather stations within the Ka’u District. Table 1 lists
the dates for the ten most significant rainfall events
since 1971.
Typically the trade winds are easterlies and because
of the orographic effects generated by Kilauea most of
the moisture precipitates on the eastern slope of the
volcano. In fact during the summer months Volcano
Village receives some rain almost every night, yet the
1981
1987
1990
1990
1994
1996
2000
2001
2
3
12
12
1
11
9
3
11
11
Pahala
Pahala Mauka
20
16.75
16.96
21
5.85
1.91
17
6.02
4.67
18
10.72
25
25
4.97
4.65
26
12.95
11.9
13
12.96
8.85
14
3.2
3.19
19
2.48
1.86
20
10.64
12.95
20
9.72
5.7
21
11.26
9.65
19
12.26
6.52
20
4.74
5.31
3
8.43
9.1
4
4.08
2.69
2
missing
16.6
3
missing
11.43
27
10.52
8.35
28
11.65
12.23
Estimating Discharge: There are a variety of
methods for estimating the maximum flow velocities
within a channel. The simplest is the depth-slope formula where the bed shear stress of a flow, or the retarding stress at the base of a flow, tb, is calculated by
the equation
SECOND WORKSHOP ON MARS VALLEY NETWORKS
14
stress, t*. The dimensionless grain parameter is defined as
!" =
Figure 3. A trench dug into sediments superposing the
December 1974 flow (orange arrow) provides evidence
that there has been 10 major storm events in the Ka’u
Desert capable of generating runoff on the Keanakako’i
tephra. An analysis of state climatic records indicate that
the storms with the heaviest rainfall amounts typically
occur during the winter months when southernly winds
create prolonged stationary disturbances.
tb = rghS
(1)
where r is the density of the fluid, g is gravitational
acceleration, h is the flow depth, which is estimated
from field observations, and S is the slope of the channel, which was measured directly using DGPS equipment. The bed shear stress can be equated to the bottom stress created by a flow, t, where
t = rCf u 2
(2)
and Cf is a dimensionless drag coefficient, and u is
the mean flow velocity. Thus, the mean flow velocity
for a channel can be calculated from
! ghS $
u=#
" C f &%
1/2
(3)
The dimensionless drag coefficient can be adjusted for
gravity by the expression
! n2 $
(4)
C f = g # 1/3 &
"h %
where n is the Manning roughness coefficient (units of
s/m1/2), which has been derived empirically from terrestrial observations. Application of an appropriate
Manning roughness coefficient, n, involves a certain
degree of subjectivity, but values ranging from 0.015
to 0.035 are typically used to describe most environments free of vegetation.
An alternative method for estimating channel flow
velocities can be made directly from the sizes of the
particles contained on the channel bed. Shields [20]
derived empirical relations for the dimensionless grain
parameter, z*, and the dimensionless boundary shear
D
3
(#s $ #) g
2
% #
(5)
where D is the characteristic particle diameter, rs is the
density of the particle, ρ is the density of the fluid, g is
the acceleration of gravity, and n is the kinematic viscosity of the fluid. As a general rule, the D84 grain size
is used to describe the largest particle transported in
saltation (as opposed to bedload) during flow. Typically, there are cross-sectional variations in the transport rate of sediments [21], so in addition to grain-size
analyses of samples collected within Sand Wash, we
also conducted particle size analyses across the channel within 1 m2 bin at regularly spaced intervals
(roughly every 1-2 m depending on channel width) to
determine a statistically representative particle size for
the channel cross-section at a particular location.
From Shields' [20] curve, values for the dimensionless boundary shear stress, t*, can be determined.
For values of z* less than ~400, Shields [20] extrapolated his curve, but empirical data can be used to determine values for t* in this range [22]. The dimensionless boundary shear stress, t* , is
! cr
(6)
!* =
"s # " gD
where tcr is the critical boundary shear stress needed to
initiate sediment motion. This is assumed to also be
the bottom shear stress, tb , during flooding. (Basically, the assumption that we make is that flow through
the channel is fast enough at the bottom to initiate
sediment transport of the D84 particle.) The shear velocity, u* (expressed in cm/sec), is
!b
! cr
(7)
u* =
=
"
"
(
)
By substituting Equation 1 for tb, Equation 7 becomes
u* = ghS
(8)
Since the other parameters are known, this approach
also allows us to solve for h, the depth of flow through
the channel, so this provides us with an independent
way of checking our field observations and verify the
results from the first approach.
As Komar [23, 24] notes, it is better to analyze the
flow in terms of u* than u due to the uncertainties in
estimating reasonable values for Cf. However, values
of u are more intuitive. These can be calculated from
the relationship
1/2 !1/6
u = C f u* = g nh
u*
(9)
Of course in order to estimate values of u
through this method reasonable values of the Manning
SECOND WORKSHOP ON MARS VALLEY NETWORKS
coefficient, n, must be used. Once again, these are
0.015 to 0.035 for channels that lack vegetation.
Using both methods (Equations 1-4 and Equations 5-9) we determined that average flow velocities
within Sand Wash were on the order of ~8-12 m/s with
corresponding discharge estimates of 115-490 m3/s.
Implications For Mars: Our observations have
several implications for the formation of martian valley
networks.
(1) Many martian valley network systems have been
described as flat-floored and often the heads have an
amphitheater shape [19]. While such characteristics
typically have been ascribed to terrestrial channels that
have been formed by groundwater sapping, our observations indicate that such characteristics can also form
in layered stratigraphy that contains an upper more
erosion resistant layer capping an underlying, more
friable layer. It is important to note that the water table
is located ~500 meters below the Keanakako’i tephra
[17, 18], so it is unlikely groundwater sapping operates
with any efficacy in this material. It is possible that
the martian regolith is friable but capped with some
more erosion-resistant material such as a thick, indurated duricrust or lava flow.
(2) The highly variable gulley width is due to the general friable nature of the Keanakako’i tephra. Valley
networks that exhibit similar morphologic characteristics most likely occur in material that are also friable
so some depth, such as basin ejecta. In general the
nature of valley network width may be a good indicator of the competency of the surface material.
(3) While there is evidence to suggest that the valley
networks formed in an early arid environment [e.g.,
25] the amount of rainfall needed to initiate runoff on
the martian surface was probably enormous and the
associated storms probably lasted several days or even
weeks. Given that the size of many of the valley networks drainage basins in the highlands are 100’s of
kilometers wide, the storm cells that carved these networks were probably several orders of magnitude
larger than any terrestrial storm. In fact, storms on
early Mars may have actually been global in nature.
Potentially the precipitation that carved the highland
valley networks was related to the collapse of the primordial atmosphere and not any kind local event.
Acknowledgements: This research is supported by
a grant from the NASA Mars Fundamental Research
Program (NNX08AN64G). An award from the Smithsonian Institution’s George F. Becker Endowment
Fund supported preliminary fieldwork.
15
References: [1] Carr, M.H., (1996) Water on
Mars, Oxford Univ. Press, New York, 229 pp. [2]
Bandfield, J.L., V.E. Hamilton, and P.R. Christensen
(2000) Science, 287, 1626-1630. [3] Malin, M. C., and
K. S. Edgett, (2001) JGR, 106, 23,429-23,570. [4]
Elias, T., A.J. Sutton, J.B. Stokes, and T.J. Casadevall,
(1998) U.S. Geol. Surv. Open File Rept. 98-462, 41 pp.
[5] Decker, R.W., and R.L. Christiansen (1984) in Explosive Volcanism: Inception, Evolution, and Hazards,
pp., 122-132, Nat. Acad. Press, Washington, D.C. [6]
Powers, S., (1916) Am. J. Sci., Fourth Series, 41, 227244. [7] Stone, J.B. (1926) B.P. Bishop Museum Bull.,
33, Honolulu, 60 pp. [8] Wentworth, C.K. (1938) Hawaiian Volcano Observ. Spec. Rep. 3, Hawaii Volcano
Research Asso., Honolulu, 183 pp. [9] Powers, H.A.
(1948) Pacific Science, 2, 278-292. [10] Malin, M.C.,
D. Dzurisin, and R.P. Sharp (1983) GSAB., 94, 11481158. [11] Easton, R.M. (1987) U.S. Geol. Surv. Prof.
Pap. 1350, 243-260. [12] McPhie, J., G.P.L., Walker,
and R.L. Christiansen (1990) Bull. Volcanol., 52, 334354. [13] Mastin, L.G. (1997) JGR, 102, 2009320109. [14] Fiske, R.S., T.R. Rose, D.A. Swanson,
and J.C. McGeehin (1999) Eos, 80, 1196-1197. [15]
Dzurisin, D., J.P. Lockwood, T.J. Casadevall, and M.
Rubin (1995) Jour. Volcan. Geotherm. Res., 66, 163184. [16] Swanson, D.A., and R.L. Christiansen,
(1973) Geology, 1, 83-86. [17] Stearns, H.T., and
G.A. MacDonald (1946) Hawaii Division of Hydrography, Bulletin 9, 363 pp. [18] Zablocki, C.J., R.I.
Tilling, D.W. Peterson, R.I. Christiansen, G.V. Keller,
and J.C. Murray (1974) GRL 1, 323-326. [19] Pieri, D.
C., (1980) Science, 210, 895-897. [20] Shields, A.
(1936) U.S. Dept. Agriculture, Soil Conservation Service Coop. Lab., Calif. Inst. Tech, 26 pp. [21] Leopold,
L.B., and W.W. Emmett (1997) U.S. Geol. Surv. Prof.
Paper 1583, 52 pp. [22] White S.J. (1970) Nature,
228, 152-153. [23] Komar P.D. (1979) Icarus, 37,
156-181. [24] Komar P.D. (1980) Icarus, 42, 317-329.
[25] Irwin, R. P., III, R. A. Craddock, and A. D. Howard (2005), Interior channels in Martian valley networks: discharge and runoff production, Geology,
33(6), 489−492.