Geo-Marine Letters(1994) 14:19-28

© Springer-Verlag 1994

K. B. Lewis

The 1500-km-long Hikurangi Channel: trench-axis channel that escapes its trench, crosses a plateau, and feeds a fan drift

Received: 23 September 1993 / Revisionreceived: 11 February 1994

AbsIraet The Hikurangi Channel, east of New Zealand, is one of the earth's major, active, sediment conduits between rising mountains and ocean basin. About 1500 km long, it uniquely incorporates most variations of canyon-channel systems worldwide. An apical canyon feeds a meandering, aggradational, trench-axis channel. This diverts, 800 km from the source, across an oceanic plateau. There, an oceanic-type channel has become incised over 500 m at the plateau-edge scarp. Beyond the scarp, distributary fan channels supply sediment to the Pacific's Deep Western Boundary Current and one distributary merges into a boundary channel.

thoroughly researched and reviewed (Carter and Carter 1993). Submarine conduits may be divided for convenience into: (1) canyons, which are steep sea valleys eroded into rock: (2) fan channels, which splay and aggrade over submarine fans at the ends of most canyons; (3) oceanic channels, which meander for hundreds or even thousands of kilometers across the great ocean basins (Carter 1988); (4) trench-axis channels, which follow the axes of a few of the world's subduction trenches (Thornberg and Kulm 1987; Shimamura 1989); and (5) boundary channels, which form where abundant sediment, orten from turbidity currents, interacts with deep boundary flows (Reed et al. 1987). Of these five types, canyons and fan channels have been the subjects of extensive research, whereas trench-axis, oceanbasin, and boundary channels are less well documented Introduction and the mechanisms that form and maintain them are less Deep-sea channels are the end member of a series of trans- clearly defined. The Hikurangi Channel has developed adjacent to subport conduits that transfer sediment from mountain ranges to ocean basins (Carter 1988). They represent a major part continental subduction that becomes increasingly oblique of the planet's sediment transport system with many char- towards the south and merges into an onshore zone of acteristics of the world's largest rivers. Yet only now are trench trench transform (Fig. 1). A late Miocene shift in their courses and mechanisms of formation being defined. the pole of rotation of the Pacific Plate has dramatically This paper records, for the first time, the course, offeastern increased the rate of convergence at the plate boundary New Zealand, of the uniquely complex Hikurangi Chan- through New Zealand (Walcott 1984), with consequent nel, which dramatically changes its character and trend rapid uplift of forearc ranges in North Island and the several times along its length. Previously, the Hikurangi collisional Southern Alps in the South Island. The increasChannel was recognized only in isolated crossings of the ing uplift has resulted in a major increase in the rate of southern Hikurangi Trough (Fig. 1), an anomalously shal- sediment supply to the Hikurangi Trough during the Pliolow, sediment-filled, structural trench at the southern ex- Pleistocene. Part of this sediment is scraped off at the edge tremity of the Tonga-Kermadec-Hikurangi subduction of an imbricate-thrust continental margin (Lewis and Petsystem. First reported by Houtz et al. (1967), the channel tinga 1993), where crust, that is much thicker than norhas merited only passing comment in later studies and has mal oceanic crust is being subducted beneath the wedge; never featured in international comparisons and reviews. (Wood and Davy 1994). To the east, this 10-15-km-thick In contrast, the nearby Bounty Channel (Fig. 1) has been crust forms the Hikurangi Plateau, a triangle of elevated (2500-3500 m deep) ocean floor between North Island and Chatham Rise. The northeastern side of the triangle steps down to the 5500-m-deep Southwest Pacifie Basin at the K. B. Lewis New Zealand Oceanographic Institute, National Institute of Watet Rapuhia Scarp, which is bathed by the massive flow of the and AtmosphericResearch Ltd.(NIWA),P.O. Box 14-901,Kilbirnie, Pacific Ocean's Deep Western Boundary Current (Carter Wellington,New Zealand and McCave 1994).

20 slope, the channel is generally incised between 100 and 500 m beneath the adjacent turbidite plains, with the left bank usually higher than the right (Fig. 3B). Both the canyon and channel typically range from 3 to 9 km wide at the lip and from 0.5 to 3 km wide in the flat-floored axis (Fig. 3C). The sinuosity of the system, which is measured as the ratio between distance along the meandering channel to the shortest distance between two points in the channel axis that are roughly 20 km on either side of each data point, ranges from 1.02 for straight segments to 1.8 in meandering segments or at sharp changes of course (Fig. 3D). On the basis of each of these characteristics, as weil as on seismic character of the walls and on the tectonic setting, the canyon/channel system is divided onto five segments (Fig. 3): (1) the steep Kaikoura Canyon, (2) a trench-axis channel, (3) a deeply incised transplateau, oceanic-type channel, (4) a distal fan channel and (5) a scarp moat or boundary channel, where turbidity currents and the Pacific Ocean's Deep Western Boundary Current interact. Fig. 1 Location of the Hikurangi Channel (small medium grey arrows) and its relationship to major topographic, oceanographic, and structural features. Stippled area is continental crust loosely defined by the 2000-m isobath and diagonal shading is thickened oceaniccrust of the Hikurangi Plateau. Large open arrows show path of the Deep Western Boundary Current (DWBC) and small light arrows the path of the Bounty Channel. Heavy flaggedlines indicate the trench-transform-trench plate boundary through New Zealand

Data sets This interpretation is based largely on 3.5-kHz and 120cubic-inch airgun profiles (Fig. 2 inset), some collected for allied projects (Wood and Davy 1994, Carter and McCave 1994). Cores up to 4 m long were obtained from both the channel and the adjaeent turbidite plains (Lewis 1985; Fenner et al. 1992).

Des©ription of a unique multiform ©hannel

Kaikoura Canyon and adjacent sediment traps Kaikoura Canyon is the most northerly of seven canyons that incise the shelf edge at the apex of the Hikurangi Trough (Fig. 4). Of the seven, it is the only one to intercept the modern nearshore sediment drift and the only one not partly infilled by sediment (Herzer 1979). The eanyon head is in water only 25 m deep and within 500 m of the rocky shore platform south of the Kaikoura Peninsula (Fig. 4 inset). Its position is unrelated to nearby fluviatile supply, the nearest substantial river being 37 km to the south. The canyon may be structurally controlled, as it is adjacent to the most active part of the transform plate boundary zone and continues the trend of the (now inactive) Hundalee Fault (Fig. 4). In the first 5 km, the canyon axis descends to over 900 m below sea level with a steep slope of about 9 ° (1 in 6). Then, it winds for 60 km down the continental slope between steep, high walls [Figs. 3 and 5 (5-34 km)], at a slope of 2 ° (1 in 26), into the head of the Hikurangi Trough at 2200 m deep; the conduit from relict canyons to the south joins the trough at the same locality (Fig. 4).

Overall route--alps to ocean basin

Rivers that rise among the 3000-m-high peaks of South Island's Southern Alps (Fig. 1) carry their sediment loads for 300 km in an asymptotic curve to the coast. There, southerly storm waves, tidal currents, and geostrophic currents carry the sediment northwards along the eastern shelf (Herzer 1979). Just south of Kaikoura, this northward nearshore drift is intercepted by the Kaikoura Canyon, which forms the "headwater" of the Hikurangi Channel (Fig. 2). The combined eanyon-channel conduit has many of the characteristics of a major river system. It is about 1500 km long with an exponential longitudinal profile (Fig. 3A), its terminal base level being the deep Southwest Pacific ocean basin. Downstream from the headwater canyon, which is incised up to 1000 m into the continental

Axial channel along Hikurangi Trough The Hikurangi Trough runs for its first 100 km as a 610-km wide, N55°E-trending, U-shaped depression between the normal-faulted northwestern corner of the Chatham Rise and the imbricate-thrust continental slope north of Kaikoura (Lewis et al. 1985). Here, the trough is at the transition from oblique subduction to intracontinental coUision and shear (Fig. 1). The deepest part of the Ushaped trough, along the toe of the Chatham Rise slope is strongly acoustically reflective and marks the start of the trench axial channel [-Fig. 5 (73-148 km)]. It is 2200-2600 m below sea level with an axial slope of 0.3 ° (1 in 170). It is 15-80 m below the adjacent trough, 1-3-km wide, and

21

1700 .Ò

38

o .

o

i:~i:i-i!~!ii,::i;~~!iiiii!ii~~i:¸i¸¸

(

./ % %

iLIiiiiiiiiii!iiTorbidite P,ai°

iõ-~-177°

17,8°c-,-_,__~_1_7~9°~-~1000~1,80

Fig. 2 Course of the Hikurangi Channel with distancesin kilometers from source at Kaikoura. Shaded are associated turbidite plains and fan-drift deposits. Also shown is a Miocene(Rekohu) drift deposit of the Deep Western Boundary Current. Bathymetryat 250-m intervals. Plate boundary deformation front indicated by heavy, broken, flagged line. Inset: Track plots of seismic profiles including 3.5-kHz profiles

almost straight (Fig. 3). The channel's left bank (looking downslope) has small irregular and parabolic reflectors [-Fig. 5 (73-148 km)], some buried, that indicate slumps and debris flows from the Kaikoura slope. At 160 km from the conduit's source at Kaikoura, the Hikurangi Trough opens out into a wide, sediment-filled structural trench at a depth of about 2650 m. After its confluence with the slump-obstructed Cook Strait Canyon (Carter 1992) at 180 km from source, the Hikurangi Channel turns right and widens to a maximum floor width of 10 km [Fig. 5 (187-198 km)], while its axial slope decreases to 0.14 ° (1 in 400). Sediments beneath the flat channel floor are parallel-bedded. The channel heads eastward for 120 km, scouring the toe of the North Chatham Slope, as it moves almost 60 km away from the slope-toe deformation front (Fig. 2). It gradually narrows to about 3 km floor width and becomes increasingly incised to nearly 300 m below the higher left bank [Figs. 3B, C and 5 (260 km)].

"~'~'~

179°

~

Fan Drift

~

Miocene Drift

,____q.__

1-78°~, ~ - 7 7 °-

Between 300 and 800 km from source, the Hikurangi Channel is a fully developed, steep-sided, axial channel of a structural trench, meandering northeastwards between volcanic knolls [Figs. 2 and 5 (320-800 km)]. Airgun profiles show that the channel floor is aggradational, having migrated vertically and laterally as the structural trench has filled with over 3000 m of Late Miocene to Recent sediment at its southern end (Fig. 6 I). No significant subsidiary channels enter the main channel; canyons that incise the adjacent upper slope all dissipate in a baffle of slope basins. These basins contain only the thin silt laminae of distal turbidites, whereas the main axial channel and its leveed overbanks have graded, sandy silt turbidites, up to 0.4 m thick (Lewis 1985). In the proximal part of this sector, the Hikurangi Channel axis slopes at only 0.07 ° (1 in 800), less steeply than the trough because of the meanders. Its sinuosity is over 1.5 with most meanders having a radius of 3-8 km. The channel floor is strongly reflective, about 2 km wide and usually flat, although tocally sloping towards the steep outsides of bends. Both banks show clear development of levees rising up to 50 m above the adjacent turbidite plain with sediment waves ranging from 5 to 15 m high and from 2 to 3 km in wavelength, migrating up the backs of the levees towards the channel-edge levee crests [-Fig. 5 (320-383 km)].

22 Progressing northward, levees become less pronounced, decreasing to less than 10 m above the turbidite plain and sediment waves are rare [-Fig 5. (703-800 km)]. Axial slope reduces to 0.05 ° (1 in 900), sinuosity decreases to 1.1, channel width reduces to less than 2 km, and depth of incision decreases to less than 100 m below the left bank and only 75 m below the right bank (Fig. 3). Thus, at 800 km from the source, the cross-sectional area is reduced to less than 20% of what it was at 300 km and the Hikurangi Channel appears to be fading into the end ofa 3400-m-deep, ponded depocenter of the Hikurangi Trough (Fig. 2).

I I I

-1000 -2000 E .c -3000 r~ ~3

-4-000 -5000 -6000 1000-

,

i

,

l

,

l

,

l

,

l

,

l

,

l

,

~

,

~

,

l

,

l

,

l

,

i

,

~ , i

[ ~Le~ benk +Right bank ]

Out across the plateau: rejuvenated oceanic-type channel

I

I

80o 2 ~ %

I I

B

I

600--

ù1:

c

/+00

(D

20o 2

18-

I~B ..... +,o0 I

16

zz

c

12

A

lO

~

8

P,

6

~

4 2 o

'

'

'

~

,

i

,

~

,

i

.

i

,

J

,

t

,

i

,

i

,

i

,

i

•

i

,

i

,

,

~

J

,

i

~

p

~

i

2.01.8 5

D

1.6 CU3

1.4

I

'

0

200

400

600

800

i

~

r

,

i

1000 1200

Distance from source [km)

~" Z

O

- -

Z

I--

Z

O

a-

r.)

,

i

.

1400

r

At 800 km from the source, the Hikurangi Channel changes trend and character. It turns at right angles through a gap between two volcanic seamounts and heads eastward for 400 km across the centrat basin of the Hikurangi Plateau (Fig. 2). Instead of dying out in a trench depocenter, it becomes much more deeply incised and apparently rej uvenated, with many of the characteristics of an ocean-basin channel (Carter 1988). After turning east, the channel becomes gradually more V-shaped; the flat floor decreases to only 0.5-2 km wide, and the banks becomes higher and more steep sided [Figs. 3 and 5 (800-1076 km)], with slumped sides and terraces being common. The gradient along the channel's meandering thalweg increases significantly to 0.12 ° (1 in 500), although this is still less than the slope of the basin into which it becomes increasingly incised. The channel floor is 230 m below the left levee at 900 km from the source and 440 m below it at 1200 km from the source (Fig. 3B). Right across the plain the left levee is 20-60 m higher than the right levee and the levees are generally 10-20 m above the surrounding turbidite plain [-Fig. 5 (980-1274 km)]. There are sediment waves on the back slopes of both levees in the eastern end of the basin [-Fig. 5 (1076 km)]. Cores from the turbidite plain have 10-100-mm-thick layers of unbioturbated, turbidite mud between bioturbated hemipelagic mud (Fenner et al. 1992), while a core from the channel axis at 1086 km from source has graded coarse silt and sand layers, up to 110 mm thick, that contain up to 67% sand near their erosional bases. Airgun profiles show that the central basin contains up to 700 ms (about 700 m) of strongly bedded sediments (Fig. 6 II, III), inferred to be turbidites resulting from late Neogene increased uplift of sediment sources. The upper half of the turbidite sequence (A 1) is markedly teveed and clearly derives from the Hikurangi Channel, whereas

1600

Fig. 3 Plots against distance from source (and channel environment) ofA depth of the channel axis below sealevel,B depth that the channel axis is incised below left and right banks, C width of the channel at its top and bottom, and D sinuosity of the ehannel axis. See Fig. 2 for location of kilometer (from source) posts. Note change in all parameters where channel leaves structural trench axis at 800 km from source

23 1692 km,

[° k=

~>
£

. ~eoinsula

m

74 o

Hikurangi ~

Fan channel

~ ~

\ L~ «

Fan 1 3 8 2 k m ~ ~ ~ ......

~

1274 km ~ :'~~"=:::"

Canyon through Rapuhia Scarp

"

.t

/

/

!1/

I

!

¢..

,

t

i'

1038 km Y'~---.~-~~~.~~ i L ~

/''/

~#) #

/

~-~==~:_~_~~

\'t, ~

\ i

/'/

980 km

:.~--~-~»

\~

~t /

920 km =~;~~.~.~~~,-ù

, :/?«.=~=.~.,5~",~~=~-~,

O~,ai"s n'ó " " " !

~~ ~~~~~;~'"

786 km

i

.

.

,t~«~

.

8o7~=

524 km

,o.~,~?.:::.

,

...... :o: -:~ "

",~

,

«

458 km ~"~ ~ ù , ~ ' 383 km .: :, : ' ~k /

320km

0

198

km "

~'i~

~

, " ~

:

» :

"~

,

.; :

~

73 km k

.. ~,

Kaikoura

~ )"m- _~ ~ -"~

NW

. «'

Slope IO0

200

34

300 ~400 ~

22 km I

500 ~

~"

8oo ~ 700 ~

)

800

r

1000 0

1 2

"

~..~ ~ _.~._. !

148 km

.0

/ .....

g

=.-.~~~» i., .}....... i ........ ~

2o0ù=

Cook Strait Canyon

:

~

900

Fig. 5. Tracings of3.5-kHz profiles across the Kaikoura Canyon and Hikurangi Channel in order of distance from the source (kilometer notations at left of each profile; for locations see Fig. 2). Note consistently higher lefl-bank levee and sediment waves (asterisks)

Seamount

. . . . . . . »: ~,~. . . . . Sea. . . . t ,~ßl!

187 km

the lower half (A 2) consists of near-horizontal, parallelbedded, sheet turbidites into which the channel is incised. In places, the channel cuts right through the parallel-bedded unit into an underlying transparent unit (B). In none of the profiles is there any indication of channel aggradation as observed in profiles across the Hikurangi Trough. The turbidites onlap an underlying more transparent sequence (B) that crops out as it drapes the plateau to north and south of the central basin (Fig. 6 II, III); dredge sam-

.~ ,-I 4

703

Banks Peninsula

Fig. 4 Feeder canyons of the Hikurangi Channel, only one of which, the Kaikoura Canyon, intercepts the modern sediment supply. The others intercepted the longshore supply during stages of lowered sea tevel and are now either partly sediment-filled or blocked by slumps (breaks in dotted lines). Bathymetric contours at 250-m intervals. Inset: detail of the Kaikoura Canyon head incising the nearshore zone south of the Kaikoura Peninsula. Contours at 100-m intervals

.

3 4

5

K i l o m e t r e s (horizontal) Vertioa[ o x a g g e r a t i o n

× 1(

Chatham Slope

0

24 2~

4S

5S

6S

5S"

6S"

5S"

6S

6S-

0 I

20km |

25 ples of drape from a knoll in the Hikurangi Trough are pelagic ooze of middle Miocene age (Lewis 1985). D S D P site 594 on the southern flank of the Chatham Rise demonstrated that a regional unconformity (Y) between a top parallel-bedded, onlapping sequence and underlying transparent drape is latest Miocene in age (Lewis et al. 1985). The Miocene drape is considerably thickened across the eastern end of the central basin, where it is inferred to represent an early drift deposit of the Deep Western Boundary Current, the Rekohu Drift, that formed after a significant boost in the southern abyssal circulation about 8.4 Ma (Carter and McCave 1994). At most places on the plateau, the drape and drift overlie an inferred late Cretaceous and Palaeogene sequence (C) that is bedded at the top and seismically transparent where it overlies opaque, pre-late Cretaceous basement (Z). At 1200 km from the source, the channel again changes trend and heads N N E to the northern end of the Rekohu Drift. It continues to become more and more canyon-like to 1274 km from the source [Fig. 5 (1274 km)], where it is incised 530 m below the 4050-m-deep left bank. Even here, where the cross-sectional area is over five times greater than at the 800 km mark, there are clearly developed levees on each bank. From 1290 to 1390 km from the source, the channel cuts through the Rapuhia Scarp at the northern end of the Rekohu Drift. As it does so, the axial gradient stays constant (Fig. 3A) so that the depth to which the channel is incised decreases rapidly (Fig. 3B). At some places, the widening channel has cut down to the strong, irregular, hyperbolic reflectors that indicate volcanic outcrop [Fig. 5 (1382 km)].

Hikurangi Fan channels At 1390 km from the source, the Hikurangi Channel reaches the toe of the Rapuhia Scarp at a depth of 4700 m. There it debouches onto the apex of a distal fan. This cornplex sediment body extends at least 120 km out onto the Southwest Pacific Basin, but it is strongly elongated towards the north by the powerful Deep Western Boundary Current; it is therefore termed the Hikurangi Fan drift (Carter and McCave 1994). Its strongly bedded sediments (A) are at least 550 m thick and overlie more

Fig. 6. Airgun profiles across the Hikurangi Channel (located in inset 3D diagram) showing I axial channel in the southern Hikurangi Trough at 250 km from source, II plateau channel in the middle of the central basin of the Hikurangi Plateau at 1076 km from the source, III plateau channel at the outer edge of the plateau at 1196km from source, and IV the main fan channel of the Hikurangi fan drift at 1411 km from source. A is interpreted as latest Miocene to Recent with A a being younger leveed channel deposits and A z being older ponded sheet turbidites of the Hikurangi Plateau. B is inferred transparent Miocene drape. C may be a condensed late Cretaceous and Palaeogene sequence marked by a strong reflector at its top. Z is probably Mesozoic basement. Profile I is Mobil 72-119. Profiles II and III are modified from Wood and Davy (1994). Profile 1V is rnodified from Carter and MeCave (1994)

transparent pelagic sediments (B) of presumed Miocene age (Fig. 6 IV). A core from the southern edge of the fan penetrated muds with silt turbidites that contain biotite from the Hikurangi Channel and muscovite and subantarctic diatoms from the Deep Western Boundary Current (Carter and Mitchell 1987). At the fan apex, the Hikurangi Channel divides into two arms with quite different characteristics. One arm heads eastward out from the scarp and across the fan (Fig. 2). It is a broad, flat-floored fan channel, about 3 km wide and incised 110-125 m beneath a sediment-waved left-bank levee [Fig. 5 (1411 km)]. The lower right bank has no sediment waves. Airgun profiles show strong reflectors that suggest buried, older courses of a laterally migrating, aggradational channel (Fig. 6 IV). The surface channel curves southeastward and, about 70 km from the scarp, it branches into several minor distributaries that merge into the 5000-m-deep edge of the Southwest Pacific Basin about 1500 km from the channel's source at Kaikoura.

Change to boundary channel The other channel on the fan is quite different. Closely spaced tracks show that it diverges from the fan apex by bending 130 ° to the left and heading northwestward along the foot of the 1000-m-high Rapuhia Scarp (Fig. 2) For the first 100 km or so, its floor is flat with parallel bedding, and it has a 15-180-m high bank levee with overbank sediment waves [Fig. 5 (1401-1446 km)]. A core from the channel is predominantly silty turbidites with datable ash horizons that indicate deposition of 3 m of sediment including over 30 turbidites in the last 10,000 years. Thus, this distributary channel has the characteristics of a turbidite, slope-toe, fan channel (Nelson et al. 1970). However, more than 100 km from the fan apex, the channel floor becomes concave without parallel reflectors, and the right bank is rounded and without the strongly waved or parallel reflectors typical of turbidite overbank deposits. The channel and mound resemble the moat or boundary channel that forms where sediment-rich, deep geostrophic currents impinge on a steep submarine slope (Reed et al. 1987). Thus, the Hikurangi turbidite channel merges into a boundary channel of the Deep Western Boundary Current. It can be traced to near the northern end of the Rapuhia Scarp, where it widens out onto the 6000-m-deep seaward flank of the Kermadec Trench.

Discussion Multiple form of Hikurangi Channel The Hikurangi Channel is unique in that it includes most of the features of the world's other submarine channel systems. There is a sediment-trapping canyon, a meandering and leveed trench-axis channel, an incised ocean-

26 basin-type channel, a distal fan channel, and a boundary channel. It is the major part of a conduit system that is actively transferring sediment from rising mountains to a deep ocean basin.

and if the volume entrained exceeds that deposited, then the flow strengthens, a process referred to as ignitive autosuspension (Parker et al. 1986). Such net loss and gain may partly explain the changing cross-sectional profile of the Hikurangi Channel.

Sediment supply and entrainment The Kaikoura Canyon intercepts a shore-parallel, sediment supply (Carter et al. 1982) of about 25 million tonnes per annum (Gibb and Adams 1982), the effectiveness of the canyon head as a sediment trap being enhanced by the groyne effect of the Kaikoura Peninsula. Metastable sediment prograding into the steep canyon head probably fails during earthquake motion associated with unlocking of the plate boundary. Holocene activity in the canyon is proved by graded gravel-sand units up to 0.64 m thick that have a structure, texture, mineral, and faunal composition indicating an origin within the surf zone and emplacement by dense turbidity currents once every 1600 years since the mid-Holocene (Carter et al. 1982). Because the record of any intervening small events would have been removed from the canyon by the gravel-laden flows, the frequency of events may be better interpreted in the axial channel to the north, where nine silt layers overlie a conspicuous ash horizon inferred to represent a 3400-year-old event that is widespread in this area (Lewis 1985). If so, then it indicates that a turbidite was deposited, on average, once every 300-400 years. A very similar Holocene frequency was evident in the slope-toe, fan-boundary channel over 1400 km from source. Although sediment input is significant today, up to 90~o of the sediment supply from some rivers draining the Southern Alps is now trapped by glacial lakes (Carter and Carter 1990). Thus, an average rate of sediment supply during glacial ages has been estimated to be two to three times higher than at present (Gibb and Adams 1982, Fenner et al. 1992). During periods of glacially lowered sea level, sediment migrating from the south around Banks Peninsula was trapped by one or more of the six other canyons south of the Kaikoura Canyon. With rising sea level, the southern canyons were each, in turn, the main sediment trap for the Hikurangi Channel (Herzer 1979). To the north, the Cook Strait Canyon was also a major supplier of sediment when its head intercepted sediment-laden tidal flows during periods of lowered sea level but major parts of this system are now blocked (Carter 1992). The extreme width of the Hikurangi Channel after confluence with the Cook Strait Canyon may indicate braiding of large flows during low sea level periods of high bedload input (Belderson et al. 1984). Further away from the source, the classical development of a meandering channel, with levees and overbank mud waves, characterizes an increasing proportion of suspended load (Damuth et al. 1988). The data imply that channelized turbidity currents travel enormous distances over very low slopes. Theoretical models suggest that this can occur so long as the volume of sediment entrained by the head of the turbidity current equals that deposited, a process known as autosuspension,

Fluviatile analogs and lefl-bank levees Like some other channel systems, the Hikurangi trenchaxis channel has many features in common with fluviatile channels, including meanders, lateral accretion, levees, overbank deposits and net aggradation. Meandering is a function of valley gradient, grain size and nature of flow (Damuth et al. 1988; Maill 1989), the channel increasing its distance of travel and reducing its gradient by meandering t o maintain an asymptotic along-axis profile. Although meandering has been inferred to imply a relatively continuous turbid flow for long periods of time (Damuth et al. 1988), this may occur only during glacial periods of increased sediment supply. The channel levees have fluviatile analogs but the consistent preferential development of levees on one bank is not evident on land. High left-bank levees in the southern hemisphere are inferred to be the result of Coriolis Force on the elevated head of the turbidity current (Carter and Carter 1988), although on the distal fan drift it may be partly a response to the boundary flow preferentially sweeping channel overflow to the north (Carter and McCave 1994). The conspicuous mud waves on some levees are lee waves, which, despite a height-to-wavelength ratio of less than 1 : 100, disrupt the overbank flow sufficiently for sediment to be deposited preferentially on the upstream side (Flood 1988). Thus, the waves appear to migrate upcurrent and upslope. In general, channels reduce gradually in height and width as the density, volume, and velocity of the turbidity currents that maintain them gradually decrease away from the source (Carter 1988). The axial channels of the Nankai Trough (Shimamura 1989) and Chile Trench (Thornberg and Kulm 1987) fade into sheet flows in trench-axis ponded basins. The axial channel segment of the Hikurangi Channel conforms to this model, but the transplateau segment does not, turbid flows overtopping the distal part despite a more than fivefold increases in depth of incision and cross-sectional area. It has been argued that turbidity currents increase in height as they slow down, perhaps with progressive dampening of turbulence, but interaction with deep currents may also be a factor in causing overflow from the incised distal ends of some channels (Carter 1988).

Cause of breakout from the trench When and why did the Hikurangi Channel first leave its structural trench and cross an oceanic plateau? Certainly, the transplateau segment of the Hikurangi Channel does not have the long history of channel and levee aggrada-

27 tion that has been reported in the similarly coast-normal Bounty Channel (Carter and Carter 1993) and in the proximal Hikurangi Channel (Fig. 6 I). Why should it take this route rather than the obvious one to the Kermadec Trench? The answer may be that the trench was blocked by the large subducting seamounts or groups of seamounts that have left massive indentations in the continental margin (Fig. 2), one such indentation occurring close to where the channel leaves the trench (Lewis and Pettinga 1993) and one about 100 km further north at the end of the sedimentflooded Hikurangi Trough (J.-Y. Collot et al. personal communication 1993). If so, then rauch of the direct evidence of channel rerouting is now lost beneath the overriding margin, but there is indirect evidence in the turbidite plains of the Central Basin. The trench must have been blocked when ponded sheet flows began to onlap the pelagic drape over the vast area of the Central Basin in latest Miocene or early Pliocene times (A 2 in Fig. 6 II and III). With the trench channel ending in a large ponded basin, why did a leveed channel incise the sheet turbidites? The answer may be that the ponded basin was overtopped or breached and the new segment began to cut down towards a base level in the Southwest Pacific Basin. Initially, the Miocene Rekohu Drift dammed the outer edge of the central basin (Fig. 2) but was overtopped at its northern end, perhaps in late Pliocene or early Pleistocene times. Thus, the plateau channel may have begun near its distal end and cut back westwards across the ponded basin, forming a nick point and change in trend where it linked with the axial channel. Down-cutting supplied extra sediment to feed turbidity currents and to build the levees and the distal fan. Turbidite to contourite At the distal fan, the fan channel perpendicular to the scarp is inferred to be the route followed by the concentrated head of each large turbidity current, whereas the scarp-toe boundary channel is inferred to be the route of diffuse turbidity currents and the tails of concentrated ones that are swept northward by the Deep Western Boundary Current. The sediment-laden boundary current produces a moat, where it is concentrated against the scarp on the left, and a drift mound, where it flows out to the right. Thus, the Hikurangi Channel is the sediment source, not only for the distal fan and deep ocean basin, but for the boundary current drift deposits, with a continuum between turbidites and contourites (Thornberg and Kulm 1987). The channel is not a major source of sediment to the Kermadec Trench. Acknowledgments Seismicdata were supplemented by profiles collected by Ray Wood and Bryan Davy of the National Institute of Geological and Nuclear Sciences (IGNS) and by Lionel Carter of National Institute of Water and Atmospheric Research (NIWA). I appreciate their generous permission to use their data. I thank Ray Wood and Rick Herzer (IGNS), Lionel Carter and Dick Pickrill (NIWA), Phil Weaver (Institute of Oceanographic Sciences, UK), Professor Bob Carter (James Cook University, Australia), Professor Roger Slatt (Colorado School of Mines), Professor Nick McCave (University of Cambridge, UK) and Arnold Bouma (editor of GeoMarine Letters) for their constructive criticisms of the manuscript.

References Belderson RH, Kenyon NH, Stride AH, and Pelton CD (!984) A ùbraided" distributary system on the Orinoco deep-sea fan. Marine Geology 56:195-206 Carter L (1992) Acousticalcharacterisation of seafloor sediments and its relationship to active sedimentary processes in Cook Strait, New Zealand. New Zealand Journal of Geology and Geophysics 35:289-300 Carter L and Carter RM (1988) Late Quaternary development of left-bank-dominant levees in the Bounty Trough, New Zealand. Marine Geology 78:185 197 Carter L and Carter RM (1990) Lacustrine sediment traps and their effecton continental sedimentation--South Island, New Zealand. Geo-Marine Letters 10:93-100 Carter L and Carter RM (1993) Sedimentary evolution of the Bounty Trough; a Cretaceous rift basin, southwest Pacific Ocean In: Ballance PF (Ed.), Basins of the Southwest Pacific. Sedimentary Basins of the World 3:51-67 Carter L and McCave IN (1994) Structure of sediment drifts approaching an active plate margin under the SW Pacific Deep Western Boundary Current. Paleoceanography, in press Carter L and Mitchell JS (1987) Late quaternary sediment pathways through the deep ocean. Paleoceanography 2:409-422 Carter L, Carter RM, and Griggs GB (1982) Sedimentation in the Conway Trough, a deep near-shore marine basin at the junction of the Alpine transform and Hikurangi subduction plate boundary. Sedimentology 29 :475-497 Carter RM (1988) The nature and evolution of deep-sea channel systems. Basin Research 1:41-54 Damuth JE, Flood RD, Kowsmann RO, Belderson RH, and Gorini MA (1988) Anatomy and growth pattern of Amazon deep-sea fan as revealed by long-range side-scan sonar (GLORIA) and highresolution seismic studies. American Association Petroleum Geologists Bulletin 72:885 911 Fenner J, Carter L, and Stewart R (1992) Late Quaternary paleoclimate and paleoceanographic change over northern Chatham Rise, New Zealand. Marine Geology 108:383-404 Flood RD (1988) A lee wave model for deep-sea mudwave activity. Deep-Sea Research 35:973-983 Gibb JG and Adams J (1982) A sediment budget for the east coast between Oamaru and Banks Peninsula, South Island, New Zealand. New Zealand Journal Geology and Geophysics 25 : 335-352 Herzer RH (1979) Submarine slides and submarine canyons on the continental slope oft Canterbury, New Zealand. New Zealand Journal Geology and Geophysics 22:391-406 Houtz RE, Ewing J, Ewing M, and Lonardi AG (1967) Seismic reflection profiles of the New Zealand Plateau. Journal of Geophysical Research 72:4713-4739 Lewis KB (1985) New seismic profiles, cores and dated rocks from the Hikurangi Margin, New Zealand. NZOI Oceanographic Field Report 22:37 49 Lewis KB and Pettinga JR (1993) The emerging, imbricate frontal wedge of the Hikurangi Margin. In: Ballance PF (Ed.), Basins of the Southwest Pacific. Sedimentary Basins of the World 3:22525O Lewis KB, Bennett DJ, Herzer RH, and Von der Borch CC (1985) Seismic stratigraphy and structure adjäcent to an evolving plate boundary, western Chatham Rise, New Zealand. In: Kennett JP, Von der Borch CC, et al. (Eds.), Initial Reports Deep Sea Drilling Project 90:1325-1337 Maill AD (1989) Architectural elements and bounding surfaces in channelised clastic deposits: notes on comparisons between fluvial and turbidite systems. In: Taira A and Masuda F (Eds.), Sedimentary Facies in the Active Plate Margin. Tokyo: Terra Scientific pp 3-15 Nelson CH, Carlson PR, Bryne JV, and Alpha TR (1970) Development of the Astoria Canyon-Fan physiography and comparison with similar systems. Marine Geology 8:259-291 Parker G, Fukushima Y, and Pantin HM (1986) Self-accelerating turbidity currents. Journal of Fluid Mechanics 171:145-181 Reed DL, Meyer AW, Silver EA, and Prasetyo H (1987) Contourite

28 sedimentation in an intraoceanic forearc system: Eastern Sunda Arc, Indonesia. Marine Geology 76:223 241 Shimamura K (1989) Topography and sedimentary facies of the Nankai Deep sea Channel. In: Taira A and Masuda F (Eds.), Sedimentary Facies in the Active Plate Margin. Tokyo: Terra Scientific pp 529-556 Thornberg TM and Kulm LD (1987) Sedimentation in the Chile

Trench: Depositional morphologies, lithologies, and stratigraphy. Geological Society America Bulletin 98:33-52 Walcott RI (1984) Reconstructions of the New Zealand region for the Neogene. Palaeogeography, Palaeoclimatology, and Palaeoecology 46:217-231 Wood RA and Davy BW (1994) The Hikurangi Plateau. Marine Geology, in press