North And Warwick 2007

Journal of Sedimentary Research, 2007, v. 77, 693–701 Research Article DOI: 10.2110/jsr.2007.072

FLUVIAL FANS: MYTHS, MISCONCEPTIONS, AND THE END OF THE TERMINAL-FAN MODEL COLIN P. NORTH

AND

GAIL L. WARWICK

Department of Geology & Petroleum Geology, University of Aberdeen, Aberdeen AB24 3UE, Scotland, U.K. e-mail: [email protected]

ABSTRACT: We propose that the so-called ‘‘terminal fan’’ facies model should be abandoned since it is flawed on several counts and it is leading to misunderstanding and poor communication. Rivers in drylands may experience excessive downstream discharge reduction such that they terminate subaerially rather than reach the sea or a lake. The facies model predicts that the distal reaches of such rivers form a network of bifurcating distributary channels producing a fan-shaped sediment body, with downstream thinning and fining of sedimentary units, ending in sand-filled ribbons encased in mud. Extensive review of modern rivers has failed to turn up convincing examples that fit the model. Rivers in drylands do not ubiquitously end in fans. Fan-shaped fluvial bodies are common wherever rivers are released from confinement and the discharge conditions promote frequent avulsion. Channels on such fans generally do not repeatedly bifurcate downstream. Where they are seen to do so, it can usually be shown they are lacustrine deltas inherited from wetter times. The term ‘‘distributary’’ is being used carelessly and is conveying incorrect understanding of sediment geometry and architecture. The proposed synonym of ‘‘fluvial distributary systems’’ is unsatisfactory as it perpetuates the same misunderstandings. Reliance on planform alone in analogue selection is highly risky. The fluvial fan is a composite sediment body resulting from frequent nodal avulsions in a setting without horizontal constraints. Channels on fans range in planform as much as any other river. The resultant sedimentary record differs little from that expected from non-fan fluvial systems except having a regionally radiating orientation when viewed over geological time scales. Contrary to the implications of the facies model, there is no distinctive ‘‘terminal fan’’ sedimentary succession.

INTRODUCTION

Friend (1978) brought the concept of a ‘‘terminal fan’’ to the widespread attention of the sedimentology community when he invoked it to explain features of some ancient fluvial successions that he felt made them distinctive from many modern river systems, as understood at that time. Each of his four examples displays a regional trend of downstream decreasing sandstone grain size, decreasing thickness of sandstone bodies (interpreted as decreasing river depth), and increasing proportions of siltstone, small-scale cross-stratification, and flat bedding. Explicitly introducing a terminal-fan model (TFM), Friend (1978, p. 539) compared these ancient examples to the ‘‘terminal fans’’ described by Mukerji (1976) on the Indo-Gangetic Plains, notably that of the Markanda River. Each of these modern rivers appears today to end in a network of distributary channels without reaching the sea or a lake. Both authors speculated this morphology is the result of discharge losses due to a combination of a drier climate at the terminus than in the headwaters and a permeable substrate. Subsequent workers have latched onto the TFM as a feature to be expected in arid and semiarid areas (collectively referred to as drylands), because of the present-day semiarid climatic setting of the Markanda River, and the inclusion of the supposedly dryland Oligo-Miocene succession of the Ebro Basin, Spain, as one of Friend’s examples. Parkash et al. (1983) produced some lithofacies descriptions for the Markanda River example by sampling the topmost sediment over the terminal few kilometers. From a study of Middle Devonian strata exposed along the northern coast of Devon, SW Britain, Tunbridge (1984, p. 713) thought

Copyright E 2007, SEPM (Society for Sedimentary Geology)

that the ‘‘decay down-slope into muddy floodbasins’’ he observed matched well to the TFM of Friend. This has to have been more by way of speculation than firm conclusion since he demonstrated neither a distributary network of channels nor a radial pattern of paleocurrents. His conjectures were based on interpreted indicators of ephemerality in the depositing flows, implying dryland climate, and some similarity of lithofacies in vertical profile to those described from the Markanda River by Parkash et al. (1983). But as a result of these speculations, the lithofacies packages Tunbridge described have implicitly become incorporated into the model as representative, and Tunbridge (1984) is typically cited along with Kelly and Olsen (1993) by those invoking the TFM for their own studies. The model was given added impetus by the review paper by Kelly and Olsen (1993), in which they evolved Friend’s hypothetical morphological model more fully into one that also predicts lithofacies character (Fig. 1). This they did through drawing at length on Devonian-age examples, including those of Tunbridge (1984), by reference again to the Markanda River example and the lithofacies described by Parkash et al. (1983), and by inclusion of another supposed modern analogue, the Gash River in Sudan as described by Abdullatif (1989). In this latest form of the model (Fig. 1), ‘‘terminal fans’’ are predicted to occur where rivers have no outlet and lose their entire discharge due to percolation and evapotranspiration (implying dryland conditions). Explicit in the model is a network of coeval distributary channels that become more widely distributed but simpler in form (straighter), reflecting down-fan bifurcation. Kelly and Olsen (1993) amplified the

1527-1404/07/077-693/$03.00

694

JSR

C.P. NORTH AND G.L. WARWICK

environments, resulting in unsafe paleogeographic and sediment-architecture reconstructions. Our analysis shows previous workers have failed to understand fully the causes of fluvial fans and the processes acting on them. Rock-record interpretations have been put forward with a certainty that fails to take account of the natural variability and ambiguity of fluvial systems. These interpretations are model-driven instead of evidence-led. We conclude the analysis by outlining what we believe to be the true picture for fluvial fan processes and their sedimentary record. RELEVANCE OF MODERN ANALOGUES

Two modern examples are usually cited to support the terminal fan model (TFM) hypothesis, yet neither provides a strong foundation. The first of these, the terminus of the Markanda River in northern India, was first invoked by Friend in his 1978 paper, picking up on the first description of it by Mukerji (1975; 1976). Additional sedimentological descriptions were provided in 1983 by Parkash and others. The second analogue is the apparent terminus of the Gash River, in Sudan. An account of this region was published by Abdullatif (1989), who made brief comparison to the Markanda River example. This was then picked up by Kelly and Olsen in their 1993 synthesis paper. Markanda River

FIG. 1.— The facies model for terminal fans as proposed by Kelly and Olsen (1993, modified from their figure 23), but which we recommend should be abandoned. 1 5 feeder zone; 2 to 4 5 distributary zone; 5 5 basinal zone.

linkage to drylands by postulating the interbedding distally of windreworked sediments with the alluvium. They included this aspect because in their own Devonian-age examples (e.g., Sadler and Kelly 1993) they had observed eolian deposits sandwiched between alluvial strata. Inclusion of the Kelly and Olsen (1993) form of the model in reviews of fluvial depositional environments and widely consulted training texts (e.g., Miall 1996, p. 249; Collinson 1996, p. 82) has given the model a stature and momentum that implies widespread acceptance. It is our experience that interpretations of ‘‘terminal fans’’ are increasingly being invoked, both formally in publications and informally by those in industry interpreting the subsurface. It has been suggested (Nichols 1987; Nichols and Hirst 1998; Nichols and Fisher 2007) that these ‘‘fans’’ should be referred to as ‘‘fluvial distributary systems,’’ though that term would additionally embrace non-terminal alluvial and fluvial fans. The aim of this article is to justify our contention that the terminal fan model should be abandoned. The TFM was a hypothetical model put together by sedimentologists to explain features observed in poorly correlated rock outcrops. Though intuitively appealing, the evidence is circumstantial. It is telling that geomorphologists do not recognize such a landform. The model does not stand up to hard scrutiny. Our view is based on critical reevaluation of the literature, and an extensive search for additional modern analogues, using satellite imagery, maps, literature, fieldwork in the western USA, and involving detailed review of 80 fluvial fan landforms (Warwick 2006). We believe Friend’s interesting speculations have been stretched by others unreasonably far, ignoring even Friend’s own explanations. The model has become like an urban myth, taking on a life of its own. We will illustrate how it is leading to poor communication and over-interpretation of past continental

The Markanda River originates in the outer Himalayan ranges (the Siwalik Ranges), and flows 112 km southwest to terminate subaerially just past the village of Jalbehpa, in the Kurukshetra district of Haryana State, India. About 10 km upstream from Jalbehpa the river is about 100 m wide, but it progressively narrows after this point. Flow is highly variable over the year, with peak rainfall in the summer months, and for much of the year the river channels are dry. Whilst the Markanda River terminus is currently in a semiarid setting (Parkash et al. 1983), its small ‘‘terminal fan’’ is not a primary phenomenon of such climates but is probably a relic of a wetter time, such as the mid-Holocene wet monsoonal period, when the Markanda was a tributary of a larger river, the Saraswati (Mukerji 1976, p. 202). It can hardly, therefore, be classed as a feature typical of dryland regions, as is implicit in the synthesis of the TFM by Kelly and Olsen. Furthermore, human abstraction of water, cultivation, and channel modification are clearly evident (Parkash et al. 1983, p. 338), so it is difficult to deconvolve human from natural processes when considering what produced the present planform. The large degree of impact caused by all of this disturbance is reflected in the absence of any fan feature on modern satellite imagery. Since Mukerji relied heavily on old maps and aerial photographs, which lack reliable temporal relationship information, there is considerable uncertainty about the claimed distributary nature of the channels (are they truly bifurcating and simultaneously active?). It is also important to keep in mind the fairly small scale of the Markanda ‘‘fan,’’ which is at the most 10.5 km from apex to toe (Mukerji 1976, p. 197). In addition to these local uncertainties it must also be remembered that the Indo-Gangetic plains have undergone a complex history of incision and aggradation due to climate change (Gibling et al. 2005), compounded by multifarious tectonic influences. The full implications of these are still not fully understood. Therefore, the current ‘‘stranded’’ Markanda River terminus is an unreliable analogue for interpreting the rock record, especially when claimed to form a basis for interpretations of fans ten times as large as this example. Gash River The ephemeral Gash River rises in southern Eritrea, near Asmara. After flowing southward, it turns west and forms the border between

JSR

END OF THE TERMINAL-FAN MODEL

695

Eritrea and Ethiopia along its middle course. It then continues into northeastern Sudan, turning northwards to pass through the town of Kassala. In most years, flow along the Gash River terminates in the region immediately north of Kassala (Fig. 2A), where natural and human-induced water losses become excessive. Rare major floods may pass beyond this region to reach and join with the Atbara River, itself an ephemeral tributary of the Nile. The Gash River has a total course of about 480 km. At Kassala it is about 100 m wide, but varies locally up to 500 m wide. Flow is usually restricted to the summer months of July to September, but even then discharge is highly variable. The Kassala region is in an arid climate, and evaporation losses from the river are high, though it is still a significant source for irrigation around the town. The Gash River terminus north of Kassala in Sudan (Fig. 2A) was described by Abdullatif (1989). As it passes through Kassala, the river is a typical sandy braided river. It is in this region, within a few hundred meters of the town, and clearly still within the apex of the ‘‘fan,’’ that Abdullatif described the river sediments, from a series of sections parallel to and across the flow. It is important to note that the lithofacies he observed are typical of the channel fill and migrating bars for many sandy braided rivers, and are in no way special. The detailed logs of Abdullatif (1989) actually tell us little to nothing about sedimentation on the ‘‘fan’’ itself, because of their location and the small size of the sampled area. It is perhaps surprising, therefore, that so much is made of this example by Kelly and Olsen (1993). They seem to be relying mostly on the regional maps and photos, and the superficial descriptions of the channel network in the introduction to the paper. But such superficial analysis is, we contend, likely to mislead. Though about 30 times larger in area, the Gash River ‘‘fan’’ suffers from limitations similar to those of the terminus of the Markanda River. A previous permanent connection to the Atbara River, which is nearby to the west, has long been seen as likely (Whiteman 1971). Records show that, in exceptional floods, discharge from the Gash still flows into the Atbara. There is strong evidence, therefore, that this feature is far from being terminal. Also in doubt is the apparent ‘‘fan’’ morphology and distributary character. Striking in planform, and easily seen in satellite and aerial imagery, is the trellis pattern of the channels (Fig. 2): almost straight segments of channel diverge westwards at just less than right angles (i.e., across the slope); shorter channel segments then flow back down the slope, at right angles to these first branches. This sort of trellis pattern is a logical irrigation morphology employed by many ancient civilizations struggling with marginal water supplies (e.g., Waters and Ravesloot 2001; Baeteman et al. 2004). Many examples can be seen across North Africa and the Middle East. All that is needed is to dig a shallow channel away from the river, with just a slight slope on it, and then at regular intervals along this channel to breach the downslope bank to allow the water to spread out and irrigate a large area of crops. Close-up aerial views (Fig. 2B) show the intimate relationship between the channels and the agricultural landscaping by humans. Furthermore, along the east side of the ‘‘fan’’ the Gash River appears to be a singlethread meandering system rather than having the braided character seen

r FIG. 2.—A) Landsat 7 satellite image of the Gash River ‘‘fan’’ to the north of the town of Kassala in Sudan in 2000, processed such that vegetation is bright green. B) Closer view of the central part of the ‘‘fan,’’ in which can be seen the strongly linear and trellis-like pattern of the channels and evidence of agricultural interference (e.g., the systems of fields bounded by channels near the top of the image). Clearly visible on the right side of the view is a primary river flowing from bottom to top—about one-quarter up the view, the river changes from braided to single-thread and meandering, which runs completely against the TFM predictions. Imagery from NASA Applied Sciences Directorate.

696

C.P. NORTH AND G.L. WARWICK

upstream towards Kassala. The possibility cannot be ruled out that there was a small fluvial fan already here, and that humans merely elaborated on the natural morphology. But with the clearly evident extensive amount of human interference, going back probably over centuries, and the mismatching planform of the main channel, combined with the likely non-terminal origin, this is hardly a reliable analogue with which to construct a generic facies model. Reliance on planform alone in analogue selection is fraught with danger. DISTRIBUTARY NETWORK

Received wisdom has it that a branching distributary pattern of channels is the norm on all fluvial fans (e.g., Leier et al. 2005). Indeed, in a well-regarded training text, Miall (1992) deliberately added this as an essential character to his definition of alluvial fans, consciously tightening up his own definitions from previous publications. The point is repeated in his much-referenced textbook on fluvial systems (Miall 1996), so further proliferating the idea. That this is still a prevalent concept is shown by Moscariello explicitly incorporating this feature into his definition of fluvial fans (Moscariello 2005, p. 599 and his figure 1), though intriguingly, and in contradiction to Miall, he chooses to omit this from his definition of alluvial fans. We believe this unqualified use of the term distributary is an unacceptably loose description of the channel network that derives from superficial planform appearance, such as from aerial or satellite imagery, or uninformed map making. It incorrectly represents causative processes, fails to take proper account of the amount of time over which the fan has been forming, and is not borne out by closer observation. On a topographic map, fan channels commonly do have a radiating form, but maps can be deceptive. Cartographers were not trained to consider age relationships, yet many studies, including those of Mukerji instrumental in the early formulation of the terminal-fan model, rely heavily on old maps. Geomorphology and Earth Science dictionaries (e.g., Allaby and Allaby 1999) are consistent in defining distributary as meaning branches from a trunk stream that distribute the water and sediment load of the main channel among many small channels that do not rejoin the trunk conduit (Fig. 3A). A straw poll of geomorphologist contacts produced an unvarying and matching definition. Critical in this definition is flow in all the branching channels at about the same time, though some branches may be inactive if the trunk channel is flowing substantially below bankfull discharge. We believe this geomorphic definition is the most relevant one, since it precisely embraces the formative physical processes responsible for sediment dispersal and deposition, and it yields the most accurate information on resultant sediment architecture. Some sedimentologists, on the other hand, take their meaning of the term directly from the more literal definitions of the verb distribute in standard English-language dictionaries. They thus use it to refer more loosely to the spreading of sediment throughout a space or area over much longer time periods than implied by the tighter geomorphic definition above. It is likely this is why the term was used for fans by Miall (1992; 1996), Leier et al. (2005), Moscariello (2005), and Nichols and Fisher (2007). In this sense, the term ‘‘distributary’’ is not being used to describe the geomorphological channel network over the timescales that the rivers locally operate (annual to decadal events), but instead is being used to refer to the accumulation of sediment over geological timescales (centuries to millennia) and at spatial scales larger than an individual channel belt. Given there are at least two different ways of understanding the meaning of the term ‘‘distributary,’’ the more specific geomorphic one and a looser sedimentological one, there is a high potential for misunderstanding when the term is used in an unqualified way. The way the channels are shown on the Kelly and Olsen (1993) TFM facies-

JSR

model diagram (Fig. 1), which is consistent with the original description by Friend (1978, his figure 6), clearly suggests that the geomorphic sense was meant. The manner of use by other workers (e.g., Moscariello 2005; Nichols and Fisher 2007) suggests they had in mind the looser sedimentological meaning, or were trying to bracket both senses with a single term (e.g., as in the Distributary Systems grouping of Gibling 2006). As we will now show, the differences between these two senses have profound implications for understanding the formation of fans and predicting detailed sediment architecture. Close inspection of the channel network on fluvial fans shows such a truly distributary channel system, in the geomorphic rather then the sedimentological sense, is the exception rather than the rule. Our survey consistently revealed evidence such as channel crosscutting, desert varnish, and soils that demonstrate the majority of channels are from different generations and were never active simultaneously (Fig. 3B). The superposition of new channels over a pattern of older ones can create the map-view impression of apparent bifurcation (e.g., points X and Y in Figure 3B). But we recommend that this should not be termed distributary, even in a looser sedimentological sense, because of the time gap between the channel generations, a time gap which can have profound implications for the lithofacies architecture. Our survey and analysis is consistent with flume studies (e.g., Schumm et al. 1987; Bryant et al. 1995) which show that fluvial fans evolve as a result of multiple channel avulsions (Fig. 3B), not channel bifurcations. This is in marked contrast to the truly bifurcating character of distributaries of river-dominated deltas (Fig. 3A). The flow of a river into a body of water produces a distinctive pattern of channel bifurcations (Olariu and Bhattacharya 2006) that scale directly with the width and depth of the channel entry point and can be modeled accurately from the theory of turbulent jets (Slingerland and Edmonds 2006). Reduction in velocity of the river flow, caused by interaction with a lake or sea, stimulates deposition of in-channel mouth bars. Eventually, such mouth bars become emergent and the flow splits to either side, so marking the completion of a bifurcation. This is an entirely different set of physical processes than the avulsions that build fans. Avulsions on fans are dominantly nodal and focused on the current fan-head trench, though local irregularities and external factors can trigger occasional down-fan avulsion. Truly distributary character on fluvial fans, in the geomorphic sense of coeval flow in multiple branches, is a localized and small-scale phenomenon unrelated to the primary drivers of fan creation. Localized distributary behavior that does occur is of two forms. One is where the channels in fact are anabranches of a braid complex or a channel network created through avulsion by annexation (Slingerland and Smith 2004, p. 262). The other is a transitory style representing imbalance between form and discharge in the channel that causes flow to spill out onto the floodplain, initially as sheetflow but over time evolving its own channel network. These latter cases are analogous in process to small-scale splays on rivers (Bristow et al. 1999; Tooth 2005). They scale with the feeder channel, and cannot in themselves be considered fluvial fans. Precise use of the term distributary is not merely semantic nit-picking. The temporal aspect must not be ignored if we are to understand fully the process-product linkage in fluvial systems. Fans are composite sedimentary bodies built over substantial time. In a network built by successive avulsions, it is highly probable that mud will have been deposited over earlier channel sands, when the older channel was in effect part of the floodplain of a newer channel. This creates a vertical separation between generations of channel-belt sand bodies. In the subsurface, fluvial-fan channel bodies are much less likely to be laterally connected than some might understand by the use of the distributary description in the TFM. And finally, a truly distributary network produces a systematic downstream reduction in channel width and depth (Fig. 3A), whereas no

JSR

END OF THE TERMINAL-FAN MODEL

697 INHERITANCE OF FORM

Our review of modern rivers that terminate subaerially did reveal several apparent contradictions to the conclusion that truly distributary patterns of channels are not a characteristic of fluvial fans. But here again planform appearance can be misleading. For every case we could investigate, across a wide range of scales, there is strong evidence to show that these fan-shaped features are intimately related to previous lake highstands (e.g., Fig. 4). The occurrence nearby of playa-type environments immediately throws up suspicion. The clinching evidence is usually the fan crossing abandoned lake shoreline features such as beach ridges (e.g., Adams and Wesnousky 1998). In other cases, fan sediments are intimately interbedded with extensive lacustrine mud, as is the case for the terminus of the Neales River to the west of Lake Eyre, Australia (Lang et al. 2004). In other words, such cases were initiated as lacustrine deltas, and have retained (inherited) their form from a previously wetter period, though there may now be no lake at all. In drylands especially, annual discharge is low and floods infrequent, so it takes a long time for a river to reform such a delta, and it may not do so before the climate again turns wetter. The result is deltas marooned in totally subaerial settings. Misunderstanding of the differences in process between fluvial fans and lacustrine deltas is disturbing, as is failure to take account of form inheritance, whereby an observed geomorphic feature was created under different conditions than prevail today. Together, these are leading to interpretations being published for the rock record that include such oxymorons as ‘‘lake plain terminal fans’’ (Pusca 2003) and ‘‘lacustrine terminal fan’’ (Abbate et al. 1991), both of which are most likely lacustrine deltas, not fluvial fans. Correct recognition of these features as lake deltas provides significant insight into past lake levels and hydrological conditions, and hence paleoclimate. But no such significance can (should) be read into them if they are sub-aerial fluvial fans as implicit in the TFM, since the presence of a lake was not required for their construction. NOT TERMINAL, NOT STRAIGHT?

FIG. 3.—The distinction between A) truly distributary channels which are active simultaneously (the geomorphic definition), and B) a radiating set of channels produced by successive nodal avulsions, but in which generally only one channel is active at any one time (T1 then T2 then T3). The former (A) produces a systematic downstream reduction in channel depth and width, because discharge is divided, and thus reduced, between channels at each bifurcation. No such systematic variation is to be expected from an avulsing channel system (B). Apparent channel bifurcations appear at the points of avulsion such as that labeled X, and at locations such as Y caused by the superposition of a channel over an older one. Both such cases might be mistaken for the true (contemporary) bifurcation of distributary channels. Multiple occurrence of crossovers such as point Y would also lead to the false impression of an anabranching channel pattern.

systematic downstream trend should be expected from an avulsing channel system on a fan (Fig. 3B). If our analysis is accurate, then a central tenet of the original TFM— a distributary breakdown of channels (Mukerji 1976 p.191; fig. 6 of Friend 1978)—is unsustainable, so undermining the entire TFM concept. This also makes invalid use of the synonym ‘‘fluvial distributary systems’’ proposed by Nichols and co-workers (e.g., Nichols and Fisher 2007). To avoid misunderstanding in future when referring to fluvial fans, we strongly recommend that use of the term distributary should either be heavily qualified, to indicate the tighter geomorphic or the looser sedimentological senses, or avoided completely and replaced with terms that communicate the intended meaning more precisely. Surely it is in the interests of all of us for geomorphologists and sedimentologists to use terms in a like and consistent manner, since these disciplines significantly overlap?

There are many well-documented examples of rivers in drylands which demonstrate that natural water losses due to infiltration and evapotranspiration can be so large that all discharge is lost before the river intersects another environment such as the sea or a permanent lake. This is common for ephemeral streams with small catchments fed mainly by local convective storms. It is also a feature of many much larger river systems fed by seasonal or cyclonic precipitation, such as Cooper Creek in Australia (Knighton and Nanson 1994) and the Okavango in Botswana (Stanistreet et al. 1993; Stanistreet and McCarthy 1993). But contrary to the impression some seem to have taken from the TFM, fans are not the omnipresent morphology for such subaerial, nondeltaic river termination. The exact morphology is much more variable than the TFM predicts, because of the dramatic effect even subtle local controls may have in drylands. On the Northern Plains of Australia, rivers carrying dominantly sand bedload may be relatively wide and shallow, almost braided, or become narrow, straight, and split into parallel anabranches where bank-lining vegetation stimulated by groundwater or local tributary input makes banks stronger (Tooth 2000; Tooth and Nanson 2004). Both types may lose definition downstream and transition rapidly through splays into sheetflow (a ‘‘floodout’’) that scales in width directly to the source channel (Tooth 2005). Neither type produce anything resembling a fluvial fan. Conversely, in the western USA, the very low gradient of streams terminating on old lake floors means that they are prone to frequent avulsion, so they do form fluvial fans. These evolve not from deposition from straight channels, as stated in the TFM, but from avulsive switching

698

JSR

C.P. NORTH AND G.L. WARWICK

Locally, a fluvial fan may be present at what appears to be the present limit of a dryland river, but detailed investigation shows it was not the terminus of that river when the fan formed. In drylands, fans are common at tributary junctions, sometimes even blocking the main channel, when the flow in a tributary is out of phase with the trunk river (Hooke and Mant 2001, p. 180; Meyer et al. 2001; Florsheim 2004). These are tributary-junction fans, well known to geomorphologists but seemingly ignored by sedimentologists working the rock record. Avulsion of the trunk above the junction and increasing aridity may leave such fans stranded in the landscape as apparently terminal features, which may explain the Markanda River fan. The situation is made even more complex because it is common for flood discharge in drylands to be insufficiently large, compared to the transmission losses, for the flow always to reach the same place in the system. Especially for larger dryland rivers, floods may terminate at many different positions and only rarely reach the most distal locations. This is the case for Cooper Creek, Australia, where it is only the largest and infrequent floods that lead to flow downstream of Innamincka, which itself is still several hundred kilometers from the ultimate terminus at Lake Eyre. When water discharge is dramatically reduced, then sediment is deposited. What looks on maps and imagery as the apparent terminus of a river is in practice only one of several places along the system that ‘‘terminal’’ depositional features are present. The most distal feature is the result of just the largest flows, which may be very infrequent. As a result, the lower reaches of dryland rivers commonly display geomorphic evidence of disequilibrium (Bourke and Pickup 1999; Tooth and Nanson 2000), so it is highly unlikely they would produce a simple or characteristic sedimentological signature of the kind portrayed in the TFM. LITHOFACIES DISTINCTIVENESS

FIG. 4.— An example of geomorphic distributary channel character that originated as a lacustrine delta but which is now entirely subaerial: the terminus of the Ruo Shui, which is at the western end of Inner Mongolia, in a Landsat 7 satellite image circa 2000. The image processing causes vegetation to appear bright green, damp playa surfaces appear light blue, lakes appear dark blue to black. A) This close-up view shows, at the left, the east end of the dry playa lake called Gaxun Nur (101u E, 42u N), with multiple abandoned shorelines, and in the lower right corner, picked out by vegetation, a distributary river termination. B) This shows the context of the example, with the lacustrine delta (in white box) occurring at the toe of the immense fluvial fan of the Ruo Shui, which displays classic nodal avulsion and many abandoned braid channels.

of highly sinuous, progressively narrower and shallower channels, often eroded into and filled by mud (Warwick 2006). These examples lack obvious splay landforms of the type seen in northern Australia. Flow does extend beyond the channels as sheetflow, but by this point it is usually carrying so little bedload sediment that no discrete depositional body results.

One of the most worrying aspects of the TFM is that implicit in the model is a presumption that there is a distinctive lithofacies suite that is diagnostic of ‘‘terminal fans.’’ In fact nothing has been proved but a lot has been assumed. The huge diversity of lithofacies types and scales in Friend’s (1978) original four examples is completely ignored. Instead, a facies pattern-matching approach is commonly taken, usually to the descriptions from Parkash et al. (1983) of the uppermost meter or so of the Markanda River ‘‘fan,’’ or to Tunbridge (1984), who himself had drawn a comparison to Parkash et al. (1983). The outcome is that a terminal fan may be interpreted (e.g., Davila and Astini 2003) solely due to superficial similarity to the thin-bedded and heterolithic successions described by Tunbridge (1984), even when both the area under investigation and Tunbridge’s analogue show clearly nonradial paleoflow indicators. In the time since Parkash et al. and Tunbridge published their studies, many more detailed accounts of the sedimentology of sandy braided rivers have appeared. It should now be evident to all that the lithofacies originally attributed to ‘‘terminal fans’’ can be found in the channels or on the floodplains of many rivers, in a wide range of climatic settings. The lithofacies observed in all cases cited to support the TFM possess no features which could not equally be produced by a non-fan dryland river, especially braided and ephemeral ones, as a perusal of any major fluvial review such as Miall (1996) would quickly show. For example, there is striking similarity between these descriptions and the lithofacies described in the frequently referenced article by McKee et al. (1967) from the valley of the modern Bijou Creek, in Colorado, USA, yet this stream is neither at its terminus nor is it part of a fan. Although not always stated explicitly, the argument in favor of a TFM interpretation seems to hinge solely on downstream fining and thinning (e.g., Kelly and Olsen 1993, p. 360), sometimes coupled to interbedding with indicators of ephemerality such as evaporites (Ciner et al. 2002), or

JSR

END OF THE TERMINAL-FAN MODEL

sometimes just a general ‘‘distal’’ nature such as thinly interbedded clay and mud (Pe´rez-Lo´pez 1996). Parkash et al. claim to have detected in ‘‘the lowermost sandy bed’’ of their sections a downstream fining of about one phi unit in grain size (Parkash et al. 1983, p. 343 and their figure 10). But closer inspection of the data suggests this may be over-interpretation of the available information. The trend is not obvious in the vertical section as a whole (Parkash et al. 1983, their figure 3). The downstream end is constrained by few data points, which probably accounts for the standard deviation of 0.8 phi units in this region. It would be difficult to be sure if the sampled bed, which varies in depth below the surface, is from a channel or an overbank context, and the channel sands are typically one phi unit coarser grained than the overbank deposits (Parkash et al. 1983, their figs. 8 and 9). Downstream fining is, in any case, a well-known fluvial phenomenon, and channel thinning is also now being recognized as a feature of some humid-region river systems (Makaske et al. 2007; Gouw and Berendsen 2007). Nothing in this kind of evidence leads conclusively to even a fan origin, let alone the special case of the TFM, yet from comparison to this example paleogeography is being reconstructed including radiating fans. Such model-driven interpretation ignores the trap of convergence (Schumm 1991, p. 58) whereby the same outcome (lithofacies) can be produced by completely different sets of processes. The TFM is an inductive model arrived at through reductivist argument from a severely limited database. SCALE

Another aspect that concerns us is the pick-and-mix approach taken to scale. Generally overlooked is the admission from Friend (1978, p. 533) that it is not possible to correlate between outcrops for each of his examples, meaning that the model is a deduction not an observation. A great deal is being argued solely from large-scale regional trends in thick successions without being able to tell if the sediments are from single or multiple drainages. To build their hypothetical model (Fig.1), Kelly and Olsen (1993, p. 342–344) used the limited available lithofacies descriptions of the topmost few meters of the 10-km-scale Markanda and Gash rivers fan-shaped features, convolved with descriptions of ephemeral stream alluvium from several non-fan cases (McKee et al. 1967; Williams 1970; Langford 1989). They then applied this model to interpretation of disjointed rock-record examples 100 km in lateral extent and many tens to hundreds of meters thick, to show they also were terminal fans and thus build support for the model. This hypothesis construction relies on dramatic extrapolations of scale. For the Markanda case, the lithofacies descriptions record only the top , 2 m of sediment over a , 10 km traverse (Parkash et al. 1983). Kelly and Olsen (1993, p. 344) claim that the Gash River case (Fig. 2) reveals the lithofacies of the distributary zone of the TFM (Fig. 1), but this is not so in our view, as explained above. Although some Gash River sections were as much as 3 m thick, all were taken in an area , 400 m across within the main channel feeding the fan (Fig. 2) in a location upstream from where the supposed distributary-channel-network system starts (Abdullatif 1989, his figures 3 and 7). The lithofacies at the apex of the fan can hardly be considered conclusively representative of the main body of the ‘‘fan’’ many tens of kilometers away. PROCESSES AND PRODUCTS OF FLUVIAL FANS

Disturbing evidence that the TFM is creating misunderstanding and unsound interpretation is the tendency for workers to argue that terminal fans must be present because the climate was arid to semiarid (e.g., Davila and Astini 2003). This clearly demonstrates the need to abandon the model and pattern-matching analysis in favor of careful process-based

699

thinking. Fluvial fans occur in all climates, not just drylands. In drylands, fluvial fans are not mandatory at river terminations. Provided sufficient horizontal space is available, fluvial fans can form anywhere along the course of a river where two criteria are met: release from confinement, together with discharge or environment conditions that promote avulsion. The same river may have more than one fluvial fan along its course. Large fluvial fans are common wherever there is strong seasonality to the river discharge, because of the instability of channels subject to large fluctuations in discharge (Leier et al. 2005). Drylands also experience wide fluctuations in discharge (McMahon et al. 1987), so it is not surprising fans are common there. Friend (1978, p. 541) suggested that a major reason his stratigraphically thick examples differ from many Quaternary fluvial successions is that they formed ‘‘under orogenic conditions.’’ Avulsion frequency is a function of sedimentation rate (Bryant et al. 1995), which increases towards orogenic belts. Avulsion is also encouraged by shallow gradients and in-channel deposition triggered by discharge loss (Jones and Schumm 1999). Wherever avulsion frequency is increased there is increased likelihood of a fluvial fan developing, provided there is sufficient lateral space available to fit it in. The sedimentary record left by fluvial fans depends primarily on the behavior of the rivers crossing the fan. Rivers do not leave distinctive deposits just because they are traversing a fan. Fluvial fans display rivers of all planform types, and planform varies both downstream and with time, as well shown by the Kosi River fan (Wells and Dorr 1987). Rivers in drylands may terminate in a fluvial fan, if the conditions are appropriate, but even where they do the lithofacies record will be little different from a non-fan river in the same climatic and tectonic setting. CONCLUSIONS

The terminal-fan model (TFM) is a hypothetical model created by sedimentologists for a depositional sedimentary environment not recognized in the modern day by geomorphologists. The limited modern analogues used to support the TFM are at best equivocal, and provide poor support for a generic facies model. The concept implicit in the model of widespread coeval distributary channels is erroneous and is leading to confusion about sandstone connectivity and geometry. The model predicts down-fan systematic decrease in channel width and depth, with connection between all the anabranches. The reality is that there is no systematic variation in channel dimensions: they may even increase downfan. Different channel generations may not act as connected subsurface fluid-flow pathways, there is a high potential for mud to occur between channel bodies, and a channel may not be linked laterally to any other channel. Unqualified use of the term distributary is resulting in misunderstanding because of different uses by the geomorphological and sedimentological communities. The proposed synonym of ‘‘fluvial distributary systems’’ is unsatisfactory because it perpetuates the same misunderstandings. Inheritance of geomorphic form and the impact of climate change have been ignored. Relict lacustrine deltas, influenced by non-fluvial processes and producing distinctly different sedimentary products, are being mistaken for purely alluvial, subaerial settings. Scale has been overlooked in a cavalier fashion. Disconnected outcrops have been used to support the model without taking into account whether the deposits are the result of a single river or multiple rivers, which is similar to the problem of telling apart a single alluvial fan from a fan bajada. Where fan-shaped sediment bodies do occur, the processes and products are no different to any other fluvial fan in that climatic setting, and such fans are not exclusive to dryland regions, nor are they always terminal. There is no diagnostic assemblage of lithofacies. We strongly recommend that the TFM should be abandoned. Superficial analysis of present-day planform without incorporating temporal relationships can be highly misleading, a point we urgently

700

C.P. NORTH AND G.L. WARWICK

wish to impress on those who draw solely on satellite and aerial imagery when constructing subsurface geological models. ACKNOWLEDGMENTS

We gratefully acknowledge the financial support for this study from the consortium members of Phase 1 of the AUDRI Project (Aberdeen University Dryland Rivers Initiative), namely Anadarko Algeria, BP, ConocoPhillips, Shell, and Total. We thank Stephen Tooth for helpful criticism of an earlier draft, and we thank Ken Adams, Stuart Archer, Simon Fagan, Alex Fordham, Adrian Hartley, Carmen Krapf, Simon Lang, and Gerald Nanson for discussions on rivers in dryland regions. Constructive reviews on behalf of JSR by S.K. Tandon, Brian G. Jones, and Martin Gibling led to significant improvements to this article. REFERENCES

ABBATE, E., BRUNI, P., AND SAGRI, M., 1991, Development of a lacustrine terminal fan and a coarse grained delta in the Tertiary Daban Basin (northern Somalia): Cuadernos de Geologı´a Ibe´rica, v. 15, p. 279–298. ABDULLATIF, O.M., 1989, Channel-fill and sheet-flood facies sequences in the ephemeral terminal River Gash, Kassala, Sudan: Sedimentary Geology, v. 63, p. 171–184. ADAMS, K.D., AND WESNOUSKY, S.G., 1998, Shoreline processes and the age of the Lake Lahontan highstand in the Jessup embayment, Nevada: Geological Society of America, Bulletin, v. 110, p. 1318–1332. ALLABY, A., AND ALLABY, M., 1999. Oxford Dictionary of Earth Sciences, Second Edition: Oxford, U.K., Oxford University Press, 619 p. BAETEMAN, C., DUPIN, L., AND HEYVAERT, V., 2004, The Persian Gulf shoreline and the Karkeh, Karun, and Jarrahi Rivers: Akkadica, v. 125, p. 141–215. BOURKE, M.C., AND PICKUP, G., 1999, Fluvial form variability in arid Central Australia, in Miller, A.J., and Gupta, A., eds., Varieties of Fluvial Form: Chichester, U.K., John Wiley & Sons, p. 249–271. BRISTOW, C.S., SKELLY, R.L., AND ETHRIDGE, F.G., 1999, Crevasse splays from the rapidly aggrading, sand-bed, braided Niobrara River, Nebraska: effect of base-level rise: Sedimentology, v. 46, p. 1029–1047. BRYANT, M., FALK, P., AND PAOLA, C., 1995, Experimental study of avulsion frequency and rate of deposition: Geology, v. 23, p. 365–368. CINER, A., KOSUN, E., AND DEYNOUX, M., 2002, Fluvial, evaporitic and shallow-marine facies architecture, depositional evolution and cyclicity in the Sivas Basin (lower to middle Miocene), Central Turkey: Journal of Asian Earth Sciences, v. 21, p. 147–165. COLLINSON, J.D., 1996, Alluvial sediments, in Reading, H.G., ed., Sedimentary Environments: Processes, Facies and Stratigraphy, Third Edition: Oxford, U.K., Blackwell Science, p. 37–82. DAVILA, F.M., AND ASTINI, R.A., 2003, Early middle Miocene broken foreland development in the southern Central Andes: evidence for extension prior to regional shortening: Basin Research, v. 15, p. 379–396. FLORSHEIM, J.L., 2004, Side-valley tributary fans in high-energy river floodplain environments: Sediment sources and depositional processes, Navarro River basin, California: Geological Society of America, Bulletin, v. 116, p. 923–937. FRIEND, P.F., 1978, Distinctive features of some ancient river systems, in Miall, A.D., ed., Fluvial Sedimentology, Canadian Society of Petroleum Geologists, Memoir 5, p. 531–542. GIBLING, M.R., 2006, Width and thickness of fluvial channel bodies and valley fills in the geological record: a literature compilation and classification: Journal of Sedimentary Research, v. 76, p. 731–770. GIBLING, M.R., TANDON, S.K., SINHA, R., AND JAIN, M., 2005, Discontinuity-bounded alluvial sequences of the southern Gangetic Plains, India: aggradation and degradation in response to monsoonal strength: Journal of Sedimentary Research, v. 75, p. 369–385. GOUW, M.J.P., AND BERENDSEN, H.J.A., 2007, Variability of channel-belt dimensions and the consequences for alluvial architecture: observations from the Holocene Rhine–Meuse Delta (The Netherlands) and lower Mississippi Valley (U.S.A.): Journal of Sedimentary Research, v. 77, p. 124–138. HOOKE, J., AND MANT, J., 2001, Morpho-dynamics of ephemeral streams, in Bull, L.J., and Kirkby, M.J., eds., Dryland Rivers: Processes and Management in Mediterranean Climates: Chichester, U.K., John Wiley & Sons, Ltd, p. 173–204. JONES, L.S., AND SCHUMM, S.A., 1999, Causes of avulsion: an overview, in Smith, N.D., and Rogers, J., eds., Fluvial Sedimentology VI: International Association of Sedimentologists, Special Publication 28, p. 171–178. KELLY, S.B., AND OLSEN, H., 1993, Terminal fans—a review with reference to Devonian examples: Sedimentary Geology, v. 85, p. 339–374. KNIGHTON, A.D., AND NANSON, G.C., 1994, Flow transmission along an arid zone anastomosing river, Cooper Creek, Australia: Hydrological Processes, v. 8, p. 137–154. LANG, S.C., PAYENBERG, T.H.D., REILLY, M.R.W., HICKS, T., BENSON, J., AND KASSAN, J., 2004, Modern analogues for dryland sandy fluvial–lacustrine deltas and terminal splay reservoirs: Australian Petroleum Production and Exploration Association (APPEA), Journal, p. 329–356.

JSR

LANGFORD, R.P., 1989, Fluvial–aeolian interactions: part I, modern systems: Sedimentology, v. 36, p. 1023–1035. LEIER, A.L., DECELLES, P.G., AND PELLETIER, J.D., 2005, Mountains, monsoons, and megafans: Geology, v. 33, p. 289–292. MAKASKE, B., BERENDSEN, H.J.A., AND VAN REE, M.H.M., 2007, Middle Holocene avulsion-belt deposits in the central Rhine–Meuse Delta, The Netherlands: Journal of Sedimentary Research, v. 77, p. 110–123. MCKEE, E.D., CROSBY, E.J., AND BERRYHILL, H.L., 1967, Flood deposits, Bijou Creek, Colorado, June 1965: Journal of Sedimentary Petrology, v. 37, p. 829–851. MCMAHON, T.A., FINLAYSON, B.L., AND SRIKANTHAN, R., 1987, Runoff variability: a global perspective, in Solomon, S.I., Beran, M., and Hogg, W., eds., The Influence of Climate Change and Climate Variability on the Hydrologic Regime and Water Resources: Wallingford, England, International Association of Hydrological Sciences, Publication 168, p. 3–12. MEYER, G.A., PIERCE, J.L., WOOD, S.H., AND JULL, A.J.T., 2001, Fire, storms, and erosional events in the Idaho batholith: Hydrological Processes, v. 15, p. 3025–3038. MIALL, A.D., 1992, Alluvial deposits, in Walker, R.G., and James, N.P., eds., Facies Models: Response to Sea-Level Change: St. John’s Newfoundland, Geological Association of Canada, p. 119–142. MIALL, A.D., 1996, The Geology of Fluvial Deposits; Sedimentary Facies, Basin Analysis, and Petroleum Geology: New York, Springer-Verlag, 582 p. MOSCARIELLO, A., 2005, Exploration potential of the mature southern North Sea basin margins: some unconventional plays based on alluvial and fluvial fan sedimentation models, in Dore´, A.G., and Vining, B.A., eds., Petroleum Geology: North-West Europe and Global Perspectives: London, Geological Society of London, p. 595–605. MUKERJI, A.B., 1975, Geomorphic patterns and processes in the terminal tract of inland streams in Sutlej–Yamuna plain: Journal of the Geological Society of India, v. 16, p. 450–459. MUKERJI, A.B., 1976, Terminal fans of inland streams in Sutlej–Yamuna Plain, India: Zeitschrift fu¨r Geomorphologie, Neue Folge, v. 20, p. 190–204. NICHOLS, G.J., 1987, Structural controls on fluvial distributary systems—the Luna system, northern Spain, in Ethridge, F.G., Flores, R.M., and Harvey, M.D., eds., Recent Developments in Fluvial Sedimentology: SEPM, Special Publication 39, p. 269–277. NICHOLS, G.J., AND FISHER, J.A., 2007, Processes, facies and architecture of fluvial distributary system deposits: Sedimentary Geology, v. 195, p. 75–90. NICHOLS, G.J., AND HIRST, J.P., 1998, Alluvial fans and fluvial distributary systems, Oligo-Miocene, northern Spain: Contrasting processes and products: Journal of Sedimentary Research, v. 68, p. 879–889. OLARIU, C., AND BHATTACHARYA, J.P., 2006, Terminal distributary channels and delta front architecture of river-dominated delta systems: Journal of Sedimentary Research, v. 76, p. 212–233. PARKASH, B., AWASTHI, A.K., AND GOHAIN, K., 1983, Lithofacies of the Markanda terminal fan, Kurukshetra district, Haryana, India, in Collinson, J.D., and Lewin, J., eds., Modern and Ancient Fluvial Systems: International Association of Sedimentologists, Special Publication 6, p. 337–344. PE´REZ-LO´PEZ, A., 1996, Sequence model for coastal-plain depositional systems of the Upper Triassic (Betic Cordillera, southern Spain): Sedimentary Geology, v. 101, p. 99–117. PUSCA, V.A., 2003, Wet–dry, terminal fan-dominated depositional sequences on the lake plain; a case study in the lower Green River Formation of southern Uinta Basin [abstract]: American Association of Petroleum Geologists, Annual Meeting Expanded Abstracts, v. 12, 141 p. SADLER, S.P., AND KELLY, S.B., 1993, Fluvial processes and cyclicity in terminal fan deposits: an example from the Late Devonian of southwest Ireland: Sedimentary Geology, v. 85, p. 375–386. SCHUMM, S.A., 1991, To Interpret the Earth: Ten Ways to Be Wrong: Cambridge, U.K., Cambridge University Press, 133 p. SCHUMM, S.A., MOSLEY, M.P., AND WEAVER, W.E., 1987, Experimental Fluvial Geomorphology: New York, John Wiley & Sons, 413 p. SLINGERLAND, R.L., AND EDMONDS, D.A., 2006, Genesis and shapes of delta distributary networks (abstract):, American Association of Petroleum Geologists, 2006 Annual Convention, Abstracts Volume, p. 99. SLINGERLAND, R.L., AND SMITH, N.D., 2004, River avulsions and their deposits: Annual Review of Earth and Planetary Sciences, v. 32, p. 257–285. STANISTREET, I.G., AND MCCARTHY, T.S., 1993, The Okavango Fan and the classification of subaerial fan systems: Sedimentary Geology, v. 85, p. 115–133. STANISTREET, I.G., CAIRNCROSS, B., AND MCCARTHY, T.S., 1993, Low sinuosity and meandering bedload rivers of the Okavango Fan: channel confinement by vegetated levees without fine sediment: Sedimentary Geology, v. 85, p. 135–156. TOOTH, S., 2000, Downstream changes in dryland river channels: the Northern Plains of arid central Australia: Geomorphology, v. 34, p. 33–54. TOOTH, S., 2005, Splay formation along the lower reaches of ephemeral rivers on the Northern Plains of arid central Australia: Journal of Sedimentary Research, v. 75, p. 636–649. TOOTH, S., AND NANSON, G.C., 2000, Equilibrium and nonequilibrium conditions in dryland rivers: Physical Geography, v. 21, p. 183–211. TOOTH, S., AND NANSON, G.C., 2004, Forms and processes of two highly contrasting rivers in arid central Australia, and the implications for channel-pattern discrimination and prediction: Geological Society of America, Bulletin, v. 116, p. 802–816. TUNBRIDGE, I.P., 1984, Facies model for a sandy ephemeral stream and clay playa complex; the Middle Devonian Trentishoe Formation of North Devon, U.K.: Sedimentology, v. 31, p. 697–715.

JSR

END OF THE TERMINAL-FAN MODEL

WARWICK, G.L., 2006, The geomorphology and sedimentology of terminal fluvial systems [Unpublished Ph.D. thesis]: University of Aberdeen: Aberdeen, U.K. WATERS, M.R., AND RAVESLOOT, J.C., 2001, Landscape change and the cultural evolution of the Hohokam along the middle Gila River and other river valleys in south-central Arizona: American Antiquity, v. 66, p. 285–299. WELLS, N.A., AND DORR, J.A., 1987, Shifting of the Kosi River, northern India: Geology, v. 15, p. 204–207.

701

WHITEMAN, A.J., 1971, Geology of the Sudan Republic: Oxford, U.K., Clarendon, 290 p. WILLIAMS, G.E., 1970, The central Australian stream floods of February–March 1967: Journal of Hydrology, v. 11, p. 185–200. Received 1 December 2006; accepted 23 March 2007.