Because of water losses, one dramatic feature of dryland channel systems is that they may reach a point downstream where they simply cease to exist. But what actually happens at this point?

The phrase "terminal fan" keeps popping up in the sedimentology literature to describe this situation. But it suffers from being weakly defined and used loosely. Examples are being interpreted in the rock record yet we have no convincing and thorough accounts of modern instances. In Australia, a feature termed a floodout is commonly described - is it the same thing?

Here, I summarise the position as I see it from the literature. Then I propose a series of questions I'd like to see discussed.

Please contact me with your own ideas. Do you agree with my summary of the current positon? Have I missed key literature? How should we proceed?

As yet, there is no good sedimentological description of a modern dryland region example of this phenomenon, though it has been recorded in passing from many continents (e.g. Glennie, 1970; Twidale, 1972; Friend, 1978; Pickup, 1986). Indeed, this aspect has also not received much attention in the geomorphological literature.

This phenomenon has been speculated upon in the sedimentological literature by Henrik Olsen and Sean Kelly (Olsen, 1987; Kelly & Olsen, 1993). From a critical review of the sparse literature, and consideration of what was then appreciated about dryland rivers, they produced the hypothetical "terminal fan" facies model which predicts broad and shallow channels in the proximal and feeder zones, grading distally into sheet-flood styles with poorly defined channels. They evaluated their model against examples from the Devonian of Greenland and western Ireland. From their analysis, they made some predictions about likely dimensions and character of sand-bodies within the fan.

This model (Fig. TFx1) is now reproduced in the standard text books as though it is proven and reliable (e.g. Collinson, 1996, p.82).

Kelly & Olsen used their model to interpret examples of Devonian successions from Ireland and Greenland, in effect trying to validate a model for modern systems with the use of Devonian examples. This is an unsafe way to proceed, since land plant types were very different, and likely to have been much less efficient at binding sediment, as noted by Kelly & Olsen (1993).

The 'modern' example commonly cited to support the terminal-fan facies model is that of the present day terminus of the Markanda River, on the Indo-Gangetic Plains of northern India. First described by Mukerji (1975; 1976), the sedimentology was summarised by Parkash and others (1983). This information has then been drawn on by others such as Olsen (Olsen, 1987) to help them interpret ancient successions.

The Markanda River today is in a semi-arid region (mean rainfall 530 mm). This is why it is quoted as a 'type' example for drylands. But this is a classic case of the phenomenon highlighted by Blair & McPherson (1992) whereby progressive extraction and extrapolation of data through a succession of papers, to produce a facies model, results in omission of key facts, and incorporation of wishful thinking unsubstantiated by observations.

In the Markanda River case, because the first account (Mukerji, 1976) included almost no sedimentological information, it is usual to draw for information about this fan on the account of Parkash and other (1983). However, though Parkash and others acknowledge the earlier work of Mukerji, they merely summarise the present day climate for the study location, and completely ignore the fact that Mukerji made it very clear there is strong evidence the Markanda 'terminal fan' was initiated in the much wetter phase after the last glaciation, and current fluvial behaviour is out of equilibrium with the climate (Mukerji, 1976 p. 202).

This was noted by Kelly & Olsen (1993), but gets ignored again by those citing Kelly & Olsen's work, perhaps because they don't read the original paper but rely on secondary reporting in syntheses by others (typically Miall, 1996, p.249; or Collinson, 1996, p.82).

Without a very careful re-examination of the raw data for the Markanda River, and probably new fieldwork as well (to allow for the significant advances in knowledge since the original study in the early 1970s), it is dangerous to rely on this single case to substantiate the 'terminal fan' model for drylands. Much more detail is needed for the Markanda Fan, as Parkash and others (1983) report just a set of widely spaced vertical profiles, and do not document the spatial arrangement of sub-environments on the fan. A "massive mud" is reported in some sections, but there is no information on pedogenesis (or lack of it) - the era in which this study was done was dominated by different concerns to today (e.g. grain size analysis).

There appears to be one other example in the literature of a dryland terminal fan, and occasionally it is cited as support for the model, but close inspection shows the paper contains too little detail to be reliably extrapolated to other regions. Abdullatif (1989) describes the termination of the River Gash just after it passes the town of Kassala in SE Sudan. Maps and photographs clearly show a distributary morphology to the river system as it dissipates into the desert over a distance of about 50 km. Abdullatif describes the distal parts as being dominated by sheetfloods.

But the only detailed sedimentology in the paper comes from four short trenches dug into the bed of the main river at the head or apex of the 'fan', right at the town, and before significant splitting of channels has occurred. It is not really, therefore, telling us about the fan itself, and the dataset is in any case too small for our needs today. It is impossible to use this example to predict facies patterns in such a setting.

When considering terminal fans, the crucial factor on the approaches to the terminus is the rate of loss of the energy to transport sediment. There are two ways to lose energy: reduce the gradient, or reduce the water in the river. The first, such as occurs at fault-controlled breaks of slope, produces the classic alluvial fans. The second might be achieved in one of two ways, each of which produces a different morphology:

• If the water is lost very quickly, then the sediment load will get dropped rapidly, and sediment must accumulate. Because of deposition along the channel, the location of the channel will be unstable, the channel will keep switching position (bifurcating), and the resultant accumulation of sediment is likely to have a crudely fan-shaped planform (though it will also depend on the topography of the region), as envisaged by Kelly & Olsen. The process mimics alluvial fan genesis, where it is the sudden decrease in gradient at the break in slope that reduces flow energy so forcing sedimentation; in drylands, there is no break in slope, the water is lost by infiltration and evaporation so reducing flow energy

• If the water is lost more slowly, over a greater distance, sediment will be deposited over a much longer reach of the river, and by the end there may be relatively little sediment left, and it would not be possible to distinguish any obvious fan. This is what has been described for some of the northward flowing rivers in the Northern Territories of Australia, which just fizzle out in the middle of nowhere (Fig. TFx2) (Tooth, 1999; Tooth, 2000). Flow energy also gets lost along the channel by frequent bank breaches and overbank floods, some triggered by in-channel vegetation, some by in-channel sedimentation. Dryland channels tend to be shallow, with low levees, so overbank flooding occurs easily. When water is lost, it is the inability of the river to transport sediment, and choking of the system with deposited material, that can force bifurcation of the channels, and produces an apparently distributary network.

In Australia, where a river becomes unconfined at its terminus is called a floodout, defined by Tooth (1999) as "a site where channelized flow ceases and floodwaters spill across adjacent alluvial surfaces". Tooth distinguishes intermediate floodouts, defined as before but with the qualifier that "… downvalley of which flow channelizes again", and terminal floodouts, as before but qualified by "… and [flow] ultimately ceases." The sedimentology of these features is known in only the most superficial detail, and their frequency on other continents is unstudied.