Geomorphology of desert sand dunes: A review of recent progress

Abstract

Through the 1980s and 1990s studies of the geomorphology of desert sand dunes were dominated by field studies of wind flow and sand flow over individual dunes. Alongside these there were some attempts numerically to model dune development as well as some wind tunnel studies that investigated wind flow over dunes. As developments with equipment allowed, field measurements became more sophisticated. However, by the mid-1990s it was clear that even these more complex measurements were still unable to explain the mechanisms by which sand is entrained and transported. Most importantly, the attempt to measure the stresses imposed by the wind on the sand surface proved impossible, and the use of shear (or friction) velocity as a surrogate for shear stress also failed to deliver. At the same time it has become apparent that turbulent structures in the flow may be as or more important in explaining sand flux. In a development paralleled in fluvial geomorphology, aeolian geomorphologists have attempted to measure and model turbulent structures over dunes. Progress has recently been made through the use of more complex numerical models based on computational fluid dynamics (CFD). Some of the modelling work has also suggested that notions of dune ‘equilibrium’ form may not be particularly helpful. This range of recent developments has not meant that field studies are now redundant. For linear dunes careful observations of individual dunes have provided important data about how the dunes develop but in this particular field some progress has been made through ground-penetrating radar images of the internal structure of the dunes.

The paradigm for studies of desert dune geomorphology for several decades has been that good quality empirical data about wind flow and sand flux will enable us to understand how dunes are created and maintain their form. At least some of the difficulty in the past arose from the plethora of undirected data generated by largely inductive field studies. More recently, attention has shifted–although not completely–to modelling approaches, and very considerable progress has been made in developing models of dune development. It is clear, however, that the models will continue to require accurate field observations in order for us to be able to develop a clear understanding of desert sand dune geomorphology.

Introduction

Although reviews of the geomorphology of desert sand dunes routinely start with reference to the work of Ralph Bagnold (e.g. 1941), Bagnold's work concentrated largely on the physics associated with the movement of individual sand grains in the wind rather than on the development of landforms. His work was ground-breaking in providing systematic empirical measurements of wind flow, particularly velocity profiles, and of sand flux, but his work on dunes was rather more descriptive and often speculative. A significant speculation was his suggestion that thermally-generated roll vortices were the origin of linear sand dunes (Bagnold, 1953). Despite his clear acknowledgment of the speculative nature of his theory, it became widely quoted, and it was not until 30 years later that it was challenged on the basis of careful field observation and measurement of linear dunes (e.g. Tsoar, 1983).

Tsoar's work was just one of a wave of single-dune studies that were undertaken in the 1970s and 1980s (Table 1). These were attempts to understand the basic controls on the development of the form of individual dunes rather than the development of groups of dunes as sand seas (e.g. Wilson, 1973) or the movement of individual grains (e.g. Bagnold, 1941). This trend was part of a wider trend in geomorphology (and even more widely in geography (Burton, 1963) and geology (Merriam, 2004)) towards the measurement of small-scale processes and the use of statistics to analyse the data collected, frequently termed the ‘quantitative revolution’. In fluvial geomorphology this started in the 1950s and manifested itself as small catchment studies (e.g. Leopold et al., 1964). Aeolian geomorphology took rather longer to catch this wave (although Coursin's work (1964) was somewhat ahead of the pace) but what may be termed the single-dune studies of the 1970s and 1980s were aeolian geomorphology's equivalent of the small catchment studies. Alongside these, there has been a number of studies based variously on satellite and aerial photographic imagery, repeated ground survey or on a combination of imagery with survey that provide measurements of dune movement.

Single-dune studies concentrated on measurements of wind flow and sand flux (the methods of the period are summarised by Knott and Warren (1981)). Wind flow was largely measured by rotating-cup anemometers, often at a single height; wind direction was measured variously by the orientation of ripple marks, paper flags or by tracking meteorological balloons or kites; and sand flux was measured by using sand traps, sometimes based on the design of Bagnold (1941). Many of these studies began to provide good fundamental basic data about how the dunes behaved. In particular, Tsoar and Livingstone were able to show that on two very different linear dunes there was no evidence to support Bagnold's roll-vortex hypothesis. Both their studies showed linear dunes responding to bi-directional wind regimes, albeit with some difference of opinion about the mechanism (e.g. Livingstone, 1988, Tsoar, 1990).

These single-dune studies spawned a series of rather more sophisticated studies (e.g. Wiggs, 1993, Wiggs et al., 1996, Lancaster et al., 1996, Frank and Kocurek, 1996a, Frank and Kocurek, 1996b, McKenna Neuman et al., 1997). This increasing sophistication was partly a consequence of rapidly improving computer technology; the data loggers used to collect and store information from anemometers improved very rapidly through the 1980s and 1990s. Loggers became able to store much more data from a larger number of devices and laptop computers were available for downloading and analysing data. More accurate anemometers were developed including hot wire anemometers although there was an issue of robustness in field situations. Equipment such as the Saltiphone (Spaan and Van den Abele, 1991) and the Sensit probe (Stockton and Gillette, 1990) were developed to count the impacts of saltating grains in transport. Often this trend for field monitoring and measurement meant that large amounts of hardware were installed for these single-dune studies (Fig. 1).

Although enabled by improving technology, the increased sophistication was driven more fundamentally by the realisation that simple wind-speed measurements at a single height were not enough to characterise the wind flow over the dune. Even though Bagnold (1941) and others had related sand transport rates to wind speed, sand transport has been more commonly related to the stress imparted at the sand surface by the wind. Using wind speed as a surrogate for shear stress was a pragmatic response to the difficulty of measuring shear stress directly. Many single-dune studies in the 1990s attempted to measure wind speeds at different heights in order to produce a velocity profile from which shear stresses could be inferred via the calculation of shear (or friction) velocity, u_⁎.

In recent years the number of single-dune studies has dropped considerably. There are two main reasons for this. The first was the realisation that links between wind and sand flow around dune forms are more complicated than perhaps was envisaged 20 or 30 years ago and instrumentation in real-world environments was not delivering the link between the essentially turbulent nature of the wind and sand flux. Even on the most fundamental of ‘free’ dunes–transverse dunes–it was not possible to make measurements of wind speed using anemometers that would give reliable values of surface shear stress. The second realisation was that single-dune studies were oversimplifying dune systems where there are frequently complicated combinations of a variety of single-dune forms. It became apparent that single-dune studies could only take us so far in our understanding of dune dynamics and development.

As a consequence the international conference on aeolian research (ICAR) held every 4 years saw a drop in the number of papers reporting single-dune studies more or less to zero in 1998 and 2002 (see editorials by Livingstone, 1999, Livingstone and Nickling, 2004). They were replaced by a burgeoning interest in dust entrainment and transport and in palaeoenvironmental studies of aeolian environments driven in part by the explosion of interest in luminescence dating which has revolutionised dating of aeolian sediments.

Furthermore, aeolian geomorphology has seen something of the wider methodological debate about whether ‘reductionist’ studies can deliver explanations of dune form and pattern. The argument has been that understanding the physical processes at the grain scale will not deliver an explanation of what is happening at the dune scale because the relationships are complex; equally dune-scale processes do not help us to understand patterns at the dunefield or sand sea scale. The complexity is a result of a number of factors, not least the non-linearity of the physical relationships. The clearest general explanation of the thinking behind non-linearity and complex systems is provided by Phillips, 1999, Phillips, 2003. Most straightforwardly, “a system is nonlinear if the outputs are not proportional to the inputs across the entire range of inputs” (Phillips, 1999, p.4).

More than a decade ago Werner (1995) suggested that dunes provided good examples of complex systems and he showed that he could model the development of dunefield patterns without reference to the small-scale grain-to-grain processes that have been the focus of many single-dune studies. More recently, Kocurek and Ewing (2005) suggested that we should view dunes as self-organised complex systems and argued that reductionism breaks down because the relationships are non-linear. Yet non-linearity in itself is not sufficient to eschew so-called reductionist studies; the most reductionist studies at the grain scale are able to deal with non-linearity in threshold equations for grain entrainment.

Kocurek and Ewing suggested that viewing dunes as complex systems represented a paradigm shift but it is probably fairer to suggest that there has been a paradigm reappraisal. It is not the case that geomorphologists have ceased to be interested in questions associated with how dunes develop as landforms. While single-dune studies continue, other approaches are being explored, in particular those in which there is greater control on the variables in the system. Modelling approaches–both hardware in wind tunnels and software–provide this option. This paper now considers how the complexity of desert dunes is being examined.