Understanding river dune splitting through flume experiments and analysis of a dune evolution model Jord J. Warmink, 1 C. Marjolein Dohmen-Janssen, 1 * Joost Lansink, 1,2 Suleyman Naqshband, 1 Olav J. M. van Duin, 1 Andries J. Paarlberg, 3 Paul Termes 3 and Suzanne J. M. H. Hulscher 1 1 Department of Water Engineering and Management, University of Twente, Enschede, The Netherlands 2 Witteveen + Bos Consultants, Deventer, The Netherlands 3 HKV Consultants, Lelystad, The Netherlands Received 22 March 2013; Revised 18 December 2013; Accepted 6 January 2014 *Correspondence to: C. Marjolein Dohmen-Janssen, University of Twente, Department of Water Engineering and Management, PO Box 217, 7500 AE Enschede, The Netherlands. E-mail: c.m.dohmen-janssen@utwente.nl ABSTRACT: Forecasts of water level during river floods require accurate predictions of the evolution of river dune dimensions, because the hydraulic roughness of the main channel is largely determined by the bed morphology. River dune dimensions are controlled by processes like merging and splitting of dunes. Particularly the process of dune splitting is still poorly understood and – as a result – not yet included in operational dune evolution models. In the current paper, the process of dune splitting is investigated by carrying out laboratory experiments and by means of a sensitivity analysis using a numerical dune evolution model. In the numerical model, we introduced superimposed TRIAS ripples (i.e. triangular asymmetric stoss side-ripples) on the stoss sides of underlying dunes as soon as these stoss sides exceed a certain critical length. Simulations with the model including dune splitting showed that predictions of equilibrium dune characteristics were significantly improved compared to the model without dune splitting. As dune splitting is implemented in a parameterized way, the computational cost remains low which means that dune evolution can be calculated on the timescale of a flood wave. Subsequently, we used this model to study the mechanism of dune splitting. Literature showed that the initiation of a strong flow separation zone behind a superimposed bedform is one of the main mechanisms behind dune splitting. The flume experiments indicated that besides its height also the lee side slope of the superimposed bedform is an important factor to determine the strength of the flow separation zone and therefore is an important aspect in dune splitting. The sensitivity analysis of the dune evolution model showed that a minimum stoss side length was required to develop a strong flow separation zone. Copyright © 2014 John Wiley & Sons, Ltd. KEYWORDS: river dune evolution; dune splitting; morphological modelling; bedform superimposition Introduction Hydraulic roughness values play an important role in the prediction of water levels in river systems (Casas et al., 2006; Vidal et al., 2007; Morvan et al., 2008), which is critical for flood management purposes. While many improvements have been made in the field of hydraulic modelling, there is still much uncertainty regarding the roughness values of the main channel and floodplains (Warmink et al., 2007, 2013). The hy- draulic roughness of the main channel is mainly determined by bedforms on the river bed. In many rivers, river dunes are the dominant bedforms; they form in beds with sediment sizes ranging from medium sand to gravel (e.g. Van den Berg and Van Gelder, 1993; Kostaschuk, 2000; Best, 2005). River dunes are rhythmic patterns caused by the interaction of the turbulent flow with the sandy bottom. The height of river dunes is in the order of 10 to 30% of the water depth and their length (distance between two consecutive crests) is approximately 10 times their height (e.g. Van Rijn, 1984). They migrate downstream and are asymmetrical, with mild stoss side slopes (around 5°) and steep lee side slopes, often reaching the angle of repose (about 30°). Recirculating eddies can develop at steep lee sides of dunes, resulting in a flow separation zone; the flow at the lee surface of these dunes is in opposite direction of the mean stream flow (see Figure 1). River beds are highly dynamic dur- ing floods: lengths and heights of dunes increase and decrease during floods as a result of the changing flow conditions. In the past, many approaches have been taken to model dune dimensions, ranging from equilibrium dune height predictors (e.g. Yalin, 1964; Allen, 1978; Van Rijn, 1984) to various forms of stability analyses (e.g. Kennedy, 1963; Engelund, 1970; Fredsøe, 1974; Yamaguchi and Izumi, 2002). Some models try to capture all hydrodynamic and sediment transport character- istics in great detail (Nabi et al., 2013). Such models are valu- able for studying detailed processes, but are computationally intensive. Simpler models exist that calculate the turbulent flow field over bedforms, in some cases in combination with morphological computations or parameterizations of important processes (e.g. Shimizu et al., 2001; Nelson et al., 2005; Tjerry and Fredsoe, 2005; Giri and Shimizu, 2006; Paarlberg et al., 2007, 2009; Shimizu et al., 2009; Nelson et al., 2011). EARTH SURFACE PROCESSES AND LANDFORMS Earth Surf. Process. Landforms 39, 1208–1220 (2014) Copyright © 2014 John Wiley & Sons, Ltd. Published online 12 February 2014 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/esp.3529