The sediment transport data are summarized in Figure 5. The sediment transport is given as well as the D₉₀ of the transported material. A hysteresis can be observed: the transport at the beginning of a series is lower than at the end of a series for the same flow velocity. Photos of the flume bed are given as well.

Photos T0, T2, T3b, T4b, T5, T7, T9 and T10. Photos of the bed taken after the experiments. Arrows denote the flow direction.

The observed bedform types are indicative of the transport condition as was argued above. The larger bedforms may modify the flow and thus influence the bedload transport rate in turn. In the presence of armour layers this system becomes more complicated. During low flow the armour layer almost inhibits bedload transport and only a very small portion of the bed is covered by bedload sediment in sand ribbons, bedload sheets or barchans. At rising stages the armour layer remains unbroken up to a certain point. In less sandy rivers this point is determined by the threshold of motion of the imbricated particles and pebble clusters in the armour layer. In more sandy rivers the bedload sediment may accumulate in larger (isolated) barchans. The turbulence in the troughs of these bedforms may break up the armour layer. As soon as the armour layer is broken or mobile, more sand becomes available and the bedforms may eventually grow together into sand dunes and in extreme cases sand gravel dunes.

In Figures 6 and 7 the bedform types that occurred are given as a function of depth-averaged flow velocity and bedform height. The flow parallel ribbons occurred during the experiments which had the lowest flow velocities: T0 and T2. The sand in the ribbons migrated downstream in the form of small ripples, which sometimes had barchanoid forms. Between the ribbons the bed surface consisted mainly of gravel. Barchans occurred in experiments T3b, T4a, T5 and T10. However, the barchans in T5 tended to grow together although the barchanoid form was maintained. Dunes occurred in experiments T4b, and T7-T9.

In experiment T4b the flow velocity and the bedform height are lower than in T4a. The bedforms in T4b became lower as a response to the lower flow velocity . Thus sand became available for extension of the width of the bedforms, which lead to a transition from barchans to dunes (see Figure 8). The same principle lies behind the difference in bedform type between T5 and T9, which have equal flow velocities. In T5 there was not enough sand available in the beginning for the formation of dunes, so the type was barchanoid. However, in T7 much more sand was in transport because it was winnowed from the bed below the armour layer. In T9 the sand of the large dunes of T7 was redistributed into the lower dunes of T9. The deposition of bedload sediment in T9 was so large that the armour layer of T7 was buried.

This picture of bedform type transition (see Figure 8) is slightly simplificated because there is no strictly defined boundary between the two classes barchans and dunes. The transition is much more gradual, as was observed in the flume experiments, from barchans to barchanoids, to barchanoids with increasing slipface lengths, to dunes with barchanoid characteristics like crescentic slipfaces and tails, to dunes with irregular slipfaces, to more or less two-dimensional dunes.

During experiments T5 and T7 a lowering of the bed level (measured in the troughs of the bedforms) of a few centimeters was observed. In experiment T9 the bed rose again. Since the flume is a closed, recirculating system, it is expected that the volume of sediment entrained from the bed is equal to the volume of sediment present in the migrating bedforms. To calculate this, the total volume of sediment in the bedforms was averaged over the length and width of the flume to calculate the average thickness of the bedload layer. The experiments with the largest bedforms (T5-T9) are therefore selected for the calculation of bedlevel change and transport layer thickness. In table 2 the lowering of the bed and the thickness of the bedload layer are given, compared to the originally installed bed level before experiment T5. The bedlevel change and bed load layer thickness are comparable in magnitude, which means that the bed level change in the experiments indeed can be explained with the amount of sediment transferred to the transport layer.

Table 2: Comparison of bed level change and the corresponding thickness of the bedload layer.

The cores that were taken of the bed were analyzed in thin layers, which allows the investigation of the FU sequence. In Figure 9 the percentages of gravel (> 2 mm) in the bedform and armour layer is shown as a function of depth below the bedform top. The curve of sand content would mirror the gravel curve, since the sum of both is 100%. It can be seen that the gravel content increases towards the base of the bedforms and reaches a local maximum at the armour layer at the base of the bedforms. In T5 the coarsest gravel fractions were not mobile, which allows a coarse armour layer to be formed. In T7 the coarsest fraction was more mobile, which lead to a much less pronounced armour layer. In T9 the buried armour layer can be discerned below the bedform trough level. The coarse material of the old armour layer and old bedform base is buried, which lead to a relatively sandy base of the bedform, and therefore no real armour layer was present. Below the base of the bedforms the cross-bedding and armour layer of T7 is preserved.

Summarizing, in the transport process during rising stages, bed sediment is transferred to the transport layer, but during waning flow stages the deposited bedload sediment is not transferred back into the bed below the armour layer. Instead the sediment is redistributed in the bedforms, resulting in a bedform change from barchan to barchanoid or dune. If (the coarser) part of the bedload sediment is deposited on the bed, it is on top of the armour layer but not below the armour layer. The bed thus becomes vertically sorted, and a sedimentological record is preserved of the 'discharge wave' in the flume (see Figure 9).

Furthermore, the bedforms in T5 (before T7 with the largest flow velocity) were smaller than in T9 (after T7) and had less sand in the bedload sediment, leading to smaller and coarser bedload rates than in T9. This partly confirms the hypothesis of anti-clockwise hysteresis and fining of bedload transport during a discharge wave. However, T5 was started with a fully mixed bed and not with a vertical sorting history of bedforms in a previous high discharge experiment.