Search

The model takes an explicit solution of the 2D diffusion wave equation, but introduces processes to represent both wetting and drying and, for situations where no channel data are available, a significantly simplified channel representation. The goal of the paper is to assess the extent to which this simplified channel representation is sufficient to provide first estimates of inundation extent. The channel simplification was based upon estimation of how much flow is conveyed in the portion of river channel not represented in the DEM (i.e. below bank top) and to remove this from the inflow hydrograph, such that the flow input represents the overbank flow only. This requires an estimate of bankfull flow and a method of partitioning the actual flow between that which is delivered to the floodplain and that which is conveyed by the channel. The simplification was based on the principle that it has a characteristic return period (QMED) which can be applied to a magnitude-frequency relationship for a given river reach and hence used to determine a first approximation of bankfull flow. The Weighted Divided Channel Method was then applied to the bankfull flow to work out the residual flow delivered to the floodplain. A user defined ratio (Qratio) was used to allow control of the weighting. Validation was based upon application to: (a) simple topography where comparisons could be made with basic calculations and/or other models; and (b) the River Thames at Buscot, U.K., which has a validation dataset based upon synthetic aperture radar data. The results show that JFLOW is capable of simulating idealised, largely 1D floodplain flows, but the real strengths lie in its ability to simulate also topographically-driven flow convergence and divergence where flow is more 2D: whilst some floodplains undoubtedly act as a single, two-stage channel, many of the more extensive and topographically-complex floodplains exhibit river-independent flow paths and/or ponding. The analysis showed that the simplified channel representation gave goods results: the greatest uncertainty was associated with the estimation of the bankfull flow itself rather than the weighting given to the flow apportionment. It was noted that deviation from QMED would be found in certain situations, notably in the presence of river defences, engineered channels and structures and clearly the channel representation should take the known bankfull flow rather than QMED in those situations.

This paper describes an application of Large Eddy Simulation methods to the flow in a laboratory-style confluence of parallel channels of unequal depths (Reynolds No. ≈ 13,500), and a natural river channel confluence that also exhibits pronounced bed discordance (Reynolds No ≈ 100,000). The aim is to investigate the formation of periodic flow features inherent in the flow dynamics in river confluences. The laboratory-style confluence predictions are compared with experimental data, and suggest that the simulation was capable of capturing some of the key flow periodicities. The model provides detailed flow and mixing data which help inform understanding of the production and evolution of large scale turbulence features in confluences.