Hydraulic modelling and flood inundation mappingin a bedrock-confined anabranching network:the Mekong River in the Siphandone Wetlands, Laos

摘要

Anabranching fluvial networks recently have become the focus of attention fromenvironmental specialists, especially in the hydraulic field. Anabranching networks can befound in different physical environments; however, the hydraulic and geomorphologicalnatures of such river networks are still not well known leading to on-going discussions on thedefinition and nature of the networks. Even though, alluvial anabranching networks generallyhave common features like vegetated islands, low water surface slope and stable channelplanform, bedrock-confined anabranching networks also have their own characteristicsinherited from the geological and structural controls imposed on the single channels thatcompose the network complex.This thesis focuses on the provision of a benchmark describing the bulk hydrauliccharacteristics of a large bedrock-confined, anabranching river network, located withinsouthern Laos. The network can be separated into: (i) the upper river network constituted bytwo bifurcations and one confluence with an interpolated bathymetry based on soundings ofcross-sections along the navigation channels; and, (ii) the downstream river networkcharacterised by a complex anabranching network with five bifurcations and five confluencesfor which there is no bathymetric survey.The river network as whole is a ‘composite’ – partly bedrock (especially the channel-bed)and partly alluvial-filled and as such it does not accord fully with any prior description orclassification of anabranching channel networks (e.g. Huang and Nanson, 1996). Tounderstand the hydraulic nature of the river network, the energy approach in a onedimensional(1D) steady-flow hydraulic model (HEC-RAS) was applied to the network.Significant challenges arose due to the lack of boundary conditions throughout the model,namely: (i) unknown splitting discharge ratios at each bifurcation; (ii) partly non-surveybathymetry; and, (iii) ungauged downstream boundary condition of one of the channeloutlets. To determine the discharge entering each channel, the splitting discharge ratio at eachbifurcation was defined originally by the ratio of the cross-sectional area of the first crosssectionof each downstream channel and then adjusted based on the Flow Optimizationfunction in HEC-RAS to minimize any rise or drop of the modelled water surface around ajunction. For the channels with non-surveyed bathymetry, a SPOT satellite image wasprocessed to construct a pseudo-bathymetry showing a range of elevations, including shallowand deep portions of channels, rather than detailed bed elevations as would be obtained froma measured bathymetry. To define the boundary condition of the ungauged channel outlet, thewater surface elevation was interpolated and validated according to predefined assumptions(i.e. the water surface slope along the ungauged channel was interpolated according to theavailable DEM and cross-sectional width extracted from a SPOT image for low dischargeconditions was assumed to be similar to the gauged channels for flooding discharges).In general, the study has helped to develop methods to model the complex river network withdata constraints (i.e. the boundary conditions). The findings include: (i) the developedpseudo-bathymetry based on a SPOT image is useful to model a large river network using theenergy approach in a 1D hydraulic model in which the cross-sectional area is important inmodelling the bulk hydraulic parameters but the influence of the cross-sectional shape issubordinate; (ii) the in-channel hydraulic roughness coefficient at each cross-section may besignificantly different from neighbouring values due to the variation in the local bedrockroughness and the roughness of intervening alluvial reaches; and, (iii) the hydraulicroughness of the riparian land cover along the floodplains does not contribute noticeably tothe modelled stage along the river network nor to the planform extent of flooding foroverbank flooding discharges. Rather, changes in land-cover, and hence the riparianroughness, are registered as small, but measureable, changes in the local velocity over theriparian floodplain and in the average in-channel velocity.Citations:Van, P.D.T., 2009. Hydraulic modelling and flood inundation mapping in a bedrockconfinedanabranching network: The Mekong River in the Siphandone wetlands, Laos.Unpublished PhD thesis submitted to the Faculty of Engineering, Science and Mathematics,University of Southampton, England.