IntroductionRiver channel networks transport both water delivered to them by catchment runoff processes and sediments acquired by various erosion processes as the runoff occurs over slopes and in the channels themselves. These networks are dynamic systems with branching structures that exhibit a high degree of complexity, but also regularity and organization; this spatial and temporal organization within river basins emerges from a large number of interconnected physical and biological processes. Sediment delivered to the river network, and then routed through it, is thus derived from many different sources within a catchment, including areal production from hillslope soil erosion (both sheetwash and gully erosion) and point and line sources, such as mass movements and riverbed and bank erosion. During transport, this sediment may be partly and temporarily stored en route, in sediment sinks such as colluvium on the lower-gradient hillslopes, alluvium (in the floodplain and river terraces) and in lakes, reservoirs and estuaries.Models developed to account for sediment transport at the catchment and river network scale have varying degrees of complexity. The simplest consist of empirical
River Confluences, Tributaries and the Fluvial Network Edited
338CH 16 SEDIMENT DELIVERY: NEW APPROACHES TO MODELLING AN OLD PROBLEM relationships for sites within the network for which sediment-yield and stream-flow data are available; these include methods combining sediment-rating relationships and flow-duration curves (Gregory and Walling, 1973;Crawford, 1991;Horowitz, 2003), and those based on surveys of lake and reservoir sedimentation (Verstraeten and Poesen, 2000). These empirical relationships based on site-specific observations and data may be useful for the particular site where the data were collected, but extrapolation to other catchments is unlikely to be generally successful. Semi-empirical relationships have been derived between sediment yield and variables defining climate forcing (rainfall properties) and catchment morphology, such as catchment area, soil type, geology, topography, vegetation and land-use management (Milliman and Meade, 1983;Restrepo and Kjerfve, 2000). Both the empirical and semi-empirical models contain weak representations of transport mechanisms, lack explicit evaluation of sediment sources, and focus on temporal variability while ignoring both the spatial heterogeneity of sources and sediment-routing processes. The earliest spatially distributed models of sediment yield were multivariate linear regression models (Anderson, 1954;Anderson, 1957;André and Anderson, 1961). These models were later developed as grid-based methods involving systematic inventories of land-surface attributes in ungauged basins (Solomon and Gupta, 1977;Cluis et al., 1979; and see below).The second category of model seeks to build on fundamental hydrologic and hydraulic processes, where the separate effects of climate forcing, catchment conditions and anthropogenic influences can all be identified (e.g. Flanagan and Nea...