DEM Based Multi-Directional Flow Path Mapping Using the Raft Method Roshan Shrestha 1 , Yasuto Tachikawa 2 and Kaoru Takara 3 1 Graduate School of Urban and Environmental Engineering. Kyoto University, Japan 2 Associate Professor, DPRI, Kyoto University, Uji, Kyoto 611-0011, Japan 3 Professor, DPRI, Kyoto University, Uji, Kyoto 611-0011, Japan ABSTRACT: Digital elevation models (DEM) are extensively used in hydrological analysis to obtain the direction of flow on a topographical surface. Knowing the direction of flow helps in determining channel networks, and in obtaining the distributed specific catchments, which are important hydrological attributes in a DEM based analysis. There are certain demerits in the known methods of DEM-based flow-path mapping, despite the ease of handling the DEM in present generation computational capacity. A major demerit is the problem of accounting multidirectional flow path mapping efficiently. In this study, a new approach of assigning the flow direction is proposed. This approach considers the flow as a vector component and argues that the direction of flow can be different from the direction of steepest slope on a planar surface depending on surface roughness and flow vectors. The direction of flow, if assigned to the direction of the maximum flow tendency, gives a better flow path mapping. In this case, the direction is not necessarily restrict either on grid or across diagonal of the grid-centers, which enables to assign the flow direction toward multiple downstream cells. This method is named as the Ranked Flow Tendency (RAFT) method. This method was tested using the DEM of Kamishiiba Reservoir Site (210 km2) as the case. The results show the RAFT method is better than the conventional D8 method and modified D8 method. The RAFT method identifies the network of channels and lakes and can incorporate the flow dispersion properly. These abilities show higher possibility of yielding a better result while used this method in sediment and pollutant flow simulation and water quality modeling. 1 INTRODUCTION Many studies in hydrological science use Digital Elevation Model (DEM) for describing the catchment topography and obtaining topographical attributes. A network of flow channels, which the DEM can produce easily, is the most widely used product of DEM in hydrological studies. In addition to that, the DEM provide the basic data for calculating the upslope area, specific catchment, and slope driven parameters in flow routing and sediment and contaminant movements (Beven et al. 1984;Beven and Kirkby 1979;Costa-Cabral and Burges 1994;Moore et al. 1991;Tarboton et al. 1991;Wood et al. 1990). It is therefore understandable the significant importance of the method, which calculated the flow direction from the DEM. The earliest and simplest method for specifying the directions of flow in a catchment using the DEM is the D8 method (O'Callaghan and Mark 1984). This method assigns the direction of flow same to the direction of steepest downward slope. For each cell the direction is assigned toward one of its eight neighbors. This method has been widely used in hydrology for flow direction mapping and to evaluate hydrological attributes Several disadvantages and limitation are reported while using D8 method (Costa-Cabral and Burges 1994;Moore et al. 1991;Tarboton 1997). Alternatives to D8 method have been investigated and tried by several researchers. Following methods can be listed as the sequential advancement in this direction. Multiple flow direction method (Quinn et al. 1991) (usually termed as MS method) recognizes Multiple Slopes (MS) to allocate fraction of flow proportional to slope downstream. Another similar method presented by (Freeman 1991) uses the slope to an exponent. The MS methods are criticized for drawbacks of too much dispersed flow. Associating a probability which refers to an aspect angle as same as that of expected flow direction, Rho8 method (Fairfield and Leymarie 1991) is suggested to obtain the flow direction, which is criticized on its disability to reproduce the result and random wiggles. Improved Rho8 method (Lea 1992), called Lea’s method, has routed the flow as a rolling ball released on a plane from the center of each grid cell. DEMON (Costa-Cabral and Burges 1994) has advanced the Lea’s method. Both Lea’s method and DEMON are questioned on plane fitting technique that may mislead the determined flow direction. A triangular facet based D∞ method (Tarboton 1997) calculates steepest descent for the triangular facets same as that of triangular irregular network (TIN) surface slope (Tachikawa et al. 1994) by constructing facet from DEM (Douglas 1986). The distribution weights are evaluated on the basis of surface angle of the steepest Proc. of Monitoring, Prediction and Mitigation of Water-Related Disasters (MPMD2005), pp. 85-90, 2005