As a two-phase anisotropic mixture, debris flowshows some complex fluid-dynamical characteristics on its motion behavior, which makes it difficult to be modelled or simulated through standard approaches. Consequently, Cellular Automaton (CA) model in the field of parallel computing, which has long been verified as efficiently applying in the simulation of complex natural process, are recently introduced to simulate flow-type phenomena. In this paper, the components of CA model for debris-flow simulation are reviewed, after that a two-dimensional cellular space is generated from the Digital Terrain Model (DTM) with 2.5m high-resolution, and the relationship of lattices in the space is defined as Moore neighborhood type. A new transition function, aiming at flow direction determination, is proposed through the way that implementing the debris-flow inertial influence into the traditional topography-based D8 algorithm, by a multiplying equation or an addition equation. We also present a three-step propagation algorithm to integrate the single flow routines, then use the constant discharge model to assign the flow depth to each routine, in this way the propagation area can be gradually delineated through each increment step. We test the developed model with CH87 Brichboden debris-flow event, and illustrate its application withYohutagawa debris-flow event in Japan, 2010. Results of both cases show that a more accurate propagation perimeter pattern is observed using addition equation in the modified flow direction algorithm, comparing to the traditional D8 algorithm and multiplying equation-based algorithm.