Abstract: Space, grade, flow and spatial relation are basic analytical conception of classical geographic process. The study shows that it has similar expressing mechanism between classical geographic process analyzing theory and cellular automata(CA) in four aspects as follows: (1) Cellular automata model express the concept of geographic space by cell, state, cellular space, etc. (2) Cellular automata represents the concept of geographic grade by the concept of neighbor. The essence of both the rule of neighbor and geographic grade is the selected object of spatial interactions, their difference is that the rule of neighbor is a kind of selective way based on geometry feature and geographic grades is a kind of selective way based on both geometry feature and non-geometry feature. (3) Cellular automata model express the concept of geographic flow effectively by local evolvement rule of cellular state, the direction of cellular flow depends on neighbor rule, and the strength of cellular flow depends on the distance between cells strongly, which is similar with the attenuation of strength of spatial interaction with distance. (4) The spatial relation of cellular automata is expressed mainly by framework and figure of cell that hides and includes spatial relation of both geometry and non-geometry. The framework and figure of cell express the absolute position of cellular space, relative position, distance, direction and topologic relation by geometry feature, and express the spatial distributing and combination of cellular state, attribute relation between spatial object. Therefore CA has ability of simulating geographic process. But as a more widely and abstractly model CA has some shortcomings of expressing geographic characteristic. The defective substance is lack in expressing geographic feature of cell, and is not able to describe the deeper geographic attribute information. It can be showed in four aspects as flows: the first is the describable restriction to geographic space cell, and not only it is not able to describe the higher spatial position, but also it is not able to describe the other geographic position that belong to attribute meaning. The second is the describable restriction to local spatial relation based on the concept of neighbor, by which local spatial interaction and spatial relation can not be represented truly because it define neighbor based on just geometry feature. The third is the restriction of presupposition that the medium of cellular flow is homogeneous. The fourth is the restriction of local evolve rule. The evolve rule of cellular state is applicable to all cells, in fact this needs an assumption that all geographic cells follow the same rule. The paper attempt to bring the geographic feature concept including geometry and non-geometry attribute into cellular automata and expend cellular automata in three aspects including cell, neighbor model and local evolvement rule: (1) selecting cellular entity concept to replace cell concept to promote the conceptual description of geographic space; (2) selecting neighbor by integrating geometric and non-geometric feature; (3) bringing attribute control into local evolve rule of cell. It makes studying flow in the condition of non-homogeneous spatial medium possible that attribute control is brought into extended-CA. Finally, the paper make a case study on land use evolvement of Shenzhen City, which draw conclusions that extended-CA model has great applied value.