As forest water conservation has become one of critical ecosystem service issues in context of global water shortage, the exploration for its quantitative methods has increasingly gained more attention. The relationship between forests and water is complex. Different geographical characteristics and forest types have various effects on the interception of rainfall redistribution in forest, surface runoff and evapotranspiration. There exist interactions among the complex hydrological and ecological processes in forest ecosystems, thus causing the spatial and temporal variations in water cycles and water balance at the watershed and landscape scales. However, the dominant factors and driving mechanisms of forest water conservation key processes at these different scales are not yet clear，making it difficult to evaluate forest water conservation at these scales and develop a scale-up method for quantifying forest water conservation. The modeling method is an efficient route towards quantitative evaluation of forest water conservation. Over the past decades, most of the quantitative researches concerning forest water conservation have focused on one dimension methods (e.g. methods of water balance, soil water storage capacity, precipitation storage, canopy interception surplus and so on). Scaling-up remains as a challenging issue in quantifying forest water conservation. In this study,the author employes the basic theories of cellular automata to developing a new modeling method for quantifying forest conservation. The model, we integrate the hydrological characteristics of rainfall redistribution by canopy, water retention of litter and water conservation of soil and define a cell as an elementary spatial unit that is rasterized from a hydrologic response unit, with every cell interior having relatively consistent characteristics of the vegetation, soil and terrain attributes. A cell neighborhood is defined as the Moore type. The range of quantitative change of water is measured in continuous real numbers. The water quantity in each cell is affected by neighboring cells and status of the previous time step. Based on the response characteristics of the underlying surfaces on different rainfall intensity and duration, the cellular responses to different rainfall events is classified by two scenarios: generation of runoff and without generation. The transformation rules of forest hydrological processes are also defined, including evapotranspiration, intercellular water transmission and deep seepage. And the hydrological processes with both horizontal and vertical directions are also considered. Taking main factors affecting forest water conservation and its spatial heterogeneity into consideration, the model overcomes the defects of most traditional methods which focus on one dimension. This new modeling will provides a useful way for solving forest water conservation scaling-up issues from small-scale (hydrological response unit) to watershed, or landscape scale; and thus improves quantitative evaluation of forest water conservation at different spatial scales. The model also provides an effective quantitative way of quantifying forest water conservation with different spatial distribution patterns of different forest types. This paper is the theoretical framework of the model described above. A case study will be presented in a follow-up paper.