Abstract: Due to global climatic change, increase in population, and scare of fresh water, forest water conservation play an important roles in interception, storage and redistribution of precipitation. Large-scale spatial estimations of forest water conservation have been mostly studied by using hydrology or remote sensing models, but there is a lack of spatial estimation of forest water conservation at a large scale based on observed data. This paper proposed the framework of the spatial prediction for forest water conservation at a large scale based on observed data at sites, and then investigated the factors affecting canopy interception, litter water-holding, soil storage and forest water conservation. Further, the methods of spatial simulation based on observed data were introduced, including regression model, machine learning and geostatistical methods and the combination of multiple methods. In addition, the problems, including method for measuring forest water conservation and the process of spatial simulation for forest water conservation based on observed data, were summarized. Finally, the future research orientation of forest water conservation in a large-scale region has been presented on the bases of observed data at three aspects, including selection of auxiliary variables, spatialization of different regions with different models and spatial analysis of dynamic forest water conservation in a large-scale region.