Effectively assessing landscape pattern characteristics and predicting their dynamics have been a basic prerequisite for more reasonably regulating and managing forest landscape, and maintaining landscape security patterns. In this study, based on three U.S. National Land Cover Databases (1992, 2001 and 2006), the landscape mosaic indicator in combination with the Markov model was adopted to analyze forest fragmentation patterns and changes in the characteristics of spatial interactions between forests and other land use types in Oregon State, USA. The results showed that conversion from the developmentdominated type  D to the single development type DD in landscape mosaic model had the highest transition probability 0.319, indicating that urbanization has been the major force responsible for the change of regional landscape patterns. In the forest security model, the highest rates of forest loss occurred in agriculture and the developed landscape mosaic type (ad), showing that in the development and agriculture dominated landscapes, encroaching upon forests was at the highest likelihood. The areal percentage of forest over the total study area was less than 50% when reaching a steadystate distribution, with an accelerating rate of forest fragmentation, and the landscape spatial distribution tended to be a mixed landscape pattern. The Kappa coefficient between the simulated values and the observed values from the 2006 landscape mosaic model was estimated at 0.82, indicating this model had a high precision. However, the accuracy of the forest security model was poor, with a Kappa coefficient of 0.21.