The loss of natural habitat and habitat fragmentation, caused by climate change and human activities is a seriously threat global biodiversity. Maintaining landscape connectivity of natural habitats is one of the most effective methods for alleviating these related problems. Landscape connectivity describes the degree to which the landscape facilitates or impedes ecological flow and movement among habitat patches. By integrating information related to landscape structure and ecological processes, landscape connectivity describes the effects of habitat fragmentation on the organisms living within the landscape. Graph theory provides powerful tools for assessing landscape connectivity, and has become a popular research method in recent years. However, some issues related to the application of graph-based methods still need to be discussed and analyzed, such as choosing a proper threshold distance and gaining an understanding of various connectivity indices. This study used the graph theory method to assess the landscape connectivity of natural forest in Minqing County, Fuzhou City, China. A comprehensive method was utilized to assign a reasonable threshold distance, which consisted of three steps:namely, determining the range of the maximum dispersal distance of a focal species, enumerating a gradient distance value, and an experiment using link thresholding. Six indices were used to assess the landscape connectivity of the study area:number of links (NL), number of components (NC), Harary index (H), area-weighted flux (AWF), integral index of connectivity (ⅡC), and probability of connectivity (PC). We harvested data related to the natural commonweal forests of Mingqing County from the data of 2010 National Forest Resource Inventory using ArcGIS 10.2 to define habitat patches in the study area. The six landscape connectivity indices were calculated using Conefor Sensinode 2.6 software. Using the comprehensive method, 1 km was found to be the optimal threshold distance that best presented the structure of the landscape. A medium-level structure, consisting of 56 components was detected using the NL and NC indices. Based on the binary (H, ⅡC) and probability (AWF, PC) indices, the level of landscape connectivity of the natural forests in Minqing is comparatively low. The patch importance indices (dIs) aided in identifying the most important patches that could potentially affect the overall level of connectivity. We determinded that the main component provided 90% of the connecting functions in maintaining the landscape connectivity; however, losing a single critical patch could significantly reduce overall connectivity. The results indicated that this new method involving evaluation of the threshold distance could help researchers determine more accurate distance values for landscape connectivity, and therefore, this method would be applicable for both scientific research and planning. In addition, binary and probability indices could be used together to reveal landscape structure and the spatial patterns of patch importance. In our analysis, we found that the values of dI are generally proportional to the area of patches; however, patch area is not the dominate element that determines the dI values of the most critical patches. Other patch features, such as patch position, may have greater influence on the dI values for the extremely important patches. This may be caused by a commonly used but inaccurate method of evaluating patch attributes. This topic required further investigation and improvement in future studies. In conclusion, this study will help promote the application of graph theory in research studies related to landscape connectivity, and can provide guidance for land managers tasked with conserving biodiversity.