Landscape pattern analysis (LPA) is an important topic of landscape ecology. The ultimate goal of LPA is to link spatial patterns of landscape with ecological processes, and detect status of processes using landscape pattern information. By analyzing spatial arrangement of relevant factors about landscape, LPA can obtain information to predict dynamic pattern of ecological processes. Landscape metrics can represent the spatial distribution of landscape and it has been used as a common tool in LPA. In the past decades, with the development of landscape ecology theories and spatial information technology, many types of landscape metrics were developed. Soil erosion is a globally critical material transport process across earth surface. It is inherently impacted by spatial arrangement of source and sink areas, flow paths and impedances of runoff and sediment transport. To account for it, diverse landscape metrics have been widely applied to delineate the relationship between landscape pattern and soil erosion. In general, a perfect landscape metrics should logically reflect soil erosion process and coordinate well with process variables. However, published literatures indicated that many common landscape metrics did not have specific correlation with soil erosion variables. Consequently, there is lack of proper interpretation for most of the common landscape metrics based on mechanism of how landscape pattern affect soil erosion process. Besides, some inherent limitations of landscape metrics exist in linking landscape pattern with soil erosion. In this review, twelve common landscape pattern metrics concerning landscape connectivity, diversity, edge/patch density and shape characteristics were presented. Their implications and limitations in linking landscape pattern with soil erosion process and indicating soil erosion status were explicitly elaborated. As most of landscape metrics were not developed on the basis of soil erosion mechanism, there were lack of specific relationship between landscape metrics and variables expressing soil erosion status. The properties of landscape data, inherence of metrics and complexity of soil erosion process across different scale caused limitations of applying landscape metrics in soil erosion research. The commonly used landscape analysis data only focused on land cover or land use type, but ignored spatial arrangement of different functionally topographical elements in soil erosion. With respect to inherence of landscape metrics, most of them were only statistical expression of geometry characteristics or spatial distribution of land use or land cover units, which lacked of ecological meaning. For those developed based on other ecological processes, they were not perfectly suitable to indicate erosion process. The combination of above three limitations made the uncertainty and inability in using metrics to describe relationship between landscape pattern and soil erosion processes. Therefore, detecting soil erosion status under specific landscape pattern by common landscape metrics was unreliable. In summary, ignorance of physical mechanism of soil erosion process was the main reason for the limitation of employing landscape metrics to evaluate soil erosion probability or risk across landscapes. Development of landscape metrics based on soil erosion process is necessary. It is a new prospect in studying interaction between landscape pattern and soil erosion.