Ecological risk assessment of wetlands is of great significance to protect natural resource and promote regional planning. Based on four phases of land use data in 2000, 2005, 2010, and 2015 in Sanjiang Plain, this study intends to construct an ecological risk assessment model of wetlands from the aspects of risk sources considering various human activities such as land use change and road construction caused by urban expansion, and risk receptors including landscape ecological pattern and ecosystem services value. To analyze the spatial agglomeration pattern of ecological risks, the distance threshold was used to determine the spatial distance weight, and the bivariate spatial autocorrelation model was applied to evaluate the spatial agglomeration and distribution characteristics of ecological risks at different time scales. Results show that from the perspective of risk sources, there is an increasing trend in the Sanjiang Plain. The Songhua River, Muling River, and Woken River areas have always been at a medium-to-high risk level. From the perspective of risk receptors, the medium to high risk areas of landscape ecological risk are concentrating on wetlands and water body areas. Low-value areas of ecosystem services are mainly distributed in the central paddy fields, dry fields, construction land, and small-scale wetland areas in the northeast and southeast. A trend of centralization in areas with medium to high risk levels resulting from wetlands, water bodies and other high-value areas of ecosystem services are occupied. The ecological risks of wetlands exist an obvious increasing trend on the time scale. The results of comprehensive ecological risk show that high-risk areas increased from 2000 to 2015, spatially scattered from the Songhua River to the Muling River. The intense of the ecological risk is affected by the spatial distance. The larger the distance threshold, the larger the relevant area centered on a certain grid, while the overall correlation index decreases. Through the comparison analysis of the adjacency matrix and the distance matrix, 5 km was selected as the distance threshold for local autocorrelation analysis. It is significant to represent a positive correlation spatial pattern of land use and ecological risk. The areas with high-to-high land use and ecological risks are mainly concentrated in the Songhua River Basin and surrounding beach areas in the study area. With the change and transfer of land use, the spatial correlation gradually strengthens and the regional distribution continues to expand. From the aspects of human activities, landscape pattern and ecosystem service value, this study provides a theoretical basis for ecological risk management in the Sanjiang Plain.