Taking the Wangjiaqiao watershed as the study region, the landscape patterns of the watershed were extracted based on remote sensing images over five periods (1995, 2000, 2005, 2010, and 2015). The temporal-spatial evolution characteristic of the source-sink landscape patterns were analyzed using transfer matrixes with the Markov model. In addition, the relationship between landscape pattern indexes and sediment yield were established based on field measured data of the Wangjiaqiao watershed for many years, and the influence of landscape patterns on the sediment yield of the watershed was analyzed. The results concluded that temporal and spatial variations in landscape patterns were obvious in the Wangjiaqiao watershed from 1995 to 2015, showing a fragmentation trend of development. The area of source landscape continued to decline, while the area of sink landscape continued to increase. Under the condition that the time-series curves of precipitation are similar, the landscape index at patch scale had a higher interpretation ability to erosion sediment yield than at landscape scale. At patch scale, the multiple correlation coefficients between sediment yield and landscape indexes of source and sink were 0.946 and 0.903 respectively, showing a significant correlation. The Proportion of Like Adjacency (PLADJ) and Patch Cohesion index (COHESION) were important landscape indexes affecting sediment yield in the watershed for the source landscape. Both PLADJ and COHESION were positively correlated with the sediment yield, and the sediment yield increased with the increasing value of the above two indexes. As for the sink landscape, Patch Density (PD), Edge Density (ED), and the Landscape Shape Index (LSI) were important indexes of soil erosion and sediment yield. Both PD and LSI were negatively correlated with sediment yield, and the sediment yield decreased with an increasing value of the two indexes. At patch scale, the influence of temporal and spatial variation of landscape patterns on sediment yield was significant. These results could provide references for establishing the relationship between variations in landscape patterns and responses of soil erosion to other watersheds.