To assess the influence of long-term fertilization on weed communities of early and late rice crops, the weed species composition was investigated in experimental plots initiated in 1981 at the Key Field Experimental Monitoring Station of the Reddish Paddy Soil Eco-Environment in Wangcheng, China. The treatments were (1) a control (CK), no fertilizer; (2) N–P, no K; (3) N–K, no P; (4) P–K, no N; (5) N–P–K; (6) N–P–K + Ca, N, P, and K plus lime; (7) N–P + S, N and P plus additional rice straw return; (8) N–P–K + S, N, P, and K plus additional rice straw; (9) N–K + M, N and K plus swine manure. The results indicated that weed flora composition and density were influenced by the different fertilization treatments. Multivariate analyses indicated that changes in the weed community composition were primarily due to soil-available N, followed by light intensity on the field surface, and soil-available P. More weed species and total weed density were observed in the control and P–K plots than in plots in which N, P, and K were applied together. Omission of N application had a greater effect on the weed community than the omission of P or K applications. Nutrients derived from synthetic fertilizers and organic manure or the additional application of lime had no obvious effect on the weed community of late rice crops.