The study of fishery community ecology depends on quality and quantity of data collected from welldesigned sampling programs. The optimal sampling design must be costefficient, and the sampling results have been recognized as a significant factor affecting resources management. In this paper, the performances of stationary sampling, simple random sampling and stratified random sampling in estimating fish community were compared based on computer simulation by design effect (De), relative error (REE) and relative bias (RB). The results showed that, De of stationary sampling (average De was 3.37) was worse than simple random sampling and stratified random sampling (average De was 0.961). Stratified random sampling performed best among the three designs in terms of De, REE and RB. With the sample size increased, the design effect of stratified random sampling decreased but the precision and accuracy increased.