Abstract: In order to solve I/O bottleneck in vector line visualization of large scale flow field, we raise two new data prefetching methods. First partition flow field to blocks; Then construct particles flowing probability graph based on the statistic results of particle flowing in data block and blocks boundary to prefetch data blocks needed in vector line computing. The results show that the new data pre-fetching method can promote 10%~25% in hitting ratio and data satisfying ratio, and can promote the efficiency in vector line visualization.