Ensemble simulation is increasingly popular in scientific domain such as climate research, weather report, mathematics and physics. Ensemble simulation data sets are usually multi-valued, multi-variate, time-variant and large in scale. Thus, analyzing such data sets is challenging. Ensemble visualization helps scientists to compare ensemble members and give overall summary to the whole data sets by utilizing visual encoding and human interaction. It thus helps scientists to explore, conclude and validate their findings. This article describes analytical tasks and strategies for organizing existing works on visualization and visual analysis on ensemble simulation data sets. The analytical tasks for ensemble simulation data sets include comparing individual members and summarizing whole ensemble, whereas the analytical strategies consist of location-based method and feature-based method. This article reviews major works in ensemble visualization. It gives explanation to their visual design, interaction approaches and application scenarios, along with a discussion of recent trends and future research directions.