Information systems with complex and nested interconnections are often represented by diagrams making these systems easier to understand and interpret. Particular such diagrams are Sankey diagrams, a visualization tool allowing to depict quantities of flow from one node set to another through a set of links or edges. We use the concept of Sankey diagrams to study the evolution of communities over time in temporal social networks. Particular emphasis is given on temporal motifs such as expansion or contraction of communities, birth or death and especially merging and splitting operations where nodes are transferred from one community to another. Such motifs give an overview on group evolution and are useful to understand social behaviors for large audience communities like online political debates or to make decisions for publishing strategies through the monitoring of co-authorship research papers. The search for an optimal layout of such diagrams can be formulated as a combinatorial optimization problem closely related to crossing minimization in multi-layered graphs. We present and discuss numerical results obtained by solving this problem with hybrid methods based on genetic algorithms and simulated annealing. We also describe the data generator which provided the test instances of different size and configuration that we used to evaluate the performance of these methods.