The increasing amount of data generated by scientific applications poses great challenges for scientists. Since traditional relational DBMSs are inadequate for the analytical workload required by these applications, it is necessary to investigate more suitable models. Multidimensional arrays are a possible alternative, because they allow efficient range querying and enforce data coherence by dividing data into chunks. However, the method used to represent data with arrays can affect performance significantly. In this work we discuss different ways of representing simulation data with arrays, along with an evaluation composed by real life queries.