Abstract: Understanding the dynamics of turbulence in the presence of a background density gradient, referred to as stratified turbulence, is a central problem in industrial and environmental fluid dynamics. In this talk we will summarize recent efforts by the STRATA research consortium to generate highly-resolved direct numerical simulations of stratified turbulence across various flow regimes, pushing over 5 trillion grid points. To efficiently extract physical insight from these massive datasets, there is an urgent need to develop algorithms that can intelligently subsample flow structures and identify distinct turbulent regimes. We here examine the utility of a nonuniform entropy-based sampling technique, in which a flow field is segmented into distinct spatial clusters, with sampling then weighted by the local cluster entropy. Due to the prevalence of extreme events in the flow, entropy-based sampling is found to significantly outperform random sampling in recovering metrics such as the cumulative scalar mixing rate. We also discuss our adoption of the Zarr data format, allowing the community to interactively download and analyze arbitrary subvolumes of flow data on demand.