Density tanks can be a teaching tool in meteorology and oceanography classes to support learning fundamental concepts by modeling how fluids of different densities behave when combined. For example, density tanks can model Mediterranean outflow, weather fronts, Atlantic Meridional Overturning Circulation, and internal waves. We worked with over 40 students in individual semi-structured, interviews and used contrasting colored water of different densities (hot v cold, and salty v fresh) to collect their predictions about fluid behavior when a barrier between two density fluids was raised, and capture their sense making after seeing the process. To analyze the interview data, we used emergent coding of participant sketches and the transcribed interviews. Our initial findings are that student predictions were almost evenly split between horizontal stratification, vertical stratification, and complete mixing. To better understand how mixing occurred and the time scale, we collected a second round of data, which confirmed students have poor mental models of persistent horizontal fluid stratification by density. Our results suggest that many students arrive in courses with inaccurate conceptions of fluid behavior and what these students grasp from common classroom demonstrations may be quite different from what instructors expect or assume. Everyday experiences with turbulent mixing, such as adding creamer to coffee, playing in swimming pools, and even popular social media seem to lead to an assumption that fluids exist as homogenous bodies with uniform characteristics throughout. Students' tendency to reason from these experiences serves students poorly when applied to large-scale fluid-Earth processes. We suggest that efforts to scaffold student learning and promote conceptional understanding before moving to more complex fluid behavior will develop deeper student understanding of fluid-Earth processes.