We have developed a pilot assessment model for improving elementary teacher preparation in earth sciences, scalable for broader systematic change. The model consists of a framework for implementing visualization learning environments (simulations, modeling, and imagery of real data sets) and a method for assessing student outcomes. We will present methods, data, and analysis from a one-year study on the use of visualization learning environments for teaching geoscience concepts in an elementary teacher preparation science course taught at the University of Northern Colorado. Data indicates students are able to scaffold their skills in interpreting the relationship among three or more variables in a complex system. For example, students are able to observe, analyze, and predict volcanic type, earthquake depth, and earthquake magnitude relationships. Students refine, apply, and defend their knowledge to a new world; complete with plate boundaries they’ve created. The entire learning environment is managed using WorldWatcher, a complete data analysis and visualization software.