Over the last several decades, visualization research has generated an extensive body of literature on the subject of color encodings for two-dimensional (2D) scalar data. When working with experts from other domains, however, visualization designers commonly encounter real-world practices that conflict with visualization guidance, especially where color is concerned. This dissertation focuses on the problem of empirically interrogating conflicting domain guidance related to color utilization in 2D scalar visualizations through a mixed-methods research program. As a primary contribution, this work uses guidance conflicts observed during a formative, qualitative design study conducted with meteorologists to motivate subsequent quantitative studies that revise and refine longstanding visualization guidance regarding color usage. As a secondary contribution, we also briefly touch on the subjects of ecological validity, generalizability, and experimental control.

The dissertation begins with two pieces of formative work. The first is our meteorological design study, which provides a grounded understanding of color usage meteorological practice. The second is a pilot study aimed at improving ecological validity in assessing the impact of uncertainty visualizations in meteorological forecasting. This formative work subsequently acts as a lens for two additional studies investigating specific color guidance conflicts between visualization and other domains: a web-based study looking into the impact of discretized color on meteorological tasks, and a laboratory study assessing the potential role of hue-based banding in implicitly discretizing 2D rainbow color-mapped datasets. Lastly, we discuss the broader implications of our mixed-methods research program and outline potential directions for future work.