Designing Dynamic Stacked Bar Charts for Alarm Semantic Levels: Hierarchical Color Cues and Orientation on Perceptual Order and Search Efficiency

In sensor-based monitoring systems, the rapid and accurate recognition of alarm semantic levels is essential for maintaining operational reliability. Traditional static visualizations often fail to communicate these distinctions effectively under time pressure, whereas dynamic stacked bar charts (DSBCs) integrate multiple semantic layers into a compact, dynamic display. This study systematically investigated how color cues applied to auxiliary visual elements (background, foreground, labels, and scale lines) and chart orientation (horizontal vs. vertical) affect users' alarm recognition performance. Thirty-two participants completed a semantic alarm recognition task involving DSBCs with various combinations of color-coded elements and orientations. Reaction time (RT) and accuracy (ACC) were analyzed using mixed-effects regression models. The results revealed that color cues in foreground and labels significantly enhanced both RT and ACC, whereas background and scale line color cues produced negligible effects. Orientation exerted a significant main effect on RT but not on ACC. Participants responded faster to horizontally oriented charts, indicating improved scanning efficiency. Moreover, increasing the number of color cues yielded higher ACC and shorter RTs, supporting a redundancy gain effect. However, no interaction was found between color cues and orientation, suggesting that these factors influence performance through distinct cognitive pathways. The findings align with theories of attentional guidance, redundancy gain, and spatial compatibility, and offer practical recommendations for alarm visualization design. Consequently, designers are advised to prioritize color coding of perceptually dominant elements, employ horizontal layouts in time-critical contexts, and implement redundant but non-overwhelming cues to enhance alarm recognition in complex sensor-based monitoring environments.