Evaluating the Impact of Touch-Table Map Visualizations on Decision Performance, Cognitive Load, and Workload in Flood Risk Management

Flood-risk interfaces must support quick spatial decisions without hiding uncertainty and without creating too much workload. This thesis reports FloodVis2026, a controlled 2 × 2 within-subject study on a 4K touch table that compares a continuous ground heatmap with 2.5D bars, each with an uncertainty cue OFF and ON, across five operational tasks (Evacuation, Route, Gate, Shelter, Warning) in a georeferenced 3D scene of HafenCity, Hamburg. All four displays share one deterministic synthetic susceptibility field, one land mask and one legend, so only the representation changes. The sample size of 30 was set by an a priori G*Power calculation; two of the 30 participants were pilot sessions that were run under the final experimental conditions and were therefore retained. The participants completed 600 trials and 120 questionnaire blocks after passing a comprehension gate on the meaning of the uncertainty cues. Outcomes are grouped by the three constructs in the title: decision performance (strict and component accuracy, completion time, calibration), cognitive load (Paas mental effort) and workload (Raw NASA-TLX), with usability, confidence, clarity, interaction traces and open responses as supporting measures. Mixed-effects models found no statistically significant effect of encoding or uncertainty on decision performance, and display-by-task interactions were not significant. Cognitive load measured with the single Paas item did not differ significantly between conditions. Workload did: showing uncertainty raised Raw NASA-TLX, and the increase was concentrated in the Effort subscale, which rose in both encodings. The cue also had a clear usability cost, lowering contextadapted SUS by about 7.5 points (7.8 within Heatmap, 7.1 within Bars), an effect that held across parametric, rank-based and Wilcoxon analyses, and it reduced perceived visual clarity. At the same time, two thirds of participants said the uncertainty information made their decisions easier, and 70 % preferred a Bars condition. Task structure explained far more variance in time and interaction than the display. The thesis contributes the empirical comparison, a quantified spatial-association property that describes how a lattice encoding trades magnitude precision for association distance, a displaymargin register fixed before data collection, a pair of fail-closed validation harnesses, and a confidence-and-provenance ledger for every substantive claim. The results support improving how uncertainty is communicated and how difficult tasks are structured rather than assuming that one encoding is universally better.