Uncertainty Visualization

Showing what is not known is as important as showing what is; uncertainty visualization keeps confidence visible in every figure.

Why show uncertainty

Every measurement and every simulation carries uncertainty. A figure that hides it invites overconfidence and bad decisions. Uncertainty visualization makes the spread, the confidence interval, or the ensemble variability part of the picture rather than a footnote.

Common techniques

Kronos motion — what is fusion

Error bars and confidence bands for one-dimensional data.
Shaded ensembles or spaghetti plots for many model runs.
Fading, blur, or reduced saturation to encode lower confidence.
Hatching or explicit no-data regions where the model is unconstrained.

Aleatory versus epistemic

Distinguish uncertainty that is inherent randomness (aleatory) from uncertainty that reflects limited knowledge (epistemic). They call for different responses: more data reduces the second but not the first. A figure should make clear which it is showing.

Pitfalls

A smooth interpolated surface implies knowledge between samples that may not exist. A single deterministic curve from a stochastic process pretends to a precision it lacks. Ensemble and probabilistic displays avoid claiming a single truth where there is a distribution.

python

import numpy as np
# 90% band from an ensemble of model runs
lo = np.percentile(runs, 5, axis=0)
hi = np.percentile(runs, 95, axis=0)
med = np.percentile(runs, 50, axis=0)
# plot med with [lo, hi] as a shaded band

Kronos use

Because Kronos physics figures are simulation output, uncertainty and honest gates are shown alongside central values; a modeled quantity is never presented as a settled measurement.