How to Visualize Uncertainty in LCA Results

By Tess Konnovitch, Scientific Visualization Advisor for EarthShift Global

Life cycle assessment is, by nature, an exercise in working with imperfect data. Background datasets carry their own variability, characterization factors come with ranges rather than fixed values, and the assumptions practitioners make to manage scope and data gaps introduce additional uncertainty into every result. None of this is a flaw in the method — it's a reflection of the complexity of the systems LCA is trying to represent. But it does raise a real communication challenge: how do you show uncertainty in a result without either hiding it or overwhelming the audience with it?

This is a question I've spent a fair amount of time on, including in a guidance document I co-authored with Gianni Guglielmi on visualizing LCA results more broadly. Uncertainty turned out to be one of the more difficult pieces of that work, and it's worth revisiting on its own.

Why uncertainty is hard to visualize well

There isn't a standard for how LCA results should be interpreted or communicated. ISO 14040 and 14044 specify mandatory elements for goal and scope, inventory analysis, and impact assessment, but the interpretation phase — where uncertainty would typically be addressed — carries essentially no requirements. Sensitivity analysis isn't required either. That leaves a lot of room for inconsistency in how practitioners choose to present (or omit) uncertainty, even though LCA results are increasingly used to support real decisions.

The risk runs in both directions. Leave uncertainty out entirely, and a result can look more definitive than it is, which sets up an audience to draw a stronger conclusion than the data supports. Include too much uncertainty detail, particularly for an audience without a technical background, and the result can become difficult to interpret correctly, which carries its own risk of misinterpretation. The right answer depends on the audience and the decision the result is meant to support, but the failure modes on either side are worth keeping in mind.

There's a related finding from research on bar charts specifically: across studies, around one in five people misinterpret a bar chart that represents mean values rather than counts. Since nearly all LCA bar charts represent means, this is a meaningful blind spot, and one more reason to think carefully about how uncertainty is layered onto that chart type.

A few visualization options, and where they fall short

Bar charts with error bars are the most familiar option, and for good reason. They're easy to read, comparable across alternatives, and they nest naturally with the bar chart format already widely used in LCA reporting. They're also a reasonable default for audiences who aren't deeply familiar with LCA, since the visual grammar (a bar, plus a range) is broadly intuitive.

Box-and-whisker plots convey more — median, quartiles, and outliers — but ask more of the reader in return. They're a solid choice for audiences who want summary statistics and are comfortable interpreting them, but they're not the easiest entry point for a general audience.

Violin plots are, in my opinion, underused outside the LCA community given how well they handle this particular problem. Rather than collapsing a distribution into a handful of summary statistics, a violin plot shows the shape of the distribution itself. At EarthShift Global, we typically draw a violin with an inner mass representing the 50% confidence interval and an outer mass representing the 95% interval, which gives a reader a genuine sense of where the data is concentrated rather than just where its edges are. The tradeoff is that violin plots take more explanation for non-expert audiences, so they tend to work best for technical readers or alongside a clear caption.

Blur charts take a different approach: instead of a hard edge at the mean, the bar fades into a gradient that represents the uncertainty range. Conceptually, this is appealing because it forces the reader to engage with uncertainty rather than anchoring on a single number. In practice, research on this approach has found that people find it harder to identify the highest-impact category or to tell where the error actually starts and stops — so it can communicate "this is uncertain" effectively while communicating the specifics of that uncertainty less effectively.

Matching the chart to the decision

The most useful framing I've come across isn't really about which chart is "best" — it's about matching the chart to what the result needs to do. A 2022 study examining comparative LCA communication described four ways a chart can mislead an audience when comparing two alternatives: making a real difference look irrelevant, making a real difference look like inaction is warranted, leaving the audience indecisive when a difference is real, or creating a misconception that a difference exists when it doesn't. All four scenarios stem from the same root issue — how clearly the chart communicates both the relative difference between alternatives and the statistical discernibility of that difference, given the uncertainty involved.

That's a useful test to apply to your own work: before finalizing a chart, ask whether someone unfamiliar with the underlying data could walk away with a meaningfully wrong impression of how confident the result actually is.

A few practical takeaways that tend to hold up across audiences:

• For broad or mixed-technical audiences, a bar chart with error bars is usually the safest default — familiar, interpretable, and still honest about the range.
• For expert audiences, a violin plot (or a box plot nested inside one) gives a more complete and accurate picture of the underlying distribution, and is worth the extra explanation it requires.
• Whatever you choose, label it clearly. Don't assume a reader will infer what an error bar or a violin's width represents — state it explicitly in the caption or alt text.
• Resist the instinct to simplify uncertainty out of the visual entirely. A clean chart that omits uncertainty isn't more honest — it's just quieter about where it's wrong.

The bigger picture

Visualizing uncertainty well is really a subset of a larger communication challenge in LCA: results are often genuinely complex, and the temptation to simplify them for an audience can quietly cross the line into distorting what they actually show. Uncertainty is one of the places where that's most likely to happen, simply because uncertainty itself is uncomfortable to sit with. But leaving it out doesn't make a result more certain — it just makes the audience more confident than the data warrants.

As we put it in the broader guidance document this post draws from: communication shouldn't be treated as the last, optional step in the scientific process. Science isn't finished until it's understood — and that's just as true of the uncertainty in a result as it is of the result itself.