This entry is part of the Chart Index, the reference library for the Chart Design Field Guide.

The stacked area chart is a line chart's contribution to part-to-whole over time. Several series are stacked on top of each other, the height at any point being the sum of all series at that point. The streamgraph is a stacked area variant where the baseline is not zero but is centred — the whole composition flows symmetrically around a horizontal axis, producing the river-like shape that gave the form its name.

Both forms answer the same question: how does the composition of a total evolve over time? The stacked area is more honest about absolute totals; the streamgraph is more elegant about shifting composition, at the cost of precise reading.

What it is

A stacked area chart maps time to the x-axis and a quantitative variable to the y-axis. Several series are summed into stacks: the bottom series sits on zero, the next series sits on the first, and so on. Each series is filled with a distinct colour. A streamgraph applies the same stacking but centres the baseline — the total flows symmetrically above and below a moving central axis. Lee Byron and Martin Wattenberg formalised streamgraphs in 2008 as part of Listening History visualisations.

The hero chart above is two years of electricity generation in the National Electricity Market, six fuels stacked with a silhouette baseline. Coal is the wide grey band at the centre, slowly narrowing; the two solar layers at the bottom swell every summer and shrink every winter; wind, in green, thickens steadily. The streamgraph's central baseline makes shifts in composition the dominant reading — the eye follows each layer's thickness rather than its height. Reading the exact total in any month is harder than it would be on a stacked area, and there is deliberately no value axis, because on a silhouette baseline no height means anything. Randomise it and every month's values jitter; the river re-flows around its centre.

When to use it

Streamgraphs and stacked areas are the right choice when:

  • The data is multi-series time-series with a part-to-whole structure.
  • The reader's question is "how is the composition changing?" more than precise totals.
  • You have 4–10 series. Fewer and a line chart suffices; more and the form becomes unreadable.
  • The chart is editorial or visualisation-led — streamgraphs especially.
  • You have substantial change in composition over the time range — flat compositions make the form decorative.

When not to use it

  • Precise comparison. Both forms make it hard to read the value of a non-bottom series at a specific time. For precision, use small-multiple line charts.
  • Many series (10+). Stacks of thin bands become unreadable. Switch to small multiples or aggregate categories.
  • Operational dashboards. The forms are exploratory and editorial. For dashboards, use simpler stacked bars or line charts.
  • Negative values. Stacking assumes non-negative contributions. Mixed signs break the form.
Twelve fuels and regions on one stream
The hero's six fuels split by region into twelve layers. The thin ones — hydro and gas in the smaller states — are ribbons a few pixels tall that the eye cannot follow, and the legend has become a twelve-entry lookup table that no longer even fits the frame. Aggregate, or split into small multiples.

Design principles

Choose the baseline deliberately

Stacked area: baseline at zero, conventional. Streamgraph: baseline centred, elegant for compositional flow. Or the wiggle baseline (Byron and Wattenberg) — algorithmically placed for minimum slope. The choice changes how the chart reads. The two figures below are the hero's data; only the baseline differs.

Stacked area: zero baseline
The same six fuels stacked from zero. The top edge is now the total, readable against a value axis — generation fell slightly across the two years — and the bottom layer, rooftop solar, can be read exactly. Every layer above it can only be read as a thickness.
Streamgraph: silhouette baseline
The same stack centred on its own midline. The total is no longer readable and there is no value axis, but the shape now says what the composition is doing: the solar layers breathing with the seasons, coal thinning, wind thickening. The vertical position of any edge means nothing; only thickness does.

Order series by stability and value

Place the most-stable series at the baseline (bottom for stacked area, central for streamgraph). The series with the most fluctuation should sit near the visual edges, where the wiggling is naturally more tolerable to the eye. In a zero-baseline stack every layer inherits the wobble of every layer beneath it, so a seasonal series at the bottom makes the whole chart seasick.

Seasonal series at the bottom of the stack
Rooftop solar, the most seasonal layer, is on the baseline. Its summer swell lifts every layer above it, so coal — which barely varies month to month — appears to surge and slump with the sun. The wobble is real; it just belongs to a different layer.
Stable series at the bottom, seasonal at the top
The same data with coal on the baseline and the solar layers at the top. Coal's slow decline is now a straight edge, and the seasonal breathing is confined to the top of the stack where it belongs. Nothing about the numbers changed.

Stacked series share semantics. Use a single hue family with varying lightness, or a desaturated categorical palette. Rainbow palettes produce bands of competing colour. The figures on this page keep the fossil fuels in the quiet greys and blues and give the renewables the warm and green accents, so the story — warm layers growing, grey layer shrinking — is carried by hue as well as thickness.

Normalise when the question is share

A 100% stacked area — every month expanded to fill the same height — removes the total entirely and shows share alone. It is the honest form when the question is what fraction, and the wrong one when the total's own movement matters.

Expanded to 100%: share of generation
Every month fills the same height, so each layer's thickness is now its share of the whole. Coal's fall from roughly half to just over forty per cent reads directly. What has vanished is any sense that total generation moved at all.

Label series directly

Each filled region should be labelled inside its own band or with a leader line. Legends force the reader's eye to shuttle between the chart and a colour key. Where the tooling offers only a legend, keep the series count low enough that the legend can be read in one pass — six is the practical limit on this page.

Show the total as a thin top line

For stacked area charts, optionally trace the top of the stack with a thin line in a different colour — this is the total, the chart's most important single value. The thin line is read as a line chart and the bands below give the composition.

Use streamgraph only when composition is the message

Streamgraphs sacrifice precise reading of totals for elegance and flow. Only use them when the chart's story is how the mix has changed over time and your audience is comfortable with the editorial form.

Annotate notable transitions

If a series surges or collapses, annotate it. Rooftop solar overtook gas in the summer of 2024. The streamgraph shows the shape; the annotation explains it.

Anatomy

The composition of a stacked area chart
An anatomical guide

A stacked area's anatomy is a set of layers summed from zero, each read as a thickness, with the top edge as the total. A streamgraph is the same stack with the baseline moved to the middle and the value axis removed: the layers keep their thickness and lose their height.

  • 100% stacked area — normalised to show composition only, as in the expanded figure above.
  • Area chart — the single-series form; the stack's bottom layer on its own.
  • Multi-line chart — for precise reading of individual series over time.
  • Area bump — ribbons ordered by rank at each step rather than a fixed stack.
  • Ridge plot — for stacked distribution shapes rather than time-series.
  • Stream + line overlay — streamgraph with a single line traced through one series; the editorial hybrid.

Reading list

  • Byron, L. & Wattenberg, M. (2008). Stacked Graphs — Geometry & Aesthetics. The original streamgraph paper.
  • Havre, S., Hetzler, B. & Nowell, L. (2000). ThemeRiver: Visualizing Theme Changes Over Time. The precursor to streamgraphs.
  • Wilke, C. (2019). Fundamentals of Data Visualization. On stacked areas, their ordering, and when to normalise.