Recurrence Plot
A square matrix visualization that reveals hidden patterns, periodicities, and regime changes in time-series data by marking when states recur in phase space.
// 01 — The chart
What it looks like
A recurrence plot showing diagonal line structures characteristic of periodic dynamics, with scattered points from noise.
// 02 — Definition
What is a recurrence plot?
A recurrence plot (RP) is a tool from nonlinear time-series analysis that visualizes the times at which the state of a dynamical system recurs — that is, returns to a previously visited region of its phase space. Introduced by Eckmann, Kamphorst, and Ruelle in 1987, it transforms a time series into a square binary matrix.
Each axis of the square represents time. A dot (or filled cell) at coordinates (i, j) means that the system’s state at time i is sufficiently close to its state at time j, given some chosen distance threshold ε. The main diagonal is always filled (every state is identical to itself).
The power of recurrence plots lies in the texture they reveal. Diagonal lines indicate deterministic dynamics and periodic or quasi-periodic behavior. Vertical/horizontal lines suggest laminar (stuck) states. Isolated points signal randomness. This makes RPs invaluable for detecting subtle structure in data that looks noisy or chaotic.
Origin: Proposed in 1987 by J.-P. Eckmann, S. Oliffson Kamphorst, and D. Ruelle as a way to visualize recurrences of trajectories in phase space, enabling the study of complex dynamical systems without requiring explicit models.
// 03 — Anatomy
Parts of a recurrence plot
// 04 — Usage
When to use it — and when not to
- Detecting hidden periodicities or quasi-periodicities in noisy signals
- Identifying regime changes, transitions, or non-stationarities in time series
- Studying chaotic or nonlinear dynamical systems where Fourier analysis fails
- Comparing the dynamical similarity of two time series (cross-recurrence plots)
- Exploring structure in short, non-stationary data where traditional methods struggle
- Your audience is unfamiliar with phase-space concepts — it will confuse more than inform
- A simple time-series line chart already reveals the pattern clearly
- The time series is purely linear and stationary — spectral analysis is more appropriate
- You need precise quantitative comparisons — RPs are primarily qualitative
- The dataset is extremely large (>10,000 points) without downsampling — the plot becomes an unreadable blob
// 05 — Reading guide
How to read a recurrence plot
Follow these steps to decode the visual textures of a recurrence plot.
Start with the main diagonal
The diagonal running from bottom-left to top-right (the line of identity) is always present. It confirms that the plot is correctly constructed — every state recurs with itself.
Look for diagonal lines parallel to the main diagonal
These indicate that the trajectory visits similar regions of phase space at regular intervals. Longer diagonals mean more deterministic, sustained periodicity. Evenly spaced diagonals reveal the period length.
Check for large-scale block structures
Square blocks of high recurrence density indicate regimes — time intervals where the system behaves consistently. Sharp boundaries between blocks reveal transitions or regime changes.
Spot vertical or horizontal lines
These laminar structures indicate that the system is stuck in (or slowly drifting through) a particular state. Common in intermittent dynamics.
Assess the density of isolated points
Scattered points without structure suggest stochastic or strongly chaotic behavior. If the entire plot is uniformly dense (or empty), the threshold ε may need adjusting.
// 06 — Pitfalls
Common mistakes
Choosing the wrong threshold ε
Too small and the plot is nearly empty; too large and everything is recurrent. The threshold should be chosen so that the recurrence rate is typically 1–5%, or guided by domain knowledge.
Ignoring embedding parameters
Before creating an RP, the time series usually needs to be embedded in a higher-dimensional phase space using a delay embedding. Wrong choices of embedding dimension or delay can produce misleading structures.
Over-interpreting small-scale textures
Tiny clusters and isolated diagonals can be artifacts of noise or finite data length. Focus on large-scale, robust structures that persist across different threshold values.
Comparing RPs with different parameters
When comparing recurrence plots of different time series, the embedding dimension, delay, and threshold must be consistent, or the comparison is meaningless.
// 07 — In the wild
Real-world examples
Cardiac rhythm analysis
Cardiologists use recurrence plots of heart rate variability data to detect arrhythmias and distinguish between healthy and pathological cardiac dynamics.
Climate science
Paleoclimatologists apply RPs to proxy records (ice cores, tree rings) to identify climate regime shifts and recurring patterns like glacial–interglacial cycles.
Engineering vibration monitoring
Mechanical engineers use recurrence plots to monitor machinery vibrations, detecting bearing faults or structural changes before they become critical failures.
// 08 — Quick reference
Key facts
// 09 — Variations
Variations of the recurrence plot
Cross-recurrence plot (CRP)
Compares two different time series by plotting recurrences between them — useful for studying synchronization or coupling between systems.
Joint recurrence plot (JRP)
Shows simultaneous recurrences in two systems, highlighting synchronized behavior more sharply than CRPs.
Unthresholded (distance) recurrence plot
Instead of binary black/white dots, uses a continuous color scale representing the distance between states — richer but harder to quantify.
// 10 — FAQs
Frequently asked questions
What is a recurrence plot?+
A recurrence plot (RP) is a tool from nonlinear time-series analysis that visualizes the times at which the state of a dynamical system recurs — that is, returns to a previously visited region of its phase space. Introduced by Eckmann, Kamphorst, and Ruelle in 1987, it transforms a time series into a square binary matrix.
When should you use a recurrence plot?+
Use a recurrence plot when detecting hidden periodicities or quasi-periodicities in noisy signals. It also works well when identifying regime changes, transitions, or non-stationarities in time series, and when studying chaotic or nonlinear dynamical systems where Fourier analysis fails.
When should you avoid a recurrence plot?+
Avoid a recurrence plot when your audience is unfamiliar with phase-space concepts — it will confuse more than inform. It is also a poor fit when a simple time-series line chart already reveals the pattern clearly, or when the time series is purely linear and stationary — spectral analysis is more appropriate.
What is another name for a recurrence plot?+
Recurrence Plot is also known as RP, recurrence matrix. The name varies between fields, but the visualisation technique is the same.
Is a recurrence plot suitable for dashboards?+
Yes — a recurrence plot can work well in dashboards as long as the panel is large enough for readers to perceive the encoded values, has a clear title, and includes the legend or axis labels needed to interpret it.
What category of chart is a recurrence plot?+
Recurrence Plot belongs to the Scientific family of charts. Charts in that family are designed to answer the same kind of question, so they often work as alternatives when one doesn't quite fit your data.