Heatmap Settings
The Heatmap card controls how the interpolated surface is computed and rendered.
Interpolation Method
The visual estimates the value at every pixel by interpolating from the sample points. Three algorithms are available:
| Method | Description | Best For |
|---|---|---|
| IDW (Inverse Distance Weighting) | Each pixel is a weighted average of all samples; weight decreases with distance raised to a configurable power | Sparse or irregular point clouds (tens to a few hundred points) |
| Bilinear | Builds a coarse IDW grid then interpolates each pixel from its 4 surrounding grid corners: v = v₀₀·(1−fx)·(1−fy) + v₁₀·fx·(1−fy) + v₀₁·(1−fx)·fy + v₁₁·fx·fy | Any point density; produces a perfectly smooth surface with no blocky or faceted artifacts |
| Triangulated | Piecewise-linear interpolation over a Delaunay triangulation of the sample points | Dense, regular grids (e.g. 50×50 or larger) where each triangle is only a few pixels wide |
| RBF (Radial Basis Function) | Gaussian-weighted Shepard interpolation with a bandwidth parameter that controls how far each sample's influence spreads | Sparse data where each sample's influence should drop off sharply beyond a certain radius |
Choosing a method
- IDW is the safe default for any point cloud.
- Bilinear gives the same smoothness as IDW but with perfectly continuous color gradients across cell boundaries — often visually cleaner on dense data.
- Triangulated produces visible angular facets at low point counts (each Delaunay triangle boundary shows as a colour crease). It is best reserved for dense regular grids (100×100+) where the triangles are sub-pixel and the facets disappear.
- RBF is useful when you want each sample's influence to drop off quickly — tune RBF Bandwidth to control the radius.
Interpolation Properties
| Property | Description | Default |
|---|---|---|
| Interpolation Method | IDW, RBF, or Bilinear | IDW |
| IDW Power | Exponent applied to distance in the IDW formula. Higher values make the influence of each point fall off faster, sharpening local peaks. | 2 |
| RBF Bandwidth | Gaussian sigma expressed as a percentage of the shorter plot dimension. Smaller values create narrower influence zones; larger values smooth more aggressively. (Only visible when RBF is selected.) | 15% |
| Grid Resolution | Side length (in pixels) of each interpolation cell. Smaller values produce a finer surface but increase render time. | 4 px |
Minimum sample count
At least 3 sample points are required to render the heatmap. With fewer points the visual displays an empty state message.