Layers: 1-D, scatter and aggregation

For ClimCanvas v1.02.1

This chapter explains the settings and caveats of each kind of layer used in 1-D plots and 2-D plots. Adding, deleting and reordering layers work as in Layers: maps and sections. The order is the drawing order and the legend order. Examples of the figures you can draw are in the Gallery. The definitions of the statistics (standard deviation, whiskers, KDE, bins) are summarized in Computational conventions.

Plot mode Available layers
1-D plot Line, line bundle, line fill, stack plot, bar chart
1-D plot (aggregation) Histogram, ECDF and line when the plot type is "Histogram / ECDF"; box plot and violin when it is "Box / violin"
2-D plot Scatter, bubble chart
2-D plot (aggregation) 2-D histogram, hexbin

Items common to all layers.

Line

Item Content Default
Variable A variable with the x-axis dimension
Automatic color (color cycle) When off, "Line color" on
Line width 0.5 to 4.0 1.5
Line style solid / dashed / dotted / dash-dot / none (markers only) solid
Marker none / ○ / ■ / ▲ / × / • / + none
Draw on the secondary (right) axis off

There is no error bar item. To add a band, use line fill or the summary lines of a line bundle.

A line can also be overlaid in a 1-D plot (aggregation) with "Histogram / ECDF" (e.g. the curve of a theoretical distribution). In that case the x-axis is not the panel's axis but the coordinate values of the dimension chosen as "Dim used as x-axis" (make its units the same as the value axis of the histogram).

Line bundle

Overlays all slices along one dimension, such as ensemble members, with the same color, line style and opacity, and merges them into a single legend entry.

Item Content Default
Variable A variable with a dimension to bundle besides the x-axis and time
Dimension to bundle Chosen from "{dim} ({n} lines)". The time dimension cannot be chosen
Line color All lines share the same color #808080
Line width 0.0 to 4.0. 0 draws no bundle lines (to show only the band) 0.8
Opacity (alpha) With many lines, a smaller value shows the density of overlaps 0.5
Line style, marker, secondary axis, legend label Same as line

Summary lines

Turning on "Summary lines (overlay statistics of the bundle)" overlays lines aggregated along the bundled dimension on top of the bundle. "Number of summary lines" (1 to 6) sets how many "Summary line {n}" items appear. NaN is excluded from the aggregation.

Statistic Content
Mean, Median, Minimum, Maximum 1 line
Min and max (2 lines)
Percentile (1 line) "Percentile (%)" (default 95)
Percentile range (2 lines) "Lower (%)" and "Upper (%)" (default 5 and 95)
Mean ± std. dev. (2 lines) "Std. dev. multiplier k" (default 1). The standard deviation is the sample standard deviation (divided by N − 1)

Two-line statistics have "Draw style": "Lines (2)", "Band (filled)" or "Band + edge lines". You set "Line color" (default black), "Line width" (default 2.0, 0 for no line), "Line style" and the "Opacity (alpha)" of the band (default 0.3). The default "Legend label" is a short name such as mean, median, min-max, p95, p5-p95 or mean±1σ; when empty, the line is left out of the legend.

line fill

Fills the area between two lines, or between a line and a reference value.

Item Content Default
Variable (first) Also set the fixed dimensions and range means
Second value "From a variable": "Variable (second)" (its fixed dimensions are independent of the first; e.g. a band of the difference between 500 hPa and 850 hPa). "Constant (baseline)": "Baseline value" (in transformed units) From a variable
Different colors above/below the baseline For a constant. Fills with "Upper color" and "Lower color" (default #1f77b4 and #d62728) off
Color The color when not split above/below #1f77b4
Opacity (alpha) 0.3
Legend label The default with a variable is "first–second"

The value transform applies the same coefficients to both variables.

Stack plot

Item Content Default
Variables (order = stacking order, bottom to top) Multiple selection the first 2 variables
Opacity (alpha) 0.8
Baseline zero (stacked) / sym (symmetric around 0) / wiggle (minimizes the wiggle) / weighted_wiggle (streamgraph) zero
Show variable names in the legend on
Set each variable's color manually When on, "Color of {var}" per variable. When off, the color cycle off

The fixed dimensions and range means are set on the dimensions of the first variable and shared by all variables. There is no normalization to a total of 100 %.

Bar chart

Item Content Default
Orientation Vertical bars (ax.bar) / Horizontal bars (ax.barh) Vertical bars
Color, Opacity (alpha) #1f77b4, 0.8
Set bar width manually When off, 0.8 times the median of the differences between adjacent x. When on, "Width (data units)" off
Show edges Edge color and width off
Apply hatching Hatch pattern (/ \ - + x o . etc.), density (1 to 6), color. When edges are on, the hatch color becomes the edge color (a matplotlib limitation) off
Show error bars "Error amount" given by a variable or a constant (symmetric). Color, width, cap size off
Draw on the secondary (right) axis Vertical bars only off

With several bar chart layers, "Bar placement" at the end of the "Layers" section selects Overlap / Side by side / Stacked (Plot modes). The value transform is not applied to the error amount, whether it is a variable or a constant. Provide it in the units shown in the figure (after the value transform) (Computational conventions).

Histogram

A layer of a 1-D plot (aggregation) first selects "Dim to aggregate" (automatic when there is only one dimension), narrows the period or interval with "Set aggregation range", and fixes the remaining dimensions with "Fix: {dim}". Only finite values are aggregated.

Item Content Default
Use equal-width bins When on: "Set value range" (off: equal division between the data minimum and maximum; on: enter Min and Max) and "Number of bins" (2 to 200). When off: enter "Bin edges (comma-separated)" directly (uneven spacing allowed) on, 20
y-axis Count / Probability density (density; the total area is 1) Count
Cumulative off
Draw style Filled (bar) / Step (line only) / Step-filled (stepfilled). Without edges, bar and stepfilled look nearly identical; with edges, bar also adds vertical lines between bins Filled (bar)
Automatic color (color cycle) When off, "Color" on
Opacity (alpha) Semi-transparent, assuming overlays 0.7
Show edges Edge color and width off

When the axis labels are empty, x gets "variable name [units]" and y gets count / probability density / cumulative count / cumulative probability automatically.

ECDF (empirical cumulative distribution)

The data are taken as for the histogram. "Complementary (1 − CDF)" draws the exceedance probability (the fraction of values at or above a threshold). You set "Automatic color (color cycle)" (when off, "Line color"), "Line width" (default 1.5), "Line style" and "Draw on the secondary (right) axis" (when overlaying on a count histogram). The automatic y label is cumulative probability, or exceedance probability for the complementary distribution.

Box plot

One layer is one series (one box); add layers for more series. The x position follows the layer order. "Orientation" below "Plot type" switches between vertical and horizontal.

Item Content Default
Series name (shown as x tick) variable name
Include in legend To distinguish series by color off
Box width 0.1 to 1.0 0.5
Whisker definition IQR multiplier ("Whisker length (whis, IQR multiplier)", default 1.5; points beyond it are outliers) / Percentile range ("Lower (%)" and "Upper (%)", default 5 and 95) IQR multiplier
Notch (median confidence-interval indent) The notch width is about 1.57 × IQR / √N. Non-overlapping notches of two series hint at a significant difference in medians off
Show outliers, Whisker end bars (caps) The cap width can also be set with "Set cap width" on, on
Show boxes When off, only whiskers and the median on
Show the mean A marker (▲) unless "Draw the mean as a line (meanline)" is on off
Fill boxes Fill color off (white)
Overlay raw data points (strip) "Jitter width" (default 0.2, reproducible with a fixed seed), "Point size (s)", "Point color", "Point opacity". They overlap the outliers, so turning "Show outliers" off is recommended off

"Appearance details" sets the color and width of the lines (box, whiskers, caps), of the median line (default #d62728, 1.5) and of the mean, and the kind, size and color of the outlier markers. The value axis gets "variable name [units]" automatically, and the series axis shows the series names as ticks.

Violin

Series and orientation are handled as in the box plot.

Item Content Default
Series name (shown as x tick), Include in legend variable name, off
Width 0.1 to 1.0 0.7
Median / Mean / Extrema (whiskers) Lines to show on / off / on
Draw quantile lines "Quantiles (0-1, comma-separated)" (default 0.05, 0.95) off
KDE bandwidth Auto (scott) / silverman / Numeric value ("Bandwidth factor", default 0.3). Smaller values show finer detail Auto (scott)
KDE evaluation points 20 to 1000. The resolution of the shape 100
Shape Both sides / Left half / Right half. Half violins compare two series facing each other at the same position Both sides
Automatic color When off, "Color" on
Opacity (alpha) Opacity of the body (the KDE shape) 0.5
Overlay raw data points (strip) Same items as the box plot off

Scatter

The x and y variables are selected in the panel's "Plot variables"; the layer sets the following.

Item Content Default
Dim to draw along A dimension with coordinates shared by x and y. Points are placed along it. "Range of {dim}" narrows it
Dims to fix for x-axis / y-axis variable Dimensions other than the one drawn along are fixed independently for x and y. The initial value is the middle index
Color Single color #1f77b4
Opacity (alpha) 0.7
Marker ○ / ■ / ▲ / × / • / + / ◇ / * ○
Marker size (s, area pt²) 1 to 200 20
Show edges Edge color and width off
Legend label empty

Selecting "x-error variable" and "y-error variable" under "Error bars" (symmetric ± errors; cropped with the same fixed dimensions and range as the main data; the value transform is not applied, so provide them in the units shown in the figure) draws error bars with the given color, width and cap width. Points whose x or y is NaN are excluded. To color by a third variable, use the bubble chart. There is no regression line or correlation coefficient.

Bubble chart

In addition to x and y, "z-axis variable (for bubble chart, optional)" in the panel's "Plot variables" selects the variable used for the size.

Item Content Default
Dim to draw along, dims to fix Same as scatter (z is also fixed independently)
Marker ○
Min size / Max size (pt²) The areas corresponding to the minimum and maximum of z. The transformed z values are normalized linearly onto this range 10 / 200
Show edges off
Color by z value (cmap) When on: colormap, reversal, value range, color discretization (number of levels / specify directly), white near zero, extend, Colorbar. When off, "Color" off
Opacity (alpha) 0.6

Error bars and the value transform (x / y / z independent) are the same as for scatter. There is no size legend.

2-D histogram

A 2-D plot (aggregation) aggregates all pairs of samples along the dimensions shared by x and y; dimensions not shared are fixed with "Fix dims of x variable" and "Fix dims of y variable". Pairs containing NaN are excluded.

Item Content Default
Bins (x) / Bins (y) 2 to 200 20 / 20
Set aggregation range When off, equal division between the data minimum and maximum. When on, enter the minimum and maximum of x and y off
Aggregation Count / Probability density (density; the total integral is 1) Count
Do not fill empty (zero) bins on
Log-scale counts (LogNorm) To see bins differing by orders of magnitude at once. Cannot be combined with color discretization (for log-like classes use "Specify levels directly" of the discretization, e.g. 1, 2, 5, 10, 20, 50) off
Colormap, value range, color discretization, extend Same items as fill
Show bin edges Width and color. Unfilled bins get no edges off
Colorbar The default label is count shown

hexbin (hexagonal bin density)

The data are taken as for the 2-D histogram. Only the number of points is aggregated; there is no reduction of a third variable per hexagon.

Item Content Default
Hexagon count (gridsize) The number of bins in the x direction (5 to 60). Larger is finer 20
Log-scale counts (bins='log') Cannot be combined with color discretization off
Fill empty bins too (mincnt=None) When off, only bins with at least "Minimum count (mincnt)" points are drawn off, 1
Colormap, value range, color discretization, extend Same items as fill
Show bin edges Only on bins where points are drawn off
Colorbar The default label is count shown