Python中实现带连续双变量调色板的散点图技术问询

Got it, let's figure out how to create that bivariate continuous color palette you need—perfect for maps where you want to encode two different continuous variables at a glance! Here's a step-by-step breakdown with code examples that build on your existing work:

Unlike single-variable palettes that map one value to a color, bivariate palettes combine two continuous variables into a single color. Typically, we map one variable to a distinct hue (like soft blue → deep blue) and the other to a complementary hue (like pale red → dark red), blending them so each color represents a unique combination of both variable values. This makes it easy to spot patterns where both variables are high/low, or one is high while the other is low.

First, let's adapt your scatter plot code to use a bivariate palette. We'll create a custom function to mix colors from two sequential palettes, one for each variable:

import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from matplotlib.colors import LinearSegmentedColormap

# Generate sample data (2 variables for color encoding)
x, y = np.random.rand(2, 50)
var1 = np.random.rand(50)  # First continuous variable (e.g., temperature)
var2 = np.random.rand(50)  # Second continuous variable (e.g., precipitation)

# Pick two distinct sequential palettes for each variable
# Swap these with any seaborn/matplotlib palette (e.g., viridis, plasma) if you prefer
palette_var1 = sns.cubehelix_palette(start=2, rot=0, light=0.7, dark=0.3, as_cmap=True)
palette_var2 = sns.cubehelix_palette(start=0, rot=0, light=0.7, dark=0.3, as_cmap=True)

# Custom function to blend colors from both palettes based on variable values
def blend_bivariate_colors(v1, v2):
    # Get color from each palette (v1/v2 should be normalized to [0,1])
    color1 = palette_var1(v1)
    color2 = palette_var2(v2)
    # Average the two colors (adjust weight here if you want one variable to stand out)
    return np.mean([color1, color2], axis=0)

# Generate colors for all data points
point_colors = np.array([blend_bivariate_colors(v1, v2) for v1, v2 in zip(var1, var2)])

# Plot the scatter plot
fig, ax = plt.subplots(figsize=(8, 6))
scatter = ax.scatter(x, y, c=point_colors, s=100, edgecolor='white')

# Add a custom bivariate legend (matplotlib doesn't have a built-in for this)
legend_ax = fig.add_axes([0.92, 0.15, 0.03, 0.7])
# Create a grid of variable values to generate the legend color grid
v1_grid, v2_grid = np.meshgrid(np.linspace(0, 1, 20), np.linspace(0, 1, 20))
legend_colors = np.array([blend_bivariate_colors(v1, v2) for v1, v2 in zip(v1_grid.flat, v2_grid.flat)]).reshape(20, 20, 4)
legend_ax.imshow(legend_colors, origin='lower')

# Label the legend
legend_ax.set_xticks([0, 19])
legend_ax.set_xticklabels(['Low', 'High'])
legend_ax.set_yticks([0, 19])
legend_ax.set_yticklabels(['Low', 'High'])
legend_ax.set_xlabel('Var1')
legend_ax.set_ylabel('Var2')

plt.tight_layout()
plt.show()

Once you're happy with the palette, scaling this to a map is straightforward. We'll use GeoPandas to load spatial data and apply the same color-blending logic to your two geographic variables:

import geopandas as gpd
from sklearn.preprocessing import MinMaxScaler

# Load sample geographic data (replace with your own shapefile/GeoJSON)
gdf = gpd.read_file(gpd.datasets.get_path('naturalearth_lowres'))

# Create two sample continuous variables (replace these with your actual data!)
gdf['temperature'] = np.random.rand(len(gdf)) * 30
gdf['precipitation'] = np.random.rand(len(gdf)) * 1500

# Normalize variables to [0,1] (critical for consistent color mapping)
scaler = MinMaxScaler()
gdf[['norm_temp', 'norm_precip']] = scaler.fit_transform(gdf[['temperature', 'precipitation']])

# Apply the same color-blending function to each geographic feature
gdf['map_color'] = gdf.apply(lambda row: blend_bivariate_colors(row['norm_temp'], row['norm_precip']), axis=1)

# Plot the map
fig, ax = plt.subplots(figsize=(12, 8))
gdf.plot(color=gdf['map_color'], ax=ax, edgecolor='white', linewidth=0.5)

# Add the same bivariate legend as before
legend_ax = fig.add_axes([0.93, 0.15, 0.03, 0.7])
v1_grid, v2_grid = np.meshgrid(np.linspace(0, 1, 20), np.linspace(0, 1, 20))
legend_colors = np.array([blend_bivariate_colors(v1, v2) for v1, v2 in zip(v1_grid.flat, v2_grid.flat)]).reshape(20, 20, 4)
legend_ax.imshow(legend_colors, origin='lower')

legend_ax.set_xticks([0, 19])
legend_ax.set_xticklabels(['Cool', 'Hot'])
legend_ax.set_yticks([0, 19])
legend_ax.set_yticklabels(['Dry', 'Wet'])
legend_ax.set_xlabel('Temperature')
legend_ax.set_ylabel('Precipitation')

# Clean up the map axes
ax.set_axis_off()
plt.tight_layout()
plt.show()

• Palette Choice: Pick colorblind-friendly palettes (e.g., sns.color_palette("colorblind", as_cmap=True) variants) to ensure your map is accessible to all viewers.
• Color Mixing: Instead of simple RGB averaging, try blending in HSV color space—this creates more intuitive hue transitions that make variable combinations easier to distinguish.
• Normalization: Always normalize your variables to [0,1] before mapping; skewed data will otherwise dominate the color palette and hide subtle patterns.

内容的提问来源于stack exchange，提问作者Suwash