Seaborn散点气泡图如何拆分颜色与大小两类图例分置不同位置

要实现拆分图例的需求，核心逻辑是关闭sns.scatterplot默认生成的合并图例，手动提取两类映射的图例元素分别添加，具体实现如下：

首先是基于tips数据集的可运行验证示例：

import seaborn as sns
import matplotlib.pyplot as plt

tips = sns.load_dataset('tips')
tips["percentage"] = tips["tip"] / tips["total_bill"] * 100

# 绘图时关闭自动生成的合并图例
ax = sns.scatterplot(data=tips, x="total_bill", y="tip", hue="time",
                size="percentage", sizes=(100, 1500), legend=False)

# 提取颜色分类（对应你场景里的大洲）的图例元素，tips中time分类共2类所以取前2个
hue_handles, hue_labels = ax.get_legend_handles_labels()
hue_handles = hue_handles[:2]
hue_labels = hue_labels[:2]

# 提取气泡大小（对应你场景里的死亡率）的图例元素，从分类之后的位置开始取
size_handles, size_labels = ax.get_legend_handles_labels()
size_handles = size_handles[2:]
size_labels = size_labels[2:]

# 添加颜色分类图例到右上角，markerscale=1保证气泡大小不被缩放
leg1 = ax.legend(hue_handles, hue_labels, loc="upper right", title="用餐时段", markerscale=1)
# 把第一个图例添加到轴上，避免被后续图例覆盖
ax.add_artist(leg1)
# 添加气泡大小图例到右下角
ax.legend(size_handles, size_labels, loc="lower right", title="小费占比(%)", markerscale=1)

plt.show()

适配你双面板场景的修改后代码：

import seaborn as sns
import matplotlib.pyplot as plt

fig, (ax1, ax2) = plt.subplots(2,1, figsize=(15,17), sharex=True)

# 绘图时关闭默认图例
sns.scatterplot(data=merged, x="gdp_per_capita", y="life_expectancy", 
                size="death_rate", hue="continent", 
                cmap="Accent", sizes=(100, 1500), ax=ax1, legend=False)
sns.scatterplot(data=merged, x="gdp_per_capita", y="human_development_index", 
                size="death_rate", hue="continent", 
                cmap="Accent", sizes=(100, 1500),  ax=ax2, legend=False)

ax1.set(xlabel='GDP per Capita', ylabel ='Life Expectancy', title='Choose')
ax2.set(xlabel='GDP per Capita', ylabel ='Human Development Index', title='Choose')

# 此处替换为你实际的大洲分类数量，比如共5个大洲就填5
CONTINENT_COUNT = 5

# 配置第一个子图图例
hue_handles1, hue_labels1 = ax1.get_legend_handles_labels()
hue_handles1 = hue_handles1[:CONTINENT_COUNT]
hue_labels1 = hue_labels1[:CONTINENT_COUNT]
size_handles1, size_labels1 = ax1.get_legend_handles_labels()
size_handles1 = size_handles1[CONTINENT_COUNT:]
size_labels1 = size_labels1[CONTINENT_COUNT:]

leg1 = ax1.legend(hue_handles1, hue_labels1, loc="upper right", title="大洲", markerscale=1)
ax1.add_artist(leg1)
ax1.legend(size_handles1, size_labels1, loc="lower right", title="死亡率", markerscale=1)

# 配置第二个子图图例
hue_handles2, hue_labels2 = ax2.get_legend_handles_labels()
hue_handles2 = hue_handles2[:CONTINENT_COUNT]
hue_labels2 = hue_labels2[:CONTINENT_COUNT]
size_handles2, size_labels2 = ax2.get_legend_handles_labels()
size_handles2 = size_handles2[CONTINENT_COUNT:]
size_labels2 = size_labels2[CONTINENT_COUNT:]

leg2 = ax2.legend(hue_handles2, hue_labels2, loc="upper right", title="大洲", markerscale=1)
ax2.add_artist(leg2)
ax2.legend(size_handles2, size_labels2, loc="lower right", title="死亡率", markerscale=1)

# 自动调整布局避免元素重叠
plt.tight_layout()
plt.show()

补充说明：

• markerscale=1参数保证图例内的气泡大小和图中实际映射大小完全一致，不会出现缩放后变形或者超出图例区域的问题
• 如果图例遮挡了图内内容，可以给legend方法添加bbox_to_anchor=(x,y)参数微调位置，也可以设置参数把图例放到图的外侧

内容的提问来源于stack exchange，提问作者Callum Marley