Master Chinese Map Visualization with Python‑geopandas: A Step‑by‑Step Guide

geopandas reads Chinese map files

geopandas provides a convenient read_file() method to load GeoJSON files. The example loads a provincial map and a nine‑dash line file, then plots them using the default WGS84 projection.

file = "中国省级地图GS（2019）1719号.geojson"
nine = "九段线GS（2019）1719号.geojson"
china_main = gpd.read_file(file)
china_nine = gpd.read_file(nine)
fig, ax = plt.subplots(figsize=(12, 8), dpi=80)
ax = china_main.plot(ax=ax)
ax = china_nine.plot(ax=ax)

The resulting map is shown below:

Projection conversion and basic settings

Switch to EPSG:2343 and adjust visual properties:

fig, ax = plt.subplots(figsize=(12, 8), dpi=80)
ax = china_main.geometry.to_crs(epsg=2343).plot(fc="white", ec="black", ax=ax)
ax = china_nine.geometry.to_crs(epsg=2343).plot(ec="black", ax=ax)

The key call is to_crs(epsg=2343) for projection transformation.

Data preparation

Scatter data is converted into a GeoDataFrame with geographic coordinates:

scattergdf = gpd.GeoDataFrame(
    scatter,
    geometry=gpd.points_from_xy(scatter.lon, scatter.lat),
    crs="EPSG:4326"
)
scattergdf.head()

After conversion, the data is re‑projected to EPSG:2343:

scattergdf_2343 = scattergdf.to_crs(epsg=2343, inplace=True)

Map visualization

The main plotting code creates the base map, adds the South China Sea inset, and customizes legends:

fig, ax = plt.subplots(figsize=(8,5), dpi=200)
plt.rcParams['font.family'] = ['Times New Roman']
ax = china_main.geometry.to_crs(epsg=2343).plot(fc="white", ec="black", linewidth=.8, ax=ax)
ax = china_nine.geometry.to_crs(epsg=2343).plot(color="gray", linewidth=.9, ax=ax)
for loc, size, class_name in zip(
        scattergdf_2343.geometry.representative_point(),
        scattergdf_2343["data"],
        scattergdf_2343["class"]):
    ax.scatter(loc.x, loc.y, s=10*size, fc=class_color[class_name], ec="black", lw=.5, zorder=2)
# hide spines
for spine in ['top', 'left', 'bottom', 'right']:
    ax.spines[spine].set_color("none")
ax.set_xlim(china_nine_2343.geometry[0].x-500000, china_nine_2343.geometry[1].x)
ax.set_ylim(china_nine_2343.geometry[0].y, china_nine_2343.geometry[1].y)
ax.set_xticks([])
ax.set_yticks([])
# custom legend points
ax.scatter([], [], c='#E21C21', s=30, label='cluster1', ec='black', lw=.5)
ax.scatter([], [], c='#3A7CB5', s=30, label='cluster2', ec='black', lw=.5)
ax.scatter([], [], c='#51AE4F', s=30, label='cluster3', ec='black', lw=.5)
# size legend
for i in range(1,6):
    ax.scatter([], [], c='white', s=i*10, label=str(i), edgecolor='black', lw=.5)
ax.legend(frameon=False, ncol=8, loc='upper center', fontsize=9, columnspacing=.2)
ax.text(.91, -0.02, '
Visualization by DataCharm', transform=ax.transAxes, ha='center', va='center', fontsize=6, color='black')
# inset map for South China Sea
ax_child = fig.add_axes([0.688, 0.125, 0.2, 0.2])
ax_child = china_main.geometry.to_crs(epsg=2343).plot(ax=ax_child, fc='white', ec='black')
ax_child = china_nine.geometry.to_crs(epsg=2343).plot(ax=ax_child, color='gray', linewidth=.9)
for loc, size, class_name in zip(
        scattergdf_2343.geometry.representative_point(),
        scattergdf_2343["data"],
        scattergdf_2343["class"]):
    ax_child.scatter(loc.x, loc.y, s=10*size, fc=class_color[class_name], ec='black', lw=.5, zorder=2)
ax_child.set_xlim(china_nine_2343.geometry[2].x, china_nine_2343.geometry[3].x)
ax_child.set_ylim(china_nine_2343.geometry[2].y, china_nine_2343.geometry[3].y)
ax_child.set_xticks([])
ax_child.set_yticks([])

Adding the inset map with add_axes() produces a focused view of the South China Sea.

The final visualization looks like this:

Summary

This issue demonstrates how to use Python‑geopandas for Chinese map drawing, covering data labeling, projection conversion, and custom legends. Note that geopandas installation can be tricky; using conda install -c conda-forge geopandas is recommended. Spatial visualizations still lack easy support for scale bars or north arrows, which are being improved.

Disclaimer: The data is for practice only and should not be used for research or publication.