Map of China

In this notebook (which covers roughly the same ground as this video) we build an annotated map of China.

We use the GeoPandas library to process map data. This is built on top of the Pandas library for general data analysis, which we will explore later. We also use the re library for regular expressions.

import geopandas
import matplotlib.pyplot as plt
import re

# Use a font that supports Chinese characters
# We could learn how to do this by asking Gemini 
# "set the default font in matplotlib to one that supports Chinese characters"
# The SimSun font is present on most Windows systems.  For non-Windows systems, 
# you might need something else.
plt.rcParams["font.family"] = "SimSun"

We first load a geojson file which has coordinates of the boundaries of all the regions of China. This gives us a GeoDataFrame object. Most of the regions are provinces, but some are municipalities, autonomous regions or special administrative regions. The file originally came from gadm.org.

china = geopandas.read_file('../data/maps/china_provinces.json')

We want to plot a map showing all the regions labelled with their names in Chinese characters. The names are mostly taken from the column NL_NAME_1 of the dataframe. However, some adjustments are needed.

• For some reason, the Chinese character names of the special administrative regions of Hong Kong (香港) and Macau (澳门) are missing, so we add them in.
• For some regions, the name is different in traditional characters (still used in Hong Kong and Taiwan) as opposed to the simplified characters used in mainland China. In these cases the NL_NAME_1 column contains the traditional characters followed by | followed by the simplified characters. We remove the traditional characters and just use the simplified ones.
• To save space on the map, we remove the characters '自治区' (meaning 'autonomous region')
• A few regions are split into several pieces and so have several rows in the dataset. In those cases we use an empty label for each piece after the first one.

china['label'] = ''
china['long_name'] = ''

# Add a column with the position of the label
# The representative_point() method tries to return an appropriate point for
# the label even if the region has a strange, highly non-convex shape.
china['label_pos'] = china['geometry'].apply(lambda x: [* x.representative_point().coords[0]])

# loop over rows
labels = []
for idx, x in china.iterrows():
    # The next line is usually the same as n = x['NAME_1'] but we add some extra code
    # to insert spaces where needed, which converts 'HongKong' to 'Hong Kong' and
    # 'NingxiaHui' to 'Ningxia Hui'.  In more detail, the regular expression
    # '([A-Z]+)' matches any sequence of capital letters.  The expression r' \1'
    # means "a space followed by the thing that was matched", so it inserts a space
    # before each capital letter.  In particular it will usually insert a space 
    # right at the beginning of the string, so we use strip() to remove it.
    n = re.sub('([A-Z]+)', r' \1', x['NAME_1']).strip()
    # We have some special cases for Hong Kong and Macau
    if x['NAME_1'] == 'HongKong':
        s = '香港'
        n = 'the Hong Kong SAR'
    elif x['NAME_1'] == 'Macau':
        s = '澳门'
        n = 'the Macao SAR'
    else:
        # If there is a bar in the name, we use the part after the bar
        # If there is no bar, we use the whole name
        s = x['NL_NAME_1'].split('|')[-1]
        # Remove the last three characters if they are '自治区'
        # (meaning 'autonomous region').  Chinese-speaking students can comment
        # on whether it would be appropriate to remove '族' as well leaving 
        # '宁夏回' instead of '宁夏回族' and '广西壮' instead of '广西壮族'.
        if s[-3:] == '自治区':
            s = s[:-3]
        if x['ENGTYPE_1'] == 'Province':
            n = 'the province of ' + n
        elif x['ENGTYPE_1'] == 'Municipality':
            n = 'the city of ' + n
        elif x['ENGTYPE_1'] == 'AutonomousRegion':
            n = 'the ' + n + ' Autonomous Region '
    if s in labels:
        # If we get here, then the relevant region has several pieces, and we 
        # already have a label for the first piece.  Subsequent pieces are 
        # always much smaller and we do not label them.
        s = ''
    else:
        labels.append(s)
    # Note that slightly strange syntax is needed to assign to a cell in a
    # dataframe.  
    china.at[idx, 'label'] = s
    china.at[idx, 'long_name'] = n

We now change the position of some labels slightly to make the map more readable.

def shift_label_pos(region, dx, dy):
    # Find the line number of the row with the given region
    i = china.loc[china['NAME_1'] == region].index[0]
    # Add dx and dy to the coordinates of the label
    v = china.at[i, 'label_pos']
    w = [v[0] + dx, v[1] + dy]
    china.at[i, 'label_pos'] = w

shift_label_pos('HongKong', 1, -1)
shift_label_pos('Macau',   -1, -1)
shift_label_pos('Hebei',    0, -1)

def show_region(region):
    """
    Plot the map of China with the specified region highlighted.  The region
    can be specified in English or Chinese, and it is not case-sensitive.
    """
    r = china.loc[china['NAME_1'].str.lower() == region.lower()]
    if len(r) == 0:
        r = china.loc[china['VARNAME_1'].str.lower() == region.lower()]
    if len(r) == 0:
        r = china.loc[china['label'].str.lower() == region.lower()]
    if len(r) == 0:
        raise ValueError(f'No region named {region}')
    fig, ax = plt.subplots(1, 1, figsize=(20, 10))
    ax.axis('off')
    ax.set_title('Map of China with ' + r['long_name'].values[0] + ' (' + r['label'].values[0] + ') highlighted')
    # Plot all the regions in light grey
    china.plot(edgecolor='darkgrey', facecolor='lightgrey', ax=ax)
    # Plot the selected region in light blue
    r.plot(facecolor='#9999ff', ax=ax)
    # Add text labels to all regions
    china.apply(lambda x: ax.annotate(text=x['label'], xy=x['label_pos'], ha='center'), axis=1)

show_region('Heilongjiang')