Last week when reading Ken Jennings Maphead I discovered the Degree Confluence project (people visiting the latitude and longitude degree integer intersections). So this weekend, I spent some time investigating where these intersections are and which is the closest one to my current location using GeoPandas.

These libraries are all you need to run my analysis:

import pandas as pd
import numpy as np
import geopandas as gpd
import matplotlib.pyplot as plt
from shapely.geometry import Point, LineString
from shapely.ops import nearest_points
import mapclassify
from scipy.spatial import cKDTree

๐Ÿ”ด Confluences

In order to get the coordinates of the intersections I created lists of latitudes and longitudes, and then permute them to get all different possibilities.

lats = [lat for lat in range(-90, 100, 10)] # A 10 degree step was set to sample the research
lngs = [lng for lng in range(-180, 190, 10)]
coords = [(lat, lng) for lng in lngs for lat in lats]

points_df = pd.DataFrame(coords, columns=['lat', 'lng'])
points_geometry = [Point(xy) for xy in zip(points_df.lng,]
crs = {'init': 'epsg:4326'}
points = gpd.GeoDataFrame(points_df, crs=crs, geometry=points_geometry)
world = gpd.read_file(gpd.datasets.get_path('naturalearth_lowres'))

fig, ax = plt.subplots(figsize=(15, 15))
basemap = world.to_crs('+proj=robin').plot(ax=ax, color='#a6a6a6', edgecolor='white')
points.to_crs('+proj=robin').plot(ax=ax, marker='o', color='red', markersize=5)
ax.set_title("Confluences", fontsize=20)

I am sure you have realized that the confluences placed on the poles are duplicated as you can see in the following plot:


๐ŸŒ Grid

I wanted to experiment further with GeoPandas, Shapely and list comprehensions, so I generate a grid or reticule using as starting points the points created earlier.

meridians = [
    (Point(lng, min(lats)), Point(lng, max(lats))) for lng in lngs
paralells = [
    (Point(min(lngs), lat), Point(max(lngs), lat)) for lat in lats

meridians_df = pd.DataFrame(meridians, columns=['origin', 'destination'])
paralells_df = pd.DataFrame(paralells, columns=['origin', 'destination'])
grid_df = pd.concat([meridians_df, paralells_df])
grid_df['lines'] = grid_df.apply(lambda row: LineString([row.origin, row.destination]), axis=1)
grid = gpd.GeoDataFrame(grid_df, crs=crs, geometry=grid_df['lines'])
fig, ax = plt.subplots(figsize=(15, 15))
basemap = world.plot(ax=ax, color='#a6a6a6', edgecolor='white')
ax.set_title("Grid", fontsize=20)


๐ŸŒก๏ธ Density

In order to know the density of these points by country, I intersected them with a world country dataset.

world['count'] = world['geometry'].apply(
    lambda row: np.sum(points['geometry'].intersects(row))
world['density'] = world['count']/world['geometry'].area

As expected, Antartica was the โ€œcountryโ€ with the highest number of confluences. But if we skip this artifact, Russia (26), Canada (17) and US (12) are the top 3. What about density? If we get rid of the countries with just one confluence in their territory, Norway (3) and Indonesia (4) are the first and second countries in the list.

fig, ax = plt.subplots(figsize=(15, 15))

world.to_crs('+proj=robin').plot(column='density', scheme="fisher_jenks", figsize=(15,10), ax=ax, legend=True)
ax.set_title("Confluences Density", fontsize=20)


๐Ÿ“ Nearest

Finally, I wanted to know the nearest cities for each of these intersections. I got Natural Earths populated places as a sample and use a function from GIS SE.

url = ''
cities = gpd.read_file(url)
# from
def get_nearest(gdf_1, gdf_2):
    n_1 = np.array(list(zip(gdf_1.geometry.x, gdf_1.geometry.y)) )
    n_2 = np.array(list(zip(gdf_2.geometry.x, gdf_2.geometry.y)) )
    btree = cKDTree(n_2)
    dist, idx = btree.query(n_1, k=1)
    gdf = pd.concat(
        [gdf_1.reset_index(drop=True), gdf_2.loc[idx, gdf_2.columns != 'geometry'].reset_index(drop=True),
         pd.Series(dist, name='distance')], axis=1)
    return gdf
nearest = get_nearest(points, cities)
nearest_df = nearest[['NAME', 'distance']].drop_duplicates(['NAME'])
nearest_df = nearest_df.merge(cities[['geometry', 'NAME', 'ADM0NAME']], on='NAME', how='left')
nearest_cities = gpd.GeoDataFrame(nearest_df, geometry='geometry')

So it looks like I need to go to Wรผrzburg, in fact, this city looks pretty nice!

fig, ax = plt.subplots(figsize=(15, 15))

xlim = [5.42, 15.95]
ylim = [48.15, 55.18]

berlin = cities.loc[cities.NAME == 'Berlin']

basemap = world.plot(ax=ax, color='#a6a6a6', edgecolor='white')


berlin.plot(ax=ax, marker='o', color='white', markersize=50)
berlin.plot(ax=ax, marker='o', color='black', markersize=5)
plt.text(berlin.iloc[0].geometry.x+0.07, berlin.iloc[0].geometry.y+0.07, berlin.iloc[0].NAME, fontsize=12)

nearest_cities.plot(ax=ax, marker='o', color='pink', markersize=40)
nearest_cities.plot(ax=ax, marker='o', color='red', markersize=5)

for x, y, country, label in zip(nearest_cities.geometry.x, nearest_cities.geometry.y, nearest_cities["ADM0NAME"], nearest_cities["NAME"]):
    if country == 'Germany':
        plt.text(x-0.75, y-0.25, label, fontsize=12)

ax.set_title("My nearest confluence", fontsize=20)


The entire Jupyter notebook can be found here.