In [1]:
import pydeck as pdk
import pandas as pd

Plotting lights at night¶

NASA has collected global light emission data for over 30 years. The data set is a deeply fascinating one and has been used for news stories on the Syrian Civil War [1], North Korea [2], and economic growth [3].

In this notebook, we'll use a deck.gl HeatmapLayer to visualize some of the changes at different points in time.

Getting the data¶

The data for Chengdu, China, is cleaned and available below. Please note this data is meant for demonstration only.

In [2]:
LIGHTS_URL = 'https://raw.githubusercontent.com/ajduberstein/lights_at_night/master/chengdu_lights_at_night.csv'
df = pd.read_csv(LIGHTS_URL)
df.head()
Out[2]:
year lng lat brightness
0 1993 104.575 31.808 4
1 1993 104.583 31.808 4
2 1993 104.592 31.808 4
3 1993 104.600 31.808 4
4 1993 104.675 31.808 4

Setting the colors¶

pydeck does need to know the color for this data in advance of plotting it

In [3]:
df['color'] = df['brightness'].apply(lambda val: [255, val * 4,  255, 255])
df.sample(10)
Out[3]:
year lng lat brightness color
106941 2001 103.933 30.292 5 [255, 20, 255, 255]
165346 2013 104.192 30.592 58 [255, 232, 255, 255]
118864 2003 104.267 31.433 3 [255, 12, 255, 255]
53787 1995 104.308 29.600 5 [255, 20, 255, 255]
316599 1999 103.833 29.567 6 [255, 24, 255, 255]
136067 2003 103.792 30.392 7 [255, 28, 255, 255]
87640 2001 104.975 31.592 4 [255, 16, 255, 255]
216237 2007 105.708 29.567 3 [255, 12, 255, 255]
67615 2009 103.592 30.775 8 [255, 32, 255, 255]
22356 1997 103.808 30.908 5 [255, 20, 255, 255]

Plotting and interacting¶

We can plot this data set of light brightness by year, configuring a slider to filter the data as below:

In [4]:
plottable = df[df['year'] == 1993].to_dict(orient='records')

view_state = pdk.ViewState(
    latitude=31.0,
    longitude=104.5,
    zoom=8)
scatterplot = pdk.Layer(
    'HeatmapLayer',
    data=plottable,
    get_position=['lng', 'lat'],
    get_weight='brightness',
    opacity=0.5,
    pickable=False,
    get_radius=800)
r = pdk.Deck(
    layers=[scatterplot],
    initial_view_state=view_state,
    views=[pdk.View(type='MapView', controller=None)])
r.show()
Out[4]:
In [5]:
import ipywidgets as widgets
from IPython.display import display
slider = widgets.IntSlider(1992, min=1993, max=2013, step=2)
def on_change(v):
    results = df[df['year'] == slider.value].to_dict(orient='records')
    scatterplot.data = results
    r.update()
    
slider.observe(on_change, names='value')
display(slider)