Nov 4, 2021

Visualising World Temperature Data

I like to start with the end result of a project, just to show if this is what you’re looking for before you read:

Some caveats:

These data are limited to what a 10 second google search yielded (I found it on Kaggle)
Data are approximately matched to country codes in order to streamline things
Year range of data is very limited
Code is written in an obtuse way in order to make clear what is happening
- There is little need for multiple lists matching data codes for example
Some countries are missing data, easily solved with better country code matching and a better data source

Libraries required for this:

pandas
plotly
numpy (probably)
pycountry (because data source lacked these codes that plotly uses)

With proper data, this should really be a country agnostic plot. I’ll update this when I have time to find data…

Okay, so this pretty much works like any data analysis/visualization. First we load in the data,

import plotly.graph_objects as go
import pandas as pd

df = pd.read_csv('temp_data.csv').melt(id_vars=['year'], var_name='country', value_name='temperature')

, because the data are column based we just have to melt it so that each country has a year and a temperature per row.

Using pycountry we can create a dictionary to match up country names in our data to their codes used in plotting per country,

import pycountry
keys = [list(pycountry.countries)[i].alpha_3 for i in range(len(pycountry.countries))]
names = [list(pycountry.countries)[i].name for i in range(len(pycountry.countries))]
conv = { v:k for k,v in zip(keys,names) }
def match_name(x):
    try:
        return conv[x]
    except KeyError as e:
        for  n in names:
            if x in n:
                return conv[n]
        return 'N/A'
df['codes'] = df.apply(lambda x: match_name(x['country']), axis=1)

Finally, using plotly sliders, we can just create a Choropleth plot for each year, set it to not being visable and create a master figure out of them all.

And thats that.

years = df['year'].unique()
years.sort()
years_sliders = []

for year in years:
    tdf = df[df['year'] == year]

    years_sliders.append(go.Choropleth(
        visible=False,
        locations = tdf['codes'],
        z = tdf['temperature'],
        text = tdf['country'],
        colorscale = 'RdBu',
        autocolorscale=True,
        reversescale=False,
        marker_line_color='darkgray',
        marker_line_width=0.5,
        colorbar_tickprefix = 'C',
        colorbar_title = 'Temperature',
        zmin=-df['temperature'].min(), zmax=df['temperature'].max()
    ))

fig = go.Figure(data=years_sliders)

steps = []
for i in range(len(years)):
    step = dict(
        method="update",
        args=[{"visible": [False] * len(fig.data)},
              {"title": "Year: " + str(years[i])}],
              label=str(years[i])  # layout attribute
    )
    step["args"][0]["visible"][i] = True  # Toggle i'th trace to "visible"
    steps.append(step)

sliders = [dict(
    active=len(years),
    currentvalue={"prefix": "Steps: "},
    pad={"t": 50},
    steps=steps,
)]

geo=dict(
        showframe=False,
        showcoastlines=False,
        projection_type='orthographic'
    )
fig.update_layout(
    sliders=sliders,
    geo=geo
)
fig.data[0].visible = True

fig.show()