About the data

Through the analysis of data made available by the European Centre for Medium-Range Weather Forecasts (ECMWF), and by harmonising historical time series with recent data obtained from a variety of sources (satellite, weather stations, buoys, weather balloons), the resulting time series should be more robust to changes on the ground (e.g. if a weather station was on the outskirt of a city, but then new buildings and roads were built around it, temperatures for more recent years would likely overestimtate the increase in temperature; not with this method). Check out this piece published on EdjNet’s website, or check out the code published as a Jupyter Notebook to find out more about all the details of data processing.

Before diving into the data, here’s a few more details that should be kept in consideration:

• the figure for each city is obtained by averaging out data points for areas of about 80 square kilometers. As a consequence, the data refer to the city and its surroundings, which likely has a substantial impact on the data, in particular for cities that are along the coastline or located next to high mountains.
• throughout the investigation, the authors compare 20th century average data with 21st century average data, so when they say “one degree warmer” they mean that the average temperature for the period 2000-2017 is one degree higher than the average temperature for 1900-1999.² This is rather unusual, as most reporting on climate breakdown refers to increase in the temparature since “preindustrial levels”, which mostly refers to the second half of the 19th century.³ While this choice likely reduces the figure on warming when compared to other studies that refer to “preindustrial levels”, this also means that having the whole time series we can decide to use other cutoff dates, and to compare - for example - the first 30 years of the dataset with the last 30. In this post, I’ll however stick with the original definition.

A quick note on contents before giving way to the code: in each and all cities included in this investigation the temperature has increased between the 20th and 21st century, in many by more than one degree. Broadly speaking, this seems to be in line with the worrying analysis presented by IPCC in early October.

Get and clean the data

A spreadsheet with temperature per year and location has been made available by EdjNet. The data for each city have been published as separate sheets, so I have first downloaded the data as an .xlsx file to speed up processing, and then merged all the data in a single data frame in the long format. Selected indicators for each of the cities has been published in a separate spreadsheet. This second file is not needed to create the graphs included in this post, but it can be useful to check if everything looks in order with the data, or to place the data on a map, as it includes also longitude and latitude for each city.

N.B.: All of the the code is available as a repository on GitHub

staticOneCity +
  transition_reveal(id = City, along = Date)

staticOneCity +
  transition_reveal(id = City,
                    along = Date,
                    range = as.numeric(c(1900, 2080)))

if(!require("av")) install.packages("av") 

animate(animatedOneCity,
        renderer = av_renderer('animation.mp4'),
        width = 1920,
        height = 1080,
        res = 250,
        fps = 25,
        nframes = 300)