In Europe One Degree Warmer we collected local temperature data from 117 years, in 558 cities across Europe, to show how much each city had warmed since last century. This allowed us to do local reporting on global climate trends.
Parallel with collecting temperature data, we summarized alla scientific data we could find on how temperature affects local society; Ranging from invasive species to effects on school exam results. That enabled us to translate the local temperature differences to more tangible things.
Because our focus was on the local stories, we published our findings as separate, robot generated texts in nine language for each city, producing in total 5,022 articles, with over 60,000 maps and charts to accompany them.
The data has been updated in april 2019 to add 2018 temperatures, and we indent to make this a living database, to provide an accessible interface to a complex dataset.
What makes this project innovative?
Finding a local angle to a global story, and doing it on a large scale was the key innovation here. By focusing on local data, and on historic data rather than predictions about the future, we also has a powerful foundation for interviewing even those local politicians who would usually deny global climate change. The scale of this project meant that we had to run all analysis in the could, using rented servers from Amazon. Using robot generated texts to present the findings was well suited to the global-data-made-local approach. Creating such texts in multiple languages is, however, challenging, and this is an area where technology is still rapidly evolving. We had to custom make a Node JS language library (working mostly language data from the Unicode CLDR database) to assist in the translation.
What was the impact of your project? How did you measure it?
EDJNet parters, including temporary partners for this particular story, published 44 articles following the release. Other news outlets published hundreds of stories, ranging from simple rewrites to their own takes on the data. After the initial peak when publishing, the 5,022 city reports has seen a significant long tail effect, with relatively small numbers but no sign of decrease in the number of visitors or page views after six months. For some countries, primarily in the Nordics, the initial publication was overshadowed by a serious data error that affected numbers for 38 cities. The error was discovered shortly after publishing, and the following week was spent: - re-calculating the data for these cities - reaching out to every outlet that we could find that has published stories with data errors, both those in EDJNet, and others. We believe that we were able to reach the vast majority of those who published, via phone and/or mail, over the following 48 hours. In the end, these errors lead us to overhauling our internal processes, and to put methods and routines in place to avoid tha anything like it happens again. Some of the conclusions have been summarized here: https://medium.com/european-data-journalism-network/europe-one-degree-warmer-how-we-got-things-wrong-and-are-working-on-fixing-them-67c9c892c13c
Source and methodology
The input data was some 100,000,000 data points made available by the European Centre for Medium-Range Weather Forecasts (ECMWF), an international organization which computes so-called “re-analyses” of weather data, based on a variety of sources such as weather stations, weather balloons, buoys and satellite observations. Such data is well-suited to study weather patterns over periods spanning over a century, because it harmonizes inputs from thousands of data sources and makes comparisons in time and space possible. Two datasets have been used for the analysis. The datasets are not directly comparable, and we have used a reconciliation algorithm to be able to make historical comparisons. A full description of the method used us available here: https://gitlab.com/edjn/onedegreewarmer_method/blob/master/method.md The result was presented as localized reports for each city.
All analysis was done in Python, and executed on Amazon EC2 instances. Reports where created with Node JS, and rendered to static web pages stored on Amazon S3, to allow virtually unlimited traffic. A custom built language-library made it possible for non-programmers to translate the reports into grammatically very different languages. Charts were created using J++' NewsworthyCharts library, and maps using the Mapnik library, and geo data from Natural Earth.
Nicolas Kayser-Bril (freelance) Leonard Wallentin (J++) Gian-Paolo Accardo (VoxEurop) Lorenzo Ferrari (OBC Transeuropa) Marzia Bona (OBC Transeuropa) Giorgio Comai (OBC Transeuropa) Anze Bostic (Pod črto) Chiara Sighele (OBC Transeuropa) Claire Alet (Alternatives Economiques)