The BBC investigated how climate change had made extremely hot days above 50C a more common phenomenon globally. Our unique analysis of 15 billion datapoints found the average number of extremely hot days in a year has doubled since the 1980s and now occurs in more areas.
Global average temperatures are rising, but abstract increases of 1C and 2C are experienced differently in different places. In the hottest regions, exceptional heat is becoming the norm.
Our study of extreme temperatures shows some of the present-day impacts of a changing climate.
This data story launched the BBC Life at 50C series.
The breaking news story of our results was published on the BBC News website and received more than a million page views on its first day. The article included an animated data visualisation showing the change in the number of days over 50C over the past four decades, which was featured on social media and had high engagement. The full story was also published on 19 BBC World Service language sites, including Arabic, Persian, Punjabi, Chinese and Spanish, with hundreds of thousands of additional readers. The English version had an above-average proportion of young readers in the 18-24 age group.
The BBC World Service’s Life at 50C series, which was kicked off the same day, ultimately spanned eight short documentary films promoted on the BBC TV, News website, and YouTube channel in the lead up to the COP26 climate summit. The films showcased people’s experiences of extreme heat in a range of environments worldwide and featured the core findings of our analysis.
On the day of publication, multiple BBC broadcasts carried our data findings and clips of the first film in the Life at 50C series, including BBC Two’s flagship current affairs programme Newsnight where it led the news agenda. TV news reports and radio programmes promoted discussion of the current impacts of more days over 50C on people across the world and how an increasing rise in extremes could change livelihoods even further in future.
The data story received some very positive feedback and was shared online by leading climate scientists. The key results were also picked up and featured by a number of specialist websites covering climate change, such as Carbon Brief, as well as other media outlets across the world, for example in Nigeria and Greece.
We used the programming language Python to download the high-resolution ERA5 dataset on maximum temperatures and stored the files in the cloud-based storage space Dropbox. The data included a temperature value for every day between 1980 and 2020 in over one million grid cells covering the globe. In total we worked with more than 15 billion datapoints.
We used another programming language, R, to compute our entire analysis. Our scripts were routinely shared in GitHub to facilitate easy collaboration.
The data comprised 492 ‘netcdf’ files, which required special tools to work with. Each file contained a grid, with each cell value representing the underlying geographic area.
For every unique location, we identified the number of days when the maximum temperature exceeded 50C in all 41 years covered by our dataset. We then imported the annual data into QGIS mapping software and scaled the points by the number of days over 50C. We animated these maps for our published story, first building a prototype in Flourish. The final product was made bespoke by BBC designers and developers with accommodations made for languages, accessibility needs and low-speed internet.
Our analysis employed several common methods used to determine climatological change over time. A 30-year average baseline, known as a climatology, mediates variability in historical temperature datasets and helps measure the significance of temperature change in a separate period. We compared the average number of days over 50C for a recent period of 10 years (2010-2019) against an earlier climatology (1980-2009). We did this for unique locations as well as the global average.
To map the global change, we again used QGIS and a colour scale developed together with a climate scientist. We applied a mask to conceal sea areas and a smoothing technique to reduce the pixelated appearance of the grid cells.
What was the hardest part of this project?
The core challenge we faced in our analysis was the lack of existing and credible research on our chosen metric of 50C. The inability to cross-reference our results with similar figures by established institutions put a heavy burden on our verification processes. This was labour-intensive and time-consuming and necessitated frequent communication with a range of climate experts.
To a lesser degree, the huge scale of the dataset was challenging. We analysed all 15 billion datapoints on our local machines by running a series of R scripts specifically coded to respect the limitations of our processing power while still producing viable results.
We repeated the entire analysis multiple times. Twice we recalculated at lower thresholds, 45C and 40C, to confirm our results at 50C were consistent with the global trend and not a unique phenomenon. As days reaching 50C is still a relatively rare event globally, an analysis focused on that high value could be susceptible to higher variability and changes risked being attributed purely to chance.
We repeated the analysis again comparing a non-standard 20-year climatology (1980-1999) against a 20-year current period (2000-2019), to confirm that the trend was not unique to the most recent years. And we calculated each decadal average individually to determine that the trends were consistent. We even cross-referenced our results for specific grid cells with weather station data, where this was available and robust enough to sense-check accuracy.
Finally, we asked non-affiliated climate scientists to check our methodology and findings with different climate datasets. They were able to achieve similar and similarly significant outcomes, which ensured we were consistent with the broader body of climate science. Through this process, we were successful in providing a robust contribution to the understanding of rising extreme temperatures around the world.
What can others learn from this project?
Collaboration with multiple experts is key to producing a story of this type. The threshold of 50C we were attempting to analyse was higher than most climate scientists typically attempt, and the ones we worked with were equally keen to know our results. Their expertise was critical to choosing the appropriate dataset, developing a strong methodology and verifying the findings.
Any journalist should be sure to use the appropriate dataset for the job. We used the ERA5 reanalysis dataset because it has the most reliable measurements for the entire globe over an extended period. Reanalysis data is a combination of actual weather observations from sources such as stations and satellites with data from modern weather forecasting models. The process fills in gaps created by poor station coverage in many parts of the world and ensures a dependable longterm history of temperature measurements in all locations.
There are strengths and weaknesses to any climate dataset. In the case of ERA5, the robust global coverage meant that the area size represented by each grid cell was larger than that of a weather station. ERA5 maximum temperature records did not precisely match the weather station maximum temperature records in a given place, but they were still reflective of the average maximum weather measurements of the area.
Large datasets, such as ERA5 or other climate data, require patience and time to analyse and to verify the findings multiple ways. There may also be a learning curve to understand how to work with the specialist filetypes. But with support from experts and a strong methodology a journalist can find unique stories even in well trodden territory like climate science.
We hope our story can encourage our fellow journalists to pursue similar intensive and groundbreaking analyses of climate data at a global or regional scale.