How Reuters covered the Sanchi oil tanker disaster in data and graphics

Insights into the Sanchi oil tanker projects created by the Reuters graphics team in Singapore

Sponsored content

On January 5, an Iranian tanker carrying almost a million barrels of oil collided with a Chinese freight ship in the East China Sea. The Sanchi tanker burst into flames and continued to burn as it drifted South without power. Nine days later, the ship eventually sank after a large explosion, leaving a large amount of oil on the surface. All 32 on board perished.

From breaking news to in-depth explainers, this is how the Reuters graphics team in Singapore responded to the story. The authors of the pieces mentioned below were Simon Scarr, Deputy Head of Graphics, and graphics journalists Jin Wu, Dawn Cai and Christian Inton.

Breaking news

The burning ship was floating around 200 km from shore with a safety perimeter around the vessel, as China’s coast guard battled the raging fire. Information from the scene was difficult to source so we got to work figuring out what material we could use.

First, we needed to find the ship, and figure out where it was likely to go. The ship was basically a drifting inferno, so we figured it would show up clearly on nighttime imagery and we were right. We turned to NASA’s VIIRS Nighttime Imagery, and used this as our base layer to track the ship’s movements. Serendipitously this also turned out to be a powerful way to show just how bright the fire was actually burning — you could see it from space at night.

As the story developed, we were later able to confirm exact locations from Greenpeace, adding more precision to the maps and vessel tracks. This imagery was updated daily until the ship sank.

The Sanchi was carrying almost a million barrels of condensate, an ultra-light form of oil that is more flammable than normal crude. It was also carrying fuel oil used to power the ship’s engines. This is the dirtiest kind of oil, extremely toxic when spilled, though much less explosive. We felt it was important to illustrate these differences and their respective hazards as well as contextualize the total volume to give the reader a sense of scale. This was the final piece that rounded out this graphic.

Click here to see the graphic ‘A drifting Inferno’

 

How the Sanchi’s oil could spread

As this first-response stage of the disaster drew to a close, we shifted focus to possible follow-on effects, the first and most obvious of which was the impact on the environment.

The obvious problem was the large slick on the water’s surface, but experts were also concerned about leaks seeping from the wreck on the sea floor. This was the first condensate spill of this magnitude and its behavior in the water was unpredictable.

South Korea and Japan maintained that contaminated water would not reach their shores. However, drift modeling from Britain’s National Oceanography Centre (NOC) and the University of Southampton predicted the opposite.

We decided the potential drift would be the basis of our next project. We reached out to the NOC, who shared their database tracking the trajectories of 6,000 virtual oil particles across a number of ocean current scenarios in January, effectively showing how fast polluted water would travel and where it might go.

The data arrived as a batch of ASCII files, one per particle.  But to make any meaning of it we needed to look at all the particles on any given day, and we wanted it QGIS friendly.  It was time to code.

First a colleague in NY made a simple script in Node.js to load the ASCII files and convert them into a single json file with all 6,000 particles regrouped by day, so we could pull them into a project and start looking at them. At this point, we were considering a programmatic rendering of the data in a browser, but the data file was just too large.  We decided instead to bring the data into QGIS and just make images, which meant the data would need to be re-organized again.  This one we scripted ourselves in python.

Python code to clean the data:

Data cleaned and converted to a CSV, we could now drop it into QGIS with oil drops from each day on individual layers and thus animate it by controlling the visibility of those layers.

From QGIS, we exported and compressed the maps into short video files with some narrative and navigation buttons. We started picking out milestones as the projected oil reached various coastlines and called them out in the animation. This formed the main section of the project.

 

That left the currents. The strength and direction of the currents, especially major flows like the Juroshio, would ultimately determine where the oil would go. We needed to show this complexity.

There are multiple sources of ocean current data at varying resolutions. They start from 1/3 degree and can be as high as 1/32 degree resolution. Taking update frequency, geographical coverage and resolution into consideration, we settled on NASA’s JPL ECCO2 Cube92 model.

This would show the current, but we needed to map the speed which involved some calculations. Using the meridional velocity (i.e. North-South speed) and zonal velocity (East-West speed) we defined the same area of interest, date and depth, so that we could accurately join the two datasets together. We used R for cleaning and analyzing the data.

Lastly, we wanted to look into damage to the region’s rich fish reserves. The nearby Zhoushan fishing ground is one of the biggest in Asia. If edible fish were contaminated, traces of pollution could eventually be consumed by humans. We added this map showing areas fished last year in the same three months following the date of sinking, according to Global Fishing Watch, an independent non-profit organisation.

With all the data in place, we turned to the design and feel of the page. We felt the content called for a muted and dark colour palette which would establish the proper mood and allow the data to stand out. We wanted a striking image for the landing page, a visual metaphor to let the reader know what was in store as they read on. The swirling spread of the risk data itself felt the most compelling, so we turned this data into an illustration, for maximum impact, which would then fade away on scroll to reveal the actual data itself. 

Months later we learned that the projections turned out to be reasonably accurate, with traces of oil washing up on over a dozen islands in southern Japan. The long-term damage to fisheries and marine life in the area remains to be seen.

Click here to see the graphic ‘How the Sanchi’s oil could spread’.

 

A history of oil spills

Having covered the breaking news event and looked at future impacts, the last area we wanted to focus on was historical context. The Sanchi spill was the largest in decades and we wanted to give readers a sense of its scope.

While researching the spread data, we also scraped and organized any historic data we came across until we had a detailed chronology of volume and location of all major historical oil spills.

We wanted the reader to have an immediate visual/emotional response to the idea of spilling oil, so instead of a standard bar chart, we inverted the scale, allowing the oil to drip down.  We got really excited about the idea of filming real oil dripping and spent a few hours with actual oil, glass and white plastic.  In the end, this didn’t work as usable material, but it provided us with a reference point to help make a convincing animation for the bars.

We used GreenSock, an SVG animation library to let the bars run viscously down the page before the headline and introduction fade in to drive the point home.

 

We rounded out this piece with a map locating the spills and charted how the industry has become safer with less spills, particularly at sea. The whole page kept with our theme of light backgrounds with heavy black blots of data, to represent the look and feel of crude.

Click here to see the graphic

Like many events we cover, we knew at the outset that this was going to be a big story with far-reaching impact.  The key for us, in these situations, is to start thinking about the wider context as soon as the event occurs, so we can gather data, germinate ideas and build the foundations for future work, while responding to the immediate breaking news aspects of the crisis.

With this foundation in place, it becomes possible to do larger contextual pieces as thoughtful follows while the story is still ongoing, rather than just reporting the events as they happen and then getting diverted to the next big event in the news cycle.

 

Brought to you by