The rise of red zones of risk

Category: Best data-driven reporting (small and large newsrooms)

Country/area: Australia

Organisation: ABC News

Organisation size: Big

Publication date: 23/10/2019

Credit: Inga Ting, Nathanael Scott, Alexander Palmer, Michael Slezak

Project description:

This project shows how changing climate risks (such as bushfire, flood, subsidence, inundation and extreme wind) are set to impact the cost and availability of insurance between now and the year 2100. 

Based on an analysis of every address in Australia (15 million), the project visualises the projected impact for almost every suburb in Australia in numerous ways.

It also includes an interactive search that allows users to explore and compare the data for their local area.

Impact reached:

By putting proprietary data into the hands of the public, this story empowered a general audience to participate in a debate that directly affects every resident of Australia, but that had previously been dominated by a handful of powerful insurance companies and big banks.

It injected a level of transparency and accountability into debates about rising insurance premiums, and enabled “ordinary Australians” to understand and explore the numbers and calculations that drive wider business decisions about their homes and neighbourhoods
By building an interactive tool that gave users the opportunity to “drill down” into detailed, 80-year projections of climate risk for their local area, this story opened up the world of climate risk, insurance and property prices to an entirely new audience.

This information is critically important not only to banks and insurance companies faced with mounting costs, but governments and homeowners making critical decisions about which parts of Australia will be safe to live in an increasingly unstable future.

It’s importance has also been highlighted by the recent unprecedented bushfire season, which has destroyed numerous properties and again focused attention on the urgent question of how to insure against extreme weather disasters.

Techniques/technologies used:

Maps of Australia pose significant problems for data visualisation because we have densely-populated coasts and a vast, sparsely-populated desert interior. This means the “zoomed out” view of the country is visually dominated by areas with almost no inhabitants, while the tiny regions that are the most important to the most people are next to impossible to see.

Due to this problem, our team built a plug-in to step users through a video of an SVG-based choropleth map. This “guided tour” allowed us to highlight Australia’s worst-affected areas even if they were geographically small.

We also created a tool that allowed users to:
1. Search the dataset for personalised projections for their local area, and 
2. Compare their local area to the projections for neighbouring areas. 

This required setting up a database in Firebase to perform a three-way match with the user’s suburb or postcode and their larger geographic region (what most users think of as their “neighbourhood”).  

Each suburb contained either partially or fully within this larger “neighbourhood” was displayed on an interactive line chart. Users could use this line chart as both a visual comparison tool or exploration tool to find detailed projections for other suburbs either in their “neighbourhood” or elsewhere in Australia.

The graphs associated with the suburb search were built using the D3 library. A custom React wrapper was used to group neighbouring suburbs. With nearly 11,000 suburbs included in the dataset, finding solutions for animating and exploring the visualisation required creative solutions that didn’t crash the browser.

Data organisation, cleaning and blending was done in Excel and Tableau Prep. Analysis and “proof of concept” visualisations (including the maps, scaterrplots and area charts) were done in Tableau Desktop.

What was the hardest part of this project?

At the heart of the story was a proprietary dataset intended for technical users in the insurance, banking and risk analysis industries. The hardest part of the project was working out how to distil this vast, multi-dimensional, technical dataset into a compelling narrative for a lay audience.

It took weeks of discussion and further analysis to come up with a “headline” measure that would immediately convey the importance of the data whilst still being meaningful and accurate.

This was achieved through a process of brainstorming what we thought readers would want to know from the data, then working out if we had the variables/calculations to provide an answer. For example:

  • The key question: “How much could my insurance premium rise?” could be answered by calculated the percentage change in the average risk fraction across all five hazards
  • “Is my house likely to become uninsurable?” could be answered by examining the number and percentage of addresses in a suburb that exceed the 1% threshold for average risk fraction across all hazards
  • “What are the climate risks in my local area?” could be answered through a breakdown of average or total risk cost for each of the five hazards

Once we had identified the what readers may want to know and the relevant measure to answer their questions, we brainstormed potential visuals for presenting the data, listing the strengths and weaknesses for each, including whether interactivity would help or hinder the readers’ search for relevant information (see attached “Visuals list”). 

This process enabled us to decide on a combination of interactive and static graphics, which first gave readers a “big picture” understanding of the issue, before allowing them to “drill down” into the data to find insights relevant to their specific location.

What can others learn from this project?

This story is an outstanding example of data-driven explanatory reporting. It showcases multiple ways of visualising and explaining complex, technical data to a general audience, including:

  1. Making data explorable, so they can find areas relevant and interesting to them

  2. Re-framing potentially dry, “boring” concepts like insurance calculations and risk analysis in ways that immediately convey their importance to individual readers

  3. Using different visual forms (e.g. maps, scatterplots, area charts, line charts, radar charts, etc.) to explore different “stories” within the data; and choosing the appropriate visual form for each story

  4. Creative approaches to handling vast amounts of data

  5. Reaching users where they are (in our case, optimising all our content for users on mobile phones. These users now make up more than half our entire audience.)

Project links: