ETC5523: Communicating with Data

Case Studies

Lecturer: Kate Saunders

Department of Econometrics and Business Statistics



Aim

  • Revise principles and practices for data based communication
  • Review four different case study examples
  • Evaluate if those case studies effectively convey intended messages to viewers

Case study

Daily maximum temperature in selected Australian locations

Data from the Bureau of Meteorology

Daily maximum temperature from beginning of 1910 to mid 2021 for 7 Australian location (Perth, Adelaide, Cairns, Brisbane, Sydney, Canberra, and Melbourne).

date max_temp station
1910-01-01 26.7 Perth
1910-01-02 27.0 Perth
1910-01-03 27.5 Perth
1910-01-04 24.0 Perth
1910-01-05 24.8 Perth
1910-01-06 24.4 Perth

Daily maximum temperature

Very long time series… what’s the intended message?

Recent maximum temperature trend

Lots of fluctuation in temperature… does it show the trend?

Recent maximum temperature

This is showing the fit from a GAM model… but we lost understanding the underlying variation in the original data

Recent maximum temperature

Some models don’t look like a good fit… let’s examine individually

Recent maximum temperature

The legend is redundant here… remove unnessary elements.

Recent maximum temperature

Using loess with span of 0.3 instead … comparison across location is harder when in data shown in different panels

Recent maximum temperature

Casting a shadow of the fitted model across locations in each panel makes it easier to compare …

Recent maximum temperature

There were missing values in data that wasn’t apparent before!

Weekly average maximum temperature

By taking weekly averages, the plot is smoother and you can focus trends in the data instead of getting distracted with daily fluctuations.

Maximum temperature by month

This plot makes it easier to see the monthly variation…

Can you think of other ways to visualise this data?

Case study

Communicating Climate Change

Line Graph

Is it clear the climate is changing?

Line Graph

Let’s add a geom_smooth() - Is the trend clear now?

Warming Stripes

Sydney #showyourstripes

Graphic by Ed Hawkins: https://showyourstripes.info/

Labelled Strips

Do we need to know what the exact temperature is to get the message?

Bars

Do bar size and summary text help convey change is happening?

Bars with scale

What are the advantages / disadvantages to making this graphic more scientifically accurate?

Climate Spiral

Interactive version shown at the Rio Olympics in 2016

Does animation help convey the climate is warming?

https://climate.nasa.gov/climate_resources/300/video-climate-spiral-1880-2022/

Case study

Brisbane Floods, February - March 2022

Predicted Extent

What works and what does not work about this warning communication? Source: https://www.brisbane.qld.gov.au/ (now expired)

Zoomed In

Does this map help people to understand if they will be flooded?

Search Addresses

Does using a table to display the data make it clearer?

Source: https://www.brisbane.qld.gov.au/ (now expired)

Probabilistic Words

People interpret probabilistic words in different ways

Likely roughly corresponds to 50% - 70%.



Source: https://hbr.org/2018/07/if-you-say-something-is-likely-how-likely-do-people-think-it-is

Gaps

Unclear

  • Will my home by impacted?

  • Will I be isolated by flood-waters?

  • Can I evacuate?

Warnings were inadequate

  • Breakdown in warning communication

  • But where exactly?

  • Clear failed data based communication was a factor

Flow of data and information

Learning from the Brisbane Floods

Warning communication

  • Needs to help people make a decision

  • Encourage people to take the right actions

Future looking

  • Scope to improve how we communicate warnings with data

  • This will help better support people during disasters

Case study

New South Wales & Victorian Flooding, October 2022

Warning Levels

https://hazardwatch.gov.au/

Decision making under stress

Difficult to make decisions in emergency situations.
People don’t always make good decisions under stress.



Categorical Data Helps

  • People are better at thinking in categories
  • Categories are less complex than probabilities
  • Categories can also help when communicating uncertainty

VicAlert - 2022-10-17

Hazard levels and related icons

VicAlert - 2022-10-17

Clear hazard messaging

Review

How do the design elements contribute to the effectiveness of the data communication? Consider

  • Colour
  • Icons
  • Polygons
  • Map
  • Pop Up Messages
  • Flashing Icons
  • Interactivity

Week 11 Lesson

Summary

  • We revised the principles and practices of data visualisation
  • We critically evaluated different examples of data communication

Contributors

  • Case study on Australian Temperature developed by Dr Emi Tanaka

  • Warming Stripes case study developed by Dr Kate Saunders

  • Natural Hazards case studies developed by Dr Kate Saunders