Using Sound to Represent Uncertainty in UK Climate Predictions 2009 Data:

Communication of Uncertainty

Nick Bearman

Phil JonesAndrew Lovett

Uncertainty

All data have some uncertaintybut it is often not representedor used in a particularly helpful way

People* generally are not comfortable with the concept of uncertainty

Mobile phones & cancer / carcinogenicity

Nuclear power stations & earthquakes

29 June 2011

It is something people deal with in their everyday lives:

Travelling (car / train / air)SmokingBettingPensions / investments

whether they are aware of it or not

So everyone can understand it….. in some situations

Uncertainty

29 June 2011

Uncertainty & UK Climate Predictions

There is uncertainty in future climate predictions

CCIRG91, CCIRG96, UKCIP98, UKCIP02, UKCP09

The first four only provided a single prediction value(for specific time, place, emissions scenario)

UKCP09 provides a range of values

29 June 2011 Temperature Increase

ProbabilityTemperature Increase

UKCP09 - Uncertainty

Uncertainty is usefulBut users have to change their workflow to make use of it

The users need to be able to understand the data

UKCIP have done a lot of training on this for users of UKCP09

They use the analogy of ‘horse race betting’ to explain how the probabilities work

Which people are familiar with

29 June 2011

2.5 Existing Workflow1.8, 2.0, 2.1, 2.1, 2.5, 2.6, 2.8

UKCP09 - Projections

UKCP09 data consists of projections of future climateList of probabilities and values

Need to view the value (e.g. temperature) and the probability distribution to fully understand how accurate the temperature prediction is

Large amount of information to showSo calculate a range instead (90th – 10th percentile)

Can do this visuallybut limits on what can be shown

29 June 2011

Temperature Increase

Probability

Representing Uncertainty

Why sound?Easy – most computers have sound cardsCheap – only need headphones / speakers

Played relevant sound as mouse moved over mapAsked users to highlight areas above specific threshold

Summer mean daily temperature for each cell50th percentile (≈ mean)Range (≈ uncertainty)

Used Google Maps interface for the evaluation

29 June 2011

Results – How does sound help represent uncertainty?

Participants (n = 71) from UEA, OS & UKCIP

Usefulness of sound varied widelySome found it very helpful, some couldn’t stand it

Knowledge of data set important (p < 0.001)

Using sound to reinforce vision increases scores (p = 0.005) for most

Choosing the correct sonification method is crucial for understanding the data

Only general findings, but there is potential to extend this sonification to show more data (e.g. the distribution of the probabilities)or for public engagement

29 June 2011

Thanks to participants

Visual Saturation

29 June 2011http://kbark.wordpress.com/2006/12/17/where-am-i/ (18/03/2009)

Finnish Town

Results

Compared users results to the correct resultsGreat variation within results

Cluster C had a very visual learning style

Awareness of data set is important (p < 0.001)

Using sound to reinforce vision increases scores (p = 0.005) for most

Using sound for different variable helps some but not othersLearning Style – Visual learners more effectiveNot significant trends:

Subject knowledge Repeated use