Oral Presentation Australasian Groundwater Conference 2017

Qualitative uncertainty analysis: why we need to slow our thinking about groundwater models (#148)

Luk Peeters 1 , Russell Crosbie 1 , Warrick Dawes 2 , Kate Holland 1 , Sreekanth Janardhanan 3 , Matthias Raiber 3 , Tao Cui 3 , David Rassam 3 , Andy Wilkins 4 , Natasha Herron 5 , Tim Evans 6 , Tim Ransley 6 , Chris Turnadge 1
  1. CSIRO Land and Water, Urrbrae, SA, Australia
  2. CSIRO Land and Water, Floreat, WA, Australia
  3. CSIRO Land and Water, Dutton Park, QLD, Australia
  4. CSIRO Energy, Pullenvale, QLD, Australia
  5. CSIRO Land and Water, Black Mountain, ACT, Australia
  6. Geoscience Australia, Canberra, ACT, Australia

Daniel Kahneman’s 2011 book ‘Thinking Fast and Slow’ popularised the notion that humans, including groundwater modellers, do not always think rationally, but often instead use a variety of mental shortcuts. Rational and systematic reasoning requires deliberate effort, which Kahneman associates with ‘slow’ thinking.

Applying a systematic, rational and critical mindset is essential when developing or evaluating numerical groundwater flow models, as each is inevitably based on a number of subjective choices. One way to identify and avoid mental biases and shortcuts, such as expert over-confidence or confirmation bias, is to carry out a qualitative uncertainty analysis.

A qualitative uncertainty analysis involves systematically listing and discussing model choices and assumptions. It formally evaluates the rationale of an assumption in terms of data and resources available, as well as technical challenges and available alternatives. Perhaps most importantly, a qualitative uncertainty analysis discusses how assumptions affect predictions.

Qualitative uncertainty analyses were routinely applied as an integral part of uncertainty analyses undertaken for the Bioregional Assessments Programme. The open and transparent listing and discussing of assumptions was found to be a daunting task. The evaluation of the potential effects of model assumptions on predictions was particularly challenging, as many groundwater models were found to feature non-linear interactions and feedback mechanisms.

While qualitative approaches to uncertainty analysis do not resolve subjective model choices, a plain English discussion of assumptions proved to be an invaluable tool when communicating the importance of various assumptions to clients and stakeholders. It also provided a useful starting point for interactions with reviewers. In addition, qualitative uncertainty analyses are a great platform to identifying and prioritising opportunities to reduce the uncertainty of modelled predictions.

  • We are offering awards for Career and Early Career presentations and posters. Please indicate length of time since highest degree completed.: 5 Years or more