Equipped to deal with uncertainty in climate and impacts predictions:lessons from internal peer review

Wesselink, Anna and Challinor, Andrew Juan and Watson, James and Beven, Keith and Allen, Icarus and Hanlon, Helen and Lopez, Ana and Lorenz, Susanne and Otto, Friederike and Morse, Andy and Rye, Cameron and Saux-Picard, Stephane and Stainforth, David and Suckling, Emma (2015) Equipped to deal with uncertainty in climate and impacts predictions:lessons from internal peer review. Climatic Change, 132 (1). ISSN 0165-0009

Abstract

The quantification of uncertainty is an increasingly popular topic, with clear importance for climate change policy. However, uncertainty assessments are open to a range of interpretations, each of which may lead to a different policy recommendation. In the EQUIP project researchers from the UK climate modelling, statistical modelling, and impacts communities worked together on ‘end-to-end’ uncertainty assessments of climate change and its impacts. Here, we use an experiment in peer review amongst project members to assess variation in the assessment of uncertainties between EQUIP researchers. We find overall agreement on key sources of uncertainty but a large variation in the assessment of the methods used for uncertainty assessment. Results show that communication aimed at specialists makes the methods used harder to assess. There is also evidence of individual bias, which is partially attributable to disciplinary backgrounds. However, varying views on the methods used to quantify uncertainty did not preclude consensus on the consequential results produced using those methods. Based on our analysis, we make recommendations for developing and presenting statements on climate and its impacts. These include the use of a common uncertainty reporting format in order to make assumptions clear; presentation of results in terms of processes and trade-offs rather than only numerical ranges; and reporting multiple assessments of uncertainty in order to elucidate a more complete picture of impacts and their uncertainties. This in turn implies research should be done by teams of people with a range of backgrounds and time for interaction and discussion, with fewer but more comprehensive outputs in which the range of opinions is recorded.