2070. Communicating model uncertainty for decision support in seasonal climate forecasting services
Invited abstract in session TB-61: Addressing behavioral issues of real-world problems 2, stream Behavioural OR.
Tuesday, 10:30-12:00Room: Maurice Keyworth G.31
Authors (first author is the speaker)
| 1. | Andrea Taylor
|
| Leeds University Business School, University of Leeds |
Abstract
Climate information services aim to support organisations in making choices about how to manage climate variability and change. At seasonal timescales climate forecasts might be used to inform a range of preparedness decisions across different sectors. However, uncertainty arises in this information due to both the probabilistic nature of the forecasting models themselves, and the fact that forecasting models do not perfectly reflect physical systems. As forecast models sometimes underperform historical observations as an indicator of future conditions, it is important that users understand both forecast probabilities and model performance (‘skill’). We report on findings of experiments with engaged climate information users (n=84) and European decision makers in climate sensitive sectors (n=122), exploring different approaches to communicating forecast likelihood and forecast skill. Participants were presented with visualisations with probabilities represented using map icons, tables, violin plots and bar graphs, and forecast skill with numeric values, verbal categories and colour-saturation. Objective understanding, perceived usefulness, preference and familiarity were assessed. We find that 1) tables are best understood with respect to likelihood information, 2) preference is positively associated with familiarity but not understanding, and 3) non-specialists struggle to interpret low-skill forecasts. We discuss practical implications for communication.
Keywords
- Behavioural OR
- Decision Support Systems
Status: accepted
Back to the list of papers