Insights into the quantification and reporting of model-related uncertainty across different disciplines

Abstract

Quantifying uncertainty associated with our models is the only way we can express how much we know about any phenomenon. Incomplete consideration of model-based uncertainties can lead to overstated conclusions with real-world impacts in diverse spheres, including conservation, epidemiology, climate science, and policy. Despite these potentially damaging consequences, we still know little about how different fields quantify and report uncertainty. We introduce the "sources of uncertainty" framework, using it to conduct a systematic audit of model-related uncertainty quantification from seven scientific fields, spanning the biological, physical, and political sciences. Our interdisciplinary audit shows no field fully considers all possible sources of uncertainty, but each has its own best practices alongside shared outstanding challenges. We make ten easy-to-implement recommendations to improve the consistency, completeness, and clarity of reporting on model-related uncertainty. These recommendations serve as a guide to best practices across scientific fields and expand our toolbox for high-quality research.

Keywords: Statistical physics.

Figure 1

Heatmap of the percentage of papers that report uncertainty from each source split by field Positive results include papers that quantified and reported uncertainty arising from each source, when uncertainty was applicable. Cases, where no uncertainty was present in a source, were removed. Only one focal model was considered per paper assessed. (N = 66 for Climate Science, 89 for Ecology, 55 for Evolution, 33 for Health Science, 88 for Neuroscience, 38 for Oceanography, and 81 for Political Science).

Figure 2

Distribution of different model types by uncertainty component and field (A) Presentation of the percentage of models in which there was uncertainty reported, uncertainty missing, or no applicable uncertainty for each source component (uncertainty was deemed not applicable if either the component was not relevant to the model or if there was no uncertainty in that component). (N = 57 for dynamical models (including mechanistic), 241 for statistical models, 12 for theoretical models or hybrid statistical/theoretical models). (B) Percentage of model type assessed by field.

Figure 3

Venn diagrams of presentation methods for uncertainty (legend in the center) by field Black edged segments are our good practice recommendation of visual + numeric or visual + numeric + text.