Abstract
Earth observation data should inform decision making, but good decisions can only be made if the uncertainties in the data are taken into account. Making sense of uncertainty information can be difficult, because uncertainties represent the statistical spread in the observations (e.g., expressed as $$x \pm y$$ x ± y ), which does not relate directly to one specific use case of the data. Here, we propose a Bayesian framework to transform Earth observation product uncertainties into actionable information, i.e., estimates of how confident one can be in the occurrence of specific events of interest given the data and their uncertainty. We demonstrate this framework using two case examples: (i) monitoring drought severity based on soil moisture and (ii) estimating coral bleaching risk based on sea surface temperature. In both cases, we show that ignoring uncertainties can easily lead to misinterpretation of the data, making any decisions based on these data unlikely to be the best course of action. The proposed framework is general and can, in principle, be applied to a wide range of applications. Doing so requires a careful dialogue between data users, to formulate meaningful use cases and decision criteria, and data producers, to provide a rigorous description of their data and its uncertainties. The next step would then be to confront the uncertainty-informed estimates of event probabilities (created by the framework proposed here) with the costs and benefits of possible courses of action in order to make the best possible decisions that maximize socioeconomic merit.