Invited perspective: Uncertainties in natural systems may be uncomfortable, but ignoring them would be absurd

Uncertainties in natural systems are pervasive, varied, and unavoidable due to inherent open system complexity and limited knowledge. Therefore, the evolution of a natural system cannot be predicted deterministically and probabilistic forecasts are commonly used to account for these uncertainties. As the Voltaire-inspired title suggests, representing and quantifying all uncertainties in hazard and risk forecasting is challenging yet essential for an effective risk management cycle and for a meaningful scientific evaluation of forecasting models. Although this paper focuses on hazard forecasting, we argue that the discussion and treatment of uncertainty apply equally to vulnerability and, therefore, to risk assessment. The most important challenges are reflected in the current absence of a common hierarchy of uncertainties, of a shared quantitative procedure to include all uncertainties in a forecast, and of effective communication and decision-making protocols, across different hazards and risks. Deepening the understanding of these distinct challenges has been the main goal of a dedicated task force of scientists from different disciplines, experts in communication, and decision-makers in the framework of a large Italian project on multirisk under NextGenEU funds – the RETURN project (https://www.fondazionereturn.it/en/, last access: 5 October 2026) – which includes eighteen Italian universities and research centers, the Italian Civil Protection Department, Italian State Railways, Assicurazioni Generali, other profit entities, and one Italian River Basin Authority. Within this initiative, we examined several examples of natural hazard forecasting and projections, from the perspectives of experts in various fields and/or users of these forecasts. The task force found that different hazards share key features and challenges regarding uncertainty definition, understanding, quantification and communication, which may be embedded in a common framework. Such a framework introduces the concept of a “complete forecast”, defined as a forecast that explicitly represents and keeps distinct all recognized uncertainties. This distinction is essential for the proper evaluation of forecasting models. This work categorizes the common key scientific and communication challenges, propose potential solutions, and intend to stimulate a deeper reflection on these issues.