Weathering Natural Disasters: Forecasting, Anticipation, and "Out-of-Model Uncertainties" in the Humanitarian Sector

While humanitarian aid is conventionally reactionary, in which aid is delivered after the occurrence of natural disasters, new anticipatory mechanisms – based on improved weather forecasts – are increasingly making their way into the humanitarian sector. Part of the predictive promise that underpins these anticipatory mechanisms is that by reducing meteorological uncertainty – through establishing thresholds and triggers and by putting standard operating procedures in place – an automated cascade of improved actions will follow. While much attention has been paid to understanding, quantifying and reducing uncertainties that are inherent to weather forecasts, much less attention has been paid to the uncertainties that emerge in the implementation chain. Drawing on examples from pilot studies in Africa, this paper addresses the complex and multifaceted ‘out-of-model’ uncertainties that are part of decision-making under uncertainty. We argue that model predictions, notwithstanding their (improved) accuracy, can never provide a unique answer to what constitutes effective action in every context. Therefore, more insight is needed into which courses of action are being privileged in current forecast-bases initiatives and which ones are being marginalised. Furthermore, predictive models can never fully capture the performative dimensions which their usage by human actors entail. We therefore call for greater attention to the complex and multi-faceted nature of uncertainties arising after the modelling stage, where human expertise, judgement, and cultural sensitivity can never be replaced by standard procedures that exclusively rely on computational model outputs.