Anticipating Weather and Climate Extremes: The Value of Early Action to Reduce Disaster Impacts

The latest IPCC report shows that weather and climate extremes are already intensifying. The world has to prepare for unprecedented extreme events in the future with severe consequences, such as floods, prolonged droughts and widespread food crises. The strong focus on forecasting these events has led to a sharp increase in forecast skill over the last few decades. This thesis contributes to building a similar understanding and evidence base for forecast-based early action. Throughout Chapters 2-5, two key research challenges have been addressed to advance early action further: (1) The design of effective action triggers: there is insufficient knowledge about the effectiveness of early action triggered by forecasts and their uncertainties in various contexts; (2) Acting on impact-based forecasts: the connection between forecasts and on-the-ground impacts is missing, which reduces the effectiveness of the actions taken. The thesis begins with an overview of the (economic) value of anticipating extreme precipitation events across Europe (Chapter 2). We used two early warning indicators of extreme rainfall from the European Centre for Medium-Range Weather Forecasts (ECMWF): the Extreme Forecast Index (EFI) and Shift of Tails (SOT). We calculated the Potential Economic Value (PEV) of the forecasts to trigger early actions, hereafter simply referred to as 'value'. We found that the forecasts can effectively trigger early action up to 2-3 days in advance across much of Europe, with the value largely disappearing at a 5-day lead. In Chapter 3, we applied the anticipation of ECMWF extreme rainfall forecasts to a specific context. Adaptation to extreme weather events in cities is a major challenge, which led the city of Amsterdam to develop 'blue-green roofs'. Using a hydrological blue-green roof model, we found that these roofs can capture 70-97% of extreme rainfall events (>20 mm/h). This can significantly reduce the pressure on the drainage system and contribute to reduce the risk of urban flooding. The potential of early action for droughts is explored in Chapters 4 and 5. In Chapter 4, we identified total seasonal rainfall as the most impactful rainfall indicator, using a unique on-the-ground impact dataset from the National Drought Management Authority (NDMA) in Kenya. We used ensemble forecasts of total seasonal rainfall from the ECMWF SEAS5 system ahead of the March-April-May (MAM) and October-November-December (OND) seasons. We found that the OND forecasts provide value for decision-making ahead of the rainy seasons. Food insecurity is one of the most important drought impacts, yet it is difficult to predict due to its many different drivers. Therefore, in Chapter 5, we developed a food security early warning model using the XGBoost machine learning algorithm. We evaluated the model predictions over the period 2019-2022, which includes a widespread (drought-induced) food crisis. We found that food security can be effectively predicted one (R2 = 0.72) to three (R2 = 0.67) months in advance. Throughout Chapters 2-5, we demonstrate the value of early action across different types of risk, socio-economic contexts, and spatial locations. Our findings lead to a straightforward conclusion: early action is effective and holds significant value in reducing the impacts of floods and droughts, but only if the triggers are carefully selected and evaluated. Chapter 6 emphasizes that it is impossible to make significant progress in the field without more and better data on impacts, including interventions. We call on the United Nations to lead this effort. The field of early action is evolving rapidly; therefore, integrating the latest research into operational procedures and guidance documents, such as the Early-Action Protocols and WMO’s Regional Climate Outlook Forums, is crucial. Achieving this can lead to a safer world through more effective early action.