Submitted by Andrew Kruczkiewicz
15 Sep 2021 , last updated 16 Sep 2021

Forecast-based Financing for flash floods: a Flash Flood Confidence Index to improve flood reporting


Flash floods: an increasingly devastating hazard


As demonstrated by recent events, flash floods are a growing concern globally. In Germany, the Democratic Republic of the Congo, Indonesia, Colombia and the USA, flash floods are having greater impacts, leading to an increased strain on services and livelihoods.

Among the many different types of flood, flash floods are the deadliest. However, relative to other types, flash floods have received significantly less interest in terms of research and policy responses. This is principally due to the lack of a universally accepted definition of ‘flash flood’. The term commonly refers to rapid onset hydrological events triggered by localized, intense rainfall events, which cause an unexpected, sudden and high velocity flow in small streams, artificial waterways or as surface runoff.

However, the term can also refer to events related to the collapse of sewage systems or river banks, dam failures, and any pluvial flood events that result in surface runoff far from the river network, whether in urban or rural environments.


Challenges of impact-based forecasting for flash floods


For impact-based forecasting, the localized and sudden nature of flash floods demands a high degree of precision, in terms of the climate, geophysical and socioeconomic data that is integrated. For this reason, flash floods are more challenging to monitor, document, characterize and predict than other flood types.

In addition, there is a lack of documentation about historical flash flood events and their impacts, and the records that exist are not always of sufficient granularity. This has contributed to less predictability around flash floods, and increased the levels of uncertainty in the overall forecasting process. For the anticipatory action and Forecast-based Financing community, these high levels of uncertainty do not bode well for developing a ‘traditional’ set of standard operating procedures. However, uncertainty should not be always interpreted in a negative light. In fact, uncertainty allows us to understand and appropriately characterize error within forecasts, allowing forecasts to exist in the first place.


An index to generate flash flood historical data: a first step towards an Early Action Protocol


While progress has been made within the anticipatory action community on many hazard types, there is a growing need to address flash flood risk. Through a NASA-funded project, the Disasters Program and the sponsorship of the Group on Earth Observations, a multi-disciplinary team of researchers have been working towards the development of the first Early Action Protocol (EAP) for flash floods. We are working alongside Dr. Humberto Vergara at the University of Oklahoma and NOAA’s Cooperative Institute for Mesoscale Meteorological Studies.

One of the primary barriers for developing this EAP was understanding historical data about flash floods. We found that while historical records existed for events labelled only as ‘flood’, there were far fewer events that were labelled as ‘flash flood’ – or labelled as any particular type of flood. While it is unlikely that all events labelled only as ‘flood’ were of a single type (e.g., flash flood), it is also very unlikely that they would not include some examples of all types.

We developed a method to assign a flash flood confidence index for each event in an historical flood dataset, based on ‘text mining’ of disaster reports and a flash flood susceptibility index extracted from the geophysical properties of the location of the events. This method was first applied for historical flood events in Ecuador. 


Application for disaster risk management practitioners


The primary objective of this flash flood confidence index is to enhance risk analysis for flash floods. The approach could be used by risk management practitioners to filter likely flash flood events from historical disaster databases that currently do not assign a flood subtype. Understanding the use of the index can help train technicians who monitor adverse events in real time to identify this type of flooding, so that the correct subtype of flooding is assigned when consolidating databases. This will help decision-makers ensure that flash flood response actions are consistent with this type of hazard. The resultant subset of likely flash flood events could then be a valuable tool in developing flash flood-specific risk assessments, climate services and early warning systems.

The flash flood confidence index can also be of valuable assistance for communities, to build capacity to support an understanding of the type of hazard to which they are exposed. Furthermore, the index will support those who help limit flood damage in other ways, such as Red Cross societies, local NGOs and the media.


Recommendations


Flash floods – due to their special characteristics and causes, and the need for improved understanding – require a multi-disciplinary, multi-sectoral approach to manage and mitigate their impacts. We recommend that resources are allocated for more precise reporting of all flood events, but most importantly for flash floods. More type-specific data will improve the development of flash flood models, forecast production, early warning systems and resilience programmes.

There is also a need for in-depth studies into the causes of flash floods, especially with a view to the role of climatic changes and human alterations to catchments. Subsequently, we must develop risk-sharing mechanisms among various levels of government and individuals to strengthen the resilience of communities affected by flash floods.

This blog was written by: Andrew Kruczkiewicz and-Agathe Bucherie from the International Research Institute for Climate and Society, Earth Institute, Columbia University; Fernanda Ayala from Cruz Roja Ecuatoriana; and Juan Bazo from the Red Cross Red Crescent Climate Centre.

For more information, please get in touch with Andrew at andrewk [at] iri.columbia.edu