۱- Introduction

۲- Method

۳- Case study area and data

۴- Results

۵- Conclusions

References

Abstract

Flash floods evolve rapidly in time, which poses particular challenges to emergency managers. One way to support decision-making is to complement models that estimate the flash flood hazard (e.g. discharge or return period) with tools that directly translate the hazard into the expected socio-economic impacts. This paper presents a method named ReAFFIRM that uses gridded rainfall estimates to assess in real time the flash flood hazard and translate it into the corresponding impacts. In contrast to other studies that mainly focus on individual river catchments, the approach allows for monitoring entire regions at high resolution. The method consists of the following three components: (i) an already existing hazard module that processes the rainfall into values of exceeded return period in the drainage network, (ii) a flood map module that employs the flood maps created within the EU Floods Directive to convert the return periods into the expected flooded areas and flood depths, and (iii) an impact assessment module that combines the flood depths with several layers of socio-economic exposure and vulnerability. Impacts are estimated in three quantitative categories: population in the flooded area, economic losses, and affected critical infrastructures. The performance of ReAFFIRM is shown by applying it in the region of Catalonia (NE Spain) for three significant flash flood events. The results show that the method is capable of identifying areas where the flash floods caused the highest impacts, while some locations affected by less significant impacts were missed. In the locations where the flood extent corresponded to flood observations, the assessments of the population in the flooded area and affected critical infrastructures seemed to perform reasonably well, whereas the economic losses were systematically overestimated. The effects of different sources of uncertainty have been discussed: from the estimation of the hazard to its translation into impacts, which highly depends on the quality of the employed datasets, and in particular on the quality of the rainfall inputs and the comprehensiveness of the flood maps.

Introduction

Numerous flash floods (FF hereafter) occur across the globe every year. In Europe, a yearly average of around 50 FF-related fatalities is recorded (Sene, 2013). Apart from posing a high danger to humans, economic losses in the range of several billion euros can be caused by individual events (CRED, 2016; EEA, 2010; Gaume et al., 2009). Typically, FFs occur in small to medium-sized catchments with steep slopes and prone to convective storms. This setting results in very short lag times between the causative rainfall and the discharge response in the stream channel – usually not more than a few hours (Georgakakos, 1986; Hapuarachchi et al., 2011). This extremely sudden onset is one of the main drivers for the devastating potential of FFs (EEA, 2010; Sene, 2013; Spitalar et al., 2014) since it leaves only little time to coordinate flood response measures (e.g. evacuations or installations of nonpermanent flood protection systems). Early warning systems (EWS) can help to extend the anticipation horizon of FFs; the European Environmental Agency (EEA, 2010) argues that the improvement of EWSs is the most effective measure to reduce FF impacts. One way to improve EWSs is to increase the forecast skill (e.g. by improved rainfall estimation and forecasting). Another way is to include additional information into the EWS to enhance the decision support for the emergency responders (e.g. by simulating the response to the rainfall in the stream network). The latter was the motivation of several methods that forecast the FF hazard (e.g. the return period or discharge at predefined river sections; see Alfieri et al., 2019, 2012; Corral et al., 2019; Gourley et al., 2014; Hapuarachchi et al., 2011 for reviews).