What Will the Weather Do? Forecasting Flood Losses Based on Oscillation Indices

Abstract

Atmospheric oscillations are known to drive the large‐scale variability of hydrometeorological extremes in Europe, which can trigger flood events and losses. However, to date there are no studies that have assessed the combined influence of different large‐scale atmospheric oscillations on the probabilities of flood losses occurring. Therefore, in this study we examine the relationship between five indices of atmospheric oscillation and four classes of flood losses probabilities at subregional European scales. In doing so, we examine different combinations of atmospheric oscillations, both synchronous and seasonally lagged. By applying logistic regressions, we aim to identify regions and seasons where probabilities of flood losses occurring can be estimated by indices of atmospheric oscillation with higher skill than historical probabilities. We show that classes of flood losses can be predicted by synchronous indices of atmospheric oscillation and that in some seasons and regions lagged relationships may exist between the indices of atmospheric oscillation and the probability of flood losses. Furthermore, we find that some models generate increased (or decreased) probability of flood losses occurring when the indices are at their extreme positive or negative phases. A better understanding of the effects of atmospheric oscillations on the likelihood of flood losses occurring represents a step forward in achieving flood resilience in Europe. For instance, improved early predictions of the indices that represent such atmospheric oscillations, or the evidence of a lagged relationship between their teleconnections and floods, can significantly contribute to mitigating the socioeconomic burden of floods.

Fluvial Flood Losses in the Contiguous United States Under Climate Change

“…Seasonal prediction of flood losses using relationships with large‐scale climate indices was proposed by Guimarães Nobre et al. (2020). Other studies forecast flood damage using heavy precipitation as predictor (Cortès et al., 2018; Pastor‐Paz et al., 2020; Van Ootegem et al., 2018; Zhou et al., 2017).…”

“…Economic losses from floods are significant; hence, many studies have focused on quantifying flood damage at scales from local to global leveraging computationally expensive numerical models or empirical models employing statistical relationships (Alfieri et al., 2017; Alipour et al., 2020; Blöschl et al., 2019; Gerl et al., 2016; Guimarães Nobre et al., 2020; Wobus et al., 2019; Zhou et al., 2017). Traditionally, flood damage forecasting employs hydrological and hydrodynamic models to quantify the extent and depth of a flood and translate these physical characteristics to economic losses using empirical depth–damage relationships (Dutta et al., 2003).…”

“…However, the application of statistical relationships to assess climate change impacts on flood-related property damage is challenging because the prediction model requires future values of variables, which are fraught by large uncertainty (i.e., land use, topography, and socio-economic conditions) (Bubeck et al, 2011) in addition to the uncertainty in climate variables (e.g., precipitation intensity and duration). Seasonal prediction of flood losses using relationships with large-scale climate indices was proposed by Guimarães Nobre et al (2020). Other studies forecast flood damage using heavy precipitation as predictor (Cortès et al, 2018;Pastor-Paz et al, 2020;Van Ootegem et al, 2018;Zhou et al, 2017).…”

Assessing the socio-economic impacts of flash floods for early warning at regional, national, and continental scales

“…For fluvial floods, several impact forecasting approaches were developed in recent years (e.g. Bevington et al, 2019;Cole et al, 2016;Dale et al, 2014;Guimarães Nobre et al, 2020). The Rapid Risk Assessment (RRA; Dottori et al, 2017) forecasts with up to 10 days ahead the economic losses, critical infrastructures, and population affected by European rivers with catchment sizes larger than 500 km 2 , based on discharge forecasts from the hydrological model 80 LISFLOOD (Roo et al, 2000;Van Der Knijff et al, 2010).…”