The challenge of unprecedented floods and droughts in risk management

The study addresses why flood and drought impacts continue to rise despite documented global declines in vulnerability, and seeks to disentangle how changes in hazard, exposure, vulnerability, and management shortcomings drive impact changes over time. The context is a world where climate change is projected to increase the frequency and severity of hydrological extremes and where economic damages from floods and droughts are expected to grow substantially. Prior work has emphasized increasing exposure and the importance of vulnerability reduction, but a lack of consistent empirical data has limited understanding of combined driver effects and overall risk management effectiveness. The purpose is to empirically assess how changes in these drivers between two events in the same area relate to changes in impact, and to identify conditions under which risk management succeeds or fails, especially for unprecedented events.

The article situates itself in literature documenting declining global vulnerability to climate-related hazards and emphasizes that impacts are nevertheless rising, with climate change likely to exacerbate them. Previous studies have highlighted exposure growth as a key driver of increasing losses and pointed to vulnerability reduction as essential for impact mitigation. However, the combined and relative contributions of hazard, exposure, and vulnerability have been difficult to assess empirically due to data limitations. The paper also references work on the role of infrastructure and non-structural measures, trade-offs and spillover effects across sectors, socio-hydrological feedbacks (e.g., levee and reservoir effects that inadvertently increase exposure), and cognitive biases that inhibit proactive risk management.

- Compiled a global dataset of 45 paired events (26 flood, 19 drought) occurring in the same area on average 16 years apart, spanning diverse socioeconomic and hydroclimatic contexts across all continents. - Quantified change using indicators-of-change for four dimensions: hazard, exposure, vulnerability, and management shortcomings. Each indicator reflects the difference between the first (baseline) and second event, categorized as large decrease (-2), small decrease (-1), no change (0), small increase (+1), or large increase (+2). Sub-indicators (e.g., standardized precipitation indices for hazard; number of houses affected for exposure; risk awareness for vulnerability) and a quality assurance protocol were used to minimize subjectivity. - Impact considered includes direct (fatalities, monetary damage), indirect (e.g., disruption of traffic or tourism), and intangible (e.g., health, cultural heritage) consequences. - Performed correlation analyses (Spearman’s rank) between changes in impact and changes in hazard, exposure, vulnerability, and management shortcomings separately for floods and droughts; assessed inter-correlations among drivers. - Stratified analyses by whether impact decreased or increased to interpret driver patterns; examined subsets with no change in hazard to understand roles of exposure and vulnerability. - Identified and discussed “unprecedented” events (hazard change +2), success stories where impact decreased despite increased hazard, and illustrative case studies. - Conducted sensitivity analyses to test robustness of correlation patterns and P values, noting reduced significance for drought due to smaller sample size.

- Risk management improvements are common: 71% of pairs show decreased management shortcomings; 80% show decreased vulnerability after the first event. - Correlations (floods): changes in impact correlate significantly and positively with changes in hazard (r = 0.64, P ≤ 0.01), exposure (r = 0.55, P ≤ 0.01), and vulnerability (r = 0.60, P ≤ 0.01). Hazard and exposure are inter-correlated for floods (r = 0.55, P ≤ 0.01). - Correlations (droughts): impact changes correlate significantly with changes in hazard and exposure but not with changes in vulnerability, indicating vulnerability reduction has been less effective in reducing drought impacts. - When impact decreased: typically associated with decreased hazard or no change in hazard, and vulnerability always decreased; all cases with a large decrease in vulnerability (-2) had decreased impacts. Decreases in impact often occurred despite increases in exposure in about 25% of such cases. - When impact increased: 85% of pairs had larger hazard; 46% of these involved unprecedented hazard increases (+2). In these cases, increases in exposure (92%) and only small decreases in vulnerability (69%) were common—insufficient to offset higher hazard and exposure. - Unprecedented events (hazard +2) almost always led to increased impacts, often exceeding design limits of infrastructure (e.g., sewer capacities, reservoir storage), revealing limits of structural defenses and the difficulty of managing extremes not previously experienced. - Success stories (2 flood cases) show lower impacts despite higher hazard due to improved governance, transnational collaboration, strong investment in structural and non-structural measures, and enhanced early warning and emergency response (e.g., Barcelona 1995 vs 2018; Danube 2002 vs 2013). - Illustrative quantitative examples: - Jakarta riverine floods: 2007 (RP ~50 yr) vs 2013 (RP ~30 yr); fatalities reduced 79→38; direct damage €1.3B→€0.76B due to reduced hazard, vulnerability, and exposure. - Central Europe droughts 2003 vs 2015: SPEI min -1.62 vs -1.18; impact reduced €17.1B→€2.2B with improved planning and restrictions, despite increased exposure. - Corigliano-Rossano (Italy) pluvial floods 2000 vs 2015: hazard increased (RP >100 yr vs 10–20 yr) and exposure higher (peak tourist season), leading to higher indirect damages. - North Carolina droughts 2000–2003 vs 2007–2009: exceptional drought extent/duration increased (30%/24 mo vs 65%/27 mo); crop losses €497M→€535M despite vulnerability reduction. - Piura, Peru floods 1998 vs 2017: hazard slightly decreased (max flow 3,367 vs 2,755 m³/s); improved forecasts and institutions reduced vulnerability; fatalities 366→159 despite increased exposure.

The analyses demonstrate that changes in hazard, exposure, and vulnerability jointly determine impact trajectories. For floods, all three drivers are equally important in explaining impact changes, challenging an exclusive focus on vulnerability reduction. For droughts, vulnerability changes did not significantly relate to impact changes, suggesting sectoral spillovers and complexities make vulnerability management less effective. Decreases in impact typically arose from combined reductions in hazard and vulnerability, even when exposure grew, underscoring the need for integrated strategies. Conversely, increased impacts were primarily driven by higher hazard and exposure, with modest vulnerability reductions unable to compensate. A key insight is the systemic difficulty of managing unprecedented events: when hazards exceed historical design thresholds, structural defenses often fail, and even improved management cannot fully prevent large impacts. Additionally, feedbacks and unintended consequences (e.g., levee- or reservoir-induced exposure growth) can erode benefits of hazard-reduction infrastructure. The observed positive correlation between flood hazard and exposure reflects the physical expansion of inundation areas with increasing hazard magnitude. Overall, effective risk reduction requires simultaneous attention to hazard moderation, exposure control (e.g., land-use planning), and vulnerability reduction, alongside proactive, anticipatory governance and early warning systems that are robust to extremes beyond design standards.

The paper contributes empirical evidence, across 45 global paired events, that risk management has generally reduced vulnerability and management shortcomings, but rising impacts often stem from increases in hazard and exposure, especially during unprecedented events that exceed infrastructure design limits. For floods, hazard, exposure, and vulnerability changes are equally important in shaping impacts; for droughts, vulnerability changes have been less effective. Two success cases show that strong governance, substantial investment in structural and non-structural measures, and improved early warning can reduce impacts even under increased hazard. Future directions include: adopting proactive, anticipatory risk management that accounts for tail risks and design exceedance; integrating land-use planning to constrain exposure growth; enhancing cross-sectoral coordination to mitigate drought spillovers; improving data and standardized indicators to reduce subjectivity; and developing adaptive, flexible infrastructure and governance to cope with unprecedented extremes under climate change.

- Data heterogeneity: availability and quality varied across cases and regions, potentially introducing inconsistency and subjectivity. - Indicator approach: relied on categorical indicators-of-change with sub-indicators; while a quality assurance protocol was used, residual subjectivity remains. - Sample size and balance: limited number of cases where only a single driver changed constrains attribution; smaller drought sample reduced statistical significance in sensitivity tests. - Unprecedented classification: designation of hazard +2 as “unprecedented” is subjective and context-dependent. - Generalizability: case-based empirical synthesis may not capture all regional dynamics or sectoral spillovers, especially for drought.