Flood exposure and poverty in 188 countries

Natural shocks cause hundreds of billions of dollars in losses annually, and floods are among the most pervasive hazards, particularly in lower-income countries where infrastructure and protection are limited. The paper’s core objective is to quantify, at high spatial resolution and with global coverage, how many people are exposed to significant flood risk (1-in-100-year events) and how this exposure intersects with poverty—an essential determinant of vulnerability and recovery capacity. The context includes rising risks from climate change (e.g., increasing flood frequency in parts of Africa, Asia, and Latin America; sea-level rise) and socioeconomic drivers (population and asset growth, urbanization, subsidence). Prior assessments have often been local, focused on high-income settings, or lacked the spatial resolution and scope to capture pluvial floods and secondary rivers, and they seldom integrated poverty data. This study addresses these gaps by combining high-resolution global flood hazard maps (fluvial, pluvial, coastal) with population density and subnational poverty estimates to identify global and regional hotspots where flood risks and poverty coincide, informing mitigation and resource allocation.

The paper reviews limitations of previous global flood risk assessments: (1) reliance on historical inventories (e.g., EM-DAT) lacking spatial coincidence between hazards and populations, preventing robust headcount estimation; (2) satellite-based event detection (2000–2018) that misses undetected events and populations at risk that did not experience flooding during the observation period; (3) coarse-resolution hazard data capturing mainly major fluvial plains and missing pluvial flooding and secondary rivers, leading to underestimation; (4) focus on single flood types (e.g., only coastal sea level rise) or subsets of countries, limiting global comprehensiveness; and (5) absence of analyses integrating poverty and exposure, despite poverty being central to vulnerability and resilience. The study positions itself to overcome these gaps by using high-resolution, multi-type flood hazard data with global coverage and integrating subnational poverty metrics.

Data and scope: A high-resolution global exposure assessment covering 188 countries and 2084+ subnational regions. Flood hazards include fluvial, pluvial, and coastal flooding. Flood data are primarily from Fathom-Global (fluvial/pluvial) and a global coastal flood risk map using LISFLOOD-FP hydrodynamics. All hazard layers are at 3 arcseconds (~90 m) resolution and based on the MERIT DEM. Population density is from WorldPop 2020 (WorldPop-PPP-2020), also at 3 arcseconds. Subnational poverty estimates come from the World Bank’s Global Subnational Atlas of Poverty (GSAP, 2020 update). Return period: All hazard layers consider 1-in-100-year floods (annual probability ~1%), using undefended flood maps (i.e., not accounting for man-made defenses). Processing steps: (1) Combine the three flood-type layers by retaining the maximum inundation depth per pixel; resample to align with WorldPop. (2) Classify continuous inundation depths into five categories: No risk; Low (≤0.15 m); Moderate (≤0.5 m); High (≤1.5 m); Very high (>1.5 m). Define “significant risk” as >0.15 m (Moderate–Very high). (3) Overlay population counts with flood categories at pixel level; aggregate exposure headcounts to ADM-1, national, regional, and global levels. (4) Poverty overlay: multiply exposure headcounts by subnational poverty rates at three thresholds ($1.90, $3.20, $5.50 per day) to estimate numbers simultaneously flood-exposed and in poverty. (5) Economic exposure: multiply exposure headcounts by subnational income (GDP per capita) to estimate economic activity sited in flood-risk areas (not losses and not distinguishing residence vs workplace). Sensitivity: Headcounts depend on the lower inundation threshold; exposure at >0.15 m totals ~1.81 billion, declining to ~1.06 billion at >0.5 m. Computation used Python; visualization via QGIS. Code and data availability are provided (GitHub and World Bank Data Hub). Assumptions include uniform poverty shares within ADM-1 and no accounting for local defense infrastructure due to lack of global inventories.

- Globally, 1.81 billion people (23% of 7.9 billion) live in areas with significant flood risk (>0.15 m inundation in a 1-in-100-year event). - Regional exposure totals: East Asia & Pacific (668 million), South Asia (576.7 million), Europe & Central Asia (145.9 million), Latin America & Caribbean (107 million), Middle East & North Africa (89.3 million), Sub-Saharan Africa (176 million), Canada & USA (46 million). - Country highlights: China (≈394.8–395 million) and India (≈389.8–390 million) together account for over one-third of global exposure. Other large absolute headcounts include Bangladesh, Indonesia, Pakistan, Vietnam, the United States, Nigeria, Egypt, and Japan. - Relative exposure leaders: Netherlands (58.7% of population in flood zones, acknowledging strong defenses), Vietnam (46%), Egypt (40.5%), Bangladesh (39.9%), among others. Drivers vary by flood type: fluvial risks in major river basins (e.g., Brahmaputra, Nile, Mekong), pluvial in mountainous/intense rainfall regions, coastal in urbanized coasts and islands. - Subnational hotspots: In India’s Bihar, Uttar Pradesh, and West Bengal, a combined 196 million people are in high-risk zones (33–53% of state populations). Punjab (Pakistan): ~48 million (38%). Dhaka Division (Bangladesh): ~61% exposed. In China, coastal and Yellow River Valley provinces have the largest exposed populations. - Economic exposure: About $9.8 trillion of economic activity (≈12% of global 2020 GDP) is sited in significant flood-risk areas; 84% of this is in high- and upper-middle-income countries. High-income countries hold 37% of exposed economic activity but only 11% of exposed people. China leads ($3.3T), followed by USA ($1.1T) and Japan ($0.7T). - Poverty intersection: Of 1.81 billion exposed, at least 170 million live under $1.90/day; 88% of these are in Sub-Saharan Africa and South Asia (≈74.7 million SSA; ≈75.0 million SAR; India ≈66 million). At $3.20/day, ≈467.4 million exposed people are in poverty; at $5.50/day, ≈779.7–780 million. Thus, roughly 4 in 10 flood-exposed people live in poverty at the $5.50 threshold. Raising thresholds shifts hotspots to include parts of Egypt, the Middle East, South/East Asia, and Latin America. In SAR, exposed-and-poor rises from 75 million ($1.90) to 464 million ($5.50); in EAP from 10 million to 81 million. - Sensitivity: Using a stricter depth threshold (>0.5 m) reduces exposed headcounts from 1.81 billion to 1.06 billion; estimates are most sensitive for pluvial flooding. - Coverage: Of 2084 subnational regions assessed, only 9 have <1% of their population exposed to flood risk.

The analysis demonstrates that flood exposure is globally widespread but disproportionately affects low- and middle-income countries when considering human vulnerability. While monetary exposure concentrates in wealthier regions (due to higher asset values and GDP density), the convergence of high exposure and poverty is centered in low-income regions, particularly Sub-Saharan Africa and parts of South Asia. This underscores that prioritizing flood protection solely by economic value at risk risks neglecting areas where floods cause the most severe, long-lasting welfare impacts due to limited defenses, weaker infrastructure and planning, constrained access to credit and savings, less-developed insurance markets, and thinner social protection systems. Integrating poverty with exposure identifies hotspots where mitigation, preparedness, and social safety nets are most urgently needed to protect lives and livelihoods. The findings address the paper’s core question by quantifying both the scale and spatial distribution of exposure and highlighting how poverty status alters risk prioritization. Given climate change, sea-level rise, subsidence, and continued urbanization, these risks are likely to intensify, making the targeted allocation of resources and resilience-building critical. The study’s high-resolution, multi-hazard approach also reveals within-country concentrations of risk, supporting subnational planning and investment decisions.

This study provides the first global, high-resolution estimates that jointly assess exposure to fluvial, pluvial, and coastal flooding and its intersection with poverty for 188 countries. It shows that 1.81 billion people (23% of the global population) live in significant flood-risk areas, with exposure concentrated in South and East Asia, and that 170–780 million exposed people live in poverty, depending on the threshold. While monetary exposure highlights higher-income regions, the coupling of exposure and poverty reveals that the gravest development risks lie in low-income regions, notably Sub-Saharan Africa and South Asia. The results inform prioritization of flood risk mitigation, including structural defenses, land-use planning, early warning systems, and strengthened social protection where vulnerability is highest. Future work could extend this framework by explicitly accounting for flood defense infrastructure where data become available, refining within-ADM-1 poverty spatial distributions, and projecting future exposure under climate and socioeconomic scenarios to guide long-term resilient development strategies.

- Undefended hazard maps: The analysis uses flood hazard layers that do not account for existing flood defenses; results likely overestimate exposure in areas with robust protection (more common in high-income countries), and underestimate effective risk reduction where defenses are present. - Poverty spatial resolution: Poverty rates are applied uniformly within subnational (ADM-1) units, not at pixel level. If poorer households are disproportionately located in riskier areas, estimates of exposed poor are lower bound. - Economic exposure measure: Monetary exposure represents economic activity located in flood zones (via headcounts × income per capita), not expected losses; it does not distinguish places of residence vs workplaces. - Threshold sensitivity: Headcounts depend on the chosen inundation depth threshold for “significant risk” (>0.15 m); results change under stricter thresholds (e.g., >0.5 m yields 1.06 billion exposed). - Data and model uncertainties: Despite high resolution, hazard and population datasets carry uncertainties (e.g., DEM accuracy, hydrodynamic modeling, population disaggregation). Coastal and pluvial exposure estimates have different sensitivities. Lack of a global inventory of protection systems prevents adjusting exposure by defense standards. - China income data limitation: Subnational mean income levels for China were not available; national averages were used.