Climate threats to coastal infrastructure and sustainable development outcomes

Climate hazards disproportionately affect poor and vulnerable communities, partly due to exposure bias where poorer households reside in more exposed locations. Repeated climate impacts can entrench poverty through reduced asset accumulation and persistent effects on health, education and employment, contributing to poverty traps and out-migration. Consequently, national development plans and progress towards the Sustainable Development Goals (SDGs) are at risk of being derailed by climate extremes. While targeted, pro-poor interventions can mitigate these dynamics, scarce resources necessitate spatially explicit analyses of climate risks, exposure and vulnerability, and their interactions across socioeconomic systems. Existing high-resolution vulnerability mapping has often been decoupled from hazard exposure and essential service access needed to evaluate SDG progress at local scales. This study addresses that gap by linking geolocated household-level socioeconomics and service access with multihazard exposure to assess risks to SDG achievement in coastal Bangladesh, a global hotspot of fluvial, pluvial and tidal flooding, cyclones and erosion. It evaluates how disruptions to critical infrastructure services (health, education, electricity, markets and cyclone shelters) threaten local SDG attainment and identifies where adaptation could most effectively safeguard development gains.

Prior research has documented the disproportionate exposure of poor populations to natural hazards and the potential for climate shocks to reinforce poverty traps. Social vulnerability mapping has been integrated with flood risk assessments at local to global scales, but often without precise household geolocation or explicit links to essential service access critical for SDG progress. Studies have examined synergies between infrastructure, energy and climate action with SDGs at aggregated scales and developed national frameworks to align adaptation with SDG priorities. However, reliable high-resolution data to distinguish local vulnerability differences remain limited, contributing to discrepancies in risk estimates in floodplains and coastal zones. Recent advances in spatial microsimulation, dasymetric modelling and remote sensing enable finer-scale assessments, and national planning exercises have begun to incorporate resilience needs across built and natural assets. Building on this literature, the present study integrates multihazard exposure with household-level service dependence to quantify localized risks to SDG progress and prioritize pro-poor adaptation.

The study assesses climate hazard threats to SDG-related infrastructure service delivery in 150 upazilas (subdistricts) across 19 coastal districts of Bangladesh using five methodological components: (1) assembling a spatial database of climate hazards and critical infrastructure; (2) linking households to nearest service-providing assets; (3) constructing a household wealth index and assigning wealth quintiles; (4) quantifying statistically significant exposure biases by wealth; and (5) downscaling SDG progress and estimating progress at risk from hazard-induced service disruptions. Climate hazards: Coastal and riverine flooding layers were sourced from the World Resources Institute Aqueduct Floods platform for baseline and projections to 2030, 2050 and 2080 under RCP 4.5 and 8.5 at ~1 km resolution (30 arc seconds) for return periods of 2–100 years. Cyclone hazard was represented by maximum wind gusts (m s−1) from ensemble resimulations of 12 historical cyclones at 4.4 km resolution, provided as return-period event maps. Erosion hazard was mapped using the Deep WaterMap model at 30 m resolution to identify pixels that experienced erosion at any time over the past 35 years. Due to model limitations, future scenarios were applied to floods only (2030, 1-in-50-year, RCP 4.5); cyclone and erosion analyses used historical characterizations. Input datasets had prior validations in Bangladesh or analogous contexts. Infrastructure data: Geolocated critical infrastructure assets and networks were compiled and verified with Bangladesh government sources and partners: electricity substations (Power Division, BPDB), market centres (WARPO), healthcare facilities (Department of Health), cyclone shelters (LGED), and educational institutions (Bangladesh Bureau of Statistics). The database includes 113 electricity substations, 2,062 market centres, 3,086 healthcare facilities, 3,777 cyclone shelters and 73,814 educational institutions. Each infrastructure layer was intersected with hazard layers to derive exposure and, for floods, estimated depth (m). Household data and service linkage: A spatially explicit synthetic household dataset (World Bank) comprising 8.2 million households in the coastal region provided geolocated clusters and socioeconomic/structural attributes (e.g., electrification, literacy, education, employment). A nearest-neighbour approach assigned each household to its nearest relevant asset per subsector, representing dependence on that asset’s service provision. Due to road network data limitations and household positional clustering, a radius-based accessibility approach was used: households reporting access in surveys were included; where survey data were insufficient, households beyond thresholds were deemed to have no access (5 km for health centres; 1.6 km for cyclone shelters). Exposure of a connected household was defined by exposure of its assigned nearest asset under a given hazard scenario. Wealth index and quintiles: A wealth index was constructed separately for urban and rural households using multiple infrastructure-access and dwelling variables. Variables were normalized and subjected to principal component analysis (PCA), following WFP-FAO guidance. Kaiser-Meyer-Olkin measures were 0.62 (rural) and 0.71 (urban); the first principal component (~40% variance explained) formed the index. Households were partitioned into quintiles Q1 (poorest) to Q5 (wealthiest). Exposure bias by wealth: For each upazila, hazard, and infrastructure type, risk ratios compared disruption probabilities of poorer (Q1–Q2) and wealthier (Q4–Q5) groups to the median (Q3). Statistical significance was assessed using two-proportion z-tests (95% confidence, |z|>1.96) given large household counts. SDG progress and progress at risk: Five nationally prioritized SDG indicators (NPTs) linked to assessed services were used: SDG 3 (health access), SDG 4 (education access), SDG 7 (electricity access), SDG 8 (employment/market access), and SDG 13 (persons protected by cyclone shelters). Household survey responses and proximity thresholds informed current upazila-level progress. Households were considered disrupted if their nearest asset exceeded hazard thresholds: flood depth >1 m (coastal/riverine), wind gust >30 m s−1 (cyclones), or erosion occurrence. For floods, the focal scenario was a 1-in-50-year event in 2030 under RCP 4.5; cyclone and erosion assessments used historical datasets. “Progress at risk” was computed as the share of current upazila-level SDG achievement potentially lost due to disruptions. Upazilas were ranked by average SDG progress at risk (population-weighted) to evaluate safeguarding benefits from prioritized resilience interventions across hazards and sectors.

- All 8.2 million households in coastal Bangladesh are exposed to some disruption of essential infrastructure services due to climate hazards. - Under a baseline 1-in-50-year hazard, average regional disruption across all services is 39.5% (coastal flooding), 22.7% (riverine flooding), and 94.5% (cyclones). Erosion has cumulatively disrupted an estimated 56% of the coastal population since 1987. - A statistically significant poverty bias in exposure to disruptions (any hazard, any service) is present in 104 of 150 upazilas (69%). For coastal flooding alone, at least one asset type shows a poverty bias in 35 upazilas (23.33%); 12 upazilas show poverty bias for two asset types, and 2 upazilas (Chakaria and Jhalokati Sadar) for three. - Relative risk by service under coastal flooding shows poorer groups more likely to face disruption for: education facilities (poor vs median 1.07; poor vs wealthiest 1.13), health facilities (1.07), market centres (1.02), and cyclone shelters (1.24). Electricity substations show a slight wealth bias (poor vs wealthy risk ratio 0.989; wealthiest 1.01× more likely), driven by exposed substations in Patiya and Hathazari (Chittagong) serving predominantly wealthier households. - The largest absolute number of poorer households (Q1–Q3) exposed to substation disruption from coastal flooding occurs in Patuakhali Sadar (≈381,680 households). - SDG progress at risk varies by hazard and indicator. Dominant hazards by indicator/upazilas: coastal flooding most impacts SDG 7 (electrification) in 51 upazilas; river flooding dominates threats to SDG 3 (health) and SDG 4 (education) in 41 and 40 upazilas, respectively; erosion dominates SDG 3 in 55 upazilas; cyclone-related disruptions broadly affect SDG 3, 4, 7 and 8. SDG 13 (shelter protection) remains vulnerable to coastal/river flooding and erosion (cyclone wind damage to shelters not considered for SDG 13). - Prioritization potential (average SDG progress safeguarded, population-weighted): • Coastal flooding: targeting 10 upazilas (6.6%) safeguards ~15–20% of progress; 50 upazilas (33%) safeguard ~60–70%; 117 upazilas (78%) needed to surpass 99% protection. • River flooding: 10 upazilas safeguard ~21–30%; 50 upazilas ~76–85%; 114 upazilas achieve >99%. • Cyclones: 10 upazilas safeguard ~15–19%; 50 upazilas ~50–56%; ~149 upazilas required to reach ~99%. • Erosion: 10 upazilas safeguard ~19–21%; 50 upazilas ~52–60%; 141 upazilas for ~99% protection. - Service access gaps: 46% of households report no electricity access; 18% live >5 km from the nearest healthcare facility. Overlaying underserved areas with hazards identifies where new infrastructure should be both added and made resilient (notably in the Meghna Estuary and southwestern coastal upazilas).

Linking household-level service dependence to multihazard exposure reveals how climate threats can undermine local SDG attainment and demonstrates a systematic exposure bias against poorer households across much of coastal Bangladesh. The findings indicate that strategic, spatially targeted resilience measures can safeguard a large share of current SDG progress by focusing on a subset of high-risk upazilas, improving the efficiency of limited adaptation resources. The poverty bias likely reflects intertwined drivers: resource-dependent livelihoods, insecure land tenure, limited mobility to higher-income urban areas, cumulative land degradation (e.g., salinity, waterlogging), and uneven provision of protective and basic infrastructure. Differences in hazard footprints and service networks lead to indicator-specific vulnerabilities (e.g., electrification more vulnerable to coastal flooding than river flooding). The prioritization curves show strong returns to focusing on top-ranked upazilas, providing a practical roadmap for adaptation planning. Integrating quantitative geospatial risk assessments with qualitative, community-level insights and evolving resilience metrics for assets (roads, power grids, water systems) can further refine local interventions. Mainstreaming such analyses into national planning processes (e.g., Bangladesh’s Mujib Climate Prosperity Plan) can help align adaptation investments with pro-poor SDG strategies and close achievement gaps under intensifying climate risks.

This study provides a high-resolution, multihazard, spatial assessment that connects household exposure and essential infrastructure service access to local SDG indicators in coastal Bangladesh. It finds that climate hazards threaten service access for all households and that poorer groups disproportionately face disruptions across most upazilas. By targeting resilience interventions to a relatively small share of the most exposed and consequential upazilas, policymakers can safeguard a substantial portion (50–85%) of progress towards key SDG targets (3, 4, 7, 8, 13). The approach—combining validated hazard layers, infrastructure inventories, and synthetic household microsimulation—offers a transferable framework for data-scarce regions to prioritize pro-poor adaptation. Future work should incorporate additional infrastructures (e.g., water supply, roads, public administration), consider compounding hazards and cascading failures, expand to other hazard types (droughts, wildfires, storms), and include uncertainty analyses and asset-specific fragility and resilience measures to better quantify disruption likelihood and duration.

- Compound hazard interactions (e.g., concurrent cyclone winds and flooding) were not modelled; hazards were assessed independently. - Temporal inconsistencies exist across hazard datasets: future scenarios were applied to floods (2030, RCP 4.5, 1-in-50-year), while cyclone and erosion hazards relied on historical characterizations. - No formal uncertainty analysis of hazard layers was conducted, though inputs have been externally validated. - Service areas were approximated via nearest-neighbour/radius methods due to road network data gaps and synthetic household clustering; exact travel-based accessibility was not modelled. - Disruption thresholds (flood depth >1 m; wind >30 m s−1; any erosion) are simplifying assumptions that may not capture asset-specific fragilities or protective measures. - The analysis assumes a worst-case where exposed assets cause household disruption and does not account for resilience measures already implemented or differential asset robustness. - Some SDG-linked sectors (e.g., water supply, local roads) were omitted due to data limitations, potentially understating cross-sectoral risks. - Household locations are synthetic (though validated), which may introduce spatial uncertainty at very fine scales.