Progress and gaps in climate change adaptation in coastal cities across the globe

Coastal cities are key economic hubs but are also hotspots of climate-related hazards (e.g., sea-level rise, tropical cyclones, floods, storms, erosion, heatwaves) and socially mediated vulnerabilities (inequality, poverty, inadequate infrastructure). Risks vary by geomorphology, climate dynamics, and urban development pathways. Despite calls for transformative adaptation in coastal cities, little is known about actual progress globally. The study argues that assessing current adaptation is important for tracking progress under the Paris Agreement’s Global Stocktake and for informing policy and practice. Prior work has examined specific adaptation types, actors, regions, or planning efforts, but a systematic global assessment of empirical evidence on implemented coastal urban adaptation—covering response types, actors, and transformation level—has been lacking. This study therefore aims to provide a global analysis of empirical evidence of adaptation in coastal cities (2013–2020) and to identify gaps and shortcomings. It is guided by four research questions: (1) How is evidence for coastal urban adaptation distributed globally? (2) Which hazards and trends of exposure and vulnerability are reported? (3) Which actors are involved in which response types? (4) What is the speed, scope, and depth of reported coastal urban adaptation? The study defines coastal cities as urban areas of central functional importance located on or influenced by coastal/tidal hydrology or within the LECZ.

The paper situates its contribution within a body of research on urban adaptation that has assessed institutional, ecosystem-based, and technological measures; actor involvement; regional adaptations; and coastal adaptation planning. The IPCC has noted that coastal cities often adapt reactively to high-impact events and that significant gaps remain. However, a comprehensive, systematic, global synthesis of empirical evidence on implemented adaptation in coastal cities, capturing response types, actor roles, and transformation levels, had not been conducted. The authors highlight overrepresentation of certain regions and types of responses in the literature and identify the need to move beyond case-specific or planning-only analyses to evidence of implementation.

The authors conducted a systematic map and content analysis following the ROSES protocol and the Global Adaptation Mapping Initiative (GAMI) approach. They combined three sources: (1) GAMI database category 'cities and settlements by the sea' (initial 361 publications); (2) systematic searches in Web of Science Core Collection and Scopus using an extended English-language Boolean search string tailored to coastal urban adaptation; and (3) consolidation and deduplication across sources. The temporal scope covered 2013–2020 (from the end of IPCC AR5 to the literature cutoff for AR6). Inclusion criteria limited the corpus to peer-reviewed, empirical studies in English focusing on coastal cities. Of 683 publications screened at full text, 183 met inclusion criteria; 501 were excluded (no empirical evidence, wrong focus, outside scope/time, regional-only); 2 were inaccessible; 7 non-English. Coding: Using SoSci Survey, coders completed one questionnaire per city covered by each paper (some papers contributed multiple city cases). In total, 183 publications yielded 284 cases across 199 cities and four unspecified urban areas. The 30-question instrument captured hazard types; exposure and vulnerability elements and their temporal treatment; actor types; response categories (technological/infrastructural, behavioral/cultural, institutional, ecosystem-based); and indicators of transformational adaptation (depth, speed, scope). Data quality: Coders received a detailed codebook and training; approximately 10% of the dataset (72 publications) was double-coded, with inter-coder variability up to 22.2% (around 80% convergence), deemed sufficient for robustness. Only medium- and high-confidence coder judgments were considered for depth, speed, and scope. Analysis: Descriptive statistics summarized frequencies and proportions across World Bank income groups and global regions. Spearman’s rank correlation tested relationships of GNI per capita and city size with actor involvement and adaptation types; χ² tests (φ coefficients) examined associations between actor and response categories. LECZ population baselines were drawn from the LECZ v3 dataset (GPWv4 Rev11 and CoastalDEM90) to benchmark coverage relative to populations at risk. Statistical processing used IBM SPSS 23.

- Uneven global coverage: Of adaptation evidence for coastal cities, Asia accounts for ~30–31% of cities studied despite hosting ~75% of global LECZ population; North America 23% (LECZ share ~5%), Europe 16%, Africa 13%, Australasia 11%, small island states 3%. High-income economies account for 56% of reported adaptation evidence though only ~16% of LECZ population lives there; upper-middle and lower-middle economies contribute 19% and 24% of evidence (vs. 34% and 43% of LECZ population), and low-income only ~1% (vs. ~8% of LECZ population). - City size: 48% of reported cases involve cities <250,000 inhabitants; mid-sized (250,000–1,000,000) are underrepresented; 35% involve >1,000,000 inhabitants, with many cases in Africa and Asia; megacities like New York, Jakarta, Manila, Lagos are frequently studied (e.g., New York 12 studies). - Hazards and risk factors: Most studies address sea-level rise and various types of flooding, with frequent multi-hazard consideration (65% consider more than one hazard), commonly combining sea-level rise with storm surge, coastal/pluvial flooding, and erosion. Hazard treatment is predominantly based on past/current events; future trends are often considered conceptually but not quantified. Exposure and vulnerability elements (population, vulnerable groups, residential buildings, coastline) are largely treated using past/current patterns; quantified future scenarios for exposure/vulnerability are rare. Consideration of exposure/vulnerability correlates weakly and positively with income level. - Response types and actors: Technological/infrastructural and behavioral/cultural responses dominate; institutional responses are also common, while ecosystem-based adaptation is least reported across regions, particularly in low- and middle-income countries. Higher GNI per capita correlates with more institutional adaptation (Spearman’s ρ = 0.23, P < 0.01) and less behavioral adaptation (ρ = −0.35, P < 0.01). City governments are the most commonly involved actors; higher GNI per capita associates with greater city government involvement (ρ = 0.30, P < 0.01) and less individual/household involvement (ρ = −0.23, P < 0.01). Larger city size correlates with less individual/household adaptation (ρ = −0.30, P < 0.01) and more city government involvement (ρ = 0.20, P < 0.01). Behavioral responses are most often linked to individuals/households; multi-actor adaptations occur, commonly involving city and national governments together. Regional patterns: Behavioral adaptation is less reported in North America and Central/South America, more in Africa and Asia; institutional and ecosystem-based measures are more reported in Europe and North America. - Transformation dimensions: Reported adaptation exhibits low depth, scope, and speed across income groups and regions, with limited evidence of risk reduction. Aggregate shares (all income groups): depth—low 60%, medium 31%, high 9%; scope—low 62%, medium 26%, high 12%; speed—low 69%, medium 17%, high 13%. By income groups (examples): lower-middle income—depth low 60%/mid 34%/high 5%; scope low 70%/mid 23%/high 6%; speed low 75%/mid 17%/high 7% (high-confidence judgments 16%). Upper-middle: depth 59/30/10%; scope 52/35/10%; speed 62/21/16% (high-confidence 14%). High income: depth 60/30/10%; scope 61/24/15%; speed 70/16/13%. Few cases show deeper changes (e.g., resettlement; robust institutionalization; integrated flood risk management; innovative design approaches; cross-sectoral planning; large-scale infrastructure complemented by ecosystem-based measures). Many cases remain incremental (traditional flood defenses, insurance uptake, reactive coping). Some measures show risk reduction (ecosystem-based and infrastructural) but also negative side-effects and instances of maladaptation. - Five synthesized insights: (1) Knowledge coverage is highly uneven, underrepresenting small/mid-sized cities and low- to middle-income regions. (2) Adaptation planning relies heavily on past/current patterns, with scarce quantified futures for hazards and especially exposure/vulnerability. (3) In lower-income contexts, individuals/households bear more adaptation burden; wealthier contexts feature more government/planned responses. (4) Larger cities emphasize technological protection; ecosystem-based approaches are underreported, raising risks of lock-in and maladaptation. (5) Transformative adaptation remains limited in depth, scope, and speed across contexts, inadequate for accelerating climate risks.

The study’s global evidence map addresses the four guiding questions by showing where adaptation is being reported, which risks are considered, who acts, how they act, and whether actions are transformative. The uneven distribution of evidence reveals research and implementation blind spots in regions with high LECZ populations (Asia, parts of Africa, Central/South America) and in small to mid-sized cities—areas likely to face rapidly intensifying risks. The prevalent reliance on past/current hazard and risk patterns, with weak consideration of quantified future exposure/vulnerability, undermines the robustness of adaptation planning for dynamic urban coastal risk. Actor-response patterns indicate inequities: lower-income settings rely more on households and behavioral changes due to limited institutional/technological support, while higher-income, larger cities feature more institutional and infrastructural responses led by city governments. The scarcity of ecosystem-based adaptation and the emphasis on technological protection, especially in larger cities, risk long-term lock-in and maladaptive trajectories if hazards intensify beyond design thresholds. Across regions and income groups, adaptation is progressing slowly, narrowly, and with low depth—insufficient for the pace and scale of coastal climate change. These findings emphasize the need for broader, faster, and deeper actions that integrate future scenarios, coordinate across sectors and levels of governance, diversify response portfolios (including ecosystem-based approaches), and address social equity in adaptation burdens.

This paper provides the first systematic, global synthesis of empirically reported adaptation in coastal cities (2013–2020), revealing that current efforts are unevenly distributed, predominantly incremental, and often anchored in past/current risk patterns with limited quantification of future exposure and vulnerability. Key contributions include: (1) identifying regional and income-group biases in the evidence base; (2) mapping dominant hazards, response types, and actor constellations; (3) quantifying the low depth, scope, and speed of adaptation; and (4) highlighting correlations between income/city size and actor/response patterns. Future research and policy directions include: expanding empirical coverage in underrepresented regions and city sizes; systematically incorporating quantified future scenarios for hazards, exposure, and vulnerability; strengthening institutional capacities and multi-actor governance, especially in lower-income settings; advancing ecosystem-based and hybrid response portfolios to avoid lock-in; evaluating long-term effectiveness and maladaptation risks; and accelerating transformative, coordinated, cross-sectoral adaptation to match accelerating coastal climate risks.

The assessment relies on peer-reviewed, English-language literature from 2013–2020 and excludes grey/non-English sources, potentially biasing geographic and thematic coverage (e.g., underrepresentation of low- and middle-income regions). Evidence reflects what is published rather than all on-the-ground activities. Coding, although supported by a detailed codebook and training with double-coding (~10%), has inherent subjectivity (approximately 80% convergence). Some categories (e.g., transformation depth/scope/speed) depend on coder judgment with confidence thresholds; limited cases in certain income groups (e.g., low-income) constrain generalizability. Quantified future exposure/vulnerability scenarios were rarely reported, limiting cross-case comparison of forward-looking planning.