Coastal Defense Systems

Coastal defence is the business of deciding where the sea is allowed to reach, and then making that decision hold. It runs from the oldest civil engineering humans do — an earth bank raised above a design water level — to the newest and strangest: a movable machine spanning a navigable estuary, closed on a forecast, with a legally specified probability of failing when asked.

Scope boundary. This brief is about defending an existing coastline in place, or deliberately giving parts of it up. Building new habitable structures on the water is floating cities. Working the sea itself — seabed resources, offshore structures, ocean energy — is ocean engineering. Moving populations, and the legal status of people who move, is climate migration planning. The generic pathology of very large public works sits with megaproject governance, and the sea-level forcing itself with climate engineering.

The thesis, in three parts. First: the hard engineering is not the constraint and has not been for fifty years. Dykes, seawalls and surge barriers are buildable, and one country has demonstrated that they can be maintained and progressively upgraded for centuries. Second: the constraint that actually binds is arithmetic that the field itself has produced and largely declined to foreground — a mainstream European economic optimisation concludes that protection is justified on roughly a quarter to a third of the coastline, and that even where it is applied it protects assets far better than it protects people. Third: the alternative to concrete is not a technology. It is an instrument. A coastline can in principle be defended by retreat, and the thing standing in the way is not sediment science or wave mechanics but the fact that coastal-protection statutes generally give an authority the power to protect and say nothing about the power to stop.

Established A storm surge barrier is not a wall. It is a machine with a licensed failure rate, and the Netherlands has written the number down. For the Maeslantkering, the barrier closing the Nieuwe Waterweg approach to Rotterdam, the closure decision threshold is a predicted water level exceeding MSL + 3.0 m at Rotterdam, or MSL + 2.9 m at Dordrecht; the barrier actually closes at MSL + 2.0 m under moderate Rhine discharge; and the critical inner-basin water level is MSL + 3.60 m (Bakker, Rovers & Mooyaart 2025). Under the Dutch Water Act the probability that a closure operation fails must be less than 1 in 100 per closure request. Secondary accounts collapse the first three of those into a single “three metre” threshold; there are three distinct levels and they do different jobs.

Frontier The 1-in-100-per-demand figure is the most consequential and least quoted number in the whole subject. It is the quantitative statement of why a barrier is a different kind of object from a dyke. A dyke fails when overtopped or breached, which is a continuous, inspectable condition. A barrier can fail on demand — the gates do not move, the ballast does not seat, the decision system does not fire — and the Dutch have priced that at one failure per hundred requests. Established Multiply a fixed per-demand failure probability by a demand count that rises with sea level and you get a rising absolute failure rate with no degradation in the machine whatsoever: hard-defence lock-in expressed as arithmetic rather than rhetoric.

Established What this brief does not tell you, and why that is a feature. The retrieval behind it could not reach rijkswaterstaat.nl, deltacommissaris.nl, gov.uk or usace.army.mil. Consequently: the Maeslantkering’s commissioning date, its gate dimensions and its operational closure count against storm surge; the Thames Barrier’s gate count, its cumulative closure total, the size of its much-cited single-winter closure cluster, and the design-life extension in the Environment Agency’s Thames Estuary 2100 plan; MOSE’s gate count, its construction cost and the date of its first full operational raising; and the Delta Fund’s annual budget — none of these is stated here. Established Every one is in wide circulation and several drift measurably between secondary sources. Frontier They are named as open gaps rather than filled from memory, and the arguments below are built only on what was read. A brief that says which numbers it does not have is more useful than one that repeats them.

Established The economics of protection have been computed for a whole continent, and the headline result is that most coastline does not get defended. Vousdoukas et al. (2020) put current European annual coastal flood losses at €1.4 billion (2015 values) with about 100,000 people affected annually. Established Without adaptation, by 2100 that becomes €209.8 bn/yr and 1.61 million people flooded under RCP4.5/SSP1 (range €29.7–844.5 bn), and €1,268.4 bn/yr with 3.9 million people flooded under RCP8.5/SSP5 (range €160.9–4,731.1 bn). Established Raising dykes returns a European mean benefit–cost ratio of 8.3 (RCP4.5/SSP1) and 14.9 (RCP8.5/SSP5), with country-level ratios spanning 1.6 to 34.3 — highest in the Netherlands, lowest in Malta. The optimal dyke-height increase averages 92 cm and 104 cm respectively, ranging by country from 31 cm in Malta to 3.43 m in Belgium, at an undiscounted investment of €1.75 bn/yr and €2.82 bn/yr over 2020–2100, discounted at 5 per cent for EU cohesion countries and 3 per cent elsewhere.

Established And the finding that inverts the paper’s popular reading: protection is economically justified along only 23.7–32.1 per cent of Europe’s coastline — a minority that nonetheless accounts for at least 83 per cent of the achievable damage reduction. Frontier This is not an argument against defence. It is a defend-oriented analysis producing an explicit map of where not to defend, and it means that retreat is not the rival hypothesis to this study. It is the study’s complement, and the study quantified how much of the coast it applies to. Established The high benefit–cost ratios exist because they are computed over the segments worth defending; quoting them without the coverage fraction converts “defend selectively” into “defend everywhere” and reverses the result.

Established The buried distributional result is the strongest quantitative argument in this subject and it comes from inside the mainstream economics. In the same analysis, optimal adaptation cuts residual damages to €8.8 bn and €24 bn — reductions of 96 per cent and 98 per cent. But residual people flooded fall only to 653,400/yr and 1,343,100/yr — reductions of 59 per cent and 66 per cent. Frontier Damage falls by 96–98 per cent; people flooded fall by 59–66 per cent. The optimisation follows asset value, because that is what a benefit–cost objective function contains, and people without assets sit in the residual. Frontier That asymmetry is a distributional result hiding inside an efficiency analysis, it is published, and it is almost never foregrounded — including by people arguing exactly this point from outside economics, who have a verified number available and do not use it.

Established Exposure is measured far better than it was, and is misreported worse than almost anything else in climate science. Kulp & Strauss (2019) applied CoastalDEM, a neural-network correction to NASA’s SRTM elevation model, and found 110 million people currently occupying land below the present high tide line and 250 million below present annual flood levels. Established The correction is large and well characterised: US coastal elevation bias falls from 4.71 m to under 0.06 m, global RMSE roughly halves, and coverage from 56°S to 60°N captures 99.7 per cent of world population. Projected: 150 million below the 2050 high tide line under RCP4.5 with a stable Antarctica; 190 million (RCP2.6) and 200 million (RCP4.5) below the 2100 high tide line; 360 million below annual flood levels in 2100 under RCP4.5; and up to 630 million under RCP8.5 with Antarctic instability. More than 70 per cent of the exposed population is in eight Asian countries, and by 2100 under RCP2.6 the shares reach 19 per cent of Bangladesh, 26 per cent of Vietnam and 17 per cent of Thailand. About 1 billion people live below 10 m above the current high tide line and 230 million below 1 m.

Established The authors’ own caveats are the most important part of that paper and are the subject of this brief’s lead correction. The analysis combines future water levels with contemporary population densities, and the authors state the results “should therefore not be taken as projected impacts”; they describe the output as a threat indicator and say they explicitly avoid projecting actual displacement outcomes. Existing coastal defences, which the authors say protect “tens or hundreds of millions” of people, are not modelled at all. They further note that the bathtub inundation model overstates episodic flooding relative to a hydrodynamic model, and that country-scale elevation error reaches ±49 per cent in the worst case. Frontier A global inventory of existing coastal defences does not exist. That absence is the largest single hole in coastal risk science, and it is why the best exposure dataset in the world cannot be turned into a risk estimate.

Established Managed realignment has a real empirical literature, it is mostly British, and it is mostly about mud. The verified corpus — Morris (2012) on realignment as sediment management, Hughes (2009) on what actually grew at two southeast England sites, Symonds (2004) at Freiston Shore, Kadiri (2011) on restored sediment characteristics, Burden (2013) and Blackwell (2010) on the carbon and nitrous-oxide balance, Cooper (2001) on twenty-five years of Essex saltmarsh erosion, and the 2021 volume Land Abandoned to the Sea — was verified at record level: authors, titles, venues, years and DOIs confirmed, texts not read. Established That supports one thing, and it is the thing that matters: a peer-reviewed performance literature on deliberate coastal abandonment exists, is two decades deep, and is dominated by a single question. Frontier The recurring finding across it is that the binding uncertainty is not whether the seawall can be breached — that is trivial — but whether the site will accrete sediment and develop into functioning saltmarsh, which depends on how far the land subsided during decades of agricultural drainage and consolidation. Established The individually named UK schemes — Medmerry, Wallasea Island, Alkborough Flats, Freiston Shore, Tollesbury, Orplands — have widely quoted areas, dates and costs that this brief does not state, because the pass behind it could not verify them and they are known to drift.

Established Nature-based defence works, and the number everyone quotes for it could not be obtained. The canonical meta-analysis is Narayan et al. (2016) in PLOS ONE, whose habitat-by-habitat attenuation percentages are the most-recycled figures in the field. Established Three attempts to retrieve it through a correctly addressed DOI URL returned, twice, the full text of an entirely unrelated paper on mouse digit-tip regeneration — confidently rendered, no error, no low-confidence signal. Its effect sizes are therefore deliberately not reproduced here. Established The verified primary literature nevertheless makes the methodological point without them: attenuation is measured over specific habitat widths, water depths, wave periods and stem stiffnesses; vegetation flexibility is a first-order control, because flexible stems reconfigure under load and attenuate far less than rigid-cylinder models predict (Anderson 2014; van Veelen 2020); storm-condition field measurement is rare (Shi 2025, over a Scirpus mariqueter marsh during typhoon Muifa); and much of the mangrove effect may be geomorphic rather than vegetative — the shallow platform the forest built and holds, rather than drag on its stems (Sánchez-Núñez 2020). Frontier An effect size quoted without its width, depth and wave condition is uninterpretable, and most of them are.

Established Two established results sit underneath all of this and are routinely left out of the frame. The first is that closing a barrier is an ecological act: barrier closures alter estuarine saltwater intrusion and stratification (Chen 2023), so as closure frequency rises the cumulative effect on salinity, water supply and navigation becomes a constraint entirely independent of the barrier’s structural condition. Established The second is coastal squeeze: a fixed defence line prevents intertidal habitat migrating landward as sea level rises, so the foreshore that dissipates wave energy narrows and the load on the structure increases (Cooper 2001, on the Essex baseline that motivated UK realignment policy in the first place). Frontier A hard line therefore degrades its own operating environment, slowly, by existing.

Frontier And there is one fully worked case of a rich country engineering a giant barrier system and then declining it. The ASCE volume Storm Surge Barriers to Protect New York City Against The Deluge appeared in 2012, the year of Hurricane Sandy, with per-structure chapters including an Arthur Kill barrier (Murphy 2012) and an East River barrier (Abrahams 2012). Speculative The US Army Corps of Engineers subsequently studied the harbour and is widely reported to have moved away from the largest barrier options; this brief flags that outcome as unverified, because usace.army.mil was unreachable. Frontier If it holds, it is the best-documented instance anywhere of a hard-defence proposal being priced, examined and turned down. Frontier Whether the Dutch precedent transfers to another estuary is itself live: antecedent geology controls what a barrier can be founded on, and a 2025 comparison poses exactly that for Maeslant against the proposed Bolivar gate system in Galveston Bay (Robbins 2025).

Frontier Storm clustering is the newest important idea in the subject and it has barely left the specialist literature. Bakker et al. (2025) show that clustered surges and multi-peak storms stress a barrier in a way single design events do not: the barrier must close, reopen and reclose, each cycle drawing on the 1-in-100 per-demand failure budget, while the enclosed basin fills from river discharge behind it. Established The authors state the effect is likely to increase with sea-level rise. Frontier This reframes barrier reliability from a question about a machine to a question about a sequence, and no design standard the pack behind this brief could reach expresses reliability that way.

Frontier Rising closure frequency, not structural fatigue, is the real end-of-life mechanism. A barrier’s design life is usually quoted structurally. Its operational life ends when closure frequency crosses the threshold at which the estuary behind it stops functioning — navigation windows, salinity regime, land drainage. Frontier On present sea-level trajectories that arrives decades before the concrete does. Established The closure-frequency projection literature exists (van den Brink 2017, on recurrence intervals for Maeslant closure) and the per-demand reliability requirement exists (Bakker 2025). Speculative Putting the two together gives the decision-relevant quantity for every surge barrier on Earth, and the pass behind this brief found no published instance of anyone doing it.

Frontier Nature-based defence is having its measurement decade. Four independent lines are open. Flexible-vegetation hydrodynamics is correcting the rigid-cylinder drag formulations that overstated attenuation (Anderson 2014; van Veelen 2020). Frontier Storm-condition field data is finally being collected, which matters because storms are exactly when vegetation lies flat, breaks, or is buried (Shi 2025). Frontier The direct-versus-indirect split — how much of the mangrove effect is vegetative drag and how much is the geomorphic platform — has been newly posed and not settled (Sánchez-Núñez 2020). Speculative If the indirect share is large, restoration does not buy back the protection on a policy-relevant timescale, because you cannot plant a platform. Frontier And modelling work continues on translating marsh geometry into wave transmission (Williams 2024).

Frontier Realignment is acquiring a predictive method, which is its central missing piece. The field’s core failure mode is that a breached site becomes mudflat rather than marsh, and nobody can reliably say in advance which it will be. Machine-learning projection of saltmarsh sub-environments for a realignment scheme is being attempted (Li 2023, a preprint). Speculative A method that predicts accretion outcome before breaching would change realignment from a hopeful intervention into a designed one, and it is the cheapest high-leverage result in this brief.

Frontier The blue-carbon case for realignment has an under-cited counterweight. Realigned marshes sequester carbon (Burden 2013), and that co-benefit is routinely used to help justify schemes. Frontier But realignment zones can also be nitrous oxide sources (Blackwell 2010), which offsets part of the claim. Speculative The net greenhouse balance of a realignment site over its first decades appears not to have been closed anywhere, and the asymmetry in how often the two papers are cited is itself informative.

Frontier Two findings inside the European economics deserve to be treated as frontier results rather than as background. First, a mainstream optimisation producing an explicit do-not-defend map covering roughly 68–76 per cent of the coastline is a policy object nobody has operationalised. Frontier Second, the 96–98 versus 59–66 per cent gap between damage protection and people protection is a measured distributional failure of the standard adaptation objective function, sitting unremarked in a highly cited paper. Speculative Both are frontier in the specific sense that they are established results whose implications have not been taken up.

Speculative The genuinely open scientific question is a reliability question, and it is answerable now. Can a fixed per-demand failure probability be maintained as demand frequency rises by an order of magnitude? Frontier The reliability-engineering apparatus is being built (Mooyaart 2023 on systematically improving closure reliability; Vrancken 2008 on the human factors that led the Maeslant closure decision to be automated so as to remove human discretion). Speculative The answer decides whether the movable barrier is a hundred-year architecture or a fifty-year one, and it is a question about maintenance regimes and failure budgets rather than about the sea.

Established Take the ambitious target seriously and the chain is ten links long. Target: a coastline defended by retreat rather than concrete, at the scale of a city region — hundreds of square kilometres, not hundreds of hectares. Established Link 1, exposure mapped on defensible elevation data: achieved globally, and the correction was large enough to matter (Kulp & Strauss 2019). Established Link 2, existing defences inventoried: not achieved, anywhere. The best exposure paper in the field states that defences protecting tens or hundreds of millions are not modelled, and no global defence inventory exists. Established Link 3, protection evaluated segment by segment: achieved for Europe; the answer is 23.7–32.1 per cent worth defending. Frontier Link 4, a distributional test applied: the method exists — compare damage reduction against people-flooded reduction — and is almost never run. Frontier Link 5, site-level sediment feasibility: accretion rate against local relative sea-level rise, site by site; the method is improving (Li 2023).

Established Link 6 is the one nobody names, and it binds. A public authority needs a statutory power to cease defending: a lawful, challenge-proof route to withdrawing protection from a place it has been protecting. Frontier Coastal-protection statutes across common-law and civil-law jurisdictions alike are typically framed as powers to protect, sometimes as duties to protect, and are silent or hostile on stopping. Speculative An authority that has read the economics, agrees with them, and wishes to withdraw may find it has no legal instrument with which to do so — and that a landowner has a colourable claim if it tries. Frontier This is the highest-value unfunded legal research in the topic and it must be done jurisdiction by jurisdiction, because it is a question about drafting rather than about doctrine.

Established Link 7 is close behind: a compensation entitlement rather than a hardship fund. Retreat schemes currently offer discretionary, post-disaster, means-tested, slow money. Frontier What is needed is a funded, non-discretionary, pre-disaster entitlement that a household can price and plan against. Speculative Link 9, social acceptance is the link everyone names and it is real (the reception of managed realignment is where schemes actually fail), but it is downstream: communities resist retreat substantially because link 7 does not exist, which is to say because the offer is bad. Established Link 8, crediting nature-based contribution inside a design standard — a method that assigns a failure probability to a marsh of width W at depth D — is absent, and it is a methodological gap rather than an ecological one. Frontier Link 10 has never been done at all. The world has executed a great deal of coastal realignment at the hundreds-of-hectares scale and no retreat at the city-region scale.

Frontier This brief waits on results in four fields, and only one of them is coastal engineering. From hydrodynamics: attenuation measured at design surge depth over defined widths, with flexibility and reconfiguration represented, and the direct-versus-indirect contribution separated (Anderson 2014; van Veelen 2020; Shi 2025; Sánchez-Núñez 2020). Frontier From reliability engineering: whether per-demand closure reliability is maintainable as demand frequency rises an order of magnitude (Mooyaart 2023; Vrancken 2008), and how clustered demands consume the budget (Bakker 2025).

Frontier From geomorphology and sedimentology: site-specific accretion rates measured against local relative sea-level rise, including subsidence, which is the single most decision-relevant number for any soft coastline and is rarely the number anyone collects. Established Antecedent land level after agricultural drainage determines realignment outcomes (Morris 2012; Hughes 2009), and antecedent geology determines what a barrier can be founded on (Robbins 2025). Frontier Both are historical facts about a specific place, which means this subject cannot be settled by a general result and will always need site data.

Frontier And from law and public finance, which is where the chain actually breaks: a drafted statutory power to withdraw protection, and a pre-disaster compensation instrument that is an entitlement rather than a grant. Speculative Neither is a research dependency in the usual sense — no experiment produces them — which is precisely why they persist. Established The pattern generalises across this cluster: in seabed mining, in floating settlement and in cross-border displacement, the binding link is likewise a missing legal or financial instrument rather than a missing measurement. Frontier That convergence is a strong hint that the useful work in coastal adaptation is instrument design, and that the impossible version is impossible for reasons that are, in principle, draftable.

Speculative Six measurements would move this subject, and every one is cheap against the cost of a single barrier.

Frontier One. Convert closure-frequency projections into absolute barrier failure rates. Take published recurrence intervals for barrier closure (van den Brink 2017) and the statutory per-demand failure requirement (Bakker 2025), propagate both forward under a sea-level ensemble, and report the expected number of failed closures per decade for each of the world’s major barriers. Speculative This is arithmetic on two existing published quantities and it appears not to have been published for any barrier.

Frontier Two. Measure wave attenuation at design surge depth, not at mean tide, over stated habitat widths and with stem flexibility characterised, reporting height and energy separately. Established Storm-condition field campaigns are the rare and valuable kind (Shi 2025), and large-flume work under surge water depths is the other route.

Frontier Three. Assign a failure probability to a habitat. Produce a method converting “marsh of width W at depth D under wave condition C” into a contribution to a system failure probability that a standards body can accept. Speculative This is the binding link for nature-based defence and it is a solvable methodological problem, not an ecological one.

Frontier Four. Build a defence inventory for one country and publish it. Position, crest level, condition, standard, ownership. Established Kulp & Strauss cannot convert exposure into risk because this does not exist. Speculative One national inventory, done properly and openly, would be the template and would immediately reprice that country’s exposure headline.

Frontier Five. Run the distributional test everywhere the efficiency test is run. Every coastal adaptation appraisal that reports damage avoided should also report people-flooded avoided, as a matter of routine, so the 96–98 against 59–66 gap is visible at the point of decision rather than buried in a supplement.

Speculative Six. Audit the statute book. For twenty coastal jurisdictions, establish whether an authority has a lawful power to withdraw protection, what standing a landowner has to prevent it, and what compensation follows. Frontier It is desk research, it is unglamorous, and it is the highest-value work in this brief.

Established The hardest components in a surge barrier are the smallest ones. The Maeslant barrier’s engineering literature is dominated not by its gates but by its ball-joints — the bearings on which the whole structure swings — with dedicated materials-characterisation studies of the composites used in them (Van Paepegem 2007; Degrieck 2006) and a design-aspects account of the crossing itself (Leendertz). Frontier These are single-point components in a system with no redundancy and a legally binding per-demand reliability target. Established That is an unusual engineering posture: most civil structures degrade gracefully and are inspected continuously; this one must work perfectly on a few hours’ notice, a handful of times a decade, increasingly often.

Established The second engineering fact is that the decision was taken away from people on purpose. The Maeslant closure decision was deliberately automated to remove human discretion, and the reason is documented in the human-factors literature the barrier generated (Vrancken 2008). Frontier A machine that operates rarely under extreme time pressure has a human-reliability problem that dominates its mechanical one, and the Dutch answer was to design the human out of the loop. Speculative Whether that generalises to barriers whose closure decision must trade off navigation, drainage and salinity against surge is not established, and the trade-off gets harder as closures get more frequent.

Established The third is that the realignment literature is already hybrid engineering and does not describe itself that way. A realignment scheme typically breaches an old seawall, builds a new embankment further inland, and lets a marsh establish in front of it. Frontier The marsh reduces the hydraulic load; the embankment carries the standard. Established That is a hybrid defence, built at scale, repeatedly, in the United Kingdom — and it is discussed in a habitat literature and a coastal-engineering literature that barely cite each other. Speculative The engineering requirement that would unify them is the crediting method: until a marsh can be assigned a failure probability, the embankment must be sized as though the marsh were not there, which is both wasteful and self-fulfilling.

Established Coastal defence is downstream of dredging and sediment logistics more than of concrete. Beach nourishment, sand engines and sediment bypassing are the same industry as port maintenance, and the material is the constraint. Frontier Realignment outcomes turn on antecedent land level after agricultural drainage (Morris 2012), so land drainage engineering, agricultural history and soil consolidation are directly adjacent and rarely in the room. Established Port and shipping infrastructure is adjacent in a harder way: a surge barrier sits across a navigation channel, and every closure is a port closure.

Frontier Reliability engineering, not coastal engineering, is the discipline the barriers actually need. Per-demand failure budgets, common-cause failure, human factors in rare high-consequence operations, and maintenance regimes for machinery that is idle by design — this is the vocabulary of nuclear and aerospace safety, and it is arriving in this field through a small Dutch literature (Vrancken 2008; Mooyaart 2023). Speculative The transfer is incomplete and the coastal profession is not yet organised to receive it. Frontier Catastrophe insurance and mortgage underwriting are adjacent in the same way and from the opposite side: they price coastal risk continuously, they revise faster than any statute, and where a state cannot decide to retreat they will decide for it by withdrawing cover.

Established Ecology is adjacent on both sides of the ledger. Realignment sites are habitat creation, blue carbon and, awkwardly, nitrous oxide sources (Burden 2013; Blackwell 2010), which connects this brief to biodiversity restoration and carbon capture at scale. Frontier Barrier closures are estuarine interventions with salinity and stratification consequences (Chen 2023). Frontier And the whole subject is coupled to sustainable megacities, because more than seventy per cent of coastal exposure is Asian and urban, and to climate migration planning, which is what happens when this brief’s instruments do not exist.

Established The jurisdictional unit is wrong almost everywhere, and this is a solvable design error. Sediment does not respect municipal boundaries; it moves along a littoral cell, a stretch of coast within which sand and mud circulate as one system. Frontier Defences, however, are typically built, funded and consented by bodies whose jurisdiction is a town, a county, or a landowner’s frontage. Speculative The predictable result is that one authority’s groyne is another’s erosion, that no body has standing to trade protection in one place against retreat in another, and that the do-not-defend map produced by the economics has no addressee. Frontier An authority constituted at the scale of the littoral cell, with a sediment budget rather than a frontage, is the institutional form this subject implies and has almost nowhere adopted.

Established The Dutch institutional answer is the counterexample to lock-in, and its details are a declared gap in this brief. The commonly described architecture — a Delta Commissioner, a ring-fenced Delta Fund, a statutory Delta Programme, and a move from exceedance-probability standards to risk-based standards expressed as individual fatality risk — is exactly the institutional machinery that makes indefinite hard defence maintainable rather than merely deferrable. Frontier This brief states none of its budget figures, dates or threshold values, because deltacommissaris.nl was unreachable and the pass behind it declined to write them from memory. Established What is verified is one component of that system and it is the load-bearing one: the statutory 1-in-100 per-demand barrier closure reliability requirement (Bakker 2025). Frontier Lock-in, on the evidence, is not a property of concrete. It is a property of concrete without a funding institution.

Established And the missing instrument is a power, not a policy. Every retreat programme that exists is discretionary. Frontier What does not exist, in most jurisdictions, is a statutory power for a coastal authority to withdraw protection — with a defined process, a defined compensation consequence, and immunity from the claim that the authority has taken property by inaction. Speculative Until it does, retreat is legally an omission, and omissions are the hardest thing for a public body to defend in court.

Established The central ethical fact in this subject is a number, and it is 96–98 against 59–66. Optimal European coastal adaptation cuts damages by 96–98 per cent and the number of people flooded by 59–66 per cent (Vousdoukas 2020). Frontier An asset-weighted objective function protects assets. Established People whose exposure is not backed by property value are systematically left in the residual, and the analysis showing this is a conventional efficiency study, not an advocacy document. Speculative The available remedies are two: put a welfare weighting in the objective function, or set a protective floor for population independent of asset value. Frontier The second is what a risk-based standard with an individual fatality-risk floor attempts — which means the Dutch answer to this critique arguably already exists and is not widely recognised as an answer to it.

Established Protection is also an intergenerational transfer, and it runs the wrong way. Defending a shoreline raises land value behind the line, attracts development, and raises the consequence of eventual failure — the levee effect. Frontier A generation that builds a defence captures the land-value gain; a later generation inherits both the maintenance obligation and a larger loss if the line goes. Speculative Coastal-squeeze narrowing of the protective foreshore (Cooper 2001) means the inherited structure is also, quietly, in a worse operating environment than the one it was designed for.

Established Retreat is where consent is most obviously owed and least well obtained. The realignment literature identifies scheme reception, not sediment, as the usual failure point. Frontier The ethical asymmetry is sharp: the household is asked to accept a permanent, irreversible loss of place, in exchange for a discretionary payment set by the body doing the asking, on a timetable the body controls. Speculative A pre-disaster, non-discretionary compensation entitlement changes the moral character of the transaction, from expropriation-by-neglect into a purchase. Frontier That is the substantive argument for the second constraint in this brief’s tech tree, and it is an ethical argument before it is a financial one.

Established Within this century the shape of most inhabited coastlines becomes a decision rather than a given. Around a billion people live below ten metres above the present high tide line, and 230 million below one metre (Kulp & Strauss 2019). Frontier The European economics say roughly a quarter to a third of coastline repays defending. Speculative If that ratio is even approximately general, then the twenty-first century coastline is an artefact selected by benefit–cost analysis, and the selection criterion is currently asset value. Frontier No civilisation has previously chosen the outline of its own continents, and the first one to do so is choosing it with a tool built to value property.

Frontier Which makes coastal defence a test case for whether societies can plan a deliberate contraction. Every other item in this category is about building something. This one is partly about unbuilding, and the historical record of states deliberately withdrawing occupied territory from habitation in peacetime, at scale, on a schedule, is nearly empty. Speculative A city region that retreated successfully — land acquired in advance, compensation paid before the disaster rather than after, defence line moved back deliberately, marsh in front of a new embankment — would be a governance demonstration far more transferable than the engineering.

Speculative And the failure mode is legible in advance. If the instruments are not drafted, the coastline will still change; it will change through unplanned loss, emergency spending, ad hoc buyouts and the slow attrition of places that stop being insurable. Handwave The choice is not between defence and retreat. It is between planned retreat and unplanned retreat, and the second is what happens by default when nobody has the legal power to choose the first.

These horizons track three things: whether barrier reliability under rising closure frequency gets measured, whether nature-based defence gets admitted to a design standard, and whether any jurisdiction drafts a power to stop defending.

• 10 yr: Frontier Closure frequencies rise measurably at existing barriers and the clustering problem (Bakker 2025) moves from a specialist result into design practice. Frontier Storm-condition attenuation measurements accumulate and the height-versus-energy conflation begins to be corrected in the grey literature, slowly. Established National defence inventories remain absent, so global exposure headlines continue to be quoted as impact projections. Speculative At least one wealthy jurisdiction publishes an explicit do-not-defend map and discovers it has no lawful mechanism to act on it.
• 25 yr: Speculative A crediting method assigning failure probabilities to habitat is drafted and enters at least one national design standard, making hybrid defence a designed rather than an incidental system. Frontier One or more major barriers reach the closure frequency at which estuarine side-effects (Chen 2023) become the binding constraint, and the resulting debate is about salinity and navigation rather than about concrete. Speculative The first statutory power to withdraw coastal protection is enacted somewhere, most probably in a small, wealthy, legally centralised state, and is immediately litigated. Frontier Retreat compensation moves from post-disaster buyout to pre-disaster instrument in at least one national programme.
• 50 yr: Speculative The coastline of every wealthy country is explicitly zoned into defended, hybrid and released, and the release category is politically survivable because the compensation instrument exists. Handwave In countries where it does not, the same zoning happens without being decided, through insurance withdrawal and mortgage refusal, which is the market performing retreat policy in the absence of a state that can. Speculative At least one first-generation surge barrier is decommissioned or superseded for operational rather than structural reasons, which would be the clearest possible confirmation of this brief’s central mechanism.
• 100 / 250+ yr: Handwave Either the great estuary barriers are replaced by outer-line successors an order of magnitude larger, committing their societies to permanent active machinery across their river mouths, or the estuaries are progressively reopened and the cities behind them relocate their vulnerable functions inland. Handwave Both are coherent; the deciding variable is whether the funding institution survives, not whether the engineering does. Speculative On the evidence of this brief, a society that keeps a Delta-Fund-class institution can hold a hard line more or less indefinitely, and a society that does not will eventually retreat whether or not it ever decided to.

• Depends on Depends on coastal geomorphology and sediment budgets at littoral-cell scale; corrected elevation data of the CoastalDEM class rather than raw SRTM, whose coastal bias runs to metres; hydrodynamic and structural modelling of overtopping and wave loading; vegetation hydrodynamics including stem flexibility and reconfiguration; reliability engineering for rarely operated high-consequence machinery, including human factors and common-cause failure; estuarine hydrodynamics for salinity and stratification response to closure; and public finance capable of funding maintenance across political cycles rather than construction alone.
• Requires (not on this map) The first constraint is a body constituted on the sediment cell rather than on the parish, the county or the frontage — and, critically, one holding a statutory power to stop defending. That second half is the part nobody names. Coastal-protection statutes almost everywhere grant authorities powers, and sometimes duties, to protect; they are typically silent or hostile on withdrawing protection from a place already protected. An authority that reads the economics, agrees that a stretch of its coast does not repay defence, and wishes to act on that finding may have no lawful route to do so, and may face a landowner claim if it tries. Cell-scale jurisdiction and a power to withdraw are one constraint because neither works alone: a body with the power but not the geography cannot trade protection here against release there, and a body with the geography but not the power can only ever add. The second constraint is financial and it is what makes the first politically survivable. Retreat compensation today is discretionary, post-disaster, means-tested and slow, which is why communities resist retreat — not from irrational attachment but because the offer is bad and arrives after the loss. What is required is a funded, non-discretionary, pre-disaster entitlement a household can price and plan against, priced against a published release map, so that a coastal property has a known exit value years before the water arrives. Every other link in this chain is technical and tractable: exposure mapping is done, segment-level economics is done for a continent, sediment feasibility is a site measurement, and assigning a failure probability to a marsh is a real but solvable methodological gap. These two are neither technical nor tractable, and they carry the entire brief.
• Enables Enables continued habitation of low-lying urban coast, which is where a large share of global GDP and more than seventy per cent of coastal exposure sits; port and navigation operation behind a movable barrier; intertidal habitat creation and blue carbon where realignment is chosen; and, in the released case, an orderly land market for coastal property rather than a disorderly one. Its absence enables nothing and forecloses a great deal, because unplanned coastal loss is the alternative.
• Adjacent Adjacent to dredging and sediment logistics, port infrastructure, land drainage and agricultural soil consolidation, estuarine ecology, blue-carbon accounting, catastrophe insurance and mortgage underwriting, and the reliability disciplines of nuclear and aerospace safety. Nature-based defence is adjacent to habitat restoration and is currently separated from coastal engineering by a missing crediting method rather than by any scientific disagreement.

Established “Three hundred million people will be displaced by sea-level rise by 2050.” The paper this is attributed to explicitly denies it. Kulp & Strauss (2019) count people living on land below a projected water level. That is exposure, not displacement. Established The authors combine future water levels with contemporary population densities and state that the results “should therefore not be taken as projected impacts”; they call the output a threat indicator and say they explicitly avoid projecting actual displacement outcomes. And the part that is always dropped: existing coastal defences, which the authors say protect “tens or hundreds of millions” of people, are not modelled at all. Frontier Three corruptions have been layered onto one paper — exposure read as displacement, one scenario’s number quoted without its scenario, and the defence caveat deleted — and the third is the worst, because it means the exposure count is an upper bound the authors themselves flagged as one. Speculative If a displacement figure is wanted, the honest answer is that no peer-reviewed global projection of sea-level-rise displacement exists, and the reason is link 2 of this brief’s workback plan: nobody has inventoried the defences.

Established “Coastal protection has a benefit–cost ratio of 8 to 15, so we should defend the coast.” The same analysis that reports European mean ratios of 8.3 and 14.9 finds protection economically justified along only 23.7–32.1 per cent of Europe’s coastline (Vousdoukas 2020). Established The ratio is high because it is computed over the segments worth defending. Frontier Quoting it without the coverage fraction inverts the paper’s policy meaning from “defend selectively” to “defend everywhere.” Established The two numbers were published together and travel separately, which is a fact about citation practice rather than about coasts.

Established “Mangroves reduce wave height by 97 per cent.” Two errors, and the first one is worth doing the arithmetic on. The large percentages in this literature are usually wave energy reductions, not wave height reductions, and energy scales as the square of height. Established A 97 per cent energy reduction leaves 3 per cent of the energy, so the residual height is the square root of 0.03, which is 17.3 per cent — that is an 82.7 per cent height reduction, not 97 per cent. Quote the energy figure as a height figure and you imply a residual wave 3 per cent as high as the incident one where the true residual is 17.3 per cent: the residual wave is understated by a factor of 5.8. Running it the other way, a 70 per cent height reduction is a 91 per cent energy reduction. Frontier This matters because residual wave height, not residual energy, drives overtopping and structural loading — so a 5.8-fold understatement is a design-relevant error, and it runs in the direction that flatters nature-based defence. Established The second error is that any such figure is measured over a specific habitat width, water depth and wave condition and does not transfer. State the width, the depth, the wave condition and whether the quantity is height or energy, or state no number.

Established “Nature-based defence can replace seawalls.” It supplements them, and in the realignment literature that is literally what it does — a marsh in front of a new embankment. Frontier The obstacle to substitution is not ecological scepticism: no accepted method assigns a habitat a failure probability, so a standards regime cannot credit it. Speculative Nature-based defence is excluded from design standards by an accounting problem, not an evidence problem — which is a solvable methodological gap and the highest-leverage target in this brief.

Established “Attenuation measured on a calm day applies during a surge.” Attenuation is strongly depth-dependent and degrades as water deepens over the canopy, which is exactly what a surge does. Frontier Storm-condition measurements are rare (Shi 2025), and that rarity is why the misconception survives.

Frontier “Restore the mangroves and you restore the protection.” If a substantial share of the effect is geomorphic — the shallow platform the forest built and holds — rather than vegetative drag, replanting does not restore it on a policy-relevant timescale (Sánchez-Núñez 2020). Speculative Contested, newly posed, and consequential for what restoration budgets are actually buying.

Established “The barrier has a design life to year X, so it is fine until then.” Structural design life is not the binding constraint. Frontier Operational life ends when closure frequency crosses the estuary’s functional threshold — navigation, salinity (Chen 2023), drainage — and when clustering erodes per-demand reliability (Bakker 2025). Speculative On current trajectories that arrives first, and it arrives without anything breaking.

Established “Barriers are proven technology.” There are very few of them, they operate rarely, and their reliability literature is dominated by human factors and single-point mechanical components such as ball-joints (Vrancken 2008; Van Paepegem 2007). Frontier A machine that must work on demand a handful of times a decade, increasingly often, has a reliability problem qualitatively different from a dyke’s. Speculative “Proven” is doing work here that a per-demand failure budget of 1 in 100 does not support.

Established “Managed realignment is retreat.” Most schemes move the defence line landward and build a new defence there. Frontier That is repositioning, and it is precisely why realignment is politically achievable where retreat is not. Established The United Kingdom has done a great deal of realignment and very little retreat, and conflating the two makes the retreat problem look solved.

Established “The Dutch defend everything, so defence always pays.” The Netherlands has the highest benefit–cost ratio in Europe at 34.3, which is why it defends (Vousdoukas 2020). Established The same analysis finds protection uneconomic for most of Europe’s coast. Frontier The Dutch case is an argument about where protection pays, not that it always does — and the transferable Dutch export is not the concrete but the institution that funds it.

Speculative “Hard defence inevitably creates lock-in.” Steelmanning the strong version: protection induces development behind the line and raises the consequence of failure; coastal squeeze narrows the dissipating foreshore and raises the load (Cooper 2001); and for barriers, rising closure frequency against a fixed per-demand failure probability raises absolute failure risk with no degradation at all (Bakker 2025). Frontier The strongest answer is the Dutch counterexample: centuries of hard defence, progressively upgraded, not catastrophically failed. Speculative The discriminator is not concrete. It is whether the maintenance and upgrade funding is institutionalised, and most coastlines have no equivalent institution.

Frontier “Barriers fail structurally.” The evidence points the other way: they fail operationally first, through closure frequency, estuarine side-effects and per-demand reliability under clustering. Speculative Structural failure would dominate only if a design event exceeded the structural envelope before the operational envelope was exhausted, and on current sea-level trajectories the operational envelope generally goes first.

Frontier “Sediment is a detail.” For soft coasts, sediment supply rather than sea level may be the binding variable, and the decisive quantity is a site-specific comparison of accretion rate against local relative sea-level rise (Morris 2012; Hughes 2009). Speculative Cheap to measure, rarely measured — and a claim with a ceiling, since above some rate of rise no plausible sediment supply keeps pace.

Established “A resolving citation is a read citation.” A fact about scholarship rather than about coasts, and it belongs here. Attempting to retrieve the field’s canonical nature-based-defence meta-analysis through a correctly addressed DOI URL returned, twice, the full text of an unrelated paper on mouse digit-tip regeneration — rendered confidently, with no error and no low-confidence signal. Frontier An agent that had asked a narrower question would have attributed regeneration content to a coastal-defence DOI, or worse, recorded “this source contains no wave attenuation data” as a finding about the source. Speculative This subject is governed by exactly the kind of large assessment report that is assembled from machine-resolved citations, and this cluster alone produced two confirmed cases of a source cited for something it does not say. Handwave Treat “everyone knows X” as a prompt to read the source, not a licence to skip it — which is why the Kulp & Strauss correction leads this section.