Underground Cities

Underground urbanism is the deliberate placement of urban functions below grade. Not the tunnels that carry water, sewage, power and trains — cities have built those for a century and a half and are good at it — but floor area, destinations, and in the exotic case dwellings. The subject has a scholarly literature, a planning-theory tradition running back decades, and two celebrated cases that are habitually treated as one thing and are in fact two unrelated propositions.

The first is the Helsinki model: deep, rock-hosted, purpose-excavated space, programmed by a public authority and written into a statutory plan. The second is the Montreal model: shallow, largely incidental space in the basements and podiums of buildings that were being built anyway, stitched into a weather-protected pedestrian network. The first needs competent bedrock and half a century of institutional continuity. The second needs a requirement to connect and almost no excavation at all. Conflating them is the most consequential error in the topic, because a city that reads Helsinki as a template without Helsinki's geology has committed to the most expensive version of the wrong programme.

A sourcing constraint, stated before anything else. Two descriptive figures a reader will expect — the extent of Montreal's RÉSO network, and the coverage and facility count of Helsinki's Underground Master Plan — could not be verified for this brief. They are named here and no number is printed for either. What would settle each is stated in section 2, and section 14 argues the failure is not an accident.

What follows. The built record and what each case actually is; excavation cost as a function of depth, and why that is the governing economic term; the third-party risk envelope that dominates it in exactly the places where the demand is; subsurface property right and three-dimensional tenure, the quiet obstacle; ventilation, egress, daylight and wayfinding; utility and transport tunnelling as the mature practice that could pay for the immature one; and the exotic end — whether a city with a substantially inhabited subsurface is achievable, and the strongest arguments that underground space is permanently a place for infrastructure rather than for people.

Established Underground space has been argued as an object of statutory land-use planning for at least thirty-five years, and specialists are still arguing for it. Bobylev's 2009 paper on mainstreaming underground space into a city's master plan appeared in Land Use Policy rather than in a tunnelling journal, and the venue is the intellectual move the whole subject rests on: the subsurface as something a city plans, rather than something engineers leave behind. Subspace Associates published a two-part mined underground space master plan for Minneapolis in 1990. Jamalludin, Zaini and Hussin were still making the case for such a plan in Malaysia in 2016. Alofe, Bastani and Tryggvason are now building quantitative geophysical proxy indicators for underground space use against Stockholm's real city plan. Frontier This brief cites all four from bibliographic records verified in the pack's session and read none of their texts, for a reason given two paragraphs down. What the sequence establishes is not a finding but a shape: three and a half decades of practitioners arguing for a planning instrument is evidence the instrument is missing, not that the argument is weak.

Established Helsinki's programme is far older than the literature that made it famous. A 1988 record in Tunnelling and Underground Space Technology describes four underground projects then planned for the city, which places an institutional programme at least three decades before the master-plan papers. Established The authoritative account of that programme is by one author: Ilkka Vähäaho, who headed the City of Helsinki's geotechnical division and published two 2016 papers on the development of urban underground space there, one in Tunnelling and Underground Space Technology and one in Energy Procedia. Frontier The important thing for a reader is what that implies about transferability. Helsinki is not a policy experiment a city could run; it is an accumulation, and the accumulation started before most of the cities now interested in the subject had a subsurface policy at all.

Established Two figures this brief will not print, and exactly what would settle them. Helsinki's Underground Master Plan is widely described as a citywide statutory instrument reserving named spaces in bedrock, with a large number of completed facilities and civil-defence shelter capacity covering much of the population. RÉSO, Montreal's ville souterraine, is widely described in kilometres of connected passage linking metro stations, towers, retail and institutions, grown incrementally from the early 1960s. Speculative Neither was verified, and no length, area, facility count or shelter fraction is stated here. For Helsinki the cheapest route is Vähäaho's Energy Procedia paper, cited below as a lead and not as a source of figures because its text was not obtained. For RÉSO no scholarly record was retrieved at all, and settling it needs a dated primary source from the network's operator or the City of Montreal, with a stated definition of what counts as connected.

Established The reason those texts were unreadable is structural, and it inverts the usual relationship between authority and access. Elsevier, Wiley, Taylor & Francis and SSRN deposit no abstracts to Crossref. Tunnelling and Underground Space Technology, Land Use Policy and the rest of this subject's most selective venues are Elsevier titles. Frontier The consequence is that the readable corpus in subsurface urbanism skews heavily toward open-abstract publishers — MDPI, Springer Open, IOP, conference proceedings — while the record-only material comes on average from more selective venues. In this subject the most authoritative literature is precisely the least readable, and any assessment assembled by machine will be systematically weighted toward the newer and less filtered end. That is worth saying plainly because it is invisible in the output: a brief built this way looks like a brief built any other way.

Established Excavation cost scales with depth, and the empirical form of that scaling is ground deformation. Cheng and colleagues monitored an ultra-deep five-level metro station excavation in Tianjin soft soil, 38.3 m deep. Maximum lateral wall displacement was about 40 mm, ranging from 0.028 to 0.184 per cent of excavation depth with a mean of 0.087 per cent; wall-top vertical displacement ran from −0.23 to 0.04 per cent of maximum depth; column rebound averaged 3.5 mm. Established The abstract reports these as outperforming comparable bottom-up excavations in Shanghai, and offers them as quantitative benchmarks. Established Deformation expressed as a fraction of depth is the whole point: it means displacement grows with the hole, and displacement is what turns into support, monitoring, underpinning, insurance and negotiation. The machines are not the term that scales.

Established The zone of influence extends far beyond the excavation, and most of the settlement is spent before anyone digs. The same study places maximum settlement at one-third of the excavation depth from the wall, a primary influence zone reaching three times the depth and a secondary zone reaching five times. Established For a 38.3 m excavation the secondary zone is roughly 190 m in every direction. Established And the finding that should govern how cities programme this work: on completion of the diaphragm wall, settlement had already exceeded 60 per cent of the final maximum. Frontier Two consequences follow that are rarely stated together. In dense fabric a single deep station touches hundreds of third-party buildings, and the negotiation with each of them is a cost scaling with depth that has nothing to do with rock or machinery. And mitigation designed around the excavation phase is aimed at the minority of the problem. Established Building settlement in the same study depended strongly on distance and foundation type, raft foundations settling much more than pile-raft.

Established In existing urban fabric the governing risk is to the neighbours, not to the hole. Zhang, Huang and Wu analysed the Erligou station extension on Beijing Metro line 16 — described in their abstract as the largest underground subway station then under construction in China — specifically to quantify construction risk to an adjacent existing station structure and track, and to identify the high-risk sub-projects within the sequence. Established They note that stations are moving toward multidimensional, large-scale, complex structures, and that this presents unpredictability, coupling amplification and high risk. Established Chen, Tsai and Wen back-analysed the St. Ignatius High School Station excavation on the Taipei MRT: 31.65 m in eleven stages using bottom-up techniques, with a water and soil burst-in at the final stage that induced very large diaphragm deflection, ground settlement and minor damage to adjacent buildings. Frontier The tail risk sits at the bottom of the excavation, which is also where the cost of stopping is highest.

Established Deep excavation resists prediction because excavating changes the ground being excavated. The Taipei back-analysis found significant changes in soil properties caused by the construction works themselves. Established Zhang, Yang and Azzam, running three-dimensional parametric studies of an underground pumping station, state the epistemic problem in the abstract: the complexity and non-repeatability of practical projects make it impossible to understand these issues thoroughly through real projects. Established Their result is that constitutive-model choice alone materially changes the predicted displacement field — Modified Cam-Clay resembling Hardening Soil with small-strain stiffness, both greatly improved on Mohr-Coulomb — and that soil-wall interface properties and soil parameters both affect excavation-induced displacement to a large extent. Established Hefny and colleagues, comparing plane-strain and three-dimensional models of the Rod El Farrag metro station excavation in Egypt against field measurement, found corner and cross-wall stiffening significantly affecting both wall deformation and neighbouring settlement. Frontier Put together: two engineers with the same site data and different defensible modelling choices produce different predictions, and the difference is not small. That is a structural source of estimate variance upstream of every cost figure in the topic.

Established The front end is longer than the construction, and it is the most useful single datum in the subject. Atzl, Ullmann and Schmidt document the Vienna U1/9 extension from Reumannplatz to Oberlaa: 4.6 km with five new stations, making U1 the longest underground line in Vienna at 19.2 km by 2017. Permit design under Austrian railway law ran through 2009 and 2010, approval came in January 2011, tenders were drafted in 2010 and published in summer 2011, the award went to Strabag, construction started in spring 2012 and tunnel excavation commenced at the end of January 2014. Established That is roughly four and a half years from the start of permit design to the first excavation, for a modest extension of a mature system, in a competent European city, with an experienced public client and no novelty anywhere in the scope. Frontier Any subsurface urbanism proposal that does not budget that front end is not a proposal. It also explains why the subject's few successes are decades-long programmes rather than projects: the fixed institutional cost is amortised over the programme or it is not amortised at all.

Established Underground environments impose measurable health penalties, and the best study in this literature says something more specific than that. Chan and Chen collected 525 surveys across eleven underground sites in Hong Kong with objective field measurement at five, before and after the COVID-19 outbreak, analysed by multigroup structural equation modelling. Their opening premise is the standard one: no natural ventilation, no sunlight, more health risk than aboveground counterparts. Frontier Their results are not the standard ones. Post-outbreak, satisfaction was significantly higher for greenery, connectivity with the aboveground environment, thermal comfort, ventilation, indoor air quality, acoustic comfort and lighting, excluding wayfinding, and both physical and psychological health were significantly better. Visual comfort's effect on health was significantly greater after the outbreak; wayfinding, greenery, and acoustic and thermal comfort mattered only before it. Established Connectivity to the aboveground environment was the only factor significant for both physical and psychological health in both periods. Frontier That is the design finding of the topic and it points away from where the money usually goes. What makes underground space tolerable is not that it is nicely finished. It is that you can tell how to get out and up.

Established The evidence base for lighting windowless space was built in the wrong climates, and colour is being recruited to solve a navigation problem. Sithravel and Ibrahim analyse five architectural lighting factors for windowless workplaces — intensity, spectrum, timing, duration, spatial distribution — and report a dearth of literature supporting psychophysiological wellbeing in the tropics, the evidence being predominantly from seasonal-climate contexts. Frontier Circadian design for underground habitation therefore rests on a base that may not transfer to the latitudes where most future urban growth is. Frontier Lv and colleagues, meanwhile, propose a colour system for underground parking dividing colours into safe, prohibited and theme categories and then into basic, auxiliary and accent tiers. Note its primary stated purpose: guidance. The one dimension Chan and Chen found had not improved is the one an entire colour framework is being built to address.

Established There is no price for subsurface volume, and barely a method for pricing the vertical dimension at all. Zheng and Wang observe that studies of the three-dimensional nature of urban space, and particularly the vertical dimension, have remained limited, and introduce vertical and horizontal land-use configuration variables into a spatially filtered multilevel hedonic model. Frontier Land markets price the surface parcel and, through air rights and height limits, implicitly price the volume above it. Almost nothing prices the volume below, and this paper is the seed of the method that would. Established Groundwater, meanwhile, has become a first-order driver rather than an afterthought: Peintner and colleagues on the Semmering Base Tunnel describe grouting as serving both rock-mass integrity and groundwater protection amid growing sensitivity toward water resources. Frontier A city's subsurface is also its aquifer, and the regulatory weight on that is increasing.

Frontier Making the subsurface legible is the live research front, and it is unglamorous. Alofe, Bastani and Tryggvason's Stockholm work develops quantitative geophysical proxy indicators for underground space use and maps them against an actual city plan — which is to say, it builds the data layer a subsurface zoning instrument would need in order to exist. Frontier From the other direction, Li, Song, Gao and Tao apply monocular depth estimation with geometric consistency constraints to non-Lambertian surfaces, constructing three-dimensional models of a corridor and an underground parking structure from a single camera, motivated explicitly by lidar cost being the obstacle to large-scale deployment. Speculative Between those two sits the proposition that most cities do not know what is beneath them and could find out cheaply. That is the least fashionable and probably the highest-return research programme in the subject.

Frontier Valuation of three-dimensional urban space is emerging as a quantitative discipline. Zheng and Wang's spatially filtered multilevel hedonic model with explicit vertical configuration variables is the first instrument in this corpus capable, in principle, of being extended downward. Speculative Extending it is not trivial: a hedonic model needs transactions, and subsurface volumes are rarely transacted separately from the surface parcel, so the data the method requires is suppressed by the legal condition the method exists to help fix. Speculative The plausible route is indirect — price the observable consequences of subsurface connection, such as retail rents and footfall at connected versus unconnected addresses, and infer the volume's value from what it does rather than from what it sells for.

Frontier Habitability research has produced one genuinely strong design and one genuinely surprising result. Chan and Chen's multigroup structural equation model across eleven sites, with a pandemic acting as a natural experiment between waves, is unusually well identified for this literature, and it returned satisfaction and health both higher after the outbreak — the opposite of what a naive account of underground space during a respiratory pandemic would predict. Speculative The mechanism is not established. A plausible reading is comparative: users judged underground environments against a surface world that had itself become constrained, which would make the result a fact about the comparator rather than about the space.

Frontier Ultra-deep excavation is generating publishable benchmarks, and nobody is aggregating them. Cheng and colleagues explicitly offer their Tianjin deformation ratios and influence-zone geometry as quantitative benchmarks. Frontier Peintner and colleagues' Semmering post-excavation grouting work is the same instinct applied to groundwater. Frontier Zhang, Huang and Wu's Erligou risk quantification adjacent to an operating station addresses the specific condition that dominates urban work. Speculative Each is a data point in a reference class that does not exist. A shared, mandatory deformation and cost database — depth, geology, method, predicted against actual — is a purely administrative object with no research content, and its absence is why every deep excavation learns privately.

Speculative Two specific numbers are missing, both tractable, and either would move the subject. The first is the connection threshold. If the subsurface's product is connection rather than floor area — which Chan and Chen's connectivity result supports — then a pedestrian network's value should grow super-linearly in connections, implying a threshold below which such networks fail and above which they become self-reinforcing. Speculative That threshold has not been estimated for any pedestrian network anywhere, and it is estimable from existing footfall and tenancy data. Speculative The second is the crossover between influence-zone cost and mined-construction risk. Cut-and-cover is cheap and disruptive; mined construction avoids the surface but, as Yang, Mei and Jun note of undercutting methods used in prosperous cities and complex surroundings, carries great risk. Speculative Nobody has published the crossover as a function of urban density and depth, and it is the single highest value-per-effort analysis in the topic.

Handwave At the far end sits a question the field has not framed. If the subsurface were surveyed, titled and priced, would anyone build in it who is not already building infrastructure? The honest answer is that nobody knows, because the counterfactual has never existed anywhere. Handwave It is possible that the entire observed pattern — subsurface space appearing only as a by-product of transport, defence and utility programmes — is a property of the missing institutions rather than of the ground, and that fixing the institutions would release a latent market. It is equally possible that the by-product pattern is the whole truth and there is no latent market at all.

Established The workback plan splits, because the target does. Substantial subsurface infrastructure and non-residential space is a near-term institutional problem. Subsurface habitation is a different problem that binds somewhere else and is much harder. Running them together is how the subject generates both false optimism and false despair.

Frontier For non-residential subsurface the ordered chain is: a subsurface cadastre; a registrable, alienable volumetric title; a valuation method; depth-band zoning in a statutory plan; a liability regime for inter-owner ground movement calibrated to the one-third, three-times, five-times influence structure; a shared deformation and cost database; the published influence-zone against mined-construction crossover; and an anchor use that pays for the void. Established The cadastre gates everything downstream: you cannot title, price, plan or regulate a volume nobody has surveyed, and most cities have no reliable record of their own subsurface. Frontier But the cadastre is cheap and tractable. The link that actually binds is the last one.

Frontier The binding link is an anchor use that accepts an oversized void. Subsurface volume has no market until something is in it, and nothing goes in until the volume exists, because nobody excavates speculatively at these unit costs against an unpriced asset. The way out is an infrastructure client — metro, district energy, stormwater detention, logistics, data, civil defence — that is going to excavate anyway and agrees to excavate larger, leaving usable surplus. The measurement is surplus volume per anchor project and whether anyone takes it up. The decision is whether an infrastructure client will accept a cost premium for space it does not need.

Established This is the fourth independent chain in this cluster to terminate on the same constraint, and that is a finding about the whole category rather than about any one brief. Automated Construction Systems binds on an anchor buyer committing multi-year volume sufficient to load one production line above about seventy per cent utilisation. Future Housing Systems binds on an anchor demand commitment decoupled from the speculative absorption rate. Arcologies binds on an independently viable first increment, which is the same requirement wearing a financing costume: something must pay for the first piece before the network effect that justifies the whole exists. Frontier Four chains, four literatures, four sets of authors who do not cite each other, one constraint. The constraint is not technological and it is not scientific. It is that no instrument exists for committing a buyer to the first tranche of a thing whose value is contingent on the tranches after it. Speculative That is a procurement and public-finance design problem, and it is the reason this brief depends on Megaproject Governance: anchor commitments are made or broken in exactly the appraisal and approval machinery that literature is about.

Established For subsurface habitation the binding link is different and worse: a life-safety strategy for sleeping occupancy below grade. The measurement is the depth and population at which performance-based fire engineering can no longer demonstrate compliance. The decision is whether a regulator accepts it. Frontier That threshold is unpublished, and section 6 argues it may not exist anywhere.

Established This brief waits on results that other fields produce and other briefs assess. The formal dependency is on Megaproject Governance, and it is not decorative. Every subsurface programme of consequence is a public megaproject appraised on a forecast, approved by people who will not hold office when the outturn is known, and delivered across a decade. Frontier Subsurface projects are structurally the worst-forecastable class of construction — Natarajan's fat-tailed overrun distributions, Liu and colleagues' formalisation of unbuildability risk, and the flat contradiction between Park's finding that reference-class forecasting cut average overrun from 38 to 5 per cent across 107 major projects and Themsen's finding on a 23.6-billion-kroner Danish megaproject that it does not improve accuracy at all. Frontier That dispute is live, both sides are peer-reviewed, and this brief declares the tie rather than averaging it.

Frontier Four non-brief dependencies sit outside the corpus. Applied geophysics must deliver cheap, repeatable subsurface characterisation at city scale, which is what the Stockholm proxy-indicator work is reaching for. Cadastral law must deliver a volumetric title that a registry can record and a lender can foreclose on. Fire engineering must deliver a defensible probabilistic case for sleeping occupancy below grade. And indoor positioning must deliver navigation that degrades gracefully, because a wayfinding system you cannot rely on in a fire is not a life-safety solution.

Speculative One dependency runs the other way. Utility and transport tunnelling is mature, continuously funded and institutionally secure. Subsurface urbanism is none of those things. The realistic path is not that urbanism funds tunnelling but that tunnelling funds urbanism, by leaving behind more void than it needs — which makes the mature programme's procurement rules, not the immature programme's ambitions, the operative variable. Speculative That inverts the usual dependency direction on this map, and it means the intervention with the highest leverage is a clause in a metro contract rather than a research programme.

Frontier Build one subsurface cadastre and publish the cost. Pick a mid-sized city, combine geophysical proxy indicators of the Stockholm kind with cheap monocular or photogrammetric capture of accessible voids and existing utility records, and report coverage, accuracy and total cost per square kilometre. The deliverable is not the map; it is the price of the map. Every downstream instrument has been blocked for thirty-five years on the presumption that this is expensive, and the presumption is untested.

Speculative Estimate the connection threshold. Take existing weather-protected pedestrian networks, measure destinations reachable without going outside against footfall, retail rent premium and tenancy survival, and fit the curve. If value grows super-linearly in connections there is a threshold, and knowing it tells a city whether its half-built network is below the point where further investment is wasted or above the point where it is self-reinforcing. Requires no fieldwork beyond data access.

Speculative Publish the influence-zone against mined-construction cost crossover. Cheng and colleagues supply the influence-zone geometry; Yang, Mei and Jun supply the method taxonomy comparing cross-diaphragm, pile-beam-arch, middle-hole and double-side-drift methods on safety, cost and time. Joining them into total delivered cost across a density gradient is analysis, not experiment, and nobody has done it.

Frontier Run the wayfinding trial under degraded systems. Measure navigation success for naive users in a real underground complex with devices, without devices, and under simulated emergency with signage obscured and the positioning system unavailable. Chan and Chen identified wayfinding as the one dimension that did not improve; Gerges, Demian and Adamu's customised-instruction results raise the governance question of what happens when the system that issues the instructions is down. The trial answers both.

Speculative Find the population threshold at which performance-based fire engineering stops being demonstrable below grade. No source in the underlying corpus addresses it. It may be genuinely unpublished, in which case saying so is a contribution and specifying how it would be measured — probably by asking practising fire engineers at what occupant load and depth they would decline to sign — is a better one.

Speculative And the terminal one: measure retention, not take-up. For any subsurface dwelling that exists, ask what fraction of residents remain when free to leave, against a matched above-grade comparator. Take-up measures price; retention measures the thing everyone actually argues about.

Established Everything underground runs on machinery that must never stop. Ventilation is mechanical and permanent. Lighting is artificial and permanent. Groundwater must be excluded or pumped, permanently, against a head that increases with depth. Structure must resist ground pressure that also increases with depth, permanently, with no possibility of external inspection of the loaded face. Frontier Every one of these is a continuous operating cost and a continuous failure mode, and each fails toward a worse condition than its surface equivalent: a failed rooftop plant is an inconvenience, a failed ventilation shaft is an evacuation.

Established Egress underground is upward egress, against smoke that also travels upward. In a tall building smoke rises away from occupants descending, and gravity assists. Below grade the two flows share a direction and a route. Frontier The vertical-egress result assessed at length in Arcologies transfers directly and is worth stating once here rather than restating in full: in a full-building phased evacuation drill in a 22-storey Tokyo office building, Kim, Park and Mizuno found that a single pedestrian ascending from Floor 1 to Floor 2 triggered a backward-propagating stop wave reaching fourteen consecutive floors. Frontier Underground, that ascending pedestrian is not an anomaly in the flow. Underground, everyone is the ascending pedestrian. Every building code in the world was written around descent, and below grade the case is reversed.

Frontier The compensating design moves are known and each has a cost. Compartmentation with defined places of safety; pressurised stairs; smoke extraction sized for buoyant flow along the egress path; short maximum distance to daylight as a hard performance requirement rather than an aspiration; and evacuation lifts, given Mossberg, Nilsson and Andrée's unannounced hotel experiment showing occupants choosing lifts over nearer stairs. Speculative A connectivity performance standard — maximum time to daylight, surface connections per unit area — would encode Chan and Chen's finding into code, and no code contains one.

Established The mature neighbour is transport and utility tunnelling, and it is mature in every respect this subject is not. It has standards, a professional community, insurance products, procurement precedent, and continuous public funding. Vienna's U1/9 sequence is what a routine subsurface project looks like when everything works. Frontier Subsurface urbanism's realistic strategy is to attach itself to that programme rather than to compete with it for legitimacy.

Frontier Mining supplies the deep-excavation methods and the risk culture. Yang, Mei and Jun's comparison of cross-diaphragm, pile-beam-arch, middle-hole and double-side-drift methods by fuzzy multi-criteria evaluation across safety, cost and time is a mining-derived way of thinking that surface construction does not have. Frontier Excavation scheme optimisation is itself a separately managed cost lever, as LI's Pudong deep-pit work argues.

Frontier Four demand-side adjacencies could each play the anchor role. District energy needs thermal storage volume. Stormwater management needs detention volume and is being forced to find it by changing rainfall statistics, which links this brief to Infrastructure Resilience. Data centres need thermally stable, secure, low-value-per-square-metre space. Urban logistics needs consolidation volume near the centre and cannot afford surface land there. Speculative Each of these is a real client with a real budget whose requirements happen to be exactly the requirements underground space satisfies best: no daylight needed, no aspect needed, high value on thermal stability and security.

Frontier And three enabling adjacencies. Computer vision and cheap depth sensing for volumetric survey; indoor positioning for wayfinding, which is also the fallback problem in an emergency; and property law's strata and mineral-rights traditions, which are the only existing legal machinery for owning a volume rather than a surface. Speculative The last is the one nobody in the engineering literature reads, and it is where the binding institutional constraint in section 9 would have to be solved.

Frontier One adjacency runs sideways rather than upstream. The megastructure family assessed in Arcologies is the same proposition rotated: both delete the gaps between buildings, both create single-owner non-parcelised space, both replace egress-to-outside with movement-within, both need an anchor buyer, and both lack a property instrument for the volume they create. Speculative The differences are that the subsurface case has a geological precondition the vertical case does not, and a life-safety inversion that is worse. Whether “megastructure” is the right shared category, with arcology as its vertical case and subsurface urbanism as its inverted one, is a question this map has not settled.

Established The quiet obstacle is that almost no jurisdiction has a coherent property right in the subsurface volume. Cities zone the surface. They regulate and in places trade the air above it. Below it there is generally mineral law, a utility easement regime, and a common-law rule about how deep ownership extends that was written for a world without tunnel boring machines. Frontier Nothing in that set produces a definable, registrable, alienable volume at a stated depth band that a bank will lend against and an insurer will cover.

Frontier Three and a half decades of practitioners asking for the instrument is the evidence. Minneapolis had a mined underground space master plan in 1990. Bobylev argued in 2009 for mainstreaming underground space into a city's master plan in a land-use policy venue. Jamalludin and colleagues argued for the same instrument for Malaysia in 2016. Speculative This brief does not know what happened to the Minneapolis plan and says so; an apparently complete, apparently unexecuted subsurface programme for a mid-sized North American city is exactly the abandoned record worth recovering, and recovering it is an archival afternoon.

Frontier The liability regime is the piece nobody mentions and everything depends on. If a deep excavation's secondary influence zone reaches five times its depth, then every subsurface owner materially affects their neighbours' ground, and the party-wall analogue for depth does not exist. Speculative Without it, no volumetric title is insurable, because a developer cannot know their exposure before they start. That single missing statute may explain more of the observed absence of private subsurface development than any engineering constraint.

Established The counter-argument deserves stating: where subsurface rights do exist, they have not produced underground cities. Mining law, mineral rights, subsurface easements and depth-limited strata title all exist in various jurisdictions and none has generated an inhabited subsurface. Frontier The right is therefore necessary and demonstrably not sufficient, which is consistent with the anchor constraint in section 4 being the real binder and the tenure gap being the thing that makes the anchor's surplus untradeable afterwards. Speculative Read that way, the two constraints are not competing explanations but a sequence: the anchor creates the volume, and the tenure instrument decides whether anyone can do anything with it once it exists.

Frontier The first question is who ends up below. Underground floor area is cheaper to occupy than equivalent surface area and worse to occupy, which means that in any market allocation the people who end up there are the people with the least choice. Speculative A city that builds substantial subsurface accommodation without an explicit allocation rule has built a mechanism for sorting by income into a daylight gradient, and the sorting will be invisible in the aggregate housing statistics because floor area is floor area.

Established Daylight has historically been treated as a right, not an amenity. Minimum daylight, aspect and ventilation provisions are among the oldest content in housing codes anywhere, written in response to nineteenth-century tenement conditions. Frontier Any serious subsurface habitation programme requires those provisions to be relaxed or reinterpreted, and the argument that artificial light and mechanical ventilation are now good enough is exactly the argument the original codes were written to refuse. Speculative That is not a knockdown objection, but a proposal that begins by asking to disapply the protections is obliged to say what replaces them, and the honest replacement — a connectivity performance standard with a maximum time to daylight — has never been drafted.

Frontier The third-party question is a compensation question. A deep excavation imposes settlement, vibration, noise and risk across a zone reaching five times its depth, on hundreds of people who receive nothing. Speculative Most jurisdictions handle this through damage claims after the fact, which is a mechanism biased against those least able to litigate. A subsurface programme at scale would need a compensation rule set before the first hole, not a claims process after it.

Speculative And the consent question about sleeping occupancy. Every successful deep underground typology — metro, retail, parking, plant — relies on waking, transient, alert occupants. Residential subsurface has none of those protections, and asking someone to sleep in a place from which egress is upward against buoyant smoke is a materially different request from asking them to shop there.

Frontier The defensible version of this subject is large and unexciting. A city that surveyed its subsurface, wrote a volumetric title, and required every metro, district energy and stormwater project to oversize its void would accumulate a substantial subsurface within a few decades, releasing surface land for the uses that actually need daylight. That is not a speculative programme. It is an administrative one, and it appears to be roughly what Helsinki has been doing since at least 1988, geology permitting.

Speculative The exotic version — a city with a substantial inhabited subsurface — is a different claim and the arguments against it are stronger than its advocates admit. The strongest is not psychological. It is that every deep typology with a good safety record achieved it by relying on occupants who are awake, alert and passing through, and habitation removes all three protections at once while inverting the direction of egress. Speculative The second strongest is economic: underground space is most valuable where surface land is dearest, which is where density is highest, which is where third-party cost is greatest — so demand and cost are driven by the same variable, and greenfield underground cities are cheap and pointless while urban ones are valuable and unaffordable.

Handwave The case for the exotic version rests on a change in the comparator, not a change in the technology. If surface conditions degrade far enough — heat, storm, radiation, air quality — then thermal stability, shielding and separation from the atmosphere stop being consolation prizes and become the product. Handwave Nothing in the current evidence base tells us where that crossover is, and a brief that pretended otherwise would be inventing it. Speculative What can be said is that the crossover is a fact about the surface, not about the subsurface, which means the honest way to track this question over the coming decades is to watch habitability above ground rather than capability below it.

These horizons track institutions, instruments and one missing regulatory decision. Nothing here waits on an invention, which makes the near horizons less forgiving than usual: a discovery can arrive early, and a cadastral statute cannot.

• 10 yr: Frontier Subsurface cadastres piloted in a handful of cities, most likely in Northern Europe and East Asia, driven by utility strike costs rather than by urbanism. Frontier Geophysical proxy indicators of the Stockholm kind maturing into a usable planning data layer. Speculative Volumetric title drafted somewhere and probably not used. Speculative The connection threshold and the influence-zone crossover both remain uncomputed unless someone chooses to, which is the frustrating thing about both: they are cheap and unowned.
• 25 yr: Speculative Depth-band zoning adopted in a small number of jurisdictions, with the anchor-oversizing rule as its operative content rather than its zoning maps. Speculative A shared deformation and cost database, if one national infrastructure client mandates deposit as a condition of contract — the only realistic route, since nobody volunteers this data. Frontier Non-residential subsurface expands substantially in dense, expensive, geologically favourable cities, and is not called underground urbanism by anyone doing it.
• 50 yr: Speculative A recognisable subsurface property market in two or three cities, with volumes changing hands separately from the surface parcel. Speculative Performance-based fire engineering for below-grade sleeping occupancy attempted, contested, and probably accepted in one permissive jurisdiction before anywhere else. Handwave Whether that jurisdiction's experience generalises depends entirely on whether it has an incident, which is an unpleasant way to say that the evidence base for subsurface habitation will be written by whatever goes wrong first.
• 100 / 250+ yr: Handwave No basis for forecasting the built environment on this horizon. The only structural argument available is that the two things which make underground space unattractive — the absence of daylight and the inversion of egress — are permanent physical facts rather than technological limitations, so the subject's ceiling is set by how much the surface degrades rather than by how good the engineering gets. Handwave A civilisation that went substantially underground would have done so because the surface stopped working, and that is a brief about something else.

• Depends on A subsurface cadastre, which is the precondition for everything and is an administrative product rather than a research one; cheap volumetric survey, from geophysical proxy indicators at city scale down to single-camera depth estimation inside existing voids; a registrable volumetric title with depth bands; a liability regime for inter-owner ground movement calibrated to the observed influence-zone geometry, without which no subsurface developer can know their exposure before starting and no insurer can price it; a valuation method for three-dimensional urban space, which today has neither transactions to fit nor a legal object to attach to; a shared deformation and cost database that would give reference-class forecasting an actual class instead of a pooled bag of dissimilar projects; and, for the habitation branch only, a life-safety strategy for sleeping occupancy with upward egress against buoyant smoke that a regulator will accept. Above all of it sits the dependency this brief carries formally: the appraisal and approval machinery in which an anchor commitment either gets made or does not.
• Enables Surface land release is the direct output: every square metre of parking, plant, storage, logistics, waste handling, thermal storage and detention moved below grade is a square metre of expensive, daylit, well-connected surface freed for something that needs those properties. Beyond that, a working subsurface tenure system is the enabling instrument for deep district energy and seasonal thermal storage, for urban stormwater detention at the volumes changing rainfall statistics require, and for the consolidation logistics that dense cities cannot site above ground. The connection-mandate model transfers immediately and cheaply to almost any city with a cold or hot season and a dense core, and requires no excavation programme at all.
• Adjacent Transport and utility tunnelling, which is the mature practice this immature one must attach itself to; mining engineering, for deep excavation methods and for a risk culture surface construction lacks; geotechnical numerical modelling, where constitutive-model choice is a live and under-acknowledged source of estimate variance; hydrogeology, since a city's subsurface is also its aquifer and groundwater protection is now a first-order design driver; property law's strata and mineral-rights traditions, which are the only existing machinery for owning a volume; fire engineering, which owns the binding constraint on the habitation branch; environmental psychology and lighting research, whose evidence base was built in seasonal climates and may not transfer to the tropics; and computer vision and indoor positioning, which supply both the cadastre and the wayfinding fix.

Handwave “Helsinki proves any city can go underground.” Helsinki sits on competent shallow bedrock. That single fact makes excavation cheap, support minimal, water ingress manageable, and civil-defence shelter requirements an asset rather than a burden, because the rock was going to be blasted anyway. In soft ground the same ambition produces the Tianjin condition: 38.3 m of excavation, 40 mm of wall displacement, an influence zone reaching five times the depth, and 60 per cent of the final settlement already spent before excavation begins. Established The depth gradient is dominated by geology, not by depth. Copying Helsinki without Helsinki's rock is copying the expense and none of the mechanism.

Established “RÉSO and PATH are underground cities.” They are climate-protected pedestrian networks, largely in the basements and podiums of buildings constructed for other reasons, at shallow depth. Their product is connection, not subsurface floor area. Frontier Calling them cities inflates the category and, worse, makes planners draw the wrong lesson: the transferable achievement is a connection mandate on new development, not an excavation programme. Speculative This brief prints no extent figure for either network, because it could not verify one, and notes that the widely circulated kilometrage for RÉSO has no source in the material behind this assessment.

Handwave “Building down is cheaper than building up because you do not buy land.” You do not buy land, but you buy the removal of everything that was there, permanent support against pressure that rises with depth, permanent dewatering, and the entire third-party risk envelope. Frontier Building up is subsidised by land you already own; building down is taxed by ground you must fight. The cost per square metre of underground floor area rises with depth while its amenity value falls, so there is a crossover, and above it building down is irrational. Frontier Two exceptions dissolve the argument locally and they are the important ones: uses with negative amenity requirements lose nothing by being buried, and competent rock removes most of the cost gradient.

Handwave “Excavation cost is falling because tunnel boring machines keep improving.” Machine cost is a minority term in urban work. The dominant terms are third-party risk, monitoring, underpinning, negotiation and disruption across an influence zone reaching five times the excavation depth, plus a front end that ran roughly four and a half years from permit design to first excavation for Vienna's 4.6 km U1 extension. Established And the residual ground risk is irreducible, because excavation changes the properties of the soil being excavated.

Handwave “Underground cities are climate-resilient and energy-efficient by default.” Ground temperature stability is real. So are permanent mechanical ventilation, permanent artificial lighting, permanent pumping against groundwater and permanent structural support, all running for the life of the asset and all failing dangerously. Frontier The energy case has to be argued on net operating energy across a lifecycle, including a great deal of embodied concrete and the pumping load, and this brief has seen no such analysis for an underground urban development. Its absence is the point.

Frontier “People will not live underground; it is psychologically intolerable.” This is the critics' overclaim and the evidence is more specific and more actionable than the slogan. Users do report health penalties, but satisfaction across greenery, thermal comfort, ventilation, air quality, acoustics and lighting can be high, and the single factor significant for both physical and psychological health in both study periods was connectivity to the surface. Established Wayfinding was the persistent exception. Frontier The finding is not that people cannot; it is that connection and legibility are the design variables that decide whether they will.

Handwave “Underground is safer, because it is shielded.” It is a life-safety inversion. Smoke rises along the egress route, occupants must climb against it, orientation is degraded, and occupants choose lifts over nearer stairs in unannounced experiments. Frontier A 22-storey drill with about 2,400 participants showed one person ascending against the flow propagating a stop wave across fourteen floors; underground, ascent is not the anomaly but the design condition. Established Every successful deep typology relies on waking, transient, alert occupants. Residential subsurface has none of those protections, and this — not daylight, not psychology — is the strongest technical argument against underground dwelling as opposed to underground everything else.

Handwave “We can model deep excavation accurately now.” Constitutive-model choice alone materially changes predicted displacement. Three-dimensional corner and cross-wall stiffening significantly changes results relative to two-dimensional plane strain. And geotechnical researchers describe their own domain as having a complexity and non-repeatability that makes thorough understanding through real projects impossible. Frontier Combine that with fat-tailed overrun distributions in megaprojects generally and subsurface work is structurally the worst-forecastable class of construction there is.

Speculative “It is just tunnels; there is nothing urban about it.” The opposite overclaim, and also wrong. Underground space has been argued as a statutory land-use planning object since at least 2009 in a land-use policy venue, with master plans proposed for Minneapolis in 1990 and Malaysia in 2016 and geophysical proxy indicators now developed against Stockholm's city plan. Frontier There is a planning discipline here, and its central complaint for thirty-five years is that cities lack the instrument.

Speculative The uncomfortable one: underground urbanism may be a civil-defence programme that acquired a peacetime justification. If the initial voids were paid for by a defence budget, subsequent commercial use is nearly free and the observed economics are a subsidy artefact rather than a demonstration. Speculative The case against is that Montreal's network had no defence rationale, and that district energy, stormwater, logistics, data and metro are real non-defence sponsors that could play the same role. Frontier Either way the mechanism is the one section 4 identifies as binding, and it should be argued as an economic structure rather than as a historical accusation.

Established Finally, a claim about this brief rather than about its subject. The material behind it came through a single bibliographic channel that failed partway through and began returning one fixed unrelated payload to every query, silently, with no error and no low-confidence signal. Several packs in this project caught the same class of failure independently, including a citation service returning organ-cryopreservation records for unrelated queries and a publisher URL serving a different article than the requested identifier. Frontier That is a real fact about how citations get corrupted in machine-assembled scholarship, and it belongs in a brief whose subject is governed by exactly the kind of assessment documents now assembled that way. Established It is also why two of this subject's most-quoted numbers appear nowhere above.