When ground conditions are screened for a hotel site, most practitioners stop at a binary judgment — shaky or not shaky. But what shows up in the pre-construction estimate is not how much the ground shakes; it is how many metres below grade the hard stratum sits. Piles do not work unless they reach the bearing stratum. The depth at which that stratum appears therefore converts directly into pile length, and from there into construction cost. Using surface-ground data from the National Research Institute for Earth Science and Disaster Resilience and the bearing-stratum depths published for the Tokyo waterfront by the Association of Tokyo Geological Survey Companies, this article quantifies how pile length can differ by as much as 4.7x within the same category of “reclaimed land”, and estimates what that means per room for a 150-room model hotel.

The first-pass screening layer — 24 submarkets ranked by microtopography and surface S-wave velocity

Start with the public data actually usable at the entrance to site screening. J-SHIS, the Japan Seismic Hazard Information Station run by the National Research Institute for Earth Science and Disaster Resilience, divides the country into a 250m mesh and publishes, for each cell, the microtopographic classification and average S-wave velocity in the top 30m (AVS30), along with the Amplification factor relative to engineering bedrock. Because a set of coordinates is all it takes, it suits first-pass screening at the stage where dozens of candidate sites are still in play.

The chart below shows the results for 24 major hotel-cluster submarkets nationwide, ordered from the lowest — that is, softest — end.

The ordering follows a clear logic. The softest is Kinshicho in Tokyo (reclaimed farmland, 151.7 m/s), followed by Makuhari New City in Chiba, and Toyosu, Ariake, Odaiba and Shibaura in Tokyo — Tokyo Bay landfill occupies the bottom of the table. At the hard end sit Sakae in Nagoya (gravelly plateau, 294.9 m/s), Sannomiya in Kobe (alluvial fan, 293.9 m/s) and Tennoji in Osaka (gravelly plateau, 272.0 m/s). On amplification factor, Kinshicho’s 2.28 against Sakae’s 1.30 means surface shaking differs by 1.75x for the same seismic input.

Notably, central Osaka and Nagoya cannot be sorted by whether a site is on the waterfront. Umeda in Osaka is inland but sits on delta and coastal lowland at 167.6 m/s, and Nagoya’s Meieki is on a natural levee at 168.8 m/s — both softer than the reclaimed land of Minato Mirai in Yokohama (174.4 m/s). A prime location in front of a terminal station has no correlation whatsoever with firm ground.

What screening flattens out — a 5.4x depth gap inside “reclaimed land”

Here is the core of it. Look only at surface S-wave velocity and the Tokyo Bay landfill looks monolithic. Among the 11 waterfront districts covered below: Tokai–Haneda in Ota 160.4, Shibaura–Konan in Minato 156.2, Takeshiba–Shibaura Pier in Minato 155.5, Ariake / Aomi in Koto and Odaiba in Minato 154.4, Shin-Kiba–Arakawa river mouth in Koto 153.3, Shiohama–Edagawa in Koto 151.4 — a max-to-min ratio of only 1.06x. Bundling these into a single “soft ground” box at first-pass screening is a natural call, and that is what happens in practice.

Overlay the waterfront bearing-stratum data published by the Association of Tokyo Geological Survey Companies (originally from the Tokyo Metropolitan Government Bureau of Port and Harbor’s “New Edition: Ground Map of the Port of Tokyo”), however, and an entirely different picture appears. The association compiles, district by district along the waterfront, the type and appearance elevation of competent bearing strata (the basal gravel bed of the alluvium, buried-terrace gravel, the Tokyo Gravel Formation, the Edogawa Formation and the Kazusa Group, all showing N-values of 50 or above). Those values run A.P. -5 to -20m for the upper sandy soil layer of the Edogawa Formation under Harumi–Kachidoki in Chuo and Toyosu in Koto, against A.P. -65 to -70m for the basal alluvial gravel bed under Tatsumi, Shiomi, Shin-Kiba and Wakasu in Koto and Shinonome 1–2 chome.

Taking representative values (the midpoint of each range), there is a 5.4x spread, from A.P. -12.5m to A.P. -67.5m. And these sites are within a few kilometres of one another, all of them “Tokyo Bay landfill”, and AVS30 cannot tell them apart. Plotted against each other, the relationship looks like this.

Every point falls within a narrow 151.4–160.4 m/s band on the horizontal axis, while the vertical axis stretches from 12.5m to 67.5m. The correlation coefficient is -0.38 — directionally a weak tendency for districts with higher AVS30 (slightly firmer surface ground) to have a shallower bearing stratum, but with almost no power to explain the variance in depth. A shakiness metric does not predict pile length. That is only to be expected: AVS30 is an average picture of the top 30m and says nothing about what lies beneath. On a site where the bearing stratum sits at A.P. -67.5m, the softness of the upper 30m is barely different from a site where it sits at -12.5m.

Mapped, the distribution of depth follows neither distance from the shoreline nor the age of the reclamation. Toyosu (A.P. -5 to -20m) and Tatsumi / Shinonome (A.P. -65 to -70m) sit adjacent across a single canal, while Odaiba and Ariake, furthest out into the bay, fall in the middle at A.P. -20 to -50m. This reflects the shape of an old buried valley beneath the fill — something unreadable from the surface.

Converting to pile length — 300m versus 1,400m in a 150-room model

To convert depth into cost, it is more accurate to pass through pile length (m), a physical quantity, first. The following model building is set out with its assumptions stated explicitly.

The table below applies those assumptions to calculate pile length and total pile length by district. Total pile length is 20 piles × pile length; the per-room figure is shown alongside.

At the shallowest — Harumi–Kachidoki and Toyosu — a 15.0m pile length gives 300m of total pile, just 2.00m per room. At the deepest — Tatsumi, Shiomi, Shin-Kiba, Wakasu and Shinonome — 70.0m piles give 1,400m in total, 9.33m per room. That is 4.67x as a ratio and 1,100m in absolute terms. Those 1,100m flow through to excavation volume, rebar tonnage, concrete volume and construction days alike.

Converting to yen — ¥440,000–880,000 per room, 3 to 6 points of total construction cost

From here, unit-price assumptions enter, so the results are presented as a sensitivity analysis rather than a single assertion. Construction unit prices for cast-in-place concrete piles swing widely with pile diameter, method, project scale and the presence of subsurface obstructions, so quoting one market price would be meaningless. Three levels — ¥60,000/m, ¥90,000/m and ¥120,000/m — are applied to all 10 districts to show how pile cost per room moves.

The gap between the shallowest and deepest districts is ¥440,000 per room at ¥60,000/m, ¥660,000 at ¥90,000/m and ¥880,000 at ¥120,000/m. Per building, that is ¥66M to ¥132M. Against a total construction cost of ¥2.12B, pile work alone moves across a 4.7-point band, from 1.27% to 5.93%.

A comparison helps put that magnitude in context. Dividing the acquisition prices of the 300 domestic properties held by seven listed hotel REITs by room count gives a median of ¥20.51M per room; narrowing to the 51 properties inside Tokyo gives ¥24.14M (ownership master as of August 31, 2026). The ¥660,000 per-room gap estimated at ¥90,000/m equals 2.7% of that Tokyo median acquisition price.

Look at it from the revenue side as well. Estimated settled ADR for Koto ward compiled by MetroEngines Research is ¥13,800 as a 12-month median from September 2025 to August 2026 (N=41 properties). At that rate, ¥660,000 per room is roughly 48 room-nights of revenue. For reference, estimated settled ADR over the same period is ¥19,100 in Chuo (N=151 properties), ¥17,900 in Minato (N=130), ¥13,000 in Shinagawa (N=35), ¥12,100 in Ota (N=57) and ¥11,300 in Edogawa (N=18). Among the waterfront wards, the relatively high-rate Chuo and Minato also sit on the shallower side of the bearing stratum — higher rates and lighter foundation costs pulling in the same direction.

In a rising construction-cost market, this gap means something different

What matters more than the absolute size of the pile cost is that it lands in the middle of a broad rise in construction costs. Hotel construction costs based on the MLIT housing-start statistics rose 41% in two years, from ¥1.38M per tsubo in 2022 to ¥1.95M in 2024, and stayed high through 2025 at an all-structures average of ¥1.952M and ¥2.405M for steel-frame construction. We have back-calculated the actual per-room construction cost from planning data in separate work, and tested from housing-start statistics how far the choice of structural system moves the unit cost. When the denominator is swelling, the extra burden of choosing a site with a deep bearing stratum grows heavier in absolute terms year after year, even at an unchanged percentage.

Put the other way round, this is also a market in which selection pays. Among candidate sites in the same waterfront area, the same zoning district and the same walking-distance band, choosing the shallower side compresses total pile length to a fraction. Unlike land price or station distance — variables everyone looks at — bearing-stratum depth is barely examined at the early screening stage. It is one of the few levers still movable before groundbreaking. Other variables change construction conditions through site selection alone, such as the parking spaces required by municipal ordinance, which we have set out separately.

Putting it into site screening — three layers

The practical point comes down to not confusing layers 1 and 2. Layer 1 looks at how the ground shakes in an earthquake, and feeds discussions of insurance premiums, structural design and business continuity planning. Layer 2 looks at the foundation method and its cost. They are different data driving different decisions. As shown here, on Tokyo Bay landfill the layer-1 metric barely moves while layer 2 moves by 5.4x — so layer 1 cannot stand in for layer 2.

Conclusion

Ground assessment for a hotel site needs to be handled as two separate layers: shakiness, and bearing-stratum depth. Rank 24 submarkets nationwide by surface S-wave velocity and the Tokyo Bay landfill clusters at almost identical values, landing in a single box at first-pass screening. Yet across those same 11 districts the bearing stratum appears anywhere from A.P. -12.5m to A.P. -67.5m, a 5.4x spread, and the correlation between the two is only -0.38.

Converting that depth gap into pile length for a 150-room, 5,400m² model gives 300m against 1,400m — a 4.67x difference. Varying the pile unit price across ¥60,000–120,000/m produces ¥440,000–880,000 per room, or 1.27% to 5.93% of total construction cost. At the ¥90,000/m mid-case, ¥660,000 per room equals 2.7% of the ¥24.14M median per-room acquisition price of Tokyo properties held by listed hotel REITs, and roughly 48 room-nights at Koto ward’s estimated settled ADR.

Land price, station distance and floor-area ratio are numbers everyone sees once the candidates are lined up. Bearing-stratum depth can be pulled from public data, and still most screening sheets have no column for it. As a cost that can be moved before groundbreaking, it is well worth one more column.