# The bridge model exposed a path conflict and produced a ten-sheet record for specialist review

> An independent study of a second Waitematā crossing, including two structural options, a ten-sheet stay anchorage package and seven open holds.

- Canonical page: https://www.landformlab.com/research/blog/waitemata-field-study
- Author: [Matt Strawbridge](https://www.landformlab.com/research/profile)
- Published: 1 September 2026
- Status: Independent concept study. Preliminary coordination basis. Not for construction or fabrication. No professional design, check, certification or approval recorded.
- Location: Waitematā Harbour, Auckland, New Zealand

## Summary

Detailing the stay anchorage exposed a material coordination problem: the first arrangement reduced the walking and cycling path from 3.0 metres to 2.2 metres. The model and drawing review made that conflict visible before the study advanced.

The study began with Nicolas Reid's May 2026 public proposal for a second Waitematā crossing. It developed the proposal's corridor and operating intent into a coordination model, two structural options and a ten-sheet stay anchorage package.

Revision P02 corrected five P01 defects, replaced an unsupported fatigue pass with a hold and carries seven open anchorage questions. The next test is whether a bridge specialist can continue a design change from that record. The package is preliminary and is not for construction or fabrication.

## The brief and the model

| Subject | The proposal | The model |
| --- | --- | --- |
| Corridor | Onewa to the Victoria Park viaduct, east of the existing bridge | Both tie-ins represented in the harbour context model |
| Length | 1,800 to 2,000 m | 2,000 m: 800 m main span, 350 m side spans, 250 m approaches |
| Deck | Around 38 m, multimodal | 38.0 m carrying walking, cycling, transit and traffic |
| Clearance | 43 m, matching the existing bridge | 43.0 m held over a 500 m navigation channel |
| Structure | Cable-stayed | Option A cable-stayed; Option B tied arch as comparator |

The 800 metre main span and A-frame towers are Landform design developments. Several parts of the proposal remain outside the model: the two-bridges-as-one-system operating allocation, Te Onewa station, and the Westhaven and St Marys Bay outcomes. The current model carries two lanes each way; the proposal assigns four southbound motorway lanes to the new bridge.

Source: Greater Auckland, "Build this brilliant bridge", 27 May 2026.

## The workflow

The public proposal, open data and standards register feed a named parameter file. That file drives the coordination model, preliminary sizing and drawing package. A scripted consistency screen, a read-only adversarial pass by a second model and human review write findings to a shared ledger. Corrections update the named inputs before affected outputs are regenerated.

The model that records a finding cannot verify its closure. A second model checks the fix on the following cycle, and the ledger accepts verification only from an actor other than the one that recorded the correction.

Models from Anthropic and OpenAI supported research, modelling and cross-model critique. Matt Strawbridge set the direction, assessed each finding and selected the published outputs.

## The model

- File: waitemata_crossing_city_context_P02.skp
- Contents: 798 groups, 165,795 faces, 21 tags, 16 scenes
- Ground: LINZ one metre elevation model, Auckland 2024
- Seabed: LINZ hydrographic soundings, inverse-distance merged into the same surface
- City: OpenStreetMap footprints, heights from tags where present
- Water depth on the alignment: about 15 m at the north tower, 6 m at the south

The context exposed an early placement error. The existing harbour bridge was first placed from a geocoded position and ended in mid-air over deep water. Its abutments were then re-derived from the ground surface at the tightest crossing recorded by the elevation model. Later coordination depends on both structures meeting the recorded terrain.

## Options

Option A develops the proposal's cable-stayed form and is the option carried into the detail package. Option B is a half-through tied steel truss arch of 550 metres, tested as a comparator on the same deck, with a total horizontal thrust of 344.4 MN, 172.2 MN per rib, 32.4 metre hanger spacing and 62 strand hangers.

Option B passed 69 of 70 checks in the current consistency screen. The unresolved check is the half-arch erection method, so the option remains a comparator and did not advance to detailing.

The two options are mutually exclusive alternatives on one corridor and are never combined into a single scene.

## Why the anchorage package exists

The model originally carried a blister anchorage on the deck plate. The arrangement occupied part of the declared three-metre active path, leaving 2.2 metres clear at each of the 32 positions per face. Detailing exposed a coordination defect that was not visible in the general arrangement.

Four families of stay-to-deck anchorage were compared. The external node outboard of the deck edge was developed, with the edge girder and transverse diaphragm retained as the fallback. Family D was selected to keep the path clear and support replacement. Its structural efficiency has not been established, and family C has more precedent. No built example of an outboard anchorage on a bridge of this class was identified in the sources reviewed, so acceptance remains hold H1.

| Quantity | Value |
| --- | --- |
| Reference stay | BACK-01, 73 strands |
| Preliminary SLS demand | 8.242 MN against a 9.165 MN threshold at 45% GUTS, ratio 0.899 |
| Preliminary ULS demand | 11.319 MN against a 13.239 MN threshold at 65% GUTS, ratio 0.855 |
| Cable GUTS | 20.367 MN |
| Stay inclination at deck | 30.6 degrees |
| Back and main fan pitch | 8,387 mm and 10,484 mm |
| Assemblies | 2 towers x 2 fans x 32 positions x 2 deck edges = 256 |
| Provisional plan gap | 0.30 m between the path outer face and the node inner face |

## The package

Package WAI-LFR-ZZZ-BRD-DR-ST-06 revision P02 contains ten A3 sheets, ST-0600 to ST-0609. The sheets record the anchorage decision, supporting evidence and seven open holds. The set is a preliminary coordination basis, stamped do not fabricate, with no designer, checker or approver recorded.

The NZTA Bridge Manual excludes cable-stayed bridges and railway loading, so a project-specific cable-stay and rail design basis, documented departures and special studies would all be required. The structure would be managed as a Category 1 structure under the Highway Structures Design Guide, including a wholly independent design review.

## What the review found

Revision P01 received a structured review. P02 records a response to each of six findings: five changes to the model or document and one withdrawn claim.

1. The register listed ten sheets and the document contained nine. P02 contains ten separate A3 sheets.
2. The assembly count omitted the second tower. 128 became 256.
3. Demand, allowable thresholds and the cable breaking force were separated and labelled.
4. Back-fan and main-fan pitch were reported separately, 8,387 mm and 10,484 mm.
5. The anchorage extended 0.85 m into the walking and cycling path. P02 moves the housing outside the path and records a provisional 0.30 m plan gap.
6. Fatigue was written as a pass and had not been analysed. The claim was withdrawn and hold H3 opened.

External validation found a separate defect in the coordinate frame. An earlier frame was left-handed, so every three-dimensional view mirrored Auckland in plan. Three adversarial review cycles and the automated checks missed the error because each relied on the same internal frame. Comparison with external ground truth exposed it. The frame was corrected and the model rebuilt. The consistency screen has a clear limit: it tests agreement with the project's selected ranges. Design adequacy and external ground truth require separate checks.

## Verification record

| Check | State | Evidence |
| --- | --- | --- |
| Anchorage typology screened | PASS | Four concept families compared on ST-0601 |
| Demand and thresholds separated | PASS | Demand 8.242 and 11.319 MN; thresholds 9.165 and 13.239 MN |
| Assembly quantity reconciled | PASS | 2 towers x 2 fans x 32 x 2 edges = 256 |
| Model path clearance | PASS | Nodes outside 19.0 m; path outer edge 18.7 m; stay axis 20.0 m |
| Project design basis accepted | HOLD | Owner-approved basis and Category 1 review not recorded. H1 |
| Built precedent for the arrangement | HOLD | None identified in the sources reviewed. H1 |
| Deck torsion from eccentricity | HOLD | Not analysed. H2 |
| Governing fatigue detail identified | HOLD | Not established. H3 |
| Fatigue categories | HOLD | Assigned by inspection, not by analysis or test. H3 |
| Supplier geometry | HOLD | System not selected. H4 |
| One stay absent | HOLD | The global load case does not exist. H6 |

## Open holds

- H1 Precedent and arrangement
- H2 3D structure and torsion
- H3 Local fatigue
- H4 Supplier envelope
- H5 Durability
- H6 Access and replacement
- H7 First article

The ten-sheet package records seven holds for this anchorage detail and remains marked Do not fabricate. Each hold names the evidence required for closure. Wider transport, global structural, foundation, geotechnical, wind, seismic, vessel impact and owner approval questions sit outside the package's scope.

## What I learned running it

The workflow produced the model, drawings and revision record. Assessing evidence, technical gaps and option credibility requires bridge-engineering knowledge I do not have. Every technical output therefore needs specialist review.

It linked sources to named inputs, rebuilt geometry after input changes and maintained a clear sequence of next actions. No conventional baseline was recorded for time, cost or review effort, so the study makes no productivity claim. The next test should be run with bridge specialists, using a firm's own project records, reference details and design criteria.

## Verdict

- Traceable: partial
- Coherent: partial
- Reviewable: partial
- Ready to continue: untested

No conventional baseline has been recorded for time, cost or review effort. Handover will be tested by asking a specialist to change the named parameters, regenerate the model and drawings, and update the affected holds without reconstructing the project history.

## Documents

- Field Study 01, Waitematā Crossing: assets/waitemata-study/docs/landform-field-study-01-waitemata.pdf
- Stay anchorage package, revision P02: assets/waitemata-study/docs/WAI-LFR-ZZZ-BRD-DR-ST-06_P02-anchorage-package.pdf

## Related studies

- [A hut on Aotea Great Barrier Island](https://www.landformlab.com/research/blog/aotea-method-study)
- [A land-bank listing at Massey read against the plan in force](https://www.landformlab.com/research/blog/chamberlain-method-study)
- [A consented townhouse site at Paparangi](https://www.landformlab.com/research/blog/paparangi-method-study)