Six jobs, including the parts that went wrong.

J‑2418 · 2026–05

Traeth Melynog realignment

Client

County council, coastal team

Extent

41 ha, surveyed twice

Disciplines

Topographic, multibeam, UAV

Achieved

±18 mm land, ±42 mm bed

Deliverable

DWG, DTM, change surface, volume report

A managed realignment scheme where the argument was about how fast the saltmarsh edge was retreating. The council had photographs going back to the nineties and two consultants with different opinions. What they did not have was two surveys on the same datum.

We surveyed the whole frontage in June 2025 and again eleven months later — beach and marsh by GNSS and UAV, the subtidal by multibeam on a rising tide, everything reduced to ODN from a gauge we levelled in ourselves. The change surface showed 1.4 m of mean retreat at the northern end and almost none at the southern, which is not what either consultant had assumed.

What went wrong: the second survey lost a full day to a gauge that flooded its housing. We caught it because the tidal reduction stopped agreeing with the predicted curve by a growing offset, re-deployed, and re-ran the affected lines rather than modelling around the gap. It cost a day and it is the reason the two epochs are comparable.

J‑2455 · 2026–06

Capel Nant, Llanberis

Client

Conservation architect

Listing

Grade II*

Capture

14 scan setups, tethered UAV

Registration

3.1 mm mean residual

Deliverable

Revit LOD 300, E57, rectified elevations

An 1841 chapel with a Welsh slate roof that nobody was willing to walk on, and a gallery that had dropped enough over 180 years to be visible from the door. The architect needed sections accurate enough to design a new roof structure that would sit on walls which are nowhere near parallel.

Scanned from fourteen setups inside and out, with a tethered UAV for the roof planes and the bellcote, all registered to a total station control network so the cloud is on the same grid as the site survey outside. Drawn to Level 3, with wall thicknesses as measured and out-of-plumb annotated rather than straightened.

What went wrong: the first registration pass came back at 6 mm because two setups in the gallery were sharing too few targets in a space that turned out to be acoustically and geometrically awkward. We went back for a morning and added three, which took the mean residual to 3.1 mm. The architect got a model they could design a roof against instead of one they had to trust.

J‑2470 · 2026–08

Cwm Aderyn access track

Client

Renewables developer

Extent

6.2 km, 41 ha corridor

Network

18 stations, 2.4 mm max residual

Ruling gradient

22%

Deliverable

As-built model, deviation report, sheet AD 148

Six kilometres of new access track to a repowering site at the head of the cwm, across ground with no mobile signal, a 22% ruling gradient and a scheduled monument the route had to stay 60 m clear of. The sheet on the front page of this site is this job’s survey.

Because network RTK was unusable for most of the corridor, the whole thing was controlled by an eighteen-station traverse observed with total stations and adjusted by least squares, then extended locally with a base-and-rover pair. Every station is witnessed to two rock features, which is the only thing that works on open hill where a peg lasts one winter.

What went wrong: nothing dramatic, which on this job was the achievement. The corridor was set out and as-built in six visits across five months, and the final deviation report closed with no section out by more than 34 mm in plan. The network was re-observed twice during the programme, both times because sheep had comprehensively destroyed a station, not because it had moved.

J‑2361 · 2025–11

Porth Wylan quay monitoring

Client

Harbour authority

Duration

14 months, 62 epochs

Targets

44 prisms, 6 datum stations

Precision

±1.2 mm 3D

Deliverable

Epoch reports, trend plots, two red notifications

A masonry quay wall with a documented lean, next to a berth that the authority wanted to keep in service while a scheme was designed. Forty-four prisms on the wall face, six datum stations set well back on rock, and an automated total station reading the lot every four hours.

Amber at 4 mm cumulative, red at 8 mm, agreed with the structural engineer before the first epoch. Over fourteen months the wall moved 5.9 mm seaward at its worst point, almost all of it in two winter storm sequences, and recovered about a third of that in the following months — which told the engineer the movement was largely elastic and changed the scheme.

What went wrong: we raised two red notifications, and one of them was ours. A datum station had been disturbed by a contractor working nearby, which put an apparent 9 mm on the whole wall at once. Uniform movement across every target on a masonry structure is not structural movement, it is a reference problem, and the call to the engineer said exactly that before anyone stopped work.

J‑2402 · 2026–03

Hen Felin mill conversion

Client

Developer, 22-unit conversion

Extent

3,100 m² GIA, 1.8 ha site

Disciplines

Measured building, topographic, PAS 128

Achieved

QL-B2 on 71% of detected runs

Deliverable

Revit, DWG, attributed services drawing

A nineteenth-century woollen mill on three levels with a leat running under part of the footprint, being converted to flats. The building survey was straightforward. The site was not: the mill had been rewired, re-served and partly demolished across four separate decades, and no two record drawings agreed.

EML and dual-frequency GPR across the yard found eleven services, of which four were not on any drawing we had been given, including a live 11 kV run at 700 mm that the developer’s programme had a foundation crossing.

What went wrong: we could not resolve depth on two runs under a reinforced slab and said so, marking them QL-B4 rather than guessing. The developer wanted a number. We recommended three trial holes instead, which cost less than a fortnight of standing time would have.

J‑2289 · 2025–07

Afon Gwyrfai flood model survey

Client

Consulting engineer, flood modelling

Extent

9.4 km reach, 118 cross-sections

Structures

14 bridges and culverts

Achieved

±20 mm bed level

Deliverable

Sections, structure schedule, HEC-RAS geometry

Cross-sections through a river reach for a hydraulic model, plus a dimensioned record of every structure that could throttle a flood: soffit levels, invert levels, wingwall geometry, and the debris screens nobody had measured since they were installed.

Waded and levelled rather than flown, because a photogrammetric model cannot see a riverbed, and a model built on a water surface instead of a bed level will under-predict a flood for the life of the scheme.

What went wrong: a July spate put 400 mm on the reach for nine days in the middle of the programme. We stood the crew down rather than surveying sections at a flow that would have made the bed unmeasurable, and picked it up eight days later. The client was told on day one, not at the invoice.