Deep Basement Monitoring: One Reading That Does Not Fit
You are the survey manager for the monitoring consultant on a 24 m deep basement excavation in Kowloon City. The site is 6 m from a 1968 six-storey residential building. The monitoring plan has precise levelling settlement points on the building at 10 m intervals, inclinometers in the retaining wall, and groundwater standpipes. Alert level is 15 mm settlement, action level 20 mm. At 07:30 on a Saturday, after 80 mm of rain in 24 hours, the technician reports that point SP-07 has settled 18 mm since baseline. All neighbouring points read 4–6 mm. The technician notes that the level was set up on a tripod that had been left overnight beside the site hoarding, and the reading was taken in strong sunlight with the staff on the shaded side. The contractor wants to resume excavation at 08:00 and asks you to 'confirm it is a bad reading'. The AP has not yet been informed.
1. How would you determine whether SP-07 is a real movement or a measurement error? Set out the checks in the order you would carry them out.
Read the worked answer
I would treat the reading as real until proven otherwise, and work from the instrument outwards.
- Instrument stability: inspect the tripod and level. A tripod left overnight beside a hoarding may have been disturbed. Check the bubble, the footscrews and the collimation error. Do not simply re-read from the same setup.
- Independent re-observation: set up a different level and staff, preferably with a different observer, and re-level SP-07 from a second benchmark. Close the loop between at least two stable benchmarks.
- Benchmark stability: check the reference benchmarks for movement, especially if they are on the building or on ground affected by the excavation.
- Environmental effects: strong sunlight on one side of the staff can cause bending or refraction. Re-observe with the staff shaded, at a different time, and with shorter sight lengths.
- Corroboration: compare SP-07 with adjacent settlement points, inclinometer readings, groundwater levels and crack monitoring. A single point moving 18 mm while neighbours move 4–6 mm is possible but demands explanation.
- Time series: plot the full history. A step change coincident with rain or a site activity is different from progressive settlement.
Only after these checks would I classify the reading as error or movement.
2. If the reading is confirmed, what immediate actions and notifications are required, and who decides whether excavation resumes?
Read the worked answer
If an independent re-observation confirms 18 mm, the alert level has been exceeded and the action level is close. I would immediately notify the AP, the RSE and the contractor's site agent, and ensure the client and monitoring team are informed. The affected area should be treated as a stop-work zone for excavation and any activity that could worsen movement.
The building should be inspected for new cracks, tilt or distortion, and the inclinometers and groundwater readings should be reviewed at once. Additional monitoring points may be needed. The AP and RSE should assess the structural and geotechnical implications and decide whether temporary support, underpinning or other remedial measures are required. The Buildings Department should be notified where the project's supervision plan or the circumstances require it.
The surveyor does not decide whether excavation resumes. The surveyor reports the measured facts, the reliability of the data and the trend. The AP, RSE and contractor's engineer decide on resumption, and that decision should be recorded with the monitoring evidence.
3. What changes would you recommend to the monitoring plan and data management to prevent a repeat?
Read the worked answer
I would recommend both technical and procedural changes.
Technically, the plan should use at least three stable reference benchmarks outside the zone of influence, with regular checks between them. Critical points should be observed by two independent methods or from two independent setups. Automatic total stations or settlement cells could be considered for high-risk points, with manual levelling as a check. Sight lengths should be controlled and staff shaded.
Procedurally, every reading should carry metadata: instrument ID and calibration date, observer, time, weather, setup location, benchmark used, and whether the reading was a repeat. Field forms should require a check reading and a closure. Any reading that exceeds a trigger should automatically generate an alert to the AP and RSE, not wait for a weekend report.
The plan should also define what counts as a valid alert: a single reading, a confirmed reading, or a trend over two or three rounds. Data should be backed up and version-controlled, and changes to baselines should be authorised and recorded. Training should cover the difference between re-reading from the same setup and independent verification.