Industries / Construction and property / Engineering and architecture / Geotechnical engineering
Monitoring
Monitoring puts instruments in the ground and on the structures built in it, to measure movement, water pressure and load, so engineers can see whether the ground behaves as the design assumed.
5
Milestones achieved, from 1931 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1931
Vibrating wires measure strain
A thin steel wire stretched across a gap sounds a note when plucked, and the note changes with its tension, so it shows how far the gap has opened or closed. The idea dates to 1888 and was tested in 1928, but commercial use began in France in 1931, when André Coyne patented what he called an acoustic indicator; from the 1930s vibrating-wire gauges became worldwide practice for monitoring dams, and many firms make them today.
LessonDunnicliff recalls that strain readings in Iraq's Dokan Dam in 1956 "turned out to be of no value" because his team had not worked out how to turn strain into stress.
Credit
André Coyne, who founded the maker Telemac in 1947, by John Dunnicliff's account
1969
Change the design as built
Ralph B. Peck of the University of Illinois described in Géotechnique the observational method that Karl Terzaghi had used: measure how the ground behaves during construction and modify the design as the work goes on. Peck wrote that it "often permits maximum economy and assurance of safety", but should not be used unless the designer has a plan of action for every unfavourable situation the readings might reveal, and that readings must be reliable and reported so as to prompt action. Eurocode 7 now lists an observational method as one of four ways to verify a design.
Credit
Ralph B. Peck, after Karl Terzaghi
2004
Wrong readings before a collapse
On 20 April 2004 a deep excavation for Singapore's Circle Line subway collapsed beside the Nicoll Highway, killing four people. The Committee of Inquiry's interim report said instrument-based monitoring "must be effective, adequately resourced and maintained unlike in C824", the contract in question, and found that strut load readings supplied by one "specialist", relied on by the main contractor and the Land Transport Authority, were wrong.
LessonReadings protect a site only if they are accurate, compared with predictions on time, and acted on by someone with the authority to stop work.
Credit
the Committee of Inquiry, with MIT's Andrew Whittle among four international experts appointed by the Land Transport Authority
2023
Europe's ground movement, checked
The European Ground Motion Service, set up by the European Environment Agency under the EU's Copernicus program, turns Sentinel-1 satellite radar images into free maps of ground movement across Europe to the millimeter, updated each year. In December 2023 a validation consortium separate from the producers (Sixense, NGI, TNO, BRGM, GeoSphere and Terrasigna) reported tests at 38 areas, and a 2024 comparison with GPS stations and radar reflectors found most differences in velocity below 2 mm a year.
LessonThe satellite's viewing direction matters: in France's Alpes-Maritimes, 54% of recorded landslide points fell inside areas flagged as moving from one direction and 25% from the other.
Credit
European Environment Agency; the validation consortium
2025
Village emptied before the collapse
On 19 May 2025 the authorities evacuated the 300 people of Blatten, in Switzerland's canton of Valais, after scientists saw more and more rock falling from the Kleines Nesthorn onto the Birch Glacier above the village. That day Geoprevent, hired by the consulting geologists Geoformer and the local authorities, set up a radar that measures movement from across the valley, and the authorities published its readings to explain the evacuation. On 28 May the glacier collapsed and buried most of the village; one man, aged 64, was reported missing.
LessonThe scientists knew something was coming, but the glacier's total collapse was not predicted, said Christophe Lambiel of the University of Lausanne.
Credit
the local authorities and their advisers Geoformer, with radar by Geoprevent
Watch for
L-band radar saw Blatten coming
After the collapse, a team in the European Space Agency's MODULATE project went back through archived L-band radar images from Japan's ALOS-2 and Argentina's SAOCOM satellites and found the Kleines Nesthorn moving from 2016–2017, speeding up to about 50 cm a year by 2023 and over 150 cm a year by August 2024. L-band radar uses a longer wavelength than Europe's Sentinel-1, and the team says it picks up movement on steep or vegetated alpine slopes where Sentinel-1 often fails. The analysis was done in hindsight.
Achieved whenL-band satellite data flag a slope before it fails, not afterwards, and an independent report confirms the warning against ground instruments or the event.
Claimed by
Tazio Strozzi (GAMMA Remote Sensing) and Andrea Manconi (WSL Institute for Snow and Avalanche Research SLF), through ESA, August 2025
Open
01
Catch readings that disagree
A checker that compares the different instruments on one excavation (strut loads, wall movement from inclinometers, ground settlement and water pressure) and flags any reading that does not fit what the others imply, plus a public test set with known faults planted in it.
The briefCredit · unclaimed02
Plans checked before digging starts
A checker that reads the instrumentation and monitoring plan for a deep excavation and lists every instrument that lacks a review level, an alarm level, a reading frequency, a named person who must act, or a planned action when a level is crossed, with each flag pointing to the line of the plan it concerns.
The briefCredit · unclaimed
Open 01 · the brief
Catch readings that disagree
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A checker that compares the different instruments on one excavation (strut loads, wall movement from inclinometers, ground settlement and water pressure) and flags any reading that does not fit what the others imply, plus a public test set with known faults planted in it.
Why it is still open
Singapore's inquiry into the 2004 Nicoll Highway collapse found that wrong strut load readings had been relied on by the contractor and the Land Transport Authority. Published machine methods check one stream of data at a time: Fang and colleagues (2023) cleaned sensor data from one deep pit foundation, and Green and Jaspan (2023) flagged unusual inclinometer readings across Britain's rail network. A web search found no open set of excavation records with labeled instrument faults.
Built with what exists
OpenSees, a free finite element program with soil models, to simulate excavations stage by stage; Python; instrument data from published excavation case histories for a final check.
First step
Model one braced excavation in OpenSees, export strut loads and wall movements at each digging stage, and plant a 20% calibration error in one strut.
Done looks like
50 simulated excavation records each with one planted fault (a drifting gauge, a stuck reading or a wrong calibration factor) and 50 without faults; the checker names the faulty instrument in at least 45 of the 50 and flags fewer than 5 of the clean records; code and data published.
Open 02 · the brief
Plans checked before digging starts
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A checker that reads the instrumentation and monitoring plan for a deep excavation and lists every instrument that lacks a review level, an alarm level, a reading frequency, a named person who must act, or a planned action when a level is crossed, with each flag pointing to the line of the plan it concerns.
Why it is still open
Ralph Peck wrote in 1969 that the observational method should not be used unless the designer has planned an action for every unfavourable situation the readings might reveal. Singapore's inquiry into the 2004 Nicoll Highway collapse gave a section of its interim report to "Clarity in Respect of Trigger, Design and Allowable Levels" and called for "pre-determined review levels" and "the timely implementation of pre-planned risk based action" when they are breached. A web search found no tool that checks a monitoring plan for missing or inconsistent levels.
Built with what exists
Monitoring plans and reports that owners and agencies have published, such as the Millennium Tower monitoring reports posted by the City of San Francisco; a language model that reads tables in PDFs; a fixed list of required fields drawn from the inquiry's findings; Python.
First step
Write the list of required fields from the inquiry's interim report and test it by hand on one published monitoring report.
Done looks like
On 30 published plans, plus 10 copies in which a person has removed or changed levels at random, the checker finds at least 90% of the planted gaps, and a person confirms or rejects every flag it raises on the untouched plans; the checker and its rule list are published.