Summary: A sea dyke is the one civil structure whose entire function is an absolute elevation. It does not hold anything up, it does not carry a load, and the stress in its body is a small fraction of what the material can take. What it does is keep the sea below a line, and the line is a number in metres above chart datum. That makes its failure modes unlike those of a wall: the ones with a visible shape, such as a slipped slope or a displaced armour unit, are not the ones that decide whether the hinterland floods. The one that decides is the crest losing height, and a crest loses height without cracking, without tilting as a whole, and without producing anything an inspector can see from the towpath. This article takes the design budget in GB/T 51015 apart, puts it against the exponential overtopping relation in EurOtop, and shows that the settlement allowance is written in millimetres while its consequence is spent in multiples.
1. Two Failure Modes, and Only One of Them Has a Shape
The first thing to establish about a sea dyke is how little structural work it does. A retaining wall resists a lateral thrust and its reinforcement or its mass is sized by that thrust. A sea dyke is a trapezoidal embankment of fill with a protective facing, and the compressive stress inside it is a small fraction of the strength of the soil it is built from. It is not designed to be strong. It is designed to be high, and to stay high while water arrives at it from the side.
That distinction is not a rhetorical one. It decides which of the things that can go wrong will actually be noticed. The failure modes with a geometry are the ones the eye is built for: a slip on the seaward face, an armour layer with units displaced out of position, a scour hole at the toe, a sinkhole in the crest road where the fill has been washed out from underneath. All of those leave a shape, and a walkover inspection is a competent instrument for finding a shape.
The failure mode without a shape is the crest being too low. Nothing about the dyke looks wrong when that happens. The slope is at the angle it was built at, the facing is intact, the joints are closed, the crest road is smooth. The only thing that has changed is a number, and the number is not written anywhere on the structure. It has to be measured against a datum that is somewhere else.
Those two families are named explicitly in the Chinese research literature on coastal defence, which describes the problem as a pair of failure modes, breach and overtopping, and treats them as requiring separate diagnostic methods. The reason they need separating is that they are measured in different quantities. A breach is measured as a volume of material moved. Overtopping is measured as a discharge per metre of crest per second, and the discharge is a function of a height difference that no photograph records.
The scale of the asset makes the choice of method consequential. China has built about fourteen thousand five hundred kilometres of sea dyke along a mainland coastline of nearly nineteen thousand kilometres. Reporting by the Nanjing Hydraulic Research Institute puts more than eighty percent of that length beyond twenty years of service and the fraction meeting current standards at forty two and a half percent, which is about six thousand two hundred kilometres. The same source states the operational conclusion directly: the practice of relying on manual patrol and repairing what keeps breaking is no longer adequate for an asset of that age and that length.
So the question this article addresses is narrow and practical. If the failure that matters is a height, what is the height supposed to be, how much of it is already committed to something else at the design stage, and what can an angle sensor contribute to knowing whether it is still there.
2. The Design Adds Three Heights and Then Books a Percentage
The Chinese sea dyke design code, GB/T 51015, computes the crest elevation as a sum of three terms. The design high water level, the wave runup for a stated exceedance frequency, and a safety height increment. All three are lengths in metres, all three are absolute, and the sum is the elevation the contractor is told to build to.
The safety height increment is tabulated by dyke grade and by whether the section is designed to prevent overtopping entirely or to tolerate a part of it. The table is worth reading because it establishes the size of the number that the rest of this article is about.
| Dyke grade | Increment, no overtopping permitted (m) | Increment, partial overtopping permitted (m) |
| Grade 1 | 1.0 | 0.5 |
| Grade 2 | 0.8 | 0.4 |
| Grade 3 | 0.7 | 0.4 |
| Grade 4 | 0.6 | 0.3 |
| Grade 5 | 0.5 | 0.3 |
Now read that beside the settlement clause. The code requires the design to include a post construction settlement allowance, and for a non soft foundation it gives that allowance as a percentage of the dyke height: three to five percent, with the smaller value where an existing dyke is being raised. For an earth dyke taller than ten metres, or for a soft foundation, the percentage is withdrawn and the allowance has to be computed under the settlement section of the code instead.
The two clauses use different currencies. The safety increment is a fixed length chosen from a table. The settlement allowance is a proportion of the structure, so it grows with the dyke. Put them side by side and the second one overtakes the first.
| Dyke height (m) | Settlement allowance at 3 percent (mm) | At 5 percent (mm) | Ratio to a 300 mm safety increment |
| 4 | 120 | 200 | 0.40 to 0.67 |
| 6 | 180 | 300 | 0.60 to 1.00 |
| 8 | 240 | 400 | 0.80 to 1.33 |
| 10 | 300 | 500 | 1.00 to 1.67 |
| 12 | 360 | 600 | 1.20 to 2.00 |
At four metres the allowance is a fraction of the safety increment, and the ordering of the budget looks sane. At twelve metres the allowance is up to twice the safety increment. That is the inversion worth carrying away. On a high dyke the amount of height reserved for ground that will compress is larger than the amount reserved for the uncertainty in the water level, and the reserved settlement is the one that is certain to be spent.
The code is aware of the difference and handles it in two places. First, the soft foundation case is pulled out of the percentage rule and sent to a computed allowance, which is the right treatment. Second, the settlement calculation carries a correction factor for soft foundations of one point three to one point six, applied to a one dimensional consolidation result. The correction is there because a soft foundation does not compress only downwards, and the extra settlement it produces has to be booked somewhere.
That factor is easy to pass over and it is not small. A one dimensional calculation returning three hundred millimetres becomes three hundred and ninety to four hundred and eighty millimetres after correction. A calculation returning four hundred becomes five hundred and twenty to six hundred and forty. The allowance is not a number that comes out of the arithmetic directly. It is the arithmetic multiplied by a factor that encodes the difference between a soil that only compresses and a soil that also squeezes sideways.
There is also a clause that names the quantity this article is really about. The settlement section requires the calculation to include not only the final and post construction settlement at the crest centreline, but also, at boundaries where geology, loading or ground treatment change, the settlement at the interface and the differential across it. Differential settlement is therefore not an accident the code forgot to mention. It is a required output, requested at exactly the places where it is largest.
One more number belongs here because it sets the floor. Where a parapet wall is present and is structurally sound, the crest elevation may be taken at the top of that wall, but the elevation ignoring the wall must still stand at least half the one percent wave height above the design high water, and never less than half a metre. That half metre is the irreducible reserve. Everything else in the budget can be argued about. That one is a floor.
3. The Response Is Not Linear in That Height
If the consequence of losing height were proportional to the height lost, the previous section would be the whole story and the arithmetic would be a subtraction. It is not proportional. The relation between the freeboard, which is the crest elevation minus the water level, and the mean overtopping discharge is exponential in a power of the freeboard, and the practical effect of that shape is that the numbers in the elevation budget do not translate into the numbers that decide the flooding.
The reference relation is the one in the EurOtop overtopping manual, which for a sloping structure with a berm, roughness, wave obliquity and a wall at the top all represented by reduction factors takes the form of a normalised discharge that falls as an exponential of the freeboard raised to the power one point three. For a smooth 1 in 3 slope, a significant wave height of two metres and a spectral period of six seconds, the breaker parameter works out at about one point seven seven when the deep water wavelength is used, and the whole relation collapses to a single curve that can be tabulated.
| Freeboard (m) | Mean discharge (litres per second per metre) | Design value, mean plus one standard deviation |
| 0.00 | 624 | 705 |
| 1.00 | 308 | 373 |
| 2.00 | 110 | 147 |
| 3.00 | 33.0 | 49.3 |
| 3.90 | 10.0 | 16.8 |
| 4.39 | 5.00 | 8.95 |
| 5.00 | 2.07 | 4.02 |
| 5.48 | 1.00 | 2.09 |
| 6.00 | 0.45 | 1.01 |
| 7.00 | 0.090 | 0.240 |
The discharge values themselves are not the numbers the guidance quotes. The manual and the national guidance that follows it specify tolerable discharges rather than tolerable heights, and the values in circulation form a ladder: a tenth of a litre per second per metre at which no damage to the surface is acceptable, one at which pedestrians remain safe on foot, ten at which both pedestrians and vehicles are at risk, fifty at which structural elements begin to suffer damage, and two hundred as the upper end at which the defence is no longer performing its function. The EurOtop manual recommends five for a maintained grass covered dyke in the wave height range of one to three metres, and at least one German coastal state designs to two.
Inverting the relation gives the freeboard each rung of that ladder requires, and the result is the argument of this article. On the same 1 in 3 slope with the same two metre wave, the rungs sit at the following heights.
| Tolerable discharge (litres per second per metre) | Freeboard required (m) | Height between this rung and the next (mm) |
| 0.1 | 6.94 | 1005 |
| 0.5 | 5.93 | 449 |
| 1 | 5.48 | 1089 |
| 5 | 4.39 | 494 |
| 10 | 3.90 | 1231 |
| 50 | 2.67 | 1223 |
| 200 | 1.45 | not applicable |
The third column is the point. The rungs are spaced at roughly half a metre to one and a quarter metres of height, and they are spaced that way whether the step between them is a factor of two or a factor of five. A designer reading a settlement report in millimetres sees a quantity that moves slowly and is reported to a tenth of a millimetre. The same designer reading that report against this table sees a ladder whose rungs are equally spaced in height and wildly unequally spaced in consequence.
At the bottom of the ladder the shape becomes a cliff rather than a slope. The relation is written for a positive freeboard, and it does not go to zero as the freeboard goes to zero, because a wave striking a crest at the water line still throws water over it. On this geometry the discharge at zero freeboard is about six hundred and twenty litres per second per metre. That is sixty two times the ten litre rung, and the height between them is three point nine metres. From the one litre rung the same cliff is a factor of six hundred and twenty over five and a half metres.
Below zero freeboard the formulation changes as well as the value. Once the water level is above the crest the dominant term is not wave overtopping but overflow, and the accepted treatment is a broad crested weir expression proportional to the three halves power of the depth of water over the crest, added to the wave term evaluated at zero freeboard. That is a different exponent applied to a different variable, and it is the reason a dyke that has lost its freeboard is not simply a dyke with a larger discharge. It is a structure behaving according to a different equation.
None of this is an argument that a dyke is fragile. It is an argument that the quantity being tracked is not the quantity being reported. Settlement is reported in millimetres because that is how it is measured. Its consequence is expressed in multiples because that is how the discharge relation is built. A record that shows the crest dropping two hundred millimetres a decade looks like a slow linear process, and the discharge it produces is not one.
4. What the Allowance Actually Buys
With the ladder in place the design allowance can be priced. Take a section designed to the five litre rung, which requires four point three nine metres of freeboard on this geometry, and then spend the settlement allowance against it a hundred millimetres at a time.
| Crest drop (mm) | Freeboard remaining (m) | Discharge (litres per second per metre) | Multiple of the design value |
| 100 | 4.29 | 5.8 | 1.2 |
| 200 | 4.19 | 6.6 | 1.3 |
| 300 | 4.09 | 7.6 | 1.5 |
| 400 | 3.99 | 8.8 | 1.8 |
| 500 | 3.89 | 10.1 | 2.0 |
| 800 | 3.59 | 15.2 | 3.0 |
The fifth row is the one to remember. Five hundred millimetres of crest loss doubles the discharge, and five hundred millimetres is the upper end of the settlement allowance for a dyke of about ten metres under the percentage rule. The four hundred row is the upper end of the allowance for a dyke of eight metres. In other words, the allowance and the step between two rungs of the tolerable ladder are the same size. A dyke that spends its settlement allowance does not move a little way down the ladder. It changes rung.
That coincidence is worth stating precisely, because it is what turns a routine maintenance question into an instrumentation question. The distance from the five litre rung to the ten litre rung on this geometry is four hundred and ninety four millimetres. The allowance for an eight metre dyke is two hundred and forty to four hundred millimetres, and for a ten metre dyke three hundred to five hundred. The allowance is between half and all of one rung. There is no version of this budget in which the settlement is absorbed without the discharge changing class.
Two things make it worse, and both are documented rather than hypothetical. The first is the soft foundation correction already described: the number that goes into the budget is the consolidation result multiplied by one point three to one point six, so a foundation whose one dimensional calculation was comfortable can produce an allowance that is not. The second is that the allowance is sometimes not carried at all. A World Bank assessment of the Vietnamese sea dyke system, which computed the required crest elevation for every dyke stretch in the country against a target discharge, included a safety increment of three to five hundred millimetres and a sea level rise provision of two to seven hundred millimetres, and stated in the method that no allowance for settlement had been included. That is a national scale assessment, not a shortcut in a student exercise.
That omission matters more than it looks, because the other two provisions are themselves being squeezed. The same assessment used a relative sea level rise rate of five to six millimetres a year under a medium scenario, which over a fifty year life is a quarter of a metre before any storm gets involved. The Vancouver shoreline design reference handles the same term explicitly and gives the arithmetic in one line: the subsidence allowance is the local rate multiplied by the number of years to the design horizon, so one millimetre a year over eighty years is eighty millimetres, and that is added to the sea level rise allowance rather than compared with it.
Assemble the budget as the documents actually write it and the composition is uncomfortable. Sea level rise contributes a quarter of a metre or more. Settlement contributes two to five hundred millimetres by the percentage rule and more where the foundation is soft. The safety increment, which is the part intended to remain at the end, is three hundred millimetres at grade four and five. The reserve and the consumption are the same size, and only one of them is optional.
5. On Soft Ground the Dyke Does Not Fail, It Sinks
There is a result from centrifuge testing that reorganises how the whole problem should be instrumented, and it is counterintuitive enough to be worth stating before the deployment is discussed.
A series of centrifuge tests on dykes with a strong stiffness contrast between the embankment and its foundation compared two arrangements: a stiff dyke on a soft subsoil, and a soft dyke on a stiff subsoil. The stiff foundation case produced slope failures inside the dyke, attributed to pore pressure building up as the embankment contracted volumetrically. The soft foundation case did the opposite. The foundation settled heavily, the stiff dyke was deformed by that settlement, and no slope failure occurred at all. Diffuse shear zones were found in the core, but the geometry that an inspector looks for, a slipped face, did not form.
Read that against the two failure families in the first section and the consequence is immediate. On the foundation type where the settlement problem is worst, the visible failure mode is the one that does not happen. The dyke keeps its shape, keeps its facing, keeps its joints closed, and loses height. An inspection regime built around finding shapes is therefore structurally incapable of finding the failure that soft ground produces, and it is not a question of diligence or frequency.
This is also why the operational practice that has grown up around sea dykes is expressed in elevation rather than in condition. One operational monitoring programme described in the published literature is built around the crest elevation and the settlement reading, with a threshold that triggers a response when the measured settlement exceeds it, a quarterly manual elevation check alongside the automated record, and a stated reason for the whole exercise: the primary damage form on a sea dyke is the void that develops under the crest road and the crest surface, which is a loss of support that shows up as a change in elevation before it shows up as anything else.
A caution belongs with that, because it would be easy to overstate. A settlement record is not a stability analysis. A dyke can settle steadily for years under consolidation and be in no danger of a slip, and a dyke can develop a slip while its crest settlement is unremarkable. The two quantities are related through the foundation response and not through a fixed ratio, which is the same reason the code asks for the settlement to be computed rather than looked up.
6. What a Rotation Can and Cannot Tell You
An angle sensor on a dyke does not measure settlement, and the honest version of this section has to begin there. It measures the inclination of whatever it is bolted to. On a dyke the mounting is either a concrete element, such as a parapet wall or a crest slab, or the fill itself, in which case the reference has to be established separately and the reading means something different.
The case it cannot see is the one the previous section described. If a dyke settles uniformly, every point on the cross section drops by the same amount, every surface stays parallel to where it was, and every tilt reading stays flat. That is not a defect of the instrument. A uniform settlement genuinely produces no rotation anywhere, and no angle sensor of any quality will report one. The failure that matters most on soft ground is therefore invisible to a single tiltmeter, and any deployment that consists of one instrument per site is measuring the wrong degree of freedom.
What rotation does respond to is the differential, and on a dyke cross section the differential has a characteristic shape. The load is largest at the crest and tapers towards the toes, so the foundation beneath the crest consolidates most and the foundation beneath the toes least. The cross section does not drop as a block. It flattens: the crest goes down further than the toes, the seaward slope becomes gentler than it was built, and the landward slope does the same. That flattening is a rotation at every point on the slope, and it is the geometric signature of exactly the mechanism the centrifuge tests found, a soft foundation squeezing laterally under a stiff embankment.
This is why the deployment pattern that appears in a published monitoring method for sea dykes has three lines rather than one. That method places a longitudinal section on the landward slope with a measuring line on the crest, a measuring line on the mid height platform and a measuring line at the toe, each carrying sensor nodes along the axis at intervals of about twenty metres, and each node carrying a vertical displacement gauge together with a tilt sensor. The three lines are not redundancy. They are the minimum needed to separate the unknowns.
The counting is worth doing explicitly because it explains a design decision that otherwise looks like over specification. One line gives a settlement profile, which cannot distinguish a uniform drop from a rotation. Two lines give the average slope change between them, which cannot distinguish a rigid rotation from a curvature. Three lines, at the crest, the platform and the toe, give two independent slope changes and therefore separate a rigid body rotation from a bending of the cross section, which is the difference between a foundation consolidating and a slope beginning to bulge. The pattern follows the number of quantities to be separated, not the length of the dyke.
The resolution requirement follows from the same arithmetic, and it is modest. An accuracy of five thousandths of a degree across a ten metre span corresponds to eight tenths of a millimetre of differential movement. Across a twenty metre span it is one and three quarter millimetres. Since the quantities being tracked on a soft foundation are hundreds of millimetres, the instrument is not the limiting factor by a wide margin, and specifying a tighter angle buys nothing that the structure can use.
| Accuracy (degrees) | Over 1 m (mm) | Over 5 m (mm) | Over 10 m (mm) | Over 20 m (mm) |
| 0.001 | 0.017 | 0.087 | 0.175 | 0.349 |
| 0.005 | 0.087 | 0.436 | 0.873 | 1.745 |
| 0.01 | 0.175 | 0.873 | 1.745 | 3.491 |
There is a second reading available that costs nothing and is often skipped. The phase of the daily cycle separates a thermal rotation from a structural one, because a concrete element on a dyke rotates with the sun and a foundation does not. A record that is flat at night and swings by day is reading weather. A record whose daily component is small and whose trend is monotonic is reading ground. The two can coexist in the same trace, which is why the interval has to be short enough to resolve a day rather than a month.
One further use of the same channel is less obvious and matters on this structure. A dyke is built in sections and the joints between them are deliberate, which means adjacent sections can move independently and the discontinuity is where the consequence lands. That is the subject of the next section.
7. The Parapet Is Where the Joint Decides
The crest of a sea dyke usually carries a parapet wall on the seaward side, and the code treats it as both a structural element and a permitted addition to the crest elevation, with conditions. The wall is placed on the seaward edge, its exposed height above the crest is limited to about one point two metres, its embedment below the crest is required to exceed half a metre, and it is to be broken by settlement joints at intervals of eight to twelve metres. A curved or recurved face is permitted where wave attack is severe.
Those two numbers, the height limit and the joint spacing, interact in a way that decides what the wall does when the foundation moves. A joint at eight to twelve metres means the wall is not a continuous beam resisting differential settlement by bending. It is a series of rigid segments, each about ten metres long and one point two metres tall, and each free to rotate about the axis of its own base. The differential settlement does not become stress in the concrete. It becomes a rotation of each segment plus a relative movement at each joint.
The arithmetic of that is small in angle and large in millimetres, which is the reverse of the usual situation. A ten metre segment with a twenty millimetre differential across its own length rotates by about one tenth of a degree. Because the joint sits at the end of that length, the same twenty millimetres is the step that appears at the joint if the neighbouring segment has not moved. The rotation and the offset are not two quantities to be reported separately. They are one quantity in two units, and the conversion factor is the segment length, which is why an angular specification can be converted into a millimetre tolerance at all, and why the joint spacing in the code fixes that conversion.
| Segment length (m) | Rotation for 5 mm differential (degrees) | For 10 mm (degrees) | For 20 mm (degrees) | For 30 mm (degrees) |
| 8 | 0.036 | 0.072 | 0.143 | 0.215 |
| 10 | 0.029 | 0.057 | 0.115 | 0.172 |
| 12 | 0.024 | 0.048 | 0.096 | 0.143 |
A joint offset of nine to seventeen millimetres on a parapet wall, which on a ten metre segment corresponds to a rotation of five hundredths to a tenth of a degree, is a maintenance item that gets noticed, because it is a step in a horizontal line and the eye is good at those, and because it is the start of a path for water into the fill behind the wall. That is a useful property and it should not be oversold: the inspection will find the offset eventually. What the tilt channel adds is the date, and on this structure the date is the whole value, because the repair decision is a comparison between the cost of repointing and the cost of rebuilding a section.
The same numbers are worth reading against the instrument, because they show the channel is overspecified here. A five thousandth of a degree accuracy across the one point two metre height of the wall corresponds to about a tenth of a millimetre at the top. The quantity that matters, the joint offset, does not appear at that scale until it is many times the instrument resolution. On a parapet wall the sensor is not limited by its specification at all, which means the sensible question to ask is not which accuracy to buy but whether the wall or the fill is the right place to put it. The wall gives a clean rigid body and a geometric consequence in the joint. The fill gives the mechanism that drives both.
There is a third reason to instrument near the joints rather than at mid segment, and it is not geometric. The settlement joints are where the fill behind the wall is most exposed, because a joint that has opened is a drain, and a drain in a dyke crest carries fines out of the embankment. That loss of fines produces its own settlement once it starts, so a joint that is being watched for offset is also, for free, being watched for the beginning of that chain. It is adjacent to rather than the same as the principal damage form described earlier, since the void under the crest surfacing is what the escaping fines leave behind.
8. What the Instrument Cannot Do
The boundary on this structure is wide and it is better set out than discovered. Four things need saying, and then a fifth that is the one people argue about.
First, an angle channel does not measure freeboard, and no arrangement of tiltmeters on a dyke can close that gap on its own. Freeboard is the difference between an absolute crest elevation and a water level, and both terms are outside what a tiltmeter reports. The instrument supplies the change in shape of a cross section. The elevation itself still needs a survey, a static level, or a GNSS reference, and on a dyke those remain the primary record rather than a legacy one.
Second, it cannot see the overtopping. The quantity that decides the flooding is a discharge averaged over a storm, produced by waves with periods of seconds. A tiltmeter with an hourly or daily interval records none of that, and a tiltmeter with a fast interval still records an angle, not a flow. Overtopping is measured by weighing or by instruments placed to catch the water, and the role of the geotechnical channel is to say whether the crest elevation that those measurements are indexed against has changed.
Third, it does not replace the underwater and subsurface inspection that this structure depends on. The dominant damage form on a sea dyke is loss of support under the crest surfacing, and the methods that find it are ground penetrating radar, coring and visual survey of the facing, not an angle. A tilt record can indicate that something has begun to move. It cannot tell whether the void is under the crest, behind the parapet, or at the toe.
Fourth, the reading is contaminated by the structure being wet, salted and working. A dyke crest carries traffic, the seaward face takes spray and abrasion, and the fill behind a parapet is at a moisture content that changes with the tide. The mounting has to survive that environment, which is a sealing and materials question rather than an angle question, and it is the reason a deployment on a dyke should be judged on its enclosure as much as on its resolution.
One more limit is worth stating because it is the one that gets argued. A tilt channel is not a stability analysis. A factor of safety against a slip circle is a ratio of resisting to driving moments computed from a soil profile, and no angle reading substitutes for it. What a continuous angle record does is tell you whether the assumption that the profile is not changing is still true. When a dyke that has been settling quietly begins to rotate at a rate the consolidation explanation does not cover, that is a signal to re run the analysis with a different question, and the value of the record is that it arrives before the slip rather than after it.
9. The Shape of a Deployment This Structure Implies
The coverage problem on a sea dyke is harder than on any other structure this series has covered, and the arithmetic makes that plain. At the twenty metre node spacing used in published methods, with three measuring lines per section, a kilometre of dyke carries a hundred and fifty nodes. Across fourteen thousand five hundred kilometres that is above two million instruments, which is not a number any programme will fund and not a number that would produce a readable dataset if it did.
So the deployment question is a selection question, and the documents point at where to select. The eight thousand three hundred kilometres currently below standard are the first cut, because a dyke that does not meet the design crest elevation is a dyke whose remaining freeboard is already the quantity in question. The eleven thousand six hundred kilometres beyond twenty years of service are the second, because post construction settlement is a curve that has largely played out by then and any movement still occurring is not consolidation. Where the two sets overlap, and where the foundation is soft, is where a continuous record changes a decision.
Three further criteria come from the physics rather than from the asset register. The first is the ground treatment boundary. The code requires the settlement calculation to report the differential at interfaces where geology, loading or treatment change, which is an instruction to look at exactly those places. In practice that means the transition between a dredged fill and an in situ soft clay, the boundary between a treated and an untreated length, and the ends of any section founded on a different stratum. Finding out which of those lengths is short of height is a survey job before it is a monitoring job, and it is the step that decides where the instruments go.
The second is the structure crossing. Where a culvert, outfall or pumping station passes through the dyke, the code requires the foundation settlement of the structure to be compatible with the dyke settlement at the same location. That is a differential settlement requirement written as a design obligation, and a crossing that has been built to satisfy it is a place where the margin is already thin by construction.
The third is where the settlement is being generated rather than where it shows. The three line pattern is the expression of this, and the reason the toe line is not optional: on a soft foundation the lateral squeeze at the toe is the mechanism, and a crest instrument alone reports its consequence with a delay and without a cause.
Two requirements follow from the earlier sections and are worth writing into a specification rather than assuming. The first is a shared time base. A differential between two nodes is only a differential if the two readings describe the same moment, because the common mode that has to be removed is thermal and it moves with the sun. A hundred and fifty nodes reporting on independent clocks produce a dataset in which the daily cycle is aliased into the trend.
The second is local evaluation rather than scheduled upload. The signal on this structure is a change of rate, and a change of rate has no schedule. A device that samples on a fixed interval and uploads on the same interval can sit on a step for hours, and over a multi year deployment those hours accumulate into a record that is missing exactly the transitions it exists to catch. A device that evaluates a threshold or a rate against its own recent history and transmits when it is crossed does not have that failure mode, and on a structure where the event is a slope beginning to move rather than a load arriving, that difference is the reason to install anything at all.
A note on what is not claimed here. No sea dyke deployment is presented as a case in this article, because the published record of this company is not on sea dykes. Its closest work is a sea reclamation project at Yangshan deep water port, where in place inclinometers monitored the lateral displacement of soft seabed sediment during dynamic compaction, in three sections twenty five metres below the seabed with eight probes each, at the shore and at three hundred and fifty and seven hundred metres from land, read daily before compaction and more often during it. That is a fill and ground improvement project rather than a dyke, and the mechanism it was watching, soft ground squeezing sideways under an imposed load, is the same one the centrifuge tests found on soft founded dykes. It is offered as evidence that the mechanism is understood and instrumented, not as a reference installation on a crest.
Frequently Asked Questions
- Can a tilt sensor tell me whether a sea dyke is safe? No, and the reason is dimensional rather than a matter of accuracy. Safety on a dyke is a comparison between a water level and a crest elevation, and a tiltmeter measures the change in shape of the cross section that carries that elevation. It can tell you that the section is flattening, which is the geometric signature of a soft foundation squeezing, and it cannot tell you where the water will be next storm. The two together are the assessment.
- Why not just survey the crest elevation? Because a survey is an interval and the failure is a rate. A quarterly elevation check, which is the interval used in operational programmes, will find a settlement that has already happened and will miss the change in rate that precedes a slip. The right arrangement is a survey for the absolute elevation and a continuous angle record for the rate, with the survey providing the datum the angle record is interpreted against. Neither replaces the other.
- Does a uniform settlement show up at all? It does not, and that is a correct reading rather than a blind spot. If every point on the cross section drops by the same amount, no surface changes its inclination and no tiltmeter reports anything. The corollary is the design rule: a single instrument per site measures nothing useful on this structure, because the quantity that matters is the differential between the crest and the toe, and a differential requires at least two points and preferably three.
- How much crest loss matters? It depends which rung of the tolerable ladder the section was designed to, and establishing that is the first step. On the geometry used here the step from the five litre rung to the ten litre rung is four hundred and ninety four millimetres of freeboard, and five hundred millimetres of crest loss doubles the discharge. Since the design allowance for an eight to ten metre dyke is two hundred to five hundred millimetres, the working answer is that a crest which has spent its allowance has probably changed class. The only way to know when it did is a continuous record indexed to a surveyed elevation.
- Does the parapet wall make the problem better or worse? It makes the consequence visible and the mechanism no different. The wall is broken into eight to twelve metre segments by settlement joints, so differential settlement becomes a rotation of each segment and a relative offset at each joint rather than bending stress in the concrete. On a ten metre segment a five hundredths of a degree rotation puts a step of nearly nine millimetres at the joint, which is something a walkover will find and a leak path that will begin carrying fines out of the fill. The wall tells you when the joint began to open. The fill tells you why.
TAG: