A rising lake gauge does not always mean the lake gained water
A lake seiche is a standing oscillation in an enclosed or partly enclosed waterbody. Water levels can rise near one shore while falling elsewhere, then reverse. A local gauge therefore records both changes in the lake’s overall water state and spatial redistribution within it.
Wind setup, changing atmospheric pressure, basin geometry, sensor corrections and sampling intervals can all affect the interpretation. The practical question is whether a water-level feature represents added volume, a moving surface, a pressure artefact or an unresolved combination.
This guide uses original simplified calculations with NOAA and USGS explanations of seiches and stage measurement. It is not a forecast of hazardous waves or a declaration that a shoreline is safe. Actual warnings and access decisions belong to the responsible authorities.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Recognise the motion
Separate storage from redistribution
Distinguish wind setup from the oscillation after forcing
Use the ideal period formula as a scale check
Measure the level correctly
Recognise nodes, antinodes and the importance of gauge position
Convert pressure to water head with the right pressure difference
Keep datum, sensor depth and physical pressure response separate
CHAPTER 1 OF 12
1. Separate storage from redistribution
A water-level gauge measures the surface elevation at its location and time under its measurement method. To infer a whole-lake volume change, the relationship between that local level and the wider water surface must be appropriate. An oscillating or tilted surface complicates a static level–volume conversion.
Imagine an idealised rectangular basin whose surface tilts linearly. One end is 0.20 metres above the equilibrium level and the other is 0.20 below. Under this symmetric small-displacement teaching geometry, the average surface displacement is zero even though either end gauge shows a substantial local change.
If the high-end reading were multiplied by the full lake area as a uniform rise, the resulting apparent storage increase would be false. The water was redistributed rather than added in this example. A real lake has more complex shape, so its spatial integration needs an appropriate geometric basis.
Conversely, a genuine inflow can raise the mean level while a seiche adds local oscillation around it. Do not force the record into a choice that only one process can occur. A useful interpretation separates the slower or spatially coherent component from the oscillatory component, with evidence supporting that separation.
Transfer check: add a uniform 0.05-metre rise to the tilted example. End readings become +0.25 and −0.15 relative to the original equilibrium. Their simple symmetric mean is +0.05, while their difference still reflects redistribution. Explain what information one gauge alone would lose.
CHAPTER 2 OF 12
2. Distinguish wind setup from the oscillation after forcing
Sustained wind can displace surface water and establish a level difference across a basin. This wind setup is a forced response. When wind changes or relaxes, the displaced water can oscillate, producing a seiche. The forcing history and basin response should be examined together.
The familiar bathtub analogy is useful if its limits are stated. A displaced free surface can slosh after the push ends. A lake additionally has irregular bathymetry, friction, shoreline geometry, variable winds, inflows and sometimes density stratification. The analogy explains restoring motion, not an exact forecast.
NOAA distinguishes wind-driven short-term level changes and seiches from true astronomical tides in its lake-level explanations. An oscillation that resembles a tidal period is not automatically a tide. Mechanism needs atmospheric and spatial evidence as well as a repeating graph.
For a fictional event, the east-shore gauge rises during a sustained westerly wind while the west-shore gauge falls. After the wind weakens, both reverse repeatedly with decreasing amplitude. That sequence is consistent with setup followed by a standing oscillation, but other observations should be checked before assigning a definitive mechanism.
Transfer check: let wind reverse before the basin settles. The subsequent record need not be a single clean decaying sine wave. Repeated forcing can add to or oppose existing motion. Fitting one curve without the wind history may conceal the process actually driving the water.
CHAPTER 3 OF 12
3. Use the ideal period formula as a scale check
For a simplified enclosed rectangular basin with uniform shallow-water depth, a common fundamental surface-seiche approximation is T = 2L divided by square root of gd. L is basin length, d is representative depth and g is gravitational acceleration. The assumptions matter as much as the expression.
Take a fictional length of 5,000 metres and depth of 10 metres. With g = 9.81 metres per second squared, square root of gd is about 9.90 metres per second. The period estimate is 10,000/9.90, about 1,010 seconds or 16.8 minutes. This is an ideal scale estimate, not a predicted period for a named lake.
Double the length while holding depth fixed and the ideal period doubles. Increase depth from 10 to 40 metres with length fixed and wave speed doubles, so period halves. These relationships provide a useful reasonableness check before accepting a fitted period.
Real basins can have several modes, varying depths, connected bays and other effects not represented by this one-dimensional idealisation. A disagreement with 16.8 minutes does not by itself prove a sensor fault. It may show that the model geometry is too simple or that a different mode is being observed.
Transfer check: attach a narrow bay to the rectangular basin. The whole-lake length and depth may no longer describe a local harbour oscillation. Identify the basin or sub-basin whose restoring motion is being modelled before choosing L and d.
CHAPTER 4 OF 12
4. Recognise nodes, antinodes and the importance of gauge position
A standing-wave pattern has places where level oscillation is comparatively large and places where it is small for a given mode. A gauge near a surface-level node can record little motion even when other parts of the lake experience larger changes. One quiet record is therefore not proof that the lake is not oscillating.
For an ideal first longitudinal mode, the two ends can oscillate in opposite phase while an interior location has smaller surface displacement. This is a simplified spatial pattern, not a universal layout for irregular lakes. Different modes can have different node positions.
Imagine a fictional three-gauge network: west shore, centre and east shore. West and east alternate around the same slowly changing baseline while the centre varies little. That pattern offers evidence of redistribution that a single gauge would not reveal. Exact equal amplitudes should not be demanded in a real basin.
Map gauge locations with bathymetry, bays, inlets and exposure. A sensor in a sheltered harbour can observe a local response rather than the open lake’s full pattern. A nearby pair of gauges may share a local effect and provide less independent spatial information than their count suggests.
Transfer check: move the centre gauge into a bay. Its amplitude may increase because it now samples a different mode or local geometry. Keep location history with the record so a new oscillation is not misread automatically as a climate or lake-volume change.
CHAPTER 5 OF 12
5. Convert pressure to water head with the right pressure difference
A submerged pressure sensor can infer water depth from hydrostatic pressure under appropriate conditions. For an absolute-pressure instrument, the measured pressure includes atmospheric pressure above the water as well as the water column. Atmospheric correction is therefore part of the depth measurement.
Under an explicitly simplified constant-density account, water head h equals the submerged pressure minus atmospheric pressure, divided by rho g. Using water density 1,000 kilograms per cubic metre and g = 9.81, a 100-pascal pressure difference corresponds to about 0.0102 metres, or 1.02 centimetres.
An uncorrected atmospheric change of 1,000 pascals could therefore imitate about 10.2 centimetres of water head in this simple conversion. That number is not a prediction of the lake’s physical atmospheric response; it demonstrates the instrument correction error when an absolute sensor is interpreted as water-only pressure.
USGS lake-level reporting describes paired underwater and atmospheric pressure observations in a real monitoring setting. A vented instrument uses a different pressure-reference arrangement, but venting does not remove every possible installation or maintenance problem. Read the actual sensor design and correction metadata.
Transfer check: replace an absolute sensor with a vented sensor midway through the record. Preserve the change date, calibration and reference handling. Do not keep subtracting a barometric series from a signal that has already been referenced appropriately, because double correction can introduce a false variation.
CHAPTER 6 OF 12
6. Keep datum, sensor depth and physical pressure response separate
Water depth above a sensor is not automatically lake elevation relative to a common datum. The sensor’s surveyed elevation connects those quantities. If the sensor moves vertically, apparent water depth can change even when the water surface does not.
For a fictional example, a sensor sits at elevation 100.00 metres and measures a water column of 2.00 metres, giving a surface elevation of 102.00. If it settles downward by 0.05 metres and its old elevation is retained, the inferred surface can be biased upward by 0.05. The pressure instrument may function correctly while the reference information is wrong.
Density also affects pressure-to-head conversion. Temperature and dissolved material can matter at the precision required by a study. Use a justified density treatment rather than assuming that every small residual is a seiche. Dynamic effects near waves or exposed installations can add further measurement considerations.
Distinguish correction of the instrument’s atmospheric pressure term from a real waterbody response to atmospheric forcing. Subtracting air pressure from an absolute sensor removes the direct reference contribution to its reading; it does not prove the lake itself has no pressure-driven motion.
Transfer check: a storm coincides with a pressure change and a mount disturbance. Develop separate evidence checks for all three. A single subtraction cannot repair a physical sensor displacement, and a surveyed offset cannot explain a genuine oscillation shared by other stations.
CHAPTER 7 OF 12
7. Choose a sampling interval that can reveal the oscillation
A periodic signal can be misrepresented when sampled too slowly or at an unfortunate phase. This is aliasing: the recorded sequence can suggest a slower pattern or no variation even when the underlying signal oscillates. Sampling design must match the timescale of interest.
Use an ideal fictional sine wave with a twenty-minute period. Sampling every twenty minutes at the same phase gives the same value each time. The record appears flat although the water moves between samples. Sampling every fifteen minutes produces a phase sequence that can resemble a sixty-minute oscillation.
The second result follows from the difference between the true frequency of 1/20 per minute and the sampling frequency of 1/15. Their difference is 1/60 per minute. This is a deliberately clean demonstration; real records have noise and multiple frequencies, making diagnosis less obvious.
The theoretical ability to distinguish a frequency is not the same as accurately measuring its amplitude and shape. Use more than the minimal number of samples per cycle when the task needs peaks, phase or transient behaviour. Preserve native-resolution data if later aggregation is used for display.
Transfer check: change only the logger interval, not the lake. If a newly reported long-period feature appears, investigate sampling and averaging before inferring a new physical mode. A monitoring-system change can alter the visible spectrum without altering the waterbody.
CHAPTER 8 OF 12
8. Understand what averaging removes and what it preserves
Averaging can reduce short-period oscillations and reveal slower variation, but it also changes the measured signal. A daily mean may be useful for water-balance context while being unsuitable for a short-term maximum-level question. No one smoothing window serves every purpose.
For an ideal sine wave observed completely over an exact whole number of cycles, its mean displacement is zero. Over part of a cycle, the mean need not be zero. Missing intervals or a window that repeatedly samples one phase can leave an oscillatory bias in an apparent baseline.
Imagine a fictional hourly average from a twenty-minute oscillation: three complete ideal cycles cancel under continuous uniform averaging. A forty-minute window also covers two complete cycles. A thirty-minute window covers one and a half and can leave a nonzero mean depending on phase. The cancellation depends on the signal and window, not merely on calling the output hourly.
For a changing seiche amplitude or ongoing wind setup, even an integer-cycle argument is only approximate. Report the filtering method and edge treatment. Do not smooth a record until the desired storage trend appears and then treat that trend as independent evidence.
Transfer check: remove observations during the highest positive phase. Compare the available-sample mean with the full ideal mean. This exposes why a high completeness percentage alone cannot guarantee an unbiased level summary when missingness is phase- or weather-dependent.
CHAPTER 9 OF 12
9. Use multiple records to distinguish candidate mechanisms
A convincing interpretation connects water-level timing across locations with wind, atmospheric pressure and other relevant evidence. Opposite-phase shore levels support a redistribution hypothesis; a coherent rise across representative stations can support a mean-level change. Real systems may contain both.
Start with a fictional pair of shore records. After removing a common slow component, one has a positive peak when the other has a negative peak. Compare their timing and persistence with the proposed basin mode. Then check whether the apparent phase relation survives reasonable timestamp uncertainty and filtering choices.
A common barometric correction error can also appear across multiple sensors if they share the same wrong reference. Spatial agreement alone is not sufficient. Keep instrument and processing dependencies visible, and use independent reference observations where practical.
Rain and inflow records help assess genuine storage changes, but their water volumes must be compatible with the lake boundary. A short local level increase cannot be converted directly into rainfall volume without considering surface geometry, inflows, outflows and redistribution.
Transfer check: introduce a tributary flood at one end while wind acts along the basin. A simple two-gauge average may not fully separate those effects. State which additional locations or hydraulic information would help rather than forcing the data into a single clean explanation.
CHAPTER 10 OF 12
10. Keep surface and internal oscillations distinct
In a stratified lake, density differences create internal interfaces that can oscillate as well as the free surface. An internal seiche concerns movement within the water column and is not characterised simply by inserting total depth into the surface-seiche formula.
A temperature sensor at a fixed depth can record alternating warmer and colder water as an interface moves past it. That does not necessarily mean the whole lake warmed and cooled at that frequency. The sensor samples changing water masses at a fixed location.
For a fictional two-layer picture, the upper layer is warmer and the lower layer colder. An interface tilts and oscillates while the free surface changes only slightly. A shoreline level gauge alone may miss much of the internal motion. A vertical temperature profile or suitable current observations address a different part of the system.
Avoid inferring oxygen, nutrient or ecological consequences from interface motion alone. Mixing, transport, chemistry and biological response require their own observations. An oscillation can motivate a hypothesis about exchange without proving a water-quality outcome.
Transfer check: remove stratification in the teaching model. The internal interface no longer exists in the same form, while surface oscillations may remain. Identify which evidence and formula belong to each mode rather than treating every periodic lake signal as one phenomenon.
CHAPTER 11 OF 12
11. Separate scientific interpretation from shoreline safety claims
Seiches and related rapid level changes can matter to shorelines, harbours and infrastructure. An educational explanation does not establish a safe location, a safe time to visit or the maximum possible event. Use current responsible-authority warnings and access restrictions for real conditions.
A record of modest past oscillations does not define a guaranteed upper bound. Monitoring duration, rare forcing, basin state and sensor location influence what was observed. A fitted average amplitude should not be marketed as the largest possible rise.
For an original planning comparison, one gauge shows a 0.10-metre oscillation while another exposed area is unmonitored. It is not defensible to apply 0.10 uniformly to every shoreline. A hazard assessment needs appropriate spatial coverage and modelling, plus the consequences of depth, timing and exposure.
Likewise, the ideal period calculation is a scale check rather than an alert clock. A real event can contain several modes and changing forcing. Do not encourage a reader to approach the water because a simple sine-wave extrapolation suggests the next low phase.
Transfer check: move a sensor from an exposed shore to a protected harbour. A lower measured amplitude may describe shelter rather than reduced basin-wide hazard. Preserve the site context when communicating changes in the record.
CHAPTER 12 OF 12
12. Build a reproducible water-level explanation
A small evidence package includes gauge locations and surveyed references, sensor types, pressure correction, density assumptions, native timestamps, averaging and filtering, missing-data treatment, wind and atmospheric records, and the basin geometry used for any period estimate.
Reproduce three independent examples: the 16.8-minute ideal period, the 1.02-centimetre pressure-head equivalent of 100 pascals, and the twenty-minute signal hidden by twenty-minute sampling. Each diagnoses a different issue: physical timescale, instrument reference and temporal support.
Then work the two-shore example with a uniform 0.05-metre mean rise superimposed on opposite 0.20-metre displacements. Explain why neither shore alone gives the whole-lake mean. State the symmetry assumptions behind the simple average rather than applying it unquestioningly to an irregular lake.
For a changed-context exercise, add a bay, a sensor relocation or a changing atmospheric reference. Predict which part of the explanation must change before recalculating. A strong answer can say that more evidence is needed; it does not invent a new mode solely to preserve an old interpretation.
End with a bounded conclusion: the observations are consistent with this oscillatory or storage component at these locations and times, under these measurement corrections and model assumptions. State remaining alternatives. This gives the reader a physical explanation that can be checked rather than an attractive wave label attached to every rise and fall.
Independent worked case and evidence checks
An independent fictional review compares two end gauges on a symmetric teaching basin. At the first observation, west is +0.12 metres and east is −0.08 relative to a common datum-based reference. Under the stipulated symmetry, their mean is +0.02 and their half-difference is 0.10. The interpretation is a possible uniform two-centimetre component plus opposite ten-centimetre displacements. That decomposition is algebraically transparent. Whether it describes the real lake depends on gauge references, geometry and spatial support, which must be checked independently.
At a later observation, west is −0.06 and east is +0.14. Their mean is +0.04 and their half-difference is −0.10. The opposite component reversed while the common component increased by two centimetres. A reader looking only at west might report an eighteen-centimetre fall; one looking only at east might report a twenty-two-centimetre rise. Both local differences are arithmetically correct, but neither alone describes the stipulated common-level change. The pair reveals information that a single trace cannot provide.
Now suppose the east sensor has an unrecognised positive datum offset of 0.02 metres throughout. Correcting it changes the first pair to +0.12 and −0.10 and the second to −0.06 and +0.12. The inferred means become +0.01 and +0.03. Their change remains two centimetres because the fixed offset cancels in this particular difference, but the absolute common levels and opposite-component amplitudes change. This shows why some comparisons can tolerate a constant offset while others cannot. A drifting or time-varying offset would not cancel in the same way.
Add a fictional absolute-pressure correction error of 200 pascals at the later time. With the simplified density and gravity used earlier, that is about 0.0204 metres of spurious head. If both sensors share the same error, their pair mean can show an apparent rise even while their opposite-phase pattern remains convincing. Spatial decomposition alone cannot remove a common reference error. The reviewer needs the atmospheric record and processing basis as well as the water-level pair.
Finally retain only observations every twenty minutes while the opposite component has a twenty-minute period. The paired record can preserve a fixed apparent tilt while missing every reversal between samples. A second network with the same slow logger schedule does not solve that temporal limitation merely by adding more locations. Better spatial support and better temporal support address different uncertainties. Design the observations around both the mode pattern and its timescale, then preserve the native records for independent inspection.
For the transfer exercise, replace the symmetric rectangle with an irregular lake and a narrow bay. Do not carry the simple two-gauge mean forward as a whole-lake volume estimate. Use a justified spatial model or additional observations, and state what remains unresolved if those are unavailable. The general lesson survives: distinguish local level, common reference, oscillatory redistribution and sampling. The convenient arithmetic shortcut does not survive automatically. A strong explanation identifies that limit before its final number is used for a water budget or an infrastructure claim.
Worked comparison
| Teaching check | Inputs | Result and limit |
|---|---|---|
| Ideal surface period | L = 5 km, d = 10 m | About 16.8 min under rectangular uniform-depth assumptions |
| Pressure-head conversion | 100 Pa, density 1000 kg/m³ | About 1.02 cm; not a lake-response prediction |
| Same-phase sampling | 20-min wave, 20-min interval | An apparently constant sampled record |
| Aliased sampling | 20-min wave, 15-min interval | Can appear as a 60-min pattern |
| Symmetric tilt plus mean rise | ±0.20 m plus 0.05 m | End levels +0.25 and −0.15 m |
Questions and practical checks
Is every repeating lake-level signal a tide?
No. Seiches, atmospheric forcing and other processes can produce periodic or quasi-periodic variation. Use mechanism and spatial evidence.
Does a higher shore gauge prove added lake volume?
Not alone. Wind setup and standing oscillations can redistribute water across the basin.
Can atmospheric pressure affect a pressure-sensor record?
Yes. Absolute-pressure measurements need the appropriate atmospheric reference treatment; the lake can also respond physically to atmospheric forcing.
Can slow sampling hide a seiche?
Yes. Sampling at the same phase can appear flat, and other intervals can alias the motion into a different apparent period.
Does this guide predict a safe shoreline visit?
No. It is a measurement and reasoning guide. Use current official warnings and restrictions for actual conditions.
Continue with the connected subjects
Reservoir water balance — distinguish local level motion from volume change
Flood risk engineering — separate water behaviour from consequences
Physical oceanography — compare waves, density structure and coastal settings
Sources and scope
NOAA lake-level FAQ — wind setup, seiches and astronomical tides
NOAA glossary — seiche definition and ideal rectangular-basin period
USGS — lake-level sensors, atmospheric pressure reference and observed seiches
USGS — stage-measurement instruments and reference methods
US National Park Service — seiche mechanisms and hazard context
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence.