Two bed maps do not yet make a sediment budget
A reservoir bed appears 12 centimetres higher than it did last year. Has sediment accumulated, has the survey reference moved, or did the second instrument identify a different acoustic surface? The answer matters before anyone multiplies that difference by a large area.
This guide follows the complete inference from repeat bathymetry to bed-volume change and, only when justified, dry sediment mass. You will calculate deposition and erosion separately, examine a common vertical offset, recognise spatially correlated uncertainty, and decide when a small change is not distinguishable from survey error.
The focus is environmental change assessment, not navigation clearance or a dredging specification. The Hydrology subject guide supplies the water-balance context; the linked Hydrographic Surveying guide explains how a trustworthy individual survey is made. Here the new question is whether two individually plausible surveys can support a reliable comparison.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Frame the comparison · Chapters 1–2
Define the quantity before opening the maps
Keep a change inventory separate from a catchment budget
Make the surfaces comparable · Chapters 3–5
Put both surveys on the same vertical reference
Align position, resolution and the meaning of the bed
Use a common footprint without hiding missing ground
Calculate volume and mass · Chapters 6–8
Calculate net, depositional and erosional volume separately
Separate bed-volume change from water-storage change
Use dry bulk density only for the mass it represents
Test uncertainty and explanation · Chapters 9–10
Distinguish local detection from total-volume uncertainty
Test competing explanations before naming the cause
Practise and report · Chapters 11–12
Practise a changed-area and changed-density comparison
Deliver a comparison that another reader can reconstruct
CHAPTER 1 OF 12
1. Define the quantity before opening the maps
Bathymetry describes the submerged bed. A repeat-survey comparison asks how the elevation of a defined bed surface changed between two observation times. It does not directly measure every grain that entered or left the reservoir during that interval. Material can enter, settle, move again and leave before the second survey.
Choose a sign convention. In this guide, bed elevation z increases upward relative to a shared vertical datum. The difference is Δz = z₂ − z₁. Positive Δz represents a higher detected bed, interpreted as deposition only after alternatives are checked. Negative Δz represents lowering, potentially erosion. A map of positive-down water depths uses a different sign, so do not subtract depth grids as though they were elevation grids.
Suppose the bed changes from elevation 42.30 m to 42.42 m. The elevation increase is +0.12 m. If water level also rises 0.50 m, measured water depth increases by 0.38 m even though the bed rose. This simple case shows why depth change alone cannot establish erosion or deposition.
Write the intended output in advance: net bed-volume change inside a fixed polygon, gross depositional volume, storage capacity at a specified water level, or an estimate of dry sediment mass. These are different products. A single total labelled sediment change conceals decisions about boundaries, signs, water levels and density.
CHAPTER 2 OF 12
2. Keep a change inventory separate from a catchment budget
A bed-difference map is an inventory of geometric change over its surveyed footprint. A catchment sediment budget includes transfers across boundaries, suspended and bed material, tributaries, banks, floodplains and storage. The inventory can constrain one storage term; it does not automatically close the entire budget.
For an original teaching example, suppose incoming dry sediment mass is independently estimated as 900 tonnes and outgoing mass as 300 tonnes over a year. A 600-tonne increase in surveyed sediment storage would be consistent with that simplified balance. It would not prove that all three estimates are accurate, because shared bias or omitted stores can also produce apparent agreement.
Now suppose bathymetry implies only 200 tonnes of storage increase. Do not repair the discrepancy by choosing a convenient density. Investigate unmeasured delta deposits above water, shoreline erosion, settling outside the footprint, export timing and consolidation. A discrepancy is evidence about the measurement system as well as the reservoir.
Choose survey dates that suit the process. Comparing a pre-flood survey with one taken months after the flood includes intervening settling, redistribution and operations. Calling that difference the flood deposit is a causal claim requiring more evidence than the two dates alone supply. Keep the observation interval visible in every caption and summary.
CHAPTER 3 OF 12
3. Put both surveys on the same vertical reference
A vertical datum defines what elevation zero means. A local gauge zero, a national height datum and an ellipsoidal GNSS height are not interchangeable labels. Transformations require the appropriate reference information, not a guessed constant taken from a nearby project.
Imagine a fixed 80,000 m² comparison area. A uniform +0.05 m bias in the second survey generates an apparent 4,000 m³ volume increase. Collecting a million soundings does not make that shared offset disappear. It affects every sounding in the same direction, unlike independent random noise.
Use surveyed benchmarks, water-level observations, vessel offsets, processing records and independent stable checks to investigate alignment. A rock surface known to be stable may help identify an offset, but only if its stability and representation are defensible. Forcing the mean difference of an actively changing reservoir to zero would erase real net change.
Keep the original and transformed values. Record the original datum, transformation method, reference epoch where relevant, units and sign. If a correction of −0.04 m is applied to the second surface, state why that correction is supported. A visually pleasing match is not an independent calibration. The corrected map should remain traceable to the observations that justified it.
CHAPTER 4 OF 12
4. Align position, resolution and the meaning of the bed
Horizontal misalignment matters especially on slopes. If bed elevation rises 0.25 m per horizontal metre, a 0.40 m positional offset can create an apparent 0.10 m vertical difference. Alternating positive and negative bands beside steep features may therefore reflect displacement rather than balanced deposition and erosion.
Resample both surfaces onto a compatible comparison grid, but do not mistake a finer grid for new information. Turning 5 m soundings into 0.25 m cells creates more interpolated values, not twenty times more independent evidence along each direction. Compare conclusions at resolutions consistent with the original coverage.
Ask whether both surveys detected the same physical interface. A soft sediment layer can present a different acoustic return from a firmer underlying surface, depending on equipment, frequency, settings and processing. A detected elevation change may partly be a change in interface identification. Cores or other independent observations can test that interpretation.
Review sound-speed corrections, draft, motion and time synchronisation alongside the final grids. Crossings and overlap checks can reveal structured errors. Keep acquisition and processing differences in the comparison report even when each survey passed its own checks: a method suitable for one-time mapping is not necessarily sensitive enough to resolve a small year-to-year change.
CHAPTER 5 OF 12
5. Use a common footprint without hiding missing ground
A fair comparison requires observations or defensible estimates for both dates at each included location. Start with the intersection of valid coverage, not the union of two attractive maps. Identify shorelines, shallow margins, vegetation, structures and acoustic shadows that were excluded or interpolated.
Suppose survey one covers 1.00 km² and survey two covers 0.85 km². The 0.15 km² difference is not zero change. It is an unobserved comparison region unless additional evidence supports an estimate. Reporting a whole-reservoir volume from the overlap alone understates the uncertainty and may miss the most active shallow delta.
Keep a fixed comparison polygon when assessing change through time. A polygon following the waterline at each survey date changes the control area as well as the bed. If shore exposure is important, combine appropriate topographic and bathymetric observations with compatible references, and identify the wet–dry seam as a potential source of error.
Interpolation across a gap is a model. Test a plausible high and low surface for that region rather than quietly filling it with the surrounding mean. If a 10,000 m² unsurveyed patch might have changed by 0 to 0.30 m, its unresolved contribution is 0 to 3,000 m³. That range may dominate a precisely calculated total elsewhere.
CHAPTER 6 OF 12
6. Calculate net, depositional and erosional volume separately
For cells of area aᵢ, signed net bed-volume change is Σ(aᵢΔzᵢ). Gross deposition is the sum of positive changes times area; gross erosion is the magnitude of the negative part. Net change equals gross deposition minus gross erosion under the chosen sign convention.
Take three fictional zones: 20,000 m² rises by 0.18 m, 15,000 m² lowers by 0.08 m, and 5,000 m² rises by 0.04 m. Their signed contributions are +3,600, −1,200 and +200 m³. Gross deposition is 3,800 m³, gross erosion is 1,200 m³ and net increase is 2,600 m³.
A report giving only +2,600 m³ hides substantial redistribution. That distinction matters if deposits approach an intake while erosion occurs elsewhere. Conversely, adding absolute changes would give 5,000 m³, which is gross morphological activity rather than net storage. Neither number should be substituted for the other.
Check the area units before multiplying. One hectare is 10,000 m². A mean rise of 0.10 m over 2 hectares corresponds to 2,000 m³, not 0.20 m³. If cells have unequal area, use their individual areas rather than multiplying an unweighted mean elevation difference by total area. Incomplete cells along boundaries also need a consistent treatment.
CHAPTER 7 OF 12
7. Separate bed-volume change from water-storage change
For a fixed water-surface elevation H, water storage can be estimated by integrating positive depth H − z over the inundated area. A rising bed generally reduces storage below that reference level. But changing operating water level can alter observed water volume independently of sediment.
In a simplified reservoir with a fixed vertical-sided footprint of 100,000 m², a uniform bed rise of 0.10 m reduces capacity at a fixed H by 10,000 m³. A simultaneous water-level rise of 0.30 m increases actual water volume by 30,000 m³ before subtracting that bed effect. The net actual-volume increase is 20,000 m³. Capacity fell while water volume rose.
Real shorelines are not vertical walls. Area changes with elevation, newly exposed or inundated margins contribute, and the comparison may need a revised elevation–area–capacity relationship. State the reference level when quoting capacity loss. The phrase reservoir lost ten thousand cubic metres is incomplete without identifying capacity, actual water or sediment volume.
Consolidation creates another distinction. A deposited layer may become thinner as pore water leaves, even when its dry mineral mass stays nearly constant. A later lower bed does not necessarily imply equivalent sediment export. Connect the geometric observations with material properties and time history before translating them into a mass balance.
CHAPTER 8 OF 12
8. Use dry bulk density only for the mass it represents
Dry bulk density is dry solid mass divided by the total bulk volume of the sampled sediment, including pore space in that bulk volume. It is not the density of individual mineral grains, and it is not wet bulk density. The correct density depends on the material, depth, compaction and sampling method.
If a reasonably characterised deposit has volume 3,800 m³ and representative dry bulk density 700 kg/m³, its estimated dry mass is 2,660,000 kg, or 2,660 tonnes. This is a teaching calculation, not a recommended density. Using 2,650 kg/m³ as though all bulk deposit volume were solid mineral would greatly overstate this fictional mass.
Do not multiply net volume by one density when deposition and erosion involve different materials. In the chapter-six example, let the 3,800 m³ deposit have dry bulk density 700 kg/m³ and the 1,200 m³ eroded material 1,100 kg/m³. The mass increase is 2,660 − 1,320 = 1,340 tonnes, not 2,600 m³ times an arbitrary compromise density.
Density sampling needs spatial and vertical representation. A shallow soft core near the inlet may not represent an older compacted deposit near the dam. Preserve water-content definitions, drying protocol, sample volume and recovery concerns. If only a plausible range is available, report the resulting mass range and avoid presenting it as a precise measured inventory.
CHAPTER 9 OF 12
9. Distinguish local detection from total-volume uncertainty
A cell difference smaller than measurement uncertainty is not automatically meaningful. If two elevation errors are independent, zero-mean and have standard deviations 0.04 and 0.03 m, the standard deviation of their difference is √(0.04² + 0.03²) = 0.05 m. Under an additional approximately normal assumption, about 1.96 times that value gives a two-sided 95% interval near ±0.10 m.
Those assumptions must be stated. Shared datum errors, spatial correlation, non-normal artefacts and interpolation can invalidate a simple independent-cell calculation. With covariance c between the two elevation errors, the difference variance is σ₁² + σ₂² − 2c. Common errors can cancel; changed errors can persist. Independence is a hypothesis, not a default truth.
Thresholding a map can help show where change is distinguishable, but setting every small cell to zero changes the estimate. Many real small deposits may accumulate over a large area. Report raw and thresholded summaries with the threshold rationale rather than treating one as unquestionable truth.
For total volume, spatial dependence is crucial. One hundred adjacent cells sharing a striping error are not one hundred independent checks. A uniform uncertain datum offset contributes area times offset uncertainty. A defensible volume interval should reflect correlated components and boundary uncertainty, using an approach appropriate to the available survey-error information.
CHAPTER 10 OF 12
10. Test competing explanations before naming the cause
A coherent positive deposit near an inlet, supported by cores and flood timing, is a stronger sedimentation interpretation than a uniform increase across every surface. Stripes aligned with survey tracks suggest acquisition or processing effects. Paired positive and negative rims along steep banks suggest horizontal misregistration. Patterns guide investigation; none alone supplies proof.
Compare independent evidence: water-level records, benchmark checks, crossing differences, core layers, sediment concentration and discharge, reservoir operations and aerial imagery. Keep these lines of evidence separate enough that their agreement is informative. Two products derived from the same flawed grid are not independent confirmation.
Avoid over-attributing a change to one storm or land-use intervention. A new sediment trap, changed drawdown, shoreline failure or survey timing can alter the result. A before-and-after map supports a measured interval change; causal attribution needs a comparison design and process evidence.
Decide what would change the conclusion. If the apparent 0.06 m rise vanishes after a supported 0.05 m datum correction, the initial sedimentation claim weakens substantially. If a metre-thick local deposit remains after every plausible correction, its existence is more robust, although density and source may remain uncertain. Separate confidence in detection, magnitude and cause.
CHAPTER 11 OF 12
11. Practise a changed-area and changed-density comparison
Begin with an overlap area of 50,000 m² and mean signed change +0.08 m. The naive net volume is +4,000 m³. Independent stable checks indicate that survey two is biased upward by 0.03 m. Applying the justified correction leaves mean change +0.05 m and net volume +2,500 m³.
Now change the context: only 40,000 m² has reliable repeat coverage. If its corrected mean is still +0.05 m, the supported overlap estimate is +2,000 m³. The remaining 10,000 m² must be shown separately. Assigning it the same change gives an extrapolation of +500 m³, not an additional observation.
Next change material properties. Suppose the supported area contains +3,000 m³ of deposition at 600 kg/m³ and −1,000 m³ of erosion at 1,000 kg/m³. Net dry mass change is 1,800 − 1,000 = +800 tonnes. Multiplying +2,000 m³ by the depositional density would yield +1,200 tonnes and overlook the denser eroded material.
Explain the transfer in words: datum alignment changes the vertical difference, coverage changes the included area, and density changes the conversion from volume to mass. These are different stages in the reasoning chain. A correction at one stage cannot justify an unsupported assumption at another. Check each stage independently before combining them.
CHAPTER 12 OF 12
12. Deliver a comparison that another reader can reconstruct
A useful report identifies both survey dates, coordinate and vertical references, equipment and processing versions, grid method, overlap polygon, exclusions and sign convention. Include a difference map with a labelled scale and comparable coverage; an impressive colour ramp without those labels is not a reproducible result.
Report gross deposition, gross erosion and net change together. State the fixed water level used for any capacity calculation. For mass estimates, give density evidence and distinguish dry from wet mass. Include uncertainty components, not only a final interval, so readers can see whether datum alignment, gaps or material sampling controls the conclusion.
Preserve raw inputs and correction history. If a later benchmark review changes the second survey by 0.02 m, users should be able to recompute the inventory without reconstructing the entire project from a presentation slide. Versioned boundaries and processing choices matter as much as versioned grids.
The practical standard is modest but demanding: do not claim sediment where you have only a difference, do not claim mass where you have only geometry, and do not claim a whole-reservoir budget where you surveyed only part of its bed. A carefully bounded result remains useful precisely because its limits are visible.
Independent worked case and evidence checks
A fictional reservoir team compares two surveys over a common 60,000 m² footprint. Zone A covers 25,000 m² and appears to rise 0.20 m. Zone B covers 20,000 m² and appears to lower 0.06 m. Zone C covers 15,000 m² and appears to rise 0.04 m. Before correction the signed contributions are +5,000, −1,200 and +600 m³, giving a net +4,400 m³. Pause here and predict what a common vertical correction will do to each zone.
Independent benchmark and stable-surface checks support an upward bias of 0.03 m in the second survey. Subtracting that bias from every difference gives +0.17, −0.09 and +0.01 m. Corrected contributions are +4,250, −1,800 and +150 m³. Net change becomes +2,600 m³. The reduction is 60,000 × 0.03 = 1,800 m³, providing an independent arithmetic check on the zone calculation.
Gross deposition is now 4,400 m³ and gross erosion is 1,800 m³. The team has not merely adjusted the net total: it has also changed the amounts assigned to deposition and erosion. If Zone C has uncertainty larger than 0.01 m, its sign is not robust. It should remain visible as a low-confidence component rather than becoming a confident new deposit in the narrative.
Suppose cores support dry bulk density 650 kg/m³ for the deposited material, while eroded older material is represented by 950 kg/m³. The estimated deposited dry mass is 2,860 tonnes and eroded dry mass 1,710 tonnes, giving a net increase of 1,150 tonnes. If the deposition density instead ranges from 550 to 750 kg/m³, while the erosional estimate is held fixed for this sensitivity test, net mass ranges from 710 to 1,590 tonnes. This is not a formal total confidence interval because other uncertainties have not been propagated.
A 5,000 m² shallow margin was absent from both-survey overlap. A separate topographic comparison would be needed to include it. If plausible change there ranges from −0.05 to +0.25 m, the missing geometric contribution ranges from −250 to +1,250 m³. Adding that scenario range to the supported net produces +2,350 to +3,850 m³ before other uncertainty, but the report should retain the observed overlap and the extrapolated margin as separate lines.
Finally, imagine the management question is intake blockage rather than net storage. Zone A lies beside the intake and Zone B lies far upstream. The positive net total is not the most useful risk indicator; local elevation, distance to the intake and operating conditions matter. Transfer means changing the reported quantity when the decision changes, while preserving the same honest survey evidence.
A second desk exercise isolates spatial correlation. Imagine 100 cells, each 100 m², with elevation-difference standard uncertainty 0.05 m. Each cell contributes volume standard uncertainty 5 m³. If all cell errors were genuinely independent, the total standard uncertainty would be √100 × 5 = 50 m³. If the entire area instead shares one uncertain vertical offset of standard uncertainty 0.05 m, the corresponding total is 10,000 × 0.05 = 500 m³. The same nominal cell uncertainty produces a tenfold difference in the volume uncertainty because its spatial structure changed.
These are limiting teaching cases, not competing recipes to select for convenience. Real survey errors may contain both shared and local components, with intermediate correlation lengths. Sketch which components arise from one benchmark, one survey line, an interpolation region or an individual sounding. The sketch helps determine which observations provide independent information. Repeating a line can test some local behaviour but does not independently verify the benchmark used by every line.
For a final capacity check, hold the water reference elevation fixed and consider a surveyed deposit that lies entirely above that elevation. It adds sediment volume to the broader landscape inventory but does not displace water storage below the chosen reference surface. If the management reference rises enough to inundate it, its relevance changes. This is why a whole-bed net-volume number should not be used as capacity loss without checking the elevation range and inundation geometry.
An independent learner can now audit a proposed result in order: identify elevation rather than depth; verify both references; mark the common footprint; calculate signed and gross volumes; separate uncertain areas; apply representative densities only to the appropriate materials; and name the fixed level for capacity. If any step is unsupported, retain the earlier defensible result rather than carrying the unsupported inference into a more impressive final total.
Worked comparison
| Quantity | Calculation in corrected case | Meaning |
|---|---|---|
| Gross deposition | 4,250 + 150 = 4,400 m³ | Positive bed-volume changes |
| Gross erosion | 1,800 m³ | Magnitude of lowering |
| Net bed-volume change | 4,400 − 1,800 = 2,600 m³ | Signed geometric inventory |
| Net dry sediment mass | 2,860 − 1,710 = 1,150 tonnes | Density-dependent material estimate |
Questions and practical checks
Can I compare two water-depth maps directly?
Only after reconciling water level, vertical reference, position, coverage and the detected interface. Depth can increase because water rises, even while sediment raises the bed.
Does a finer grid make a more accurate change map?
Not by itself. Interpolation adds cells, not independent observations. Use resolution appropriate to the original surveys and test whether the conclusion survives reasonable gridding choices.
Should every change below a detection threshold become zero?
That is a reporting choice with consequences, not proof of no change. Preserve the raw signed inventory and explain any thresholded alternative.
Can net volume be converted with one density?
Only when that representation is justified. Separate deposition and erosion if their dry bulk densities differ, and keep consolidation distinct from export.
What would make me postpone a sedimentation claim?
An unresolved datum mismatch, incompatible acoustic interfaces, missing active margins, or errors as large as the inferred change. Resolve the dominant ambiguity before attaching a precise total.
Continue with the connected subjects
Build the survey evidence
Hydrographic Surveying: references, sound-speed correction, coverage and uncertainty
Hydroacoustics: what an underwater echo does and does not identify
Connect geometry with transport and storage
Watershed Sediment Budgets: inputs, outputs and changing stores
Suspended-sediment loads: transport observations are not bed inventories
Reservoir residence times: geometry, flow and incomplete mixing
Dredging Engineering: separate environmental evidence from authorised works
Connect bed change with continuous discharge
Index-velocity ratings: maintain the surveyed stage-area relationship as channels change
Sources and scope
USGS: bathymetric uncertainty and spatial survey artefacts
NOAA: bathymetric data quality and source information
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. Sources were checked on 29 October 2026.