Designing event sampling for annual suspended-sediment load
SN-2026-076 · Methods note · Land and freshwater
By Elín Bjarnadóttir · · 3 pages · 1,492 words
Fictional demonstration publication. Not peer reviewed. Still North Arctic Institute is a fictional organisation. This paper is a teaching synthesis or illustrative practice document, not a report of original institute fieldwork. Published sources are identified below; illustrative examples are not measured results.
Abstract
Monthly sediment sampling can span an entire year while missing its dominant transport events. This methods note uses US Geological Survey estimation guidance and an empirical river investigation to examine the design of annual suspended-load monitoring. A fictional year demonstrates how two event days can dominate the total and why continuous discharge alone cannot recover their unmeasured concentrations. The discussion considers event limbs, routine and triggered samples, cross-section representativeness, turbidity calibration, sensor failures, and validation that withholds whole events. It distinguishes model intervals from uncertainty caused by unobserved conditions and changing sediment sources. No actual river load or Arctic conversion factor is estimated. The practical recommendation is a coverage account tied to the annual result, allowing subsequent sampling to address the largest demonstrated blind spots rather than simply filling the calendar.
Download the complete PDF (3 pages)
Full paper
Annual coverage is not annual representation
Twelve monthly sediment samples can cover every month and still miss the hours carrying most of a river's annual suspended load. Increasing laboratory precision does little to address this temporal gap. The design question is not simply how often to sample, but which parts of the transport history are likely to remain unobserved. A successful annual-load programme must connect event dynamics, representative channel sampling, and the model used between samples.
This methods note focuses on temporal allocation and the vulnerability of annual suspended-sediment estimates to missed events. It draws on US Geological Survey guidance for turbidity-based time series, the LOADEST report, and an empirical investigation of Minnesota rivers. Those studies are methodological sources, not Arctic observations. The numerical examples are fictional, and no monitoring programme, flood, or sediment budget is attributed to Still North.
Define the load and the sampling unit
Suspended-sediment load is the mass carried in suspension past a cross-section over a stated interval. It does not include all bedload or every source of erosion within the catchment. Eroded material can be stored before reaching the section, and transported material can be deposited downstream. Annual outlet load is therefore a boundary flux, not a direct measure of all soil loss or ecological exposure across the basin.
The time-series calculation combines representative suspended-sediment concentration with simultaneous discharge and duration. Concentration in milligrams per litre multiplied by discharge in cubic metres per second gives grams per second. Integrating through time gives mass. If concentration and discharge change sharply, averaging them separately can lose important covariance. A short interval is useful only if its concentration estimate actually represents the water passing through the section.
A fixed sampler or optical sensor occupies a small part of that section. Its relationship with cross-section concentration must be evaluated under relevant flows and sediment mixtures. An automatic intake that performs well at ordinary stage may become unrepresentative during a flood. Temporal density does not automatically provide spatial representativeness. This distinction is especially important when comparing continuous sensor output with occasional depth-integrated or cross-section samples.
Worked example: the two days that change the year
Fictional worked example. Imagine a simplified year with 363 ordinary days, each at 10 milligrams per litre and 1 cubic metre per second. Two event days each have 200 milligrams per litre and 20 cubic metres per second. Treat these as exact daily means for the illustration, not as realistic constant hydrographs. The ordinary-day flux is 10 grams per second, while the event-day flux is 4,000 grams per second.
The ordinary days carry 313.632 tonnes in total. The two event days carry 691.2 tonnes. Annual load is therefore 1,004.832 tonnes, about 69% of it moved in less than 1% of the days. A monthly schedule that misses both events might see only the ordinary concentration. If that concentration were multiplied by the complete discharge record, the resulting estimate would be 348.192 tonnes, roughly 65% below the constructed annual total.
The complete flow record reduces one kind of ignorance, but it cannot supply the missing event chemistry by itself. A model might infer higher concentration at high flow if independently calibrated evidence supports that relationship. It might also fail if a slope failure suddenly changes sediment availability, or if the event has no analogue in the calibration data. A plausible-looking rating curve is not an observation of the missing event.
Under an idealised random-day design, twelve distinct sampling days would have only about a 6.5% chance of hitting at least one of the two event days. Fixed monthly visits need not behave like that random design, and events can be seasonally predictable. The calculation is a reminder that sparse temporal coverage can remain poor even when visits are distributed neatly through the calendar. It is not a prescription for random visits during dangerous flows.
Rising and falling water can require different evidence
At the same discharge, concentration may differ on the rising and falling limbs. Nearby sediment can be flushed early; distant material can arrive late; exhaustion or changing channel connection can alter the relationship during an event. Sampling only near the peak does not necessarily capture the integrated load. Event sampling should aim to constrain the trajectory, including onset, rise, peak vicinity, and recession, within safe operational arrangements.
A practical design combines routine samples with triggered sampling. Routine observations preserve seasonal and low-flow coverage, while triggers target rapid stage changes, unusual turbidity, rainfall, or snowmelt periods. The trigger variable and threshold should be evaluated against missed events and false triggers. A fixed trigger may behave differently after channel change or in winter. Automatic equipment needs adequate bottles, power, timing, and a plan for documenting failures.
The design should also reserve capacity for independent calibration samples rather than using every resource for sensor installation. A dense proxy record can reveal when events occurred, but the conversion from proxy to sediment mass requires paired measurements. Field teams must never enter unsafe water or cross unstable banks merely to satisfy a sampling schedule. A documented gap is evidence about coverage; an unsafe measurement is not an acceptable solution.
A continuous surrogate is a model, not a balance
Rasmussen and colleagues' USGS guidance describes developing and evaluating sediment estimates from turbidity and streamflow. It emphasises site-specific relationships and continued calibration because sediment sources and properties can change. Turbidity is an optical response, influenced by the particles present, not a direct weighing of sediment. A new particle mixture can change the relationship even when the sensor is functioning normally.
The Minnesota investigation by Ellison, Savage, and Johnson found that streamflow alone related poorly or not at all to suspended-sediment concentration at several sites, while turbidity generally provided a better surrogate. It also documented differences between total suspended solids and suspended-sediment concentration. Those findings should not be imported as Arctic conversion factors. They show why method choice and local calibration can matter more than a familiar variable name.
A sensor may saturate, foul, freeze, lose power, or become exposed during exactly the conditions of interest. Report such periods explicitly. Replacing them with an ordinary low-flow relationship can systematically suppress annual transport. A defensible reconstruction can use alternative observations and bounded scenarios, but should distinguish direct measurements, model predictions within calibration range, and extrapolations beyond it.
Evaluate the annual estimate where its mass occurs
A model fitted to many ordinary samples can achieve a reassuring overall error while missing the largest loads. Examine residuals by discharge, season, event limb, and sediment source where identifiable. Compare the fraction of water volume and estimated load represented by the calibration range. Consider validation that withholds entire events. Randomly separating closely spaced event samples can exaggerate apparent independence and leave almost identical conditions in both calibration and validation sets.
The LOADEST report provides regression-based estimation methods and diagnostics, including approaches for censored data. Its relevance is the explicit link between the calibration model and interval load estimates. Using a recognised program does not justify a weak sampling design. A model's statistical interval may not include uncertainty from an omitted event type, unrepresentative intake, or changed sediment source.
A useful sensitivity analysis asks how the annual result changes under credible alternative reconstructions of major gaps. If the event concentration in the worked example were substantially uncertain, most effort should go into constraining those two days rather than adding decimal places to ordinary samples. This is an allocation principle, not permission to discard routine observations: the programme still needs to detect when its supposedly ordinary regime changes.
Limits and a sampling plan that can learn
Annual suspended load does not describe all ecological consequences. Duration of high concentration, particle size, deposition, and timing relative to organisms may matter more for some decisions. Those targets can require different sampling intervals. Glacier-fed, thaw-affected, and rainfall-dominated catchments also differ in event structure and sediment supply. There is no universal number of bottles per event or visits per year that guarantees accuracy across them.
The strongest annual report pairs the load with a coverage account: when events were sampled, which flow ranges calibrated the estimate, where sensors failed, and how missing intervals were treated. The next year's design can then address the largest demonstrated uncertainty. More samples are worthwhile when they reduce an identified blind spot. Without that link, a fuller calendar can still leave the river's decisive hours almost entirely unseen.
References
- Rasmussen, P. P., Gray, J. R., Glysson, G. D., and Ziegler, A. C. (2009). Guidelines and procedures for computing time-series suspended-sediment concentrations and loads from in-stream turbidity-sensor and streamflow data. USGS Techniques and Methods, book 3, chapter C4.
- Runkel, R. L., Crawford, C. G., and Cohn, T. A. (2004). Load Estimator (LOADEST). USGS Techniques and Methods, book 4, chapter A5.
- Ellison, C. A., Savage, B. E., and Johnson, G. D. (2014). Suspended-sediment concentrations, loads, total suspended solids, turbidity, and particle-size fractions for selected rivers in Minnesota, 2007 through 2011. USGS Scientific Investigations Report 2013–5205.
Keywords: suspended sediment; event sampling; annual load; turbidity calibration; temporal coverage.
Browse and filter the research library · Download learning guides