A dye peak marks one part of a transport history, not the arrival time of every water parcel
A downstream dye record rises, reaches a peak and fades into a long tail. That curve contains several different timing questions. When was dye first detected? When was concentration greatest? When did most recovered tracer mass pass? How much was never observed within the sampling window? One reported travel time cannot answer all four.
This guide concerns interpretation of an already authorised tracer study. It is not a field protocol or permission to release dye into water. Real studies require the relevant authority, environmental review, qualified personnel and instrument-specific procedures. All numerical examples below are fictional and are not release quantities or predictions for an actual river.
The existing tracer-dilution guide asks how tracer observations can estimate discharge. Here the central questions are transport timing, mass recovery and the meaning of a late tail. The methods share conservation reasoning, but their outputs and assumptions are not interchangeable.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1: Define timing and observations · Chapters 1–2
Route 2: Calculate transport summaries · Chapters 3–5
Route 3: Challenge tracer behaviour · Chapters 6–8
Interpret recovery without identifying every missing pathway
Route 4: Test windows and scale · Chapters 9–10
Route 5: Interpret and transfer · Chapters 11–12
Compare reaches and conditions without assuming a universal speed
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Define timing and observations · Chapters 1–2
Define first arrival, peak and centre of recovered mass
Convert the instrument signal into a background-corrected record
Calculate transport summaries · Chapters 3–5
Turn concentration into tracer mass flux
Integrate the curve with its actual time support
Calculate a mass-weighted passage time
Challenge tracer behaviour · Chapters 6–8
Interpret recovery without identifying every missing pathway
Keep sorption separate from hydrodynamic storage
Consider light, optical interference and changing response
Test windows and scale · Chapters 9–10
Treat a long tail as a hypothesis-generating feature
Use detection limits and observation windows honestly
Interpret and transfer · Chapters 11–12
Compare reaches and conditions without assuming a universal speed
Report a transport result without promising contaminant behaviour
CHAPTER 1 OF 12
1. Define first arrival, peak and centre of recovered mass
First arrival means the earliest detected signal under a stated detection rule. Peak time is when the measured concentration reaches its maximum under the sampling resolution. A mass-weighted mean passage time uses the timing of recovered tracer mass. These statistics describe different aspects of a breakthrough curve. None should be labelled simply “the travel time” without saying which definition was used.
A sharply rising curve with a long low tail can have an early peak and a substantially later mean. A small early detection may precede both. The leading edge can matter for an early-warning question, whereas the mass-weighted mean may be more useful for describing bulk transport. The tail can matter for persistence, but its physical interpretation depends on tracer behaviour and observation completeness.
For a transfer check, keep peak time fixed and add a late pulse carrying substantial recovered mass. The mean becomes later even though the peak is unchanged. Then lower the reporting limit enough to detect an earlier faint signal: first detected arrival can become earlier without changing the underlying water movement. Definitions and measurement rules both affect the reported timing.
CHAPTER 2 OF 12
2. Convert the instrument signal into a background-corrected record
A fluorometer observes an optical signal that requires calibration and appropriate corrections before it is interpreted as tracer concentration. Background fluorescence, temperature, turbidity or other optical influences may matter for a particular instrument and water matrix. Preserve the calibration relation, units and relevant checks. A smooth fluorescence trace is not automatically a clean tracer-concentration record.
Subtract a justified background rather than assuming every signal above zero is introduced dye. A background estimated before the event may become inadequate if water conditions change during passage. Keep raw and corrected records distinguishable. If a corrected value falls below the reporting limit, treat it according to the stated rule rather than silently converting uncertain small values into precise positive concentrations.
A fictional background of 0.4 concentration units and observed signal equivalent to 1.1 units gives a corrected 0.7 units under an additive model. If actual background rose to 0.6, the corrected tracer contribution would instead be 0.5. The arithmetic is simple, but choosing the applicable baseline requires evidence. Its importance often grows in a weak late tail where tracer signal approaches background.
CHAPTER 3 OF 12
3. Turn concentration into tracer mass flux
Concentration is tracer mass per water volume. Multiplying concentration C by discharge Q gives a tracer mass flux, provided units and observation support are compatible. Integrating Q(t)C(t) over the observation interval estimates recovered mass. Concentration alone is not a mass-passage rate when river discharge changes. A low concentration at high discharge can carry more mass per hour than a larger concentration at low discharge.
Use a unit example: 2 micrograms per litre equals 2 milligrams per cubic metre. At Q = 3 m³/s, tracer flux is 6 mg/s. If those values held for 100 s, the passing mass would be 600 mg. The conversion works because litres and cubic metres were reconciled. Confusing micrograms per litre with micrograms per cubic metre creates a thousand-fold mass error.
For transfer, double discharge while halving concentration over an equal interval. Mass flux is unchanged under the product model. The concentration peak may look smaller even though the same tracer mass passes per unit time. This is why comparisons between events should distinguish dilution from a changed amount of recovered tracer. A graph plotted in concentration units cannot establish that distinction by appearance alone.
CHAPTER 4 OF 12
4. Integrate the curve with its actual time support
Discrete observations approximate a continuous passage curve. A simple trapezoidal calculation averages adjacent flux observations and multiplies by their elapsed interval. The method is an approximation, especially around a narrow peak or an unobserved pulse. Preserve actual timestamps rather than assuming evenly spaced samples when the record contains gaps or manual changes in sampling frequency.
Suppose tracer flux is 0, 4, 2 and 0 mg/s at times 0, 100, 200 and 400 s. Trapezoidal masses are 200, 300 and 200 mg, totalling 700 mg. The last interval lasts 200 s, twice the earlier intervals. Treating every interval as 100 s would lose 100 mg in that final contribution. The error arises from time weighting, not the concentration calibration.
The calculation does not prove that linear interpolation was correct between observations. A brief high pulse could be missed. For a changed-context check, insert an additional observation halfway through a long interval and recalculate. A changed integral can reveal the sensitivity to temporal resolution. Report the approximation and coverage rather than describing a sampled curve as a complete record by default.
CHAPTER 5 OF 12
5. Calculate a mass-weighted passage time
Let w_i represent recovered mass assigned to a small time interval and t_i its representative time relative to the agreed start. The discrete mean passage time is the sum of t_i w_i divided by the sum of w_i. For continuous records, the corresponding expression weights time by Q(t)C(t). A concentration-weighted mean is equivalent only when the relevant discharge treatment makes it so.
For a teaching example, 20, 50 and 30 mass units pass around 10, 20 and 40 minutes. The mean is (20 × 10 + 50 × 20 + 30 × 40)/100 = 24 minutes. The largest mass interval is centred at 20 minutes, but the mean is later because of the last interval. State the weighting scheme and the meaning of each representative time.
For transfer, remove half the late mass because the observation window ended too soon. The mean of the recovered portion becomes earlier. That shift does not establish faster transport. It is a change in which mass was included in the statistic. Call it an observed-window or recovered-mass mean where appropriate, and assess whether unobserved late mass could materially alter the intended inference.
CHAPTER 6 OF 12
6. Interpret recovery without identifying every missing pathway
Recovery fraction compares observed recovered mass with the relevant introduced mass under a consistent accounting definition. If 80 of 100 mass units are recovered, the fraction is 0.80. The missing 20 units are not automatically destroyed. They may remain in unsampled storage, leave by another path, be affected by sorption or transformation, or reflect errors in source mass, discharge, calibration or sampling.
A recovery estimate can exceed one if measurement or accounting errors are important. That is a diagnostic problem, not evidence that tracer mass was created in the river. Investigate the introduced-mass definition, background, concentration units, discharge, integration and observation boundary. Do not force the fraction into the zero-to-one range by an undocumented adjustment that hides the original inconsistency.
The changed-context check moves the sampling section upstream of a tributary junction or adds a side-channel outlet. The observation boundary changes, so the meaning of recovered fraction can change. Keep the physical pathways mapped. A percentage reported without a section, time window and mass definition cannot be interpreted reliably, even if its arithmetic is precise.
CHAPTER 7 OF 12
7. Keep sorption separate from hydrodynamic storage
A tracer can interact with surfaces rather than remaining entirely in the water phase. Temporary retention and later release can alter an observed curve; incomplete return within the window can reduce apparent recovery. These effects can resemble transport through slower water pathways. A single tail is therefore not sufficient to distinguish physical exchange from tracer-specific behaviour.
Runkel’s USGS-hosted review and modelling analysis specifically cautions against using rhodamine WT tails to quantify hyporheic-zone processes because nonconservative behaviour can affect that portion of the signal. Bulk travel-time information and detailed tail interpretation have different requirements. The lesson is not that every dye observation is useless; it is that validity depends on the property being inferred and the part of the curve used.
For a teaching comparison, imagine two tracers travelling with the same flow, but only one experiences appreciable retention by the material encountered. Their different tails cannot be attributed solely to different water residence times. An appropriate reference tracer and independent evidence can help separate explanations. Do not turn a convenient transport model into proof that the tracer itself behaved conservatively.
CHAPTER 8 OF 12
8. Consider light, optical interference and changing response
Measured signal can change through a combination of water transport, tracer chemistry and instrument response. Exposure and water conditions may affect a fluorescent tracer or its measurement. The relevant controls depend on the tracer and equipment. Preserve laboratory and field checks that test those mechanisms, rather than treating a signal decline as a universal measure of dilution.
In a conditional example, a known calibration check changes signal response by 10% between two periods while concentration is held fixed. If the old calibration is retained, the computed concentration and mass flux can change even though tracer mass did not. This is a measurement-chain problem. A conservative transport assumption cannot fix it, because the error entered when signal was translated into concentration.
For transfer, compare an open sunlit reach with a shaded reach. Do not assume identical tracer-specific losses merely because flow and length match. Equally, do not assert that sunlight explains a difference without suitable evidence. The useful approach lists plausible mechanisms, identifies which controls address them and limits the resulting transport claim to what those controls support.
CHAPTER 9 OF 12
9. Treat a long tail as a hypothesis-generating feature
Late signal can be consistent with a mixture of travel paths and exchange with slower domains, but the curve does not map those domains uniquely. Surface pools, subsurface exchange, tracer interactions and a prolonged input can all affect late passage. A fitted storage parameter is a model-dependent summary, not a directly observed underground volume or a map of its location.
An upstream input that lasts longer than assumed also broadens the downstream curve. To interpret reach transport, preserve the input history or an appropriate upstream observation. A downstream tail cannot be assigned entirely to the reach if the upstream boundary already contained one. Comparing sections requires accounting for the signal that entered, not only the signal that left.
For transfer, keep the reach unchanged but replace a short input pulse with an extended input of equal total mass. The downstream concentration history can broaden even though reach storage properties are unchanged. Explain which observed change belongs to the boundary condition. This example exposes why a visually longer tail is not by itself evidence of more hyporheic storage.
CHAPTER 10 OF 12
10. Use detection limits and observation windows honestly
The first detected arrival depends on sensitivity, background and the rule for accepting a signal. A more sensitive method can detect a weak leading edge sooner. The last detected tail likewise depends on the reporting limit. Report those limits when comparing durations. An instrument’s inability to distinguish a late signal from background is not evidence that no tracer remains anywhere in the system.
Suppose the last quantified observation occurs at 90 minutes, with a weak uncertain signal at 120 minutes. A mean calculated only to 90 minutes is a defined-window result. Extending the curve to zero at an invented time introduces an assumption that should be tested. Where the missing tail matters, provide scenarios or a limitation rather than pretending that a neat integration endpoint was observed.
A changed-context check raises the reporting limit while keeping the physical curve fixed. First detected arrival may become later and apparent tail duration shorter. The measured passage can appear narrower without any hydrological change. Keep analytical limits with the data so that improvements or changes in equipment are not accidentally interpreted as changes in the river’s transport regime.
CHAPTER 11 OF 12
11. Compare reaches and conditions without assuming a universal speed
Travel time depends on the flow and storage conditions during the observation. A result at one discharge is not automatically transferable to a drought, flood or changed channel. Nor should peak travel velocity over a long reach be assumed equal to a locally measured surface velocity. The statistics have different spatial support and weighting. Link them through a justified model rather than equating their labels.
If a 3 km reach has a mass-weighted travel time of 100 minutes, a bulk ratio of length to time is 0.5 m/s. That ratio is a summary of the observed transit under its definitions. It does not establish that every parcel moves at 0.5 m/s, or that doubling discharge doubles the ratio. Changing storage access and flow paths can alter the relationship.
For transfer, repeat the conceptual calculation after adding a connected slow side pool. The leading edge may change little while the recovered-mass mean shifts later. A single multiplier applied to every timing statistic would miss that difference. Choose the output appropriate to the question and preserve the flow conditions under which it was obtained.
CHAPTER 12 OF 12
12. Report a transport result without promising contaminant behaviour
A dye can inform aspects of dissolved transport, but another substance may sorb, react, settle or degrade differently. A tracer travel time is not automatically the arrival history of every pollutant. For decisions involving a real release, use the substance-specific properties, current flow conditions and qualified response framework. An educational curve cannot certify water safety or prescribe an emergency action.
A reproducible report includes input history, sampling sections, time references, calibration, background, discharge, quality flags, integration rule, detection limits and observation window. Distinguish observed statistics from model-fitted parameters and explain uncertain recovery. State which part of the curve supports each inference. This makes the result useful for later analysis without overstating what was directly measured.
The independent transfer task changes input duration, discharge, tracer retention and reporting limit separately. Explain which observed timing statistic can change under each modification and why. If the same explanation is used for every curve change, the interpretation is collapsing distinct mechanisms. A good transport account keeps boundary conditions, water movement, tracer behaviour and measurement response separate long enough to test them.
Independent worked case and evidence checks
Work an original fictional record with all tracer masses expressed in arbitrary consistent units. The introduced mass is 200 units at a defined reference time. Downstream analysis divides the corrected record into four intervals represented by 10, 20, 40 and 70 minutes. The corresponding recovered masses are 20, 80, 50 and 10 units. Assume these interval masses already include appropriate discharge and time integration.
Total recovered mass is 160 units and observed recovery fraction is 160/200 = 0.80. The recovered-mass mean time is (20 × 10 + 80 × 20 + 50 × 40 + 10 × 70)/160 = 4,500/160 = 28.125 minutes. The largest mass contribution is at 20 minutes, but the mean is later. Neither statistic is the first detected arrival, which needs its own observation rule.
Suppose first accepted detection was at 6 minutes and the concentration maximum was observed at 18 minutes. These results are compatible with the four-interval summary. First detection, concentration peak and recovered-mass mean are 6, 18 and 28.125 minutes. Listing them together is clearer than selecting one and calling it the travel time without qualification.
Now end observation before the last interval. Recovered mass becomes 150 units and the numerator of the mean becomes 3,800 mass-minutes. The observed-window mean is about 25.333 minutes and recovery is 75%. The apparent faster mean comes from losing late observations. It does not establish a change in the reach between the two calculations.
Consider an additional 20 units eventually recovered around 100 minutes. Total observed recovery would become 180/200 = 90%, and the new mean would be (4,500 + 2,000)/180 ≈ 36.111 minutes. A relatively small late contribution shifts the mean materially. The calculation demonstrates sensitivity to window closure; it does not justify inventing that contribution when no evidence supports it.
Check the concentration-to-mass conversion independently. At 1.5 micrograms per litre and 4 m³/s discharge, tracer mass flux is 6 mg/s. Over a 300 s interval with those constant values, mass passage is 1,800 mg. If concentration halves while discharge doubles, the product remains 6 mg/s. A lower concentration curve is therefore not sufficient evidence for lower recovered mass.
Next use a simple trapezoidal flux record: 0, 8, 4 and 0 mg/s at 0, 60, 180 and 300 s. Interval masses are 240, 720 and 240 mg, giving 1,200 mg. The time steps are unequal. Using a simple average of the four flux values multiplied by total elapsed time would give a different result and would not represent the stated trapezoidal integration.
An uncertain background contributes 0.2 micrograms per litre throughout a 20-minute window at 2 m³/s. The corresponding mass is 0.2 mg/m³ × 2 m³/s × 1,200 s = 480 mg. A small background offset integrated over enough water and time can materially affect recovery. This calculation is why a low tail needs careful baseline treatment, not merely more decimal places.
For a conditional retention scenario, assume two conservative water pathways: 70% of tracer mass exits around 15 minutes and 30% around 60 minutes. Their mean is 28.5 minutes. If a particular dye’s late-path observed mass is reduced by half through unresolved nonconservative behaviour, the observed weights become 70 and 15, giving a mean of about 22.941 minutes. The water pathways did not become faster.
The missing late dye in that conditional example could not be diagnosed from the altered mean alone. Independent tracer behaviour evidence is needed. The numerical example only shows how selective loss can change the weighting. It does not establish that every dye loss occurs in slow pathways or supply a field correction factor for rhodamine WT.
Now change the input history rather than the tracer. A short upstream pulse is replaced by an input spread across 20 minutes, with the same total mass. Downstream passage generally includes this input-duration contribution as well as reach transport. To compare the reach itself, use the measured boundary history and a suitable transport model. Comparing only downstream tail length would confound the two.
For a detection exercise, the leading signal crosses a low reporting threshold at 5 minutes and a higher threshold at 8 minutes. Both are first-detection times under their respective rules. Neither is an exact observation of the fastest water parcel. Report the thresholds and background uncertainty before concluding that one event had a genuinely later onset than another.
A 2.4 km reach and the original 28.125-minute recovered-mass mean give a length-to-time summary of approximately 1.422 m/s. This is not a surface velocity measurement. Its meaning inherits the mass weighting, observation window and tracer assumptions. An image-based surface estimate elsewhere in the reach can be compared only with those distinctions retained.
Finally imagine that an emergency planner wants to substitute the dye curve for a substance that strongly reacts with bed material. The transport evidence can inform the pathway model, but the substance’s concentration history needs additional properties and current conditions. The correct conclusion is not a reassuring single arrival time. It is an explicit account of which physical transport information transfers and which chemical behaviour remains unresolved.
Your independent solution should recover the original 80% recovery, 28.125-minute mean, the earlier 25.333-minute truncated-window mean and the 480 mg baseline contribution in the separate unit exercise. Then explain why each result can be calculated correctly while a larger interpretation still fails. That distinction turns a breakthrough curve into usable evidence rather than a visually persuasive story.
Worked comparison
| Statistic | Definition | Important limit |
|---|---|---|
| First detected arrival | Earliest accepted signal | Depends on threshold and sampling |
| Peak time | Maximum measured concentration time | Not the mean passage time |
| Recovered-mass mean | Time weighted by observed mass flux | Sensitive to missing or altered tail |
| Recovery fraction | Recovered divided by relevant introduced mass | Does not identify missing pathways |
| Fitted storage parameter | Model-dependent explanation of the curve | Not a directly mapped underground volume |
Questions and practical checks
Does the dye peak mark arrival of all water?
No. It is one feature of the concentration record. First detection, peak time and recovered-mass mean describe different parts of passage.
Is missing dye necessarily destroyed?
No. Unsampled storage, alternate pathways, tracer interactions and measurement errors can all affect apparent recovery.
Can a long rhodamine tail quantify hyporheic storage?
Not safely from the tail alone. Tracer-specific nonconservative behaviour can confound that inference; the linked USGS review provides the specific caution.
Why multiply concentration by discharge?
Their product represents tracer mass flux. When discharge varies, concentration alone does not give the mass weighting needed for recovery or mean passage time.
Can I carry out a dye release from this guide?
No. This is interpretation of fictional data, not a release protocol or authorisation. A real study requires qualified planning and the applicable permissions.
Continue with the connected subjects
Keep transport, discharge and exchange distinct
Tracer-dilution streamflow and mixing checks
Reservoir residence times and the difference between short-circuiting and long tails
Groundwater–surface-water exchange and independent heat-tracing evidence
Separate physical transport from reaction
Reactive tracers: separate residence times, reference recovery and reaction attribution
Sources and scope
USGS: Measurement of time of travel in streams by dye tracing
Runkel (2015): rhodamine WT, nonconservative behaviour and limitations of tail interpretation
USGS field study: travel-time definitions and incomplete observed dye recovery
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 30 October 2026.