Why a receiver detection is not a location fix
Fish acoustic telemetry can tell us when an electronic tag was detected by a receiver, but a row in the detection file is not a continuous record of the fish’s location. The receiver hears an acoustic message within an environment that changes with noise, depth, distance, obstacles and equipment condition. Translating detections into ecological movement claims therefore requires more than drawing lines between receiver positions. It requires an explicit account of the probability of hearing a tag that was present.
In this guide we will read fish acoustic telemetry data from first principles. We will distinguish a coded acoustic transmission from a decoded observation, calculate detection percentages with defensible denominators, inspect range-testing evidence, recognise false-positive and false-negative possibilities, and work through sample records before assigning habitat or passage interpretations. The numerical data in the exercises are invented learning examples, not actual movement tracks or field observations.
The first question to ask about any detection table is ‘what exactly was observed?’ Perhaps a stationary hydrophone decoded a transmitter’s identifier at 14:07:31. That observation says a compatible acoustic message reached that receiver and passed its decoding conditions at that time. It does not identify the fish’s precise coordinates, prove the animal was alive, or show where it went during the preceding silence. Those are separate hypotheses, supported by additional design or evidence.
Tags, transmissions, receivers and records
A transmitter produces coded acoustic messages according to its programmed schedule. Some tags transmit at a fixed interval; others use randomised delays designed to reduce repeated collisions among nearby tags. A fixed receiver listens and attempts to decode the messages, often logging an identifier, time and receiver details. Its data represent successful reception events, not every transmission the tag emitted.
A detected tag may be attached to a fish, but its state can change: a fish can move, the tag can be shed, the animal can die, or a detached transmitter can be carried elsewhere. Different study designs have different safeguards, including accelerometers, depth sensors, mortality indicators, recapture records or arrays of receivers. Never treat a decoded identifier as direct visual confirmation of a living fish without discussing those limitations.
Even at close range, not every transmitted ping is necessarily detected. Acoustic energy may be obstructed, masked by ambient noise, attenuated by water conditions or corrupted by collisions. The instrument can also be temporarily unavailable. The observation process is therefore conditional, and a good movement analysis records that condition instead of treating all non-detections as known absences.
The shape of a typical detection table
A basic table may contain receiver identifier, transmitter code, timestamp, station coordinates, deployment identifier and a quality or decoding field. Some systems also provide signal measurements, sensor payloads or environmental records. Tag metadata, including transmission interval and battery life, often live in a separate table. Without an accurate join between these tables, the apparent track can inherit the wrong fish identity or tag schedule.
Time fields deserve particular care. A timestamp may be stored in local time, UTC, receiver clock time or an adjusted laboratory time. Daylight-saving rules can affect some sites, while instruments can drift from an initial synchronisation. A five-minute clock mismatch may reverse apparent arrival order between receivers separated by a small distance. Translate the clock basis explicitly before interpreting transit time.
Coordinates identify the receiver deployment location, not the position of the fish at detection. If the acoustic detection radius is broad or strongly anisotropic, the fish could be in several possible locations when the record appears. Map labels should say ‘receiver station’ rather than ‘fish point’ unless a separate positioning method, such as a multilateration system with validated geometry, supplies an actual position estimate.
What detection probability means
Detection probability is the chance that a receiver or receiver array records an available tag under specified conditions and an observation rule. The denominator is crucial. If a known test tag transmitted one hundred times during a test, and seventy-five messages were correctly decoded, the observed per-transmission detection proportion is 75 divided by 100, or 0.75. This is a useful measurement of that tested circumstance, not a universal property of the receiver.
A second measure is the proportion of passing tags recognised as present during a defined window. A receiver might miss many individual pings but still register nearly all fish that pass through its useful detection area because each passing tag transmits repeatedly. This is why a paper may report modest per-ping probability yet high passage-level detection efficiency. Replacing one denominator with the other changes the meaning of the result.
A reported ‘90% detection rate’ should always say what the 90% refers to: transmissions, test trials, tagged animals, passes or receiver-hours. The estimate also needs its range of depths, environmental conditions, tag orientations and receiver settings. Without these details, the percentage looks portable but may fail outside the tested domain.
Detection range is a probability curve, not a perfect circle
The phrase ‘receiver range of 300 metres’ is often used as though a boundary is drawn on water. Real detection generally changes gradually with distance and with conditions. A tag may be heard at 350 metres on a quiet day and missed at 80 metres in a noisy or obstructed setting. A plotted range curve should relate distance to a measured detection metric with a clearly defined observation interval.
Bathymetry, substrate, vegetation, boats, bubbles, stratification, wind, wave state and ambient biological sound can alter the acoustic pathway. Tag orientation can also matter. The geometry of a harbour, river bend or reef can make reception highly directional. For that reason, a single radial range estimate is often too simple for a spatial model, especially when a study spans habitats with different sound propagation.
A well-designed range test places or moves a test transmitter at known distances and depths while recording reception. Repeated trials at different times and conditions reveal how stable the curve is. The resulting figures support a statement such as ‘detection probability exceeded 0.8 within the tested distance band under these conditions’, not ‘all fish inside this circle were recorded’.
A worked range test
Imagine a stationary test transmitter that sends fifty valid transmissions at each distance. At 50 metres, forty-eight are heard, giving 48/50 = 96%. At 150 metres, thirty-six are heard, or 72%. At 250 metres, eleven are heard, or 22%. The observed decline is consistent with decreasing reception as distance grows, but these three proportions are merely outcomes of the invented experiment.
Before using the numbers as a general curve, ask whether transmissions are independent, whether the test lasted long enough to span conditions, and whether tag orientation and water depth resemble real fish paths. A run of failed pings caused by one vessel passing can cluster failures, so the naïve binomial standard error may be overoptimistic. Repeated sessions help expose that temporal correlation.
An explanatory caption might say: ‘Percent of known test transmissions decoded at each stated distance during the range trial; uncertainty and conditions are shown separately.’ It should not call the x-axis fish displacement or the percentages species survival. The experiment measured the observation system, not the animals’ behaviour.
False negatives are not proof of absence
A false negative occurs when the tag is present in a relevant detection opportunity but no usable record is produced. This can happen because no ping was emitted during a brief passage, because the receiver was down, or because emitted signals were not decoded. Non-detection therefore does not alone establish that a fish avoided the area. The probability of a missed passage depends on the observation window and the tag’s transmission schedule.
If a fish spends only twenty seconds in a monitored corridor and its tag transmits approximately once every sixty seconds, it may pass without transmitting at all. More receivers cannot recover a message that was never emitted in the accessible interval. Passage models should consider tag delay distribution, speed, detection geometry and receiver uptime rather than assuming each crossing creates an equal opportunity to observe.
The converse caution is also important: silence does not prove mortality. A tag can be out of range, its battery can expire, the fish can move into shielding habitat, or a deployment can have data gaps. Ecological conclusions about disappearance need alternative explanations and, where possible, independent observations.
False positives and unexpected identifiers
A false positive is a record that appears to indicate a particular tag when that tag was not actually present. Coded telemetry systems can sometimes produce spurious decodes through collisions, noise or other processing artefacts, depending on their design. Other apparent anomalies come from data-handling mistakes, such as a code reused by another project, a receiver moved without updating station metadata, or a timestamp assigned to the wrong deployment.
USGS methodological literature warns that both false-positive and false-negative observations affect telemetry studies. Filters may consider implausible isolated detections, repeated confirmations, physical travel constraints, tag inventories and receiver quality. The exact filter should be justified by the system and study objectives. An overly permissive filter creates fictional movements; an overly strict filter deletes rare but real appearances.
Keep raw records available. A filtered analysis should identify the rules and the number of observations removed for each reason. If a detection was excluded because no neighbouring confirmation occurred, report that operational rule, not a claim that the animal definitely was absent. The rule controls error risk but cannot produce perfect knowledge.
Example: an isolated tag hit
Suppose receiver A logs one occurrence of tag 407 at 02:11. No other receivers log it within the surrounding three days, and the tag was expected hundreds of kilometres away. One interpretation is a genuine but rare long-distance movement; another is false decoding or a metadata error. The single event is insufficient to choose between those stories without more evidence.
Check whether code 407 was programmed into another local transmitter, whether the receiver clock is valid, whether the detection’s signal characteristics are plausible and whether neighbouring stations had operating coverage. Examine route feasibility using distance and elapsed time rather than imposing a universal maximum speed. An extremely implausible transition is a warning, not a mathematical proof of a bad record.
A clear report would say ‘one unconfirmed detection of this identifier was logged and is flagged for review under the study’s criteria’. If later evidence supports the movement, it can be restored with documentation. Translating the event as ‘the fish migrated to receiver A’ before that review would turn an uncertain observation into a categorical biological claim.
Receiver uptime and exposure time
A receiver’s operational history is part of the sampling design. Records should show when it was deployed, recovered, serviced, moved, disconnected or otherwise unavailable. Analyses of passage or detection opportunity must use actual monitoring periods. A station installed for one hundred days but recording successfully for only eighty should not be treated as having one hundred complete days of exposure.
Battery failure, memory saturation, environmental damage and clock resets can create gaps. Some gaps are documented directly, while others require inference from diagnostic logs or known sentinel detections. Distinguish a verified outage from an unexplained quiet period. The former justifies excluding effort; the latter may be biologically meaningful but remains ambiguous.
If receiver A operates for forty hours and receiver B for twenty, comparing their raw detection totals without adjusting for effort can be misleading. A detections-per-hour metric may help describe activity, yet it still mixes animal abundance, transmitter behaviour and reception probability. Effort standardisation is necessary but not sufficient to create a comparable animal-density estimate.
Arrays are not simply collections of independent stations
A receiver array can increase detection coverage and allow movement sequences to be inferred, but its stations often have overlapping fields. A fish may be recorded simultaneously by several receivers. Counting each record as a separate fish inflates abundance, and treating redundant receiver observations as independent detections exaggerates statistical evidence. The unit of inference must be defined explicitly.
An array across a river mouth may support a passage question: did a particular tag generate a qualifying detection at the entrance? A grid in an estuary may support residence patterns within broad areas. A precisely synchronised positioning array might estimate location through time differences of arrival. These are different designs with different requirements, and the term ‘acoustic tracking’ should not imply they offer equivalent spatial precision.
Receiver geometry also changes what a non-detection means. A narrow channel covered across its width may yield stronger passage evidence than a coastal bay monitored by sparse stations. Without discussing geometry and environmental range tests, an apparently neat track map can overstate sampling completeness.
Worked passage-level detection
Imagine twenty tags are independently known to have passed an array during a validation trial. Eighteen are detected according to a pre-defined rule of at least two accepted pings within the passage window. The observed passage-level efficiency is 18/20 = 90%. The denominator is known passages, not emitted pings and not all tags released into an entire river system.
Now suppose those twenty tags transmitted four hundred pings during their passes and the array decoded only one hundred and eighty. The per-transmission detection proportion is 180/400 = 45%. Both values can be correct because they answer different questions. The array detects most passing tags despite missing more than half their individual transmissions.
The translation should preserve both labels. ‘Ninety percent of passage events detected’ does not mean ‘ninety percent of signals received’. Whenever a report compares two systems, check that both use the same event definition, observation window and denominator before comparing percentages.
Tag schedules and collision risk
A tag’s nominal delay determines how often it transmits, but nominal frequency is not the same as a guaranteed inter-ping interval. Some devices randomise intervals deliberately. Battery life depends on transmission power, frequency and operating mode. Missing tag-programming metadata can make a silence appear mysterious when the fish simply traversed the receiver zone between scheduled transmissions.
When many tags transmit within a common reception field, overlapping messages can interfere. The resulting probability of successful decoding is therefore partly a function of tag density and timing. Range tests performed with one quiet transmitter may overestimate detection during dense releases. A study should record whether validation tested realistic code-collision conditions.
It is also possible for sensor payloads to modify transmission schedules or packet lengths. A data field that looks like a second tag identifier may actually encode temperature, depth or another measurement depending on the manufacturer. Translate the technology using the actual codebook. Never infer that every number in the row is a location or a count of animals.
Travel time between stations
Suppose a tag is first recorded at station A at 10:00 and later at station B at 12:30. The interval between those detections is two and a half hours. It does not prove the fish departed A at exactly 10:00 or arrived B exactly at 12:30. There may have been earlier missed pings, travel within broad detection fields, pauses, detours or clock offsets.
If station centres are five kilometres apart, dividing five by 2.5 gives a station-to-station apparent speed of 2 km h−1. That quantity is not necessarily the fish’s actual swimming speed because the true path can be longer than straight-line distance, and the time interval includes detection uncertainty and possible residence near a station. Label it ‘minimum centre-distance over first-detection interval’, not biological swim velocity.
For movement ecology, comparisons are stronger when the same definition is applied consistently across individuals and environmental conditions. Where a study needs speed, incorporate independent positional data, explicit path assumptions or a movement model rather than treating a pair of receiver timestamps as a complete continuous route.
A filter that respects uncertainty
A useful quality-control workflow can assign each observation to accepted, suspect and rejected categories. Start with receiver deployment windows, valid tag inventory, clock quality and documented sensor faults. Then evaluate isolated detections, plausible movement times and nearby confirmations. Record whether suspected errors are due to acoustics, timing, metadata or data processing. Each reason is an auditable diagnosis.
Do not tune filters until the desired biological outcome appears. A filter selected because it maximises passage estimates or creates a tidy map can introduce bias. Where possible, develop rules from known-control tags, controlled experiments and simulation. Test how the ecological conclusion changes under stricter or more permissive assumptions; report the sensitivity if it materially affects a result.
A good filter can increase trust in retained detections while losing some true events. That trade-off should be explained. If the aim is a rare migration event, discarding all single detections may systematically remove genuine rare observations. If the aim is reliable survival estimation, false positives may be particularly damaging. The right rule depends on consequences and validation evidence.
Avoid mixing detection probability with survival
Survival, migration and detection are distinct probabilities. A fish not detected downstream could have died, chosen a different route, lost its tag, passed unheard or entered unmonitored habitat. Statistical models such as mark-recapture frameworks can attempt to separate survival and detection, but doing so requires suitable release groups, station design and model assumptions. A simple ratio of downstream detections to upstream releases is not automatically survival.
If one hundred tagged fish are released and sixty are subsequently detected, ‘60% detected downstream’ is a valid descriptive statement given the design and exclusions. Calling it ‘60% survival’ assumes detection and other fates are perfectly known. With a verified passage detection probability below one, a model may estimate a different passage quantity, but its confidence interval and assumptions must be reported.
Before interpreting an estimate, ask whether the denominator is tagged fish released, fish known alive at an upstream site, fish available to pass a station, or fish with functioning tags. Changing denominators changes biological meaning. Translation should not replace these nuanced quantities with the vague word ‘success’.
The uncertainty of a proportion
For a simple illustrative binomial model, where each of n independent opportunities has the same detection probability p, the estimated standard error of a proportion is the square root of p(1−p)/n. If 80 of 100 known test transmissions are decoded, p-hat is 0.80 and the approximate standard error is sqrt(0.8×0.2/100) = 0.04. That does not automatically imply a 4% uncertainty for every real deployment.
Acoustic trials often violate independence because waves, vessel passages and noise can cluster failures. Detection probability also varies with distance, depth and tag orientation. A pooled standard error can conceal these differences. Blocked trials, mixed models or stratified estimates may better describe uncertainty if the available design supports them. Present sample sizes and conditions, not only percentages.
Confidence intervals quantify sampling uncertainty under a chosen model. They do not include every systematic bias, missing deployment record or false-positive filter decision. Readers should distinguish uncertainty in the measured trial proportion from uncertainty in extrapolating that proportion to freely swimming fish.
What happens when receivers move
A receiver may be redeployed, dragged by currents or temporarily displaced. Station coordinates then need timestamps and deployment identifiers. If an analyst plots detections using one fixed coordinate for all years, the map can quietly move older records to a newer location. A movement inference based on such a map is a metadata error rather than a biological discovery.
A practical record associates each detection with the station’s location during its actual deployment. Recovery and servicing logs may document deviations, while anchor movements can remain unobserved. If the position is uncertain, represent that uncertainty and avoid precise travel distances. Do not use GPS coordinates from the equipment inventory as though they were exact positions at every moment.
A similar issue arises when a receiver identifier is reused or renamed. Reconcile deployment IDs before calculating occupancy and transition matrices. Biological conclusions depend on station provenance just as chemical conclusions depend on specimen provenance.
Designing a readable figure
A detection raster can place time on the horizontal axis, fish identifiers on rows and receiver detections as marks or colours. It is compact, but readers need to know whether each mark means one ping, a filtered detection window, or a summarised passage event. Aggregating to hour-long bins can make sparse events look continuous. Mark receiver outages and tag expiration rather than representing unknown exposure as silence.
A route network diagram should show receiver locations and monitored corridors, and it should visually distinguish confirmed detections from inferred paths. Connecting every station visited with a straight line can be a helpful schematic but must not imply a measured swimming route. Where movement models are used, present their uncertainty or plausible alternatives.
For cross-language reports, translate ‘detection’, ‘presence’, ‘passage’, ‘residence’, ‘range’, ‘sensitivity’ and ‘specificity’ consistently. These words are not synonyms. Keep time zone, unique tag code, unit of distance, filtering rule and observation interval beside the figure, where the reader can audit claims directly.
Transfer exercises and worked repairs
Exercise A. A range test yields 72 successful decodes from 90 known transmissions at 100 metres. The observed per-transmission detection proportion is 72/90 = 0.80, or 80%. This statistic is conditional on the tested distance, orientation, time and equipment. It cannot be applied automatically to all fish passages at other depths. A defensible caption explicitly identifies the ninety test transmissions as its denominator.
Exercise B. Twenty tagged fish are independently confirmed to traverse a monitored channel; nineteen generate at least one accepted receiver event. The passage-level detection efficiency is 19/20 = 95%. Do not compare it directly with a 60% per-ping decoding proportion without identifying different event units. One probability concerns individual passage recognition; the other concerns individual acoustic messages.
Exercise C. A fish passes a receiver’s useful field in thirty seconds and the transmitter may remain silent for up to seventy seconds. Non-detection is possible even when the fish passes close to the station. The correct diagnosis is a possible absence of transmission opportunity. Compare expected residence time with the tag’s actual delay distribution before assigning avoidance or mortality.
Exercise D. A report shows zero detections during a known receiver battery failure lasting twelve hours and calls the area empty. Remove that interval from verified monitoring effort, describe the outage, and avoid ecological inference from the resulting silence. A missing observation system cannot generate valid absence data, irrespective of how cleanly the timeline graphic displays the gap.
Exercise E. A test transmitter emits fifty messages at 50, 150 and 250 metres. The receiver decodes 48, 36 and 11. The percentages are 96%, 72% and 22%. This suggests declining reception with distance in the test. It does not define a unique cutoff radius; a detection model should account for conditions and uncertainty rather than turning the largest observed distance into a perfect boundary.
Exercise F. An isolated identifier appears at a station far from its previous accepted location, with no intervening confirmations. Classify it as needing review, not immediately as either a spectacular migration or confirmed acoustic error. Verify tag-code allocations, clock basis, receiver logs, physical travel constraints and surrounding detections before deciding how to handle it.
Exercise G. Two receivers share a coverage area and both record the same tag within one minute. This is likely two observations of one transmitter, not two fish. Deduplicate or model overlapping detection opportunities according to the research question. Preserve the original records because multiple receivers can provide useful confirmation and spatial constraints.
Exercise H. A data export contains local timestamps at one station and UTC timestamps at another. The analyst computes negative transit time and concludes that fish moved backward through time. Harmonise time zones and clock corrections first; only then compare event order. Document the conversion so a future analyst can reproduce the chronology.
Exercise I. One hundred tagged animals are released; sixty are detected downstream. The only unqualified fraction established is 60% of released tags detected under the chosen definition. Calling this a survival rate assumes complete observation and ignores movement, tag loss and mortality uncertainty. If a survival model is used, name its design and assumptions.
Exercise J. A receiver recorded 240 events over forty valid hours, while another recorded 150 over fifteen valid hours. The raw totals suggest the first is busier; effort-normalised event rates are 6 and 10 per hour. Neither rate alone is animal abundance because ping frequency, overlapping coverage and detection probability differ. The exercise diagnoses denominator choice without assigning ecology prematurely.
Exercise K. A tag has a theoretical battery duration of one year but is absent after month nine. Theoretical lifespan is not evidence that the transmitter was still working, and silence is not proof that the fish died. Check empirical battery performance, activation time, programming and independent detections. Describe fate uncertainty until stronger observations are available.
Exercise L. A report maps every reception as an exact fish coordinate at the receiver. Replace those points with receiver detections and indicate the likely acoustic coverage area where validated. A fixed hydrophone generally localises itself, not the animal. Exact positioning needs a specialised array, synchronisation and geometry tested for location error.
Exercise M. Four tags transmit in a confined basin, and the number of successful decodes drops when all are active. Signal collisions or noise could explain the reduction. Validate tag-density conditions before extrapolating a quiet single-tag range test. Avoid diagnosing fish departure from reduced decoding without examining the acoustic observation process.
Exercise N. Of 100 known test pings, 80 are decoded. With a simplified independent Bernoulli model, the approximate standard error of p-hat is 0.04. If transmissions cluster during high-noise intervals, the effective number of independent opportunities may be smaller. The formula demonstrates mathematical precision under assumptions, not a guarantee of field-level reliability.
Exercise O. A figure joins two stations five kilometres apart with a straight line and labels a 2.5-hour detection interval ‘2 km/h fish swimming speed’. The arithmetic 5/2.5 = 2 is correct but the quantity is an apparent station-centre travel rate. The actual animal path, entry and exit times, and intermediate behaviour are unknown.
Exercise P. A filter removes all detections that occur only once within a day. This may reduce false positives, but it can also erase genuine brief or peripheral visits. Test the filter against known controls, and report how counts or movement conclusions change. A strict rule is not automatically a scientifically superior rule.
Exercise Q. A report contains 3,000 ping records from twenty fish and says ‘sample size 3,000 fish’. Correct the unit of replication. It may have twenty tagged animals, multiple detection events per animal and repeated measurements per receiver. A model may need to accommodate within-fish dependence rather than treating every message as independent.
Exercise R. A station’s deployment coordinates changed midway through a study but its receiver identifier remained the same. The correct procedure is to join each record to the appropriate dated deployment location. Combining all records under the latest coordinate creates false movement distances and could change inferred river passage sequences.
Exercise S. A researcher observes no downstream detection and concludes the fish did not cross an array. With validated passage-level efficiency of 85% under comparable conditions, a missed crossing is possible. The observation should be reported as ‘not detected under available monitoring’; inference about crossing probability requires a model, and may also depend on route choice.
Exercise T. A translated report describes ‘95% detections’ without naming whether it counted pings or animals. Repair it by finding the source numerator and denominator, then writing the complete unit in the sentence. If the source does not supply those definitions, flag the ambiguity rather than inventing a precise-looking rate from context.
Frequently asked questions
Can a single ping confirm a fish was alive? It can confirm a decoded tag message under accepted conditions, not automatically the condition of the fish. Tags can be shed, and mortality or movement sensors require their own validation. Biological claims should use the design’s independent evidence rather than treating every coded transmission as direct observation of a living animal.
Why does a tag disappear and reappear? It may leave and return to coverage, experience noise or orientation changes, occupy sheltered water, or encounter a temporarily unavailable receiver. The pattern should be analysed with deployment records, range tests and tag programming before being assigned a behaviour label. Apparent absences sometimes reflect the observation process.
Is a receiver’s advertised range enough for publication? A manufacturer specification describes equipment capability under conditions that may not match the study. Local range testing and explicit environmental conditions give a stronger basis for movement inference. If range tests are unavailable, the limitation should be disclosed and uncertainty in coverage retained.
Should suspicious detections always be deleted? Not automatically. Retain raw records, define audit rules and keep a record of excluded observations. A suspicious event may be valid, and removing every isolated event can bias studies of rare travel. Filtering should match the scientific objective and be tested against known-control evidence where possible.
Further learning and connected methods
The nearby eduKateSG guide to ocean hydrophone soundscapes and calibrated acoustic measurements explains the physical language of acoustic signals. The guide to fish otolith element transects describes a different kind of movement evidence: retrospective chemical histories rather than electronic reception events. Use these complementary approaches without pretending that their uncertainties or sampling units are interchangeable.
Primary sources and validation references
Source 1. U.S. Geological Survey, Beeman and Perry (2012), ‘Bias from false-positive detections and strategies for their removal in studies using telemetry’, https://www.usgs.gov/publications/bias-false-positive-detections-and-strategies-their-removal-studies-using-telemetry . The method distinguishes false negatives, false positives and their analytical consequences.
Source 2. NOAA repository, study of detection probability and range in acoustic telemetry for juvenile salmon, https://repository.library.noaa.gov/view/noaa/68165 . The investigation demonstrates that tag orientation, direction and distance affect detection, and distinguishes ping reception from recognising passage.
Source 3. NOAA CalFishTrack real-time receiver information and detection-filter descriptions, https://oceanview.pfeg.noaa.gov/CalFishTrack/pageREAL.html . These filters are an example tied to a particular programme, not universal rules for all telemetry systems.
Source 4. Google Search Central, helpful reliable people-first content guidance, https://developers.google.com/search/docs/fundamentals/creating-helpful-content . This editorial source supports explicit method, accountability, uncertainty and useful explanations; it does not prescribe a ranking word count.