Crown-of-Thorns Starfish Survey Counts and Outbreak Bands

A clear answer before the technical detail

Crown-of-thorns starfish manta-tow surveys describe animals observed during a standardised search, not the exact number of animals living on an entire reef. To read the evidence, identify the number seen, the number of valid observation tows and the survey protocol. Divide sightings by valid tows to obtain starfish per tow; only then compare that number with the outbreak bands used by the monitoring organisation. A raw sighting total without effort is difficult to interpret because a team searching sixty times has had more opportunities to see animals than a team searching six times.

The Australian Institute of Marine Science, or AIMS, describes two-minute manta-tow observations from a snorkeller moving around a reef slope. Its current crown-of-thorns starfish outbreak categories are based on mean sightings per tow: below 0.1 means no recent outbreak, 0.1 up to below 0.22 indicates potential, 0.22 through 1 indicates established, and above 1 indicates severe. These are terms belonging to a particular observational method, not global biological facts and not a guarantee that hidden animals or other habitat zones were absent.

The practical reading sequence is: check reef identity and survey date; find the exact observation unit; confirm valid effort; calculate a raw-count-to-effort ratio; use the underlying unrounded value to assign the stated category; then examine visibility, coverage, spatial clustering and comparison with previous surveys. Keep the numerator, denominator and scope visible in the final explanation. A reef status label never replaces the evidence supporting it. The worked examples below are invented learning datasets, clearly separated from primary monitoring results, so their arithmetic can be checked without mistaking them for field observations.

What a standardised search records

A manta tow can efficiently survey a considerable portion of a reef slope while a snorkelling observer records starfish and other reef features. It does not require the observer to inspect every crevice or turn every coral plate. This makes it valuable for broad surveillance but imperfect for hidden adults and small cryptic juveniles. The method’s strength is comparable repeated observation under defined conditions, not omniscience. If someone translates ‘surveyed’ into ‘counted every animal’, the sentence has made the evidence stronger than the method can support.

During a standard two-minute tow, the unit is one survey pass. Even though the duration is known, observations per tow are not automatically interchangeable with observations per diver-hour or animals per square metre. The travel path, swimming perspective, observer skill and portion of habitat seen all affect what the tow measures. Arithmetic conversion of minutes does not turn a visual index into a calibrated population-density estimate. When a scientific table says ‘COTS per tow’, retain that unit. Explain the abbreviation COTS once as crown-of-thorns starfish and use it consistently thereafter.

The starfish index and live hard-coral cover are different variables, even if collected during the same trip. Coral cover describes the proportion of a visually assessed habitat surface occupied by living hard corals under the protocol; starfish per tow describes the number observed within survey effort. Coral cover may decline due to heat stress, cyclones, disease or predation, among other processes. It is therefore wrong to state that a coral-cover percentage was ’caused by’ the starfish count without independent evidence. Report each measurement separately, then discuss possible associations with appropriate caution.

Sampling design determines how widely a finding can be generalised. Tow observations concentrated on a reef slope cannot automatically describe a lagoon or a deep offshore wall. An entire perimeter search is different from ten passes at a hotspot. Check the definition of the survey area, tow placement, weather windows and excluded locations. A sample drawn for rapid surveillance of a whole reef is not necessarily designed to estimate local density within a particular cove. Responsible explanation identifies precisely which part of the ecosystem the reported metric represents.

Calculate the ratio before interpreting a label

In an invented classroom example, twenty crown-of-thorns starfish are recorded across fifty valid tows. The average is 20 divided by 50, or 0.40 starfish per tow. Under AIMS terminology, this falls within the established-outbreak band. A careful report says that observers saw an average of 0.40 starfish per tow at the surveyed reef during the stated monitoring visit. It does not say the reef contains 0.40 animals, that forty percent of tows necessarily contained an animal, or that the population was exactly forty percent of some earlier value.

The proportion of tows with a sighting is not the same as the mean number of starfish per tow. Twenty animals might have been concentrated in five tows, leaving forty-five tows with no sightings. That would mean ten percent of tow units had sightings but the mean would still be 0.40. If each of twenty tows contained one animal, the sighting frequency would be forty percent instead. Both patterns have the same count and effort denominator, but they imply different spatial clustering. A manager investigating local hotspots needs the per-tow records, not only the overall mean.

Now compare two fictional reefs. Reef A records eight animals across twenty tows, giving 0.40 per tow. Reef B records twelve animals across sixty tows, giving 0.20 per tow. Reef B has the larger raw count, but Reef A has the larger observation index. In AIMS terms, A would be established and B potential. The misleading headline ‘B has more starfish because twelve exceeds eight’ ignores search effort. The repaired sentence states both ratios and notes that the comparison assumes comparable survey methods and sighting conditions. Ratios do not erase every ecological difference.

Do not overtrust a ratio based on tiny samples. If one animal appears during two tows, the arithmetic yields 0.50 per tow and a nominal established-band value. But two standard tows cannot support the same spatial confidence as many tows distributed around a reef. State the sample size and ask whether the survey met the protocol’s minimum coverage. The correct response is not to secretly replace 0.50 with an invented lower value, but to give its true numerical result together with the limitation of the evidence.

Missing or aborted effort also changes the calculation. Suppose twenty-five passes were attempted but two were aborted and excluded from valid observations, while seven animals were recorded in the remaining twenty-three. The index is 7 divided by 23, approximately 0.304 per valid tow, not 7 divided by 25. A proper field log must clarify whether observations from an aborted tow were retained. If the numerator includes encounters during invalid passes but the denominator excludes their effort, the ratio is biased upwards. Audit both sides of the fraction together.

Thresholds, labels and rounding

Current AIMS monitoring names four outbreak categories. A mean below 0.10 is ‘no recent outbreak’; 0.10 to below 0.22 is ‘potential outbreak’; 0.22 to 1.00 is ‘established outbreak’; above 1.00 is ‘severe outbreak’. The important point is that these are explicitly defined numeric bands on an observational index. They are not a legal designation for every jurisdiction, nor a universal instruction to begin an intervention. A management team may combine the band with ecological, logistical and risk information before acting. Read the methods page associated with the dataset rather than guessing from a colour legend alone.

To classify boundary cases correctly, calculate first and round afterwards. Eleven sightings across fifty tows give exactly 0.22, the beginning of the established band. Ten sightings across fifty give 0.20 and remain potential. A computed result of 0.219 may print as 0.22 when rounded to two decimal places, but its classification should still follow the unrounded ratio. If a dashboard displays too few digits to justify its label, seek the underlying numerator and denominator. Precision in formatting is no substitute for precision in reasoning.

At the other end, sixty sightings over sixty tows give exactly 1.00 per tow, while sixty-one over sixty give approximately 1.0167. Under the stated AIMS threshold, the latter is severe and the former remains established. If both are rounded to one decimal place, a table may show 1.0 for each, obscuring the distinction. Keep enough digits or the original counts to let a reader reproduce the category. The existence of a sharp administrative boundary does not imply nature changes suddenly when one additional animal is sighted.

Terminology may change across editions. AIMS explains that in 2024 it updated crown-of-thorns category names to harmonise with those used by the Great Barrier Reef Marine Park Authority. Historical reports may call a band ‘incipient’ where newer material calls it ‘established’, or use ‘active’ in the place now occupied by ‘severe’. A translator should not silently swap old words inside an archival quote. State the source-era label, numeric boundary and modern equivalent when relevant. That keeps a change in vocabulary separate from a change in population evidence.

Observability and uncertainty

Crown-of-thorns starfish often shelter beneath coral ledges or within reef architecture, so visual surveys have detection limits. Visibility, daylight, observer experience, animal size and habitat complexity can alter what is seen. A low survey index is a low index of observations under the chosen method, not proof that no animals exist in any unsurveyed habitat. If an observer records turbidity or an incomplete route, retain that limitation in the summary. Do not multiply the recorded index by an arbitrary ‘hidden animal’ correction factor unless a validated model and data support it.

Survey uncertainty is not captured by printing extra decimal places. A mean of 0.30 can arise from many different sets of per-tow counts; clustered observations and dispersed observations will not have the same variance. Without underlying data and a justified sampling model, a confidence interval cannot honestly be manufactured. Distinguish numerical exactness, such as an integer divided by the number of valid tows, from statistical precision and ecological representativeness. The best sentence may supply a perfectly checked ratio while explicitly saying the distribution and detection uncertainty remain unknown.

Survey time is another denominator. A web dashboard may display today’s access date, the most recent survey date and earlier years of observations in a plot. Do not turn a page timestamp into the date on which the animals were counted. When comparing years, align each reef’s monitoring date, number of tows, method and conditions. Missing years are gaps in observation, not evidence of an unchanging population. Where a storm or bleaching event occurred between surveys, preserve its timing without automatically attributing a subsequent count change to that single disturbance.

Readers can safely use the number as a surveillance index if they preserve its limitations. A first-pass summary should carry the observed total, valid tows, ratio, category and survey date. A second sentence can note whether effort was representative, whether visibility was constrained and whether a detailed dive investigation followed. A third sentence can separate the starfish indicator from coral condition. This architecture is useful in English or another language because the quantities and relationships remain unchanged when the vocabulary changes.

Five fully worked survey readings

Training example one begins with a fictional reef where four animals are counted in forty valid tows. The calculated mean is 4 ÷ 40 = 0.10 per tow, placing the record at the potential-outbreak threshold under AIMS terminology. A misleading headline would say ‘Only four were seen, so there is no outbreak risk.’ The sound explanation retains the denominator, unit and category: four COTS observed during forty standard tows, equivalent to 0.10 per tow and a potential-outbreak index. It also makes clear that this is a measurement of sightings during the field visit, not a verified total population or a prognosis.

Training example two uses a fictional report listing nine animals across thirty standard tows and the comment ‘one starfish in thirty percent of tows’. The average is indeed 9 ÷ 30 = 0.30 COTS per tow, an established-band value. But the claim about thirty percent of tows cannot be checked. Nine animals could all have been recorded during three tows, or one each on nine different tows. Ask for a tow-by-tow record if encounter frequency matters. This example shows why a mean count and the fraction of survey units with any detection are different statistics even when someone uses the word ‘frequency’ for both.

Training example three compares a fictional survey of sixty valid tows with twelve sightings to another of forty valid tows with sixteen sightings. The indices are 0.20 and 0.40, respectively, even though the count difference is only four. It would be correct to say the second survey had twice the starfish sightings per standard tow, assuming identical protocols. It would not be correct to say twice as many starfish inhabited the entire reef, because detectability and spatial distribution are unmeasured. State dates and methods next to the ratios; if the surveys covered different reef sectors, avoid claiming a population doubling at all.

Training example four is designed to expose rounding problems. A fictional dataset contains thirteen starfish observations over sixty standard tows, giving 0.216666… per tow. Rounded to two decimal places, it becomes 0.22, but the unrounded mean is below 0.22 and thus belongs to the potential band. A second dataset with eleven sightings in fifty tows equals 0.220 exactly and belongs to the established band. When a figure caption shows both as 0.22, the labels may differ for a legitimate arithmetic reason. Recover raw counts before accusing the source of inconsistency; then explain the precision issue directly.

Training example five considers 37 sightings in thirty valid tows, a ratio of about 1.233 per tow. That sits in the severe band under the stated AIMS categories. Do not write that the reef has 1.233 animals per tow ‘at every location’; a mean can contain zeros and clusters. Suppose ten tows recorded zero, ten recorded one, and ten recorded 2.7 on average: that latter phrasing already shows how an average may conceal an integer-count distribution. The correct record needs whole-animal counts for each pass. An overall mean can be useful for monitoring classification while being insufficient for deciding exactly where divers should search.

Reconstruct an honest trend rather than a dramatic story

Consider a fictional two-year comparison. In year one, 36 starfish were observed across forty valid tows, giving 0.90 per tow. In year two, twelve were seen across forty valid tows, giving 0.30. The index difference is 0.60 sightings per tow, and the proportional change relative to year one is (0.90 − 0.30) ÷ 0.90 = two-thirds, or about 66.7 percent. These are changes in observed average sightings under the survey method. They do not demonstrate that exactly two-thirds of the animals died, moved away or were removed. No individual fates were followed in the arithmetic.

Suppose those same fictional visits straddled a culling operation. The decline may be consistent with successful suppression of visible adults, but its attribution requires additional evidence. Dates of intervention, numbers recorded as culled, coverage of treated habitat, repeated observation, weather conditions and untreated comparison reefs can all matter. If neighbouring reefs also experienced an unrelated disturbance, a simple before-and-after graph cannot isolate the effect of culling. State the observed index decline first and describe the management activity separately. Stronger causal language requires a design capable of distinguishing intervention effects from natural changes and changing detectability.

Another hypothetical reef has exactly twenty sightings on both visits, but the first used twenty tows and the second used one hundred. The means change from 1.00 to 0.20 per tow. A superficial report says ‘population unchanged’ because the totals are the same; an equally superficial report says ‘an eighty-percent collapse in population’. Neither follows from the available evidence. The observation rate declined under a larger effort denominator. Before drawing ecological conclusions, investigate why the later survey included more tows, whether the route expanded beyond a hotspot, and how closely the protocols and sighting conditions matched.

Index comparisons require special care when reefs are grouped. Suppose five reefs have many tows each and one smaller reef has only a few. Averaging six reef means without weights answers a different question from dividing pooled sightings by pooled tow counts. Neither computation is universally right; they estimate different quantities. A reef-weighted result represents the average of reef-level means, while the pooled ratio represents average sightings per observed tow across all included effort. Read the monitoring organisation’s aggregation method before translating a regional dashboard. Do not combine a regional status band with a single reef’s raw count.

If a chart includes a category change, describe which numeric values drove it. For example, a move from 0.19 to 0.25 per tow crosses the 0.22 threshold from potential to established. The absolute change is 0.06 per tow and the relative change is approximately 31.6 percent of the earlier index. This is a modest numerical difference at a formal boundary, not an automatic transition from harmless to disastrous. The category is useful as a decision shorthand but should be accompanied by the actual values and their context so readers understand what changed and what did not.

The problems hidden by maps, photographs and dashboards

Maps require a question about scale. A coloured point may mark the centre of a reef, the location of a specific tow, or a summary statistic for an entire administrative region. If a red dot means the whole reef’s severe index, it does not locate every observed starfish. If the colour means a single tow’s count, it should not be applied to every slope around the reef. Find the legend, geographic definition and aggregation level. Translate spatial labels such as ‘reef’, ‘sector’, ‘zone’ and ‘site’ according to the original source rather than treating them as interchangeable pieces of coastline.

Photographs add confidence that an organism was present in a photographed spot, but they cannot substitute for a documented sampling design. A photograph of many animals on one coral head may be striking, yet it says nothing about the total valid tows around the reef. A photograph of apparently clean coral cannot prove there are no cryptic starfish beneath ledges. Attach imagery to the claim it actually supports: appearance, behaviour or a documented observation. Do not use a close-up to illustrate a numeric average unless the caption makes clear the image is descriptive rather than the denominator for the calculation.

A dashboard may show coral-cover history beside the COTS index. A fall in coral cover at the same time as high starfish observations is compatible with predation contributing to coral loss, but other drivers need consideration. Thermal bleaching, cyclones, disease, sediment impacts and sampling variation can overlap. The correct interpretation tracks each signal independently and looks for mechanistic or controlled evidence before assigning causality. A data plot’s proximity is not causal proof. In translation, terms such as ‘associated with’, ‘consistent with’, ‘following’ and ’caused by’ express distinct strengths of inference and should never be swapped casually.

Historical series can contain missing periods. A blank interval in a dashboard usually means no comparable observation or no reported value, not that the index remained at zero. Revisions can also change category names without changing previous recorded counts. The reader should note the source date, latest survey date and any methods update. A table combining older ‘incipient’ labels and newer ‘established’ labels needs a legend explaining the transition. This is particularly important when someone searches decades of monitoring data and tries to compare the same reef before and after definitions were updated.

One monitoring unit might conduct manta tows for surveillance while another deploys divers for detailed search swims. Both count starfish, but their denominators differ. A diver can inspect hidden sites that a towed observer cannot easily examine, and detailed diver effort often varies by habitat and visibility. The correct account might say that surveillance indicated an established outbreak, after which divers conducted targeted searches and removals. Do not pool animals sighted by different methods and divide by only the manta-tow count. The resulting number would be mathematically valid-looking but methodologically incoherent.

Translate technical language into usable decisions

Some reports use the word ‘density’ for the COTS-per-tow indicator. In ordinary language density often suggests animals per square metre or hectare, but here the practical unit may remain sightings per tow. Preserve that actual unit and define it before adding any explanatory synonym. When translating into another language, a sentence like ‘average visible starfish per standard survey pass’ is usually safer than a term readers will mistake for physical area density. If the source also provides area-based estimates, show them separately along with their sampling or calibration methods.

The abbreviation COTS should be defined once and not silently interpreted as ‘catches’ or harvested animals. A manta-tow survey records sightings, whereas control-program reports may separately list animals culled. Those numbers answer different questions. For example, ‘fifty observed over forty tows’ is an observation index; ‘fifty culled during ten diver-hours’ is an operational removal tally with different effort. Counting an animal once when sighted and again when removed would double-count across datasets if the two sources were simply summed. Keep detection, removal and estimated population size as distinct nouns.

An outbreak band is not an unconditional instruction to intervene. Managers may prioritise reefs based on ecological value, outbreak trajectory, connectivity, coral communities, dive access and available resources. A threshold can identify a concern requiring examination, but it is not proof that a particular action is safe or effective at every site. Explain the band, then distinguish the technical classification from the later management decision. A reader who sees ‘severe’ should understand why the index was labelled that way without believing the word, on its own, contains an approved treatment plan.

Public communication needs accurate hedging rather than evasive hedging. ‘Observed an average of 0.30 per tow’ is specific and auditable. ‘The reef may perhaps contain something’ is too vague to help. ‘There are exactly 0.30 starfish in every two-minute area of the reef’ is falsely precise. Pair a well-supported fact with a clearly named limitation. If you know visibility was low, say so. If data are insufficient to assess a cause, say causal attribution was not established. Honesty consists of preserving the right amount of certainty for each statement, not weakening every sentence equally.

Comparing survey records from different languages requires also preserving numerical conventions. Some languages use a decimal comma; 0.22 may appear as 0,22. A thousands separator or spreadsheet delimiter can change how these numbers are parsed. Keep the original numerator and denominator in a validation table and check that 11 divided by 50 still equals the translated boundary value. Keep the slash in ‘COTS/tow’ intelligible: it means per, not a ratio of two distinct species. If a translated infographic contains labels without units, correct the infographic before distribution.

Independent exercises, answers and repairs

Exercise one: an invented survey records eight sightings across eighty valid tows. The ratio is 8 ÷ 80 = 0.10 per tow, which sits at the beginning of the potential band. If the report says ‘no recent outbreak because the number is small’, it has ignored the formal boundary. If it says ‘eight percent of tows contained a starfish’, it has confused animal count with tow-level presence and may also have the arithmetic wrong. To repair the statement, report the index, band and the fact that the per-tow distribution has not been supplied.

Exercise two: a fictional log has thirty completed tows, five animals observed, three additional attempted tows that were cancelled before any recorded observations, and a dashboard showing 0.17. Five divided by thirty is approximately 0.1667, so 0.17 is a plausible rounded display and the potential category fits. If someone divides by thirty-three, they get approximately 0.1515, which happens to remain in the same band but uses an invalid denominator. The correct result is more than a label: it is a reproducible chain from valid search effort to classification.

Exercise three: a team reports two reef years, 0.09 per tow then 0.12 per tow, but the earlier work comprised one hundred standard passes while the later work comprised ten. A category change from no recent to potential is arithmetically compatible with the reported means, yet the very different effort requires caution. Ask for raw counts, survey coverage and field notes; note that a 0.12 average cannot come from an integer animal count over exactly ten standard tows, so perhaps the displayed values are aggregated or modelled differently. This small numerical inconsistency is a signal to examine how the report calculated its estimate.

Exercise four: one analyst translates 1.00 per tow as severe and another uses established. Under the current AIMS description, severe is strictly above one; an exact 1.00 belongs in established. If the original source used another historical or local system, establish that system’s boundaries rather than forcing AIMS terminology onto it. The model answer must cite which rules it applies. There is no value in a correct-looking label when the wrong protocol was used. An expert reader wants the unit, threshold and version, not confidence in an unsupported guess.

Exercise five: a dive team removes eighty animals, and a separate manta-tow record contains a mean of 0.50 over forty tows. Is it reasonable to say one hundred starfish were recorded at the reef by adding eighty removals and 0.50 times forty? The arithmetic 0.50 × 40 = 20 visible sightings is correct, but the datasets may include the same individuals. The total one hundred would be an unsupported combination. Report twenty survey sightings and eighty recorded removals in distinct fields and explain whether the sampling and removal windows overlap.

Exercise six: an invented chart indicates 32 percent hard-coral cover and 0.6 COTS per tow. Can a reader infer that 0.6 percent of coral was consumed by starfish? No. One measurement is a percentage of assessed reef surface; the other is observed animals per search pass. The denominators have different dimensions, and no conversion factor links them. The repaired statement names two separate observations and, if justified, says monitoring teams would evaluate whether predation contributed to observed coral condition. A percentage sign should never be added to an organism-per-effort index.

Exercise seven: an author writes that a very clear survey of sixty tows found no COTS, so the species was extirpated from the site. The correct conclusion is narrower: zero crown-of-thorns starfish were detected on the surveyed route during those sixty valid passes. Even clear conditions cannot ensure every hidden animal or unsurveyed area was observed. The report may reasonably call the measured index no recent outbreak under the category system, but a declaration of local extinction would require much stronger evidence over appropriate habitat and time.

Exercise eight: a fictional reef has twenty sightings in ten tows at a single hotspot, then twenty sightings in fifty broadly distributed tows a month later. The calculated indices are 2.00 and 0.40 per tow, but these observations do not establish that the whole reef’s population fell eighty percent. The spatial designs are different. A responsible analyst reports the numbers, notes that the first survey deliberately targeted a concentration and the second had broader coverage, and declines to calculate a population trend from incomparable sample frames. Such restraint prevents dramatic but unsupported reporting.

Frequently asked questions and closing guidance

Why not publish only the outbreak colour? Because different readers need to know what was counted, where it was counted and whether the label follows the documented rules. A colour without its numerator, denominator and date can conceal sparse effort, outdated terminology or a rounding boundary. A useful one-line summary can remain short: reef, date, sightings, valid standard tows, sightings per tow and category. Adding one sentence about visibility or coverage may prevent a much larger misunderstanding than the extra words cost.

Does a no-recent-outbreak classification guarantee no future outbreak? No. It describes observations in a particular monitoring round, not an ecological promise about future recruitment, migration or favourable conditions. Does a severe classification mean a reef has no living coral left? No. The starfish index does not directly measure total coral cover or the proportion already eaten. Can the index be calculated from a photograph or an eDNA test? Not without a validated method connecting that other evidence to the standard tow observations. Keep measures from different methods in separate fields.

The transferable skill is to preserve measurement structure when simplifying specialist prose: a count belongs to a search unit; a search unit belongs to a method; a category belongs to a threshold definition; and a trend belongs to two genuinely comparable observations. Anyone who can check these four relationships can read an unfamiliar monitoring dashboard far more reliably. For the next task, try taking a single public reef survey row, writing its numerator and denominator without copying the source’s conclusion, then verifying the label independently. That is translation as disciplined understanding, rather than merely replacing words.