Mud crab nursery sheets often describe density, shelter and survival as if the three numbers tell one simple story: more crabs in a tank, some shelter supplied, then a final percentage recovered. That summary can hide the important questions. Was density measured per litre or per square metre? Was shelter measured as floor coverage, usable refuge or number of structures? Did missing crabs die, escape, hide or get removed? Translating the record accurately begins with separating those claims.
This guide explains how to read Scylla mud crab nursery data from the megalopa and early crablet phases, especially when growth, aggression, claw damage and cannibalism become important. It is not a prescription to use a universal stocking rate. Published experiments differ by crab species, size class, facility, nursery duration and habitat, and a real farm must apply its own validated animal-health and husbandry procedures.
Our worked figures are transparent teaching scenarios, not a disguised farm trial. The purpose is to turn vague terms such as “high density,” “better shelter,” “lost crabs” and “nursery survival” into statements that specify the animal population, physical unit, observation method and endpoint. Once those foundations are in place, a reader can interpret a result without attributing every missing crab to cannibalism.
Why nursery density has more than one denominator
Density is an amount of animals divided by a relevant quantity of space. In a tank containing free-swimming larvae, animals per litre can be useful because water volume describes the sampled environment. For benthic juveniles that settle, occupy surfaces and interact with one another, animals per square metre of usable bottom or settlement area may better describe encounter pressure. Neither denominator is automatically superior in all stages.
Suppose a nursery tank contains 2,000 megalopae in 1,000 L of water and has 5 square metres of bottom area. Its volume density is 2 megalopae/L and its bottom-area density is 400 megalopae/m². Those are two correct descriptions of the same tank, not competing estimates. A sheet that merely says “density 400” has omitted the unit required to interpret it.
The problem becomes more serious when two nursery designs have different depths. If tanks A and B both have 5 m² of floor but water depths of 20 and 50 centimetres, the nominal water volumes are 1,000 and 2,500 L. Stocking 2,000 animals in each gives the same floor-based density but different volume densities, 2/L versus 0.8/L. A comparison can be biologically meaningful only when it identifies the stage, habitat and denominator relevant to the question.
Calculate area correctly before dividing
For a rectangular tank, plan-view bottom area equals length times width in consistent units. A 2.5 m by 1.6 m tank has 4.0 m² of bottom area. Stocking 800 juvenile crabs makes the starting nominal density 800/4.0 = 200 animals/m². Dividing by volume instead would answer a different question and should not be silently substituted in a report.
An irregular pond needs a measured or mapped wetted area rather than a guess from fence length. A shallow shelf may be available at high tide and dry at low tide, so the effective habitat area changes with water level. A document that quotes pond area from a construction drawing should distinguish that design figure from actual accessible nursery habitat during the observation period.
Be precise about whether the floor-based unit is a gross tank footprint, water-covered footprint, or a sum of both floor and vertical structures. Adding the surface area of netting to bottom area without a clear protocol can make density appear much lower. If animals do not use a surface equally, its geometric area is not necessarily equivalent habitat. Explain the measurement convention so comparisons can be replicated.
Interpret shelter quantity without treating all structures as equal
Shelter can be netting, seaweed, tubes, partitions or other refuge structures. A shelter count is not a shelter area unless every item has a specified dimension. Twenty tubes and two square metres of fine mesh are not interchangeable quantities. In reports, “shelter coverage” may mean projected shadow on the floor, material area, percent covered by algae or the number of available hiding positions.
For example, six rectangular refuge panels each cover 0.25 m² of tank floor in plan view, and a tank has 5.0 m² of accessible floor. If panels do not overlap, their nominal projected coverage is 6 times 0.25 divided by 5.0, or 30%. This is a geometric figure, not proof that 30% of crabs were sheltered or that refuge quality is sufficient for all sizes.
Now suppose the panels overlap over 0.40 m². Adding panel areas would give 1.50 m², but the union of unique covered floor is 1.10 m², or 22%. A translation should specify whether coverage is summed material footprint or non-overlapping ground coverage. Otherwise two teams can report different percentages for the same physical setup without either having made an arithmetic error.
Shelter is a mechanism, not an outcome
A refuge may reduce direct encounters, interrupt line of sight, provide edges for settlement or help smaller crabs escape contact with larger ones. But geometry alone does not confirm the effect. A dense structure can also accumulate feed residues or restrict cleaning, and a shelter that helps one size class may be unsuitable for another. The report should preserve the difference between installed refuge and observed animal behaviour.
Primary research on mud crab nursery systems has found that stocking density and artificial habitat can influence survival, claw loss and yield, with effects varying by phase. For example, an intensive nursery study on Scylla serrata reported different density sensitivity across later juvenile stages and benefits from added net structure. This supports measuring these variables carefully, but it does not justify transplanting its exact settings into every nursery.
An honest translation might say “Refuge netting was installed at the beginning of the trial, and recovered C1 crabs were counted at the endpoint.” It should not say “The netting prevented cannibalism” unless the study design and evidence support that specific causal interpretation. A useful technical article teaches where observation ends and inference begins.
Recognise cannibalism without inventing a body count
Cannibalism is a real concern in mud crab culture, especially when animals of different stages or sizes encounter each other and during vulnerable moulting periods. However, disappearance from a tank is not a direct count of cannibalism. Crabs can escape, hide under structures, die from other causes or be lost through sampling and transfer. The final number recovered is not a complete cause-of-death record.
Suppose 1,000 juvenile crabs are stocked and 650 are found alive after two weeks. Apparent recovered survival is 65%, and unrecovered input is 350, or 35%. If 40 dead bodies are actually found and documented, the known observed dead fraction is 4% of input. The other 310 are unaccounted relative to those measured categories. It would be misleading to translate that 31% as “eaten by other crabs.”
What evidence can strengthen a cannibalism hypothesis? Direct observations or suitable video, injuries consistent with attacks, a patterned size-class disparity, carcass remains, controlled comparisons and a transparent search method may all contribute. Even then, a diagnosis must distinguish compatible signs from proof. Use words like “suspected,” “observed,” or “experimentally supported” according to the evidence, not according to how strongly a manager feels about the result.
Read size distributions before interpreting stocking density
Two tanks can have the same mean body mass and profoundly different size distributions. A tank with ten animals around 1.0 g each is not the same interaction environment as a tank with five at 0.2 g and five at 1.8 g, even though both have mean mass 1.0 g. Large individuals may prey on smaller ones or dominate refuges. The spread must accompany the mean when size disparity is relevant.
For a simple nursery sort, classify 300 recovered animals into 180 small, 90 medium and 30 large, using predefined carapace-width limits. The distribution is 60%, 30% and 10% of recovered measured animals. It is not a distribution of the original stocked population if mortality, escape or selective sampling has changed the group. The size boundaries and measurement tool should be reported alongside the categories.
A grading event changes the composition of each unit. If the large animals are moved to another tank, an improvement in small-crab survival cannot automatically be credited to added shelter that was installed on the same day. Multiple operational changes create confounding. Translate the actual timeline, including the size sorting, before suggesting what caused any change.
The difference between survival and production
Survival counts how many animals remain or are recovered relative to an appropriate starting population. Production can refer to a number of juveniles, their total biomass, or output per unit area and time. A nursery that produces fewer but larger crabs may have lower numerical survival but higher weight output. One metric does not replace the others.
Consider nursery A: 1,000 crabs stocked, 700 recovered, mean final weight 2 g. Recovered biomass is 1.4 kg and number-based recovery is 70%. Nursery B: 1,000 stocked, 600 recovered, mean final weight 3 g. Recovered biomass is 1.8 kg and number-based recovery is 60%. The result depends on whether the goal is live seed numbers, stockable size or biomass.
For operational comparisons, initial biomass matters too. Weight gain should usually be distinguished from gross final biomass. If A started with 1.0 kg and ended at 1.4 kg, its net increase is 0.4 kg, assuming transfers, losses and weighing definitions are accounted for. If B started with 0.3 kg and ended at 1.8 kg, its apparent net increase is 1.5 kg. Report the inputs before celebrating an output number.
Calculate claw-loss incidence with a clear denominator
A nursery quality sheet may report “claw loss 12%.” Was that 12% of recovered crabs with at least one missing claw, 12% of all claws expected, or 12% of animals ever injured during the period? Each has a different denominator. A crab with both claws missing might contribute one affected animal or two lost appendages depending on the chosen definition.
Suppose 200 recovered crabs are examined. Twenty have one missing claw and four have two missing claws. The proportion of animals with at least one missing claw is (20+4)/200 = 12%. The number of missing claws is 20 + 2 times 4 = 28. Because 200 crabs could have 400 claws, the missing-claw proportion is 28/400 = 7%. Neither percentage is incorrect; they measure different quantities.
An endpoint injury count cannot tell us exactly when injuries occurred or whether limbs were lost through aggression, handling or other processes. Some crabs may regenerate appendages later. Label the measure as an observed endpoint prevalence if that is what the procedure assessed. Avoid translating “incidence” in its epidemiological sense when the method is actually a single cross-sectional inspection.
Why stocking density changes after mortality or transfer
A density printed on the first day is an initial stocking density, not necessarily the density experienced throughout the trial. With 800 animals initially in 4 m², density begins at 200/m². If 200 are removed to another nursery, the nominal remaining population drops to 600 and density becomes 150/m² before other losses. If a subsequent count estimates 500 alive, the estimated current density is 125/m².
The report should say whether analyses use starting density, measured time-varying density, peak density or average density over the period. Each captures a different exposure. For animals that interact through space, crowding may be intense early and ease later as mortality reduces numbers. A result that correlates final survival with “density” requires careful thought about whether density was an assigned treatment or a consequence of survival itself.
Do not calculate an average of arbitrary measurements and claim it represents time-weighted exposure. If density was 200/m² for 12 days and 100/m² for two days, the simple average of the two readings is 150/m², but the duration-weighted mean is (200 times 12 + 100 times 2)/14 = about 185.7/m². Neither figure by itself captures local encounters, but at least the time basis is explicit.
Distinguish floor area from shelter access
A tank can have abundant material area and little functional refuge. A stack of folded netting might offer many microspaces; a flat sheet on the floor may not separate individuals at all. Conversely, a tall complex structure might increase usable surface without changing measured projected coverage. Report dimensions, mesh aperture, vertical arrangement and how animals use the structure when those observations exist.
The same shelter can become crowded as individuals grow. A juvenile that fit under a narrow partition initially may be unable to enter after several moults. Thus “one shelter per crab” is not a stable biological exposure unless the shelter definition and body-size distribution are fixed. A useful report records shelter condition at each relevant stage rather than copying the first-day installation inventory across every observation.
Shelter maintenance is part of the evidence. If fouling, collapse or removal occurred during the trial, a nominal treatment labelled “with shelter” may not describe the full exposure. Record maintenance events and any temporary absence. This is an ordinary data-integrity issue, not proof that a management method succeeded or failed.
What a fair two-tank comparison looks like
Suppose two tanks each receive 600 juveniles, with 3 m² bottom area and the same planned duration. Tank R has added refuges; tank C is a comparison without them. At harvest, R yields 450 alive and C yields 390 alive. Their observed recoveries are 75% and 65%, a difference of ten percentage points. Their recovered counts differ by 60 animals.
Those numbers alone are not enough to claim the refuge caused the difference. One tank per treatment confounds the shelter intervention with any unmeasured tank difference: location, water delivery, illumination, operator handling, starting crab condition or disease exposure. An experiment needs replication and an appropriate design. A translation should call this a two-tank comparison rather than a statistically established shelter effect.
If the original log also shows size sorting in R but not C, the difference becomes even less attributable to shelter alone. The proper conclusion is “The refuge-and-sorting tank had a higher recovered fraction under these conditions.” Do not trim the phrase to “Refuge raised survival by ten percent.” Even that percentage wording could confuse an absolute ten-percentage-point difference with the relative improvement of 10/65, about 15.4%.
A compact decoding method for reports
For every density claim, write out the organism stage, count, area or volume and date. For every shelter claim, write the unit and whether it is installed material, projected coverage or observed occupancy. For every survival claim, write exactly what was recovered and what was stocked. For every mortality claim, state whether dead animals were observed or inferred from inventory difference.
Then keep a column of events that alter interpretation: grading, transfer, water exchange, cleaning, shelter removal, large moulting episodes, harvest and changes in counting protocol. A record with these entries can be understood by a reader who was not present. A list of bare percentages cannot.
The result is a report that remains honest across languages and audiences. “Two hundred juvenile crabs were initially stocked per square metre; the cage also received added refuge netting; seventy percent of stocked animals were recovered alive after fourteen days” contains a clear experimental frame. It leaves the causal explanation open until supporting evidence is reviewed.
Measure shelter occupancy separately from coverage
Coverage describes the equipment; occupancy describes the animals. An observer might record 30% of floor under structures but find 70% of visible crabs inside or alongside refuges. That does not mean refuge coverage is 70%. It means a separately defined fraction of observed animals was located in or near the shelter at the observation time. If crabs inside opaque shelters cannot be counted, even that occupancy fraction needs a visibility qualification.
Suppose a morning inspection sees 45 crabs in refuges and 15 outside. The observed occupancy proportion is 45/(45+15) = 75% among 60 visible animals. If 40 other crabs are estimated to be hidden and not individually classified, you cannot claim 75% occupancy of all 100. You can report 75% of visible, classified observations and explain the missing observation frame. This protects the report from falsely increasing precision.
Animals move, so an instantaneous occupancy rate is not a duration of refuge use. A time-lapse programme or repeated observations may be more informative for behaviour. Record observation times relative to feeding, cleaning, light and moulting, because occupancy can change rapidly. Do not translate “observed in shelter” into “spent most of the day sheltered” when that duration was not measured.
The unit problem when a study uses cages inside a pond
A research paper may call a nursery cage a tank, a hapa, a net enclosure or a pond unit. The words are not synonyms for the denominator. A two-square-metre cage installed in a 200-square-metre pond may stock 500 animals. The cage density is 250/m². Dividing 500 by the entire pond area would yield 2.5/m² and misleadingly imply that all animals had access to the whole pond.
A large water body may supply shared temperature, oxygen and plankton conditions while the cage wall limits each cohort’s horizontal movement. The experimental unit may be the cage rather than the pond, but shared water conditions can make cage observations correlated. A translation should keep experimental unit, water body and crab habitat separate. Otherwise a report might overstate the number of independent replicates.
When a paper lists a stocking range from a particular study, check whether it uses the internal cage footprint, gross enclosure area or actual settlement surface. Do not quote the numeric rate alone as a recommendation. The FAO practical manual discusses nursery methods with a range of stocking practices and explains why crab size and density affect cannibalism risk. It is context, not a universally approved target for every species or operation.
How to interpret density over time with removals
Imagine a five-square-metre unit stocked with 1,000 juveniles. After seven days, 100 are removed for grading and 700 remain in the counted inventory. The original stocking density is 200/m²; after removal and unaccounted losses, the current estimated density is 140/m². Both should appear in a readable longitudinal report rather than one generic column labelled “stocking density.”
If the recorded 700 comes from a sample extrapolation rather than a full count, say so. If 100 animals were moved but the source tank was sampled before the move, do not deduct them twice. Reconciling the chronology is essential. A seemingly simple “remaining density” can be wrong because a spreadsheet formula uses the previous day’s volume or transfers as well as mortality to update the count.
If the nursery is divided by adjustable mesh partitions, accessible floor area can change too. A compartment containing 250 juveniles over 1.0 m² has 250/m². Expanding it to 1.5 m² without moving any crab changes nominal density to 166.7/m². This is a real change in spatial availability without a change in population. Translate the structural intervention, not just the lower density.
A worked example of grade-specific losses
A hatchery sorts 600 crablets into 400 small and 200 large individuals. After ten days, it recovers 280 small and 180 large alive. Small-group recovery is 280/400 = 70%; large-group recovery is 180/200 = 90%. The overall recovery is (280+180)/(400+200) = 460/600, approximately 76.7%. This is a weighted result, not the mean of 70% and 90%, which would be 80%.
The higher large-group recovery is an association in this scenario. It does not demonstrate that small animals were consumed by larger ones, because the groups may differ in physiology, health, handling, initial condition and detection. If the groups were in separate enclosures, a direct large-on-small cannibalism explanation is especially inappropriate. The report needs the housing arrangement before it can support that causal story.
If some animals changed size grades during the interval, final grade counts are not identical to survival by initial grade. An initial-small animal that grows into the large category should still belong to its original cohort when calculating survival by initial category. A record that only sorts by final size answers a different question. Labels such as “initial size class” and “final harvested size class” help avoid accidental substitution.
A worked example of inconsistent shelter denominators
Nursery A uses 12 shelters in a four-square-metre tank. Nursery B uses 20 shelters in an eight-square-metre tank. Comparing shelter counts makes B look more supplied, but A has 12/4 = 3 structures/m² and B has 20/8 = 2.5 structures/m². The absolute and area-standardised rankings differ. Neither necessarily captures actual refuge capacity because the structures might have different dimensions.
Now suppose each A shelter allows one crab to hide separately, while each B shelter has four suitable cavities. The calculated refuge places are 12 for A and 80 for B. A measure of refuge spaces per stocked crab may be more useful, provided the cavities are independently accessible and appropriately sized. A translated report should explain this added operational assumption rather than silently treating a shelter as a fixed unit of biological capacity.
If only geometric dimensions are known, do not invent cavity capacity. Record footprint, depth, aperture size and material as available. The word “shelter density” itself can mean structures per floor area or structures per animal. Stating the ratio in full makes the distinction survive both translation and subsequent spreadsheet analysis.
A worked example of injury interpretation
In a 250-crab final census, 30 animals have one missing claw and 10 have both claws missing. The proportion of crabs affected is 40/250 = 16%; the total number of missing claws is 30 + 20 = 50 out of 500 expected claws, or 10%. If a translator calls both numbers “claw injury rate,” a reader may reasonably think there is a contradiction where none exists.
There is a third possible metric: the number of new claw-loss events per crab-day at risk. To calculate this correctly one needs repeated animal-level observations or a defined event-logging system, plus exposure time. A one-day endpoint census does not supply that rate. It also may not separate loss from aggression, handling or self-amputation under stress. That is why “prevalence of visibly missing claws at harvest” is often the more defensible phrase.
For animal welfare and quality, the distribution of severity can matter more than a single percentage. Report one versus two missing claws, other damage, and uncertainty in identification. If an animal with a regenerated claw is classified as intact under the method, say so. Never infer a precise causal pathway from an endpoint description alone.
A worked example in which larger biomass hides lower survival
Unit A starts with 500 individuals averaging 0.8 g and recovers 400 at an average 2.0 g. Initial biomass is 400 g, final recovered biomass 800 g, apparent gain 400 g and numerical recovery 80%. Unit B starts with 500 at 0.8 g and recovers 300 at 3.0 g. Its final recovered biomass is 900 g, apparent gain 500 g and numerical recovery only 60%.
A report headlined “B had higher output” may be true for final mass but unhelpful if the business needs 400 viable crablets for stocking. Likewise “A had higher survival” does not establish better feed conversion or economic return. Define whether the result is yield as animals, final kilograms, net gain, marketable units or another outcome. Do not combine them into a vague success score.
An apparent biomass gain also depends on whether animals removed earlier are included. Suppose Unit B sold an additional 100 juveniles at 1.5 g before final harvest. Cumulative recovered biomass becomes 900+150 = 1,050 g, while final biomass remains 900 g. Keep cumulative harvest and remaining inventory separate. Without the ledger, two reasonable calculations can appear to disagree simply because one includes an earlier removal.
A worked example of a percentage-point error
Trial A recovers 72% of stocked juveniles and trial B recovers 60%. The absolute difference is 12 percentage points. Relative to B, A’s recovery is (72-60)/60 = 20% higher. These are mathematically different descriptions. A report that says “12% higher” is ambiguous and can be read incorrectly as either. Use “12 percentage points higher” for the difference between percentage levels.
Suppose the trial labels are reversed in a translation. Calling 60% a 12-point improvement over 72% is an obvious mistake; calling 60% “only slightly worse” is an editorial judgment that needs a defined materiality threshold. Technical translation should preserve numbers and measured comparisons first, leaving value judgments to a clearly stated decision context.
Further, the uncertainty around each proportion matters. Seventy-two percent from 25 counted animals is a very different precision situation from 72% from 2,500, though other sampling features remain relevant. Report numerator, denominator, replication and count method. A neat percentage may conceal inadequate sample size, nonindependence or selection of a convenient endpoint.
A worked example of cross-tank pooling
Three cages receive 100, 200 and 700 juvenile crabs and recover 80, 120 and 490. Cage recoveries are 80%, 60% and 70%. Their unweighted mean is 70%, while pooled animal recovery is (80+120+490)/(100+200+700) = 690/1,000 = 69%. The difference is small here but can be larger with unequal cage numbers and more divergent recoveries.
Which is right? If the target is the fraction of all stocked animals physically recovered, 69% is right. If the target is the average cage-level percentage, equal weighting produces 70%. However, when cages represent independent experimental units, further inferential analysis must respect that structure rather than treating all thousand animals as independent replicates. The reporting question governs the estimator.
If the three cages are all inside one pond and share water, their independence may be limited. A translation should not claim “three independent environmental trials” without the design to justify it. Record the hierarchy: animals within cages, cages within ponds, ponds within sites. Correct scientific language includes the observational unit, not only the numerical result.
An exercise on missing animals and cause assignment
Start with 800 juveniles. At harvest, 510 are recovered alive; 35 dead bodies are observed; and 25 had previously been transferred alive to another confirmed unit. The total observed-accounted number is 510+35+25 = 570, leaving 230 unaccounted, or 28.75% of the initial input. Final live recovery in the original unit is 510/800 = 63.75%, while confirmed live recovered or transferred is (510+25)/800 = 66.875%.
It would be incorrect to say “230 crabs were cannibalised.” The records only establish an accounting gap. That gap might be biologically important, but it cannot be assigned to a cause by subtraction alone. If the question is whether the originally stocked cohort survived, the 25 transferred crabs need downstream follow-up; transfer alive at an intermediate date does not prove survival to the original endpoint.
A revised report could state: “Of 800 stocked juveniles, 510 were recovered alive in the original unit, 25 were transferred alive earlier, 35 dead were observed, and 230 remained unaccounted at reconciliation.” That phrasing is longer but contains the actual knowledge instead of an imagined cause.
How to design an observation sheet that translators can trust
A strong observation form begins with species and cohort, life stage, counting date and time, cage or tank ID, accessible area, water depth or volume where relevant, and number stocked. It then separates animal counts by measurement method: direct total, subsample estimate, transferred in, transferred out, observed dead and final recovered. The form also has explicit fields for refuge type, dimensions and changes.
Include a simple injury classification with the number of animals examined and the definition of damage, such as one claw missing or both missing. Record size measurements with instrument precision and whether animals were sampled randomly, fully measured or selected for marketing. A field marked “not examined” is better than an empty cell later mistaken for zero injuries.
Finally, give every event a unique timestamp and a note of who made the observation. This supports independent audit and allows readers to reconstruct the sequence. Translation works best when the original sheet already respects distinctions between observation, calculation and explanation. Clear data architecture reduces pressure on the translator to guess what the producer intended.
Frequently asked questions about nursery comparisons
Is “per tank” a useful density unit? It can identify a total stock count, but without tank dimensions it does not standardise density. Two tanks can have different areas and volumes. A comparison needs the physical denominator appropriate to the developmental stage and mechanism, and the exact housing arrangement.
Does higher shelter coverage guarantee better survival? No. Refuge effect depends on shape, access, stage, size distribution, water conditions and maintenance. Published evidence can support an intervention in a specific setting, but a percentage of floor covered by material is not a direct measure of effective protection.
Can unrecovered animals be labelled mortality? Only with caution and a declared definition. Some studies use apparent mortality calculated by subtraction, but this should not be presented as observed deaths or attributed to cannibalism without evidence. If the counting process misses hiding animals, even apparent mortality is sensitive to detection.
Should we average the percentages from different cages? Decide whether the question concerns mean cage performance or the fraction of all animals recovered. Equal-weight averages and pooled recovery use different denominators. Keep both only when they genuinely answer different reporting needs, and describe that distinction.
Does a change in mean body weight show growth for every animal? No. Selection can change the mean when smaller animals die or are removed. Individual growth requires repeated measurement of the same animals or a justified study design. A before-and-after average may reflect both growth and changing composition.
Is claw loss proof of cannibalism? No. Aggression is one possible cause, but losses can also result from handling and other events. An endpoint exam can document damage prevalence, not automatically its cause or exact timing. Use direct observation or suitable controlled evidence to support causal claims.
Can we infer welfare from productivity alone? No. High yield may coexist with injuries or stress, and low yield may have several explanations. Report survival, condition, injury measures and environmental evidence separately. Ethical husbandry decisions need qualified oversight, not a single score.
What if a study’s stocking rate is much higher than ours? Check the life stage, species, area definition, substrate, feeding, duration and facility design before comparing. A quoted experimental range is not a blanket recommendation. A number without context may be technically accurate and operationally dangerous to generalise.
Final check: turn abbreviations into a statement that survives translation
For every reported percentage, write the full numerator and denominator. For every area, specify whether it is usable bottom, enclosure footprint or material coverage. For every survival claim, distinguish observed live recovery from missing inventory. For every shelter treatment, record what was actually installed and whether its configuration changed. These steps make a report understandable even to someone who has never seen the nursery.
If the reader can reproduce the calculations and can tell which statements were directly observed, the translation has done its job. A carefully limited conclusion is not weaker science; it is what allows separate teams to compare their evidence without confusing a cage design, a biological process and a mathematical percentage.
Sources and further reading
FAO, Mud crab aquaculture: a practical manual. Research-based husbandry context and nursery considerations.
FAO AGRIS, Stocking density and artificial habitat in Scylla serrata nursery systems. Specialist research record comparing survival, size and claw loss.
SEAFDEC, Improved production of mud crabs in marine pens with shelters. Research example of shelter comparisons in a specific setting.
FAO AGRIS, Stocking density and seaweed shelters in juvenile Scylla paramamosain. Stage- and species-specific context for nursery comparisons.
eduKateSG, How to Read Mud Crab Zoea, Megalopa and First-Crab Counts. Companion explanation of cohort accounting and life-stage conversion.
eduKateSG, How English Works: Vocabulary and Meaning. Precision and contextual language.
eduKateSG, How to Translate Easily series index. Other technical translation guides.