What a fish growth result actually says
A growth report becomes useful only when its reader can reconstruct what grew, over which interval, and according to which denominator. A fish that gains twenty grams during a month has not necessarily performed better than a larger fish gaining sixty grams in the same month. Starting weight, developmental stage, temperature, feeding regime and survival all alter the interpretation. This guide explains specific growth rate and thermal growth coefficient as different tools, not interchangeable labels for a single biological truth.
Consider a hypothetical tank trial beginning with a mean fish weight of 40 grams and ending at 62 grams after 28 days. The absolute gain is 22 grams per sampled fish and the simple gain is 0.786 grams per day. That is a defensible account of mean-weight change, provided the beginning and ending samples represent the same group. It does not yet say what happened to the total biomass, what individual fish experienced or whether heavier fish were preferentially removed.
The most reliable reading begins with the original measurement table and asks five questions: whether weights are wet or dry, whether observations refer to individual fish or tank totals, whether each interval contains the same number of elapsed days, how mortalities and transfers were treated, and whether the sampling method changed. Answers to these questions often matter more than an impressive percentage displayed on the final chart.
Define absolute gain without hiding its denominator
Absolute weight gain is final weight minus initial weight, normally expressed in grams per animal when weights are individual means. Dividing by elapsed days produces mean daily gain in grams per fish per day. These are additive measures: a change of ten grams followed by fifteen grams yields a total gain of twenty-five grams if the same defined individual or representative average is being followed. They are easy to audit but can be misleading across very different starting sizes.
If a cohort starts at 40 grams and reaches 62 grams over 28 days, absolute gain equals 62 minus 40, or 22 grams. Mean daily gain equals 22 divided by 28, or about 0.786 grams per day. The word mean here is important. Even a precisely calculated average is not a statement that every fish gained the same amount. A wide distribution could include slow-growing fish and very fast-growing fish whose outcomes cancel in the mean.
An absolute gain chart should display start and end dates, sampling time, sample size, the weighing method and the water conditions near sampling. Fish may have different gut contents or surface moisture depending on handling and feeding schedules. A defensible report controls or records these sources of variation instead of treating the scale reading as a pure measure of tissue growth.
Read percentage growth before interpreting daily rates
A cumulative percentage increase uses the initial weight as denominator: 100 multiplied by final minus initial, divided by initial. In the example, 100 times 22 divided by 40 equals 55 percent over the full 28-day interval. The percentage is neither a daily growth rate nor a forecast of the next month. A report that omits the interval quietly invites readers to compare numbers that are not comparable.
Suppose a second group rises from 200 grams to 250 grams. It gains 50 grams, much more than the first group, but its cumulative percentage gain is only 25 percent. Neither comparison alone establishes which stock is healthier or better fed. Body size alters maintenance requirements, and feeding strategies often seek different outcomes at juvenile and market stages. The denominator tells us what question a percentage answers.
Two common errors are dividing by final rather than initial weight and treating a 55 percent cumulative increase as 55 percent per day. The first gives 35.5 percent in this example, a different question about the fraction of the final weight represented by gain. The second is simply a time-unit mistake. A careful translator preserves both the original reference mass and time frame instead of replacing them with a more familiar sounding percentage.
Compounded specific growth rate and the logarithmic convention
Specific growth rate, or SGR, is not perfectly standardized across every report. A commonly used logarithmic expression is 100 times natural logarithm of final weight minus natural logarithm of initial weight, divided by the number of days. This gives a continuously compounded percentage rate per day. Some technical material instead computes the equivalent discrete daily compound percentage as 100 times the quantity final divided by initial raised to the reciprocal of days, minus one.
For the 40-to-62-gram example over 28 days, the logarithmic rate is approximately 100 times ln(1.55) divided by 28, which is 1.565 percent per day. The equivalent discrete compounded rate is 100 times exp(ln(1.55) divided by 28) minus 100, approximately 1.578 percent per day. Both summarize exactly the same hypothetical endpoints, but their numerical values are slightly different. Label which formula the author used.
The daily percentages are not the same as simple cumulative gain divided by days. That shortcut gives 55 divided by 28, about 1.964 percentage points of initial mass per day. It assumes the initial denominator remains fixed throughout the interval, whereas compounding allows the implied reference mass to grow. The choice matters increasingly with faster growth or longer intervals; the source formula is therefore part of the result, not a disposable footnote.
Check the logarithm and its domain
The natural logarithm requires positive weights. Zero recorded weight, negative weight or a missing observation coded as zero makes the calculation undefined or meaningless. Before typing a spreadsheet formula, filter invalid values and inspect why they occurred. An escaped fish, unweighed fish or dead fish is not an animal with zero living body weight. Replacing a missing endpoint with zero converts a data-management problem into an impossible growth measurement.
Because the logarithm of a ratio equals the difference of logarithms, 100 times ln(final divided by initial) divided by days is algebraically identical to 100 times the difference in natural logarithms divided by days. This equivalence is a useful cross-check when a spreadsheet uses one expression and a report prints another. Both require the same measurement unit for initial and final weight so that the ratio itself has no unit.
When a fish loses weight, final divided by initial is positive but below one, so its logarithm is negative. Negative SGR is therefore mathematically possible. It may indicate fasting, stress, disease, reproductive allocation, inaccurate sampling or changes in cohort membership. A negative result is not automatically a failed experiment, but the interpretation requires evidence about context. The equation reports change; it does not identify its cause.
Why daily growth is not a promise of constant growth
SGR compresses an entire observation interval into one equivalent daily rate. It does not demonstrate that the fish followed that rate on each calendar day. Appetite, oxygen supply, temperature, handling and feeding schedules can vary within a month. Growth might have slowed during a poor-water-quality episode and accelerated after recovery. Endpoints alone cannot reconstruct that time path, no matter how many decimals are printed.
For example, a fish moving from 40 to 62 grams in 28 days could move through 50 grams at day 14 or through 56 grams at day 14. Both paths have the same final SGR but different early and late growth. If the management question concerns a mid-cycle feed change, an endpoint-only average may conceal the timing that matters. Additional representative sampling dates are needed to evaluate the interval-specific response.
The word specific means relative to existing size under the particular formula, not specific to a species or proof of biological optimality. Labels such as excellent growth must be grounded in species, age, temperature and defensible benchmarks. A universal SGR target across fish sizes is particularly unreliable because percentage growth tends to change as animals become heavier and move through developmental stages.
Thermal growth coefficient answers a different question
Thermal growth coefficient, commonly shortened to TGC, attempts to relate size change to time and thermal exposure. One widely used form is 1,000 multiplied by the difference between the cube roots of final and initial individual weights, divided by elapsed days times average water temperature in degrees Celsius. The cube-root transformation is not decorative. It changes how growth scales with body size before dividing by the chosen thermal exposure.
Using weights of 40 and 62 grams, the cube roots are approximately 3.420 and 3.958. If the interval is 28 days at a recorded mean temperature of 22 degrees Celsius, TGC is about 1,000 times 0.538 divided by 616, or roughly 0.87 under that equation’s unit convention. This value cannot be read as 0.87 percent per day. It uses a different transformation and dimensional construction.
Before using TGC, confirm the author used grams rather than kilograms, the exact coefficient factor, and whether temperature means daily average, sum of daily values or a temperature corrected for a biological threshold. The standard formula above uses the arithmetic thermal exposure expressed as days multiplied by mean degrees Celsius. Other thermal-growth formulations or species-specific models use different assumptions. They should not be silently interchanged.
Temperature exposure must come from relevant records
A single thermometer reading on the sampling date is not a valid estimate of the entire interval’s thermal history. In a 28-day study with daily readings, an arithmetic mean of 22 degrees Celsius implies a total of 616 degree-Celsius days for the simple formula. If some days were much colder or hotter, the same arithmetic total could conceal biologically important stress. Temperature is not always linearly beneficial for growth.
Consider two 14-day periods, one averaging 18 degrees and the other 26 degrees. The overall arithmetic mean is 22 degrees if each period covers the same number of days, and the total exposure is 14 times 18 plus 14 times 26, or 616 degree-days. A report might calculate the same TGC as a perfectly constant 22-degree environment, yet animal feeding and maintenance may respond differently to excursions outside a comfortable range.
Use temperatures recorded where the animals actually lived, not merely the air temperature at a weather station or a shallow surface reading from a stratified pond. Location and sensor depth matter. Calibration, logging intervals and sensor gaps should be documented. TGC offers a useful standardized comparison under suitable conditions, but it does not magically remove all effects of temperature, oxygen or species physiology.
The surprising problem of identical final averages
A tank mean is a summary of sampled fish, not an identity assigned to an individual fish. Suppose a group contains ten fish starting at 40 grams each, and five are 50 grams while five are 74 grams at day 28. The final mean is 62 grams, exactly the value used in the worked example. Another group could contain ten fish each at 62 grams. Both produce the same group-level mean-based SGR even though size uniformity differs markedly.
These groups may need different management. Large size differences can influence feed access, grading decisions and sampling design, although the consequences depend on species and husbandry. The average can mask a subgroup that is not prospering. Reading growth with a distribution, such as median, lower and upper quantiles, or coefficient of variation, gives more insight than reporting the mean alone.
Never claim that computing SGR from a mean equals the mean of SGR values computed fish by fish. The logarithm is nonlinear. When each individual can be tracked reliably, individual growth rates can be summarized directly, but the cohort mean-of-endpoints approach answers another question. Selection and fish identity matter: tagged-individual growth, tank-level growth and population-level turnover are separate analytical designs.
An individual example with uneven growth
Imagine three individually identifiable fish weighing 20, 40 and 80 grams on day zero, then 28, 60 and 100 grams after 20 days. Their respective weight ratios are 1.4, 1.5 and 1.25. Their logarithmic SGR values are about 1.682, 2.027 and 1.116 percent per day. The arithmetic mean of these three rates is approximately 1.608 percent per day, which directly summarizes equal-weighted individual trajectories.
The arithmetic initial mean is 46.67 grams and the final mean is 62.67 grams. The ratio of those means is about 1.343, giving a mean-based logarithmic SGR of approximately 1.473 percent per day over twenty days. Neither figure is an arithmetic mistake. They weight fish and growth differently. Calling both the average fish growth rate without explaining their computation would create an avoidable interpretive ambiguity.
The difference also reveals why extreme or skewed body sizes matter. A large fish contributes more grams to group mean weight, while equal averaging of individual log rates gives each tagged fish one vote. A well designed study chooses the summary that matches its question before looking at which result looks larger. The denominator and aggregation level should appear next to every rate in a published table.
Survival and mortality change what a cohort means
A cohort does not necessarily contain the same live individuals at two dates. If larger fish are harvested early or smaller fish die before the final sample, the remaining mean can change even without proportional growth across the original stock. A survivor mean is conditional on survival. That distinction matters particularly when health events, grading or harvest occur during the interval.
Suppose a stocked population begins with one thousand fish averaging 40 grams, for 40 kilograms of initial live biomass. After the interval, nine hundred fish average 62 grams, for 55.8 kilograms of remaining live biomass. Mean weight rose by 55 percent, but live biomass rose by 15.8 kilograms, or 39.5 percent of its original total. The one-hundred-fish loss changes the production interpretation even though per-survivor growth looks strong.
Additional biomass from dead fish removed during the period, graded-out transfers or partial harvests belongs in a reconciled production account, but should not be casually inserted into mean individual weight. Keep two ledgers: an individual size-growth ledger and a population biomass ledger. Link them with date-stamped counts, removals and transfers. This prevents a good SGR from concealing poor survival or a misleading FCR denominator.
Sampling error is not solved by a longer equation
Random error arises when only some animals are weighed. If a tank holds thousands of fish, a random sample may estimate the mean with useful precision, but the result depends on sample selection, sample size and size variability. A net may preferentially catch active, large or near-surface animals. A scale may be precise to a tenth of a gram while selection bias exceeds several grams.
Consider 25 representative fish whose sample mean is 62 grams and sample standard deviation is 10 grams. Under simple random independent sampling assumptions, the approximate standard error of the sample mean is 10 divided by the square root of 25, or 2 grams. A rough 95 percent interval based on a suitable t multiplier is about 62 plus or minus 4.1 grams. That interval addresses sampling variation, not every possible systematic bias.
Repeated weighing of the same individual might instead provide paired data, often giving a clearer picture of individual change. Paired design is useful only when identification is reliable and handling effects are controlled. Tank replication also matters. Thirty fish from one tank are not thirty independent tanks for a feed trial; the treatment may have been assigned at tank level. Avoid false precision created by pretending subsamples are independent experimental units.
Propagate a realistic endpoint uncertainty
For logarithmic SGR, small independent errors in mean initial and final weight can be approximated through derivatives. If s0 and s1 are standard errors of initial and final mean weight, the approximate variance of ln(final divided by initial) is (s1 divided by final) squared plus (s0 divided by initial) squared, minus any covariance term when the estimates are paired. Divide the resulting standard error by the number of days and multiply by 100.
Take initial mean 40 grams with standard error 1 gram and final mean 62 grams with standard error 2 grams, treated as independent estimates for illustration. The relative standard-error terms are 0.025 and about 0.03226. Their squared sum is about 0.001666; square root gives 0.04082. After multiplying by 100 and dividing by 28 days, SGR standard error is approximately 0.146 percentage points per day. This quantifies a component of uncertainty, not the entire truth.
Positive covariance between paired endpoint estimates can reduce the variance of the difference, while unrecognized selection bias can dominate it. If days or temperature exposure are uncertain, additional terms enter the error budget. An interval that prints SGR as 1.5649 percent per day without mentioning four or five percent uncertainty in mean mass is over-precise. A defensible report rounds after calculation and states what its interval covers.
Measurement bias and calibration
Scale calibration and handling practices can distort growth even when samples are numerous. A systematic five-percent shift in all final measurements would raise the apparent weight ratio. If the true final mean is 62 grams but the final scale reads 65.1, the computed logarithmic SGR becomes approximately 100 ln(65.1 divided by 40) divided by 28, or 1.74 percent per day instead of about 1.57. Sampling more fish cannot repair a biased instrument.
Calibration records should cover the useful measurement range and the conditions under which weights were taken. Wet baskets must be tared; batch weights require subtracting container mass; devices should stabilize before reading. For very small larvae, specimen counts, water retained in containers and drying protocol can matter more than nominal digital resolution. Each method needs appropriate operational checks.
Translation of a report must preserve whether a weight was measured on a whole animal, an eviscerated body, a dry specimen or a bulk lot. These are different denominators. If two laboratories disagree, revisit procedures before concluding that biological growth changed. The correct response to uncertain data may be remeasurement or qualified interpretation rather than a more complicated equation.
Distinguish TGC from growing-degree-day development
TGC describes growth by relating a cube-root mass change to thermal exposure. Growing-degree-day models in other settings often subtract a base temperature before accumulating positive thermal units, and some are designed for developmental milestones rather than mass gain. The words thermal and degree-day are not permission to substitute those equations. A report should identify the exact convention and its biological justification.
If an author uses 22 degrees Celsius for 28 days, multiplying gives 616 degree-Celsius days under an unadjusted TGC denominator. A separate model might use an explicitly justified base temperature of 10 degrees and sum daily max(temperature minus 10, zero), producing 336 degree-days in the constant-temperature hypothetical example. Dividing by 336 would yield a different numeric coefficient. Neither denominator can be silently substituted for the other.
Both approaches may fail during excessive temperatures because stress, feeding suppression or mortality can make response nonlinear. Read the domain over which a coefficient was calibrated. A growth coefficient measured in juveniles at safe temperatures does not necessarily project to broodstock, a different species or a heatwave. The proper use of temperature normalization is conditional comparison, not universal biological correction.
Compare groups fairly before claiming a feed effect
A high SGR in a feed trial might reflect a smaller starting fish, warmer water or different stocking density rather than feed formulation alone. Useful comparisons pair the growth metric with a study design: randomized tank allocation, matched starting weights, measured feeding amounts, environmental monitoring and replication at the treatment-assignment level. A rate cannot replace those design features.
Imagine feed A produces growth from 40 to 62 grams at 22 degrees while feed B produces growth from 50 to 75 grams at 24 degrees over the same 28 days. Logarithmic SGR is about 1.565 percent per day for A and 1.448 for B. Feed A has the higher relative rate, while feed B produces 25 grams rather than 22 grams of absolute gain. The temperature and initial-size differences prevent a clean causal conclusion.
TGC can provide an additional lens, but it cannot repair confounding if the groups were systematically different. Any comparison should report starting mass distributions, temperatures and survival, not only two coefficient values. If the intervention was assigned to tanks, estimate variation across tanks rather than treating every fish as an independent replicate. Genuine uncertainty is more valuable than overstated certainty.
Check a growth forecast by reversing the equation
A useful arithmetic audit is to solve the logarithmic equation backwards. Final weight equals initial weight multiplied by exp(SGR divided by 100 times elapsed days). For an initial weight of 40 grams, SGR of approximately 1.565 percent per day and 28 days, the expression returns about 62 grams. If it returns 620 grams, the most likely problem is mishandling percentages or a decimal place.
For a hypothetical forecast at the same rate for another 14 days, 62 multiplied by exp(0.01565 times 14) is approximately 77.2 grams. This is a mathematical extrapolation, not a biological prediction. Fish can slow as they grow; environmental conditions, feed and health can change. A reliable forecast states that it assumes the recent mean rate continues and gives a plausible range rather than pretending to know future performance.
Any forecast should be stress-tested with lower and higher rates. At 1.2 percent per day for 14 days, a 62-gram fish would reach about 73.4 grams; at 1.8 percent it would reach about 79.8 grams. These scenarios show the leverage of the growth-rate assumption. They are illustrative calculations only and should not be presented as species-specific expected outcomes.
Worked comparison of three independent units
Suppose three hypothetical tanks each start at 40 grams per fish and are sampled again after 30 days. Their final means are 54, 60 and 66 grams. For tank one, ln(54 divided by 40) times 100 divided by 30 is approximately 1.00 percent per day. For tank two, the same calculation with 60 grams gives 1.35 percent per day. Tank three gives approximately 1.67 percent per day.
These are tank-level point estimates, not a statistically credible treatment comparison if all tanks received different conditions. A single tank per condition has no replication at the unit that experienced the treatment. If all three instead received the same protocol, their variability is informative about ordinary tank-to-tank spread. Each tank’s survival and sampling uncertainty should still be reviewed before averaging.
A simple arithmetic mean of the three tank SGR estimates is about 1.34 percent per day, with each tank given one vote. Pooling all sampled fish before calculating a single start-to-end mean rate may weight tanks by different sample sizes or population sizes. State which estimand matters: mean independent-unit performance or population-weighted performance. Mixing them without disclosure can reverse the apparent ranking of production strategies.
Report the correct experimental unit
In aquaculture trials, animals within one tank share water, aeration, stocking density, feed delivery and exposure to many disease agents. When the treatment is applied to the entire tank, the tank is usually the experimental unit for testing treatment effects. Collecting fifty fish from one tank improves the estimate of that tank’s mean but does not manufacture fifty independent treatment replicates.
An illustrative design could allocate four tanks to each of two feeds, with comparable starting mass and recorded temperature. Within each tank, fish might be sampled for mean body weight at several dates. A valid comparison can then use the four tank-level outcomes per feed, with methods selected for the design. It should not use all sampled fish as if they were unrelated observations.
A technical translation that turns tank replicates into fish replicates changes the implied evidence strength. Preserve phrases such as n equals four tanks, twenty sampled fish per tank, and repeated measurements of the same cohort. If the source is ambiguous, flag the ambiguity instead of silently resolving it. The reader needs to know the basis of precision and inference.
Separate growth performance from feed conversion
SGR measures relative individual mass change under a selected formula. Feed conversion ratio instead compares feed supplied with biomass gain under a stated accounting convention. A feed may support rapid growth but also produce higher feed consumption and a less favorable feed-use metric. Conversely, restricted feeding may improve some efficiency numbers while reducing growth and welfare. There is no guarantee that optimizing one index optimizes all others.
For context, a hypothetical tank might increase live biomass from 40 to 55.8 kilograms while receiving 25 kilograms of feed, giving a simple economic feed conversion ratio of 25 divided by 15.8, or about 1.58 kilograms feed per kilogram net live biomass gain. That value depends on the treatment of mortalities, transfers and harvests. It is not derived from individual SGR alone.
Read our related feed conversion and mortality guide alongside growth results, and use the biomass sampling and uncertainty guide for population-level checks. These links offer neighboring questions, not substitutes for defining the denominator in the present calculation.
Growth and disease are connected but not equivalent
Poor growth can be a useful early warning, but it is not a disease diagnosis. Food availability, hierarchy, fluctuating oxygen, temperature, water chemistry, genetics, maturation, parasite burdens and handling can all affect mass gain. A growth rate that declines deserves a structured investigation, not an immediate leap to a single cause. Daily behavior and feeding records often provide critical context.
Conversely, strong mean growth does not guarantee good health across the entire cohort. A subset of fish can grow well while others display injury, abnormal behavior or failure to thrive. A mean can even improve following the loss of smaller individuals. Combine growth monitoring with survival, health observations, water-quality records and, where appropriate, professional aquatic animal health assessment.
An honest explanation preserves a clear boundary: growth metrics describe measured changes in mass and their statistical uncertainty. Causes require separate evidence. A rate is a clue in a diagnostic story, not a laboratory confirmation of a pathogen, an approval of husbandry or a guarantee that welfare standards are being met.
Interpreting graphs and axes
A growth chart should have an unambiguous vertical axis: mean individual wet weight in grams, total standing biomass in kilograms, or an index such as SGR percent per day. These axes are not interchangeable even when they all rise with performance. The horizontal axis should show dates or days since a stated start, not unlabeled sampling occasions if intervals differ.
Error bars must say whether they are standard deviations, standard errors, confidence intervals or variation among independent tanks. A standard deviation describes spread among sampled animals; it does not directly quantify uncertainty about a treatment effect. Confidence intervals depend on the model and design. A graph can be visually persuasive while conveying an entirely different scientific statement from the one the caption claims.
If the chart uses a logarithmic weight axis, equal vertical distances represent equal ratios rather than equal gram changes. A straight trend in log weight suggests approximately constant proportional growth under the interval considered. That is useful diagnostically, but it does not prove the underlying growth mechanism. Always check whether an apparent acceleration was created merely by the axis scale.
Translation traps around percent and percentage points
When SGR changes from 1.2 to 1.5 percent per day, the absolute change is 0.3 percentage points per day. The relative increase in the numerical SGR is 0.3 divided by 1.2, or 25 percent. Writing that the growth rate improved by 0.3 percent is ambiguous and often wrong. State which comparison the author intends.
The phrase daily growth can also describe grams per day, a log-compounded rate, a discrete compounded rate or a percentage of the original mass spread across days. Translators must attach the formula and denominator, not just choose the smoothest sounding words. If one language omits units casually, restore them from the source table only when supported by the author, and do not invent a missing time basis.
Likewise, thermal coefficient should not be paraphrased as temperature-corrected biological efficiency without caveats. The equation normalizes selected changes by a chosen temperature exposure. It does not verify causal independence from temperature. A translation that turns an analytical index into a physiological mechanism can materially change what the data warrant.
A compact audit checklist for any growth paper
Start with the biological population: species, life stage, size range, stocking date, survival and any transfers. Then inspect how individuals were selected for weighing, whether they were fasted consistently, and whether measurements were paired. A defensible mean is inseparable from its sampling procedure. It is possible to calculate a flawless rate from an unrepresentative set of fish; arithmetic correctness does not imply representativeness.
Next reconstruct the equations and denominators. For every reported SGR, record the exact formula, initial and final weight units, day count, and whether the rate is logarithmic or discrete. For TGC, confirm the cube-root weight term, coefficient factor, temperature units, averaging interval and any base-temperature modification. Rerun at least one result with independent arithmetic and check it against the reverse equation.
Finally ask what decision the numbers can support. A monitoring question needs accurate within-farm trends. A controlled feed comparison requires replication and reasonable confounding control. A production forecast needs explicit assumptions and uncertainty. If the data are too thin for a causal claim, a good report says what remains unresolved. Such caution is a sign of technical competence, not weakness.
Independent transfer problem: a short interval
Try this before reading the solution. A tank’s representative sampled mean rises from 30 grams to 36 grams in 12 days. Calculate absolute gain, mean daily gram gain, cumulative percentage gain and logarithmic SGR. Then explain why none of these measures, by itself, proves that the feed was responsible for growth. The data are hypothetical and the same sampling definition is assumed at both dates.
Answer: absolute gain is 6 grams and daily gram gain is 0.5 grams per day. Cumulative percentage gain equals 100 times 6 divided by 30, or 20 percent across twelve days. Logarithmic SGR equals 100 ln(36 divided by 30) divided by 12. Because ln(1.2) is approximately 0.18232, the rate is about 1.519 percent per day. The discrete daily compound value is about 1.531 percent per day.
The results describe observed endpoint change. They do not isolate feed effects because there is no comparison group, no randomized treatment, and no temperature or health context. A high-quality translation retains the qualifier representative sampled mean; it does not convert the result into every fish gained 20 percent. The phrase supports a population estimate, not individual certainty.
Independent transfer problem: thermal exposure
A hypothetical batch begins at 64 grams and ends at 125 grams after twenty days. Its daily mean water temperature is constant at 25 degrees Celsius for this exercise. Calculate the unadjusted TGC using grams, cube roots and a factor of 1,000. Then explain why a paper using a base-temperature degree-day equation would not necessarily obtain the same coefficient.
Answer: the cube root of 64 is 4, and the cube root of 125 is 5. The difference is 1. The denominator is 20 times 25, or 500 degree-Celsius days. TGC is 1,000 times 1 divided by 500, which equals 2 under this definition. No percent sign belongs on the result. The value is not 2 percent growth per day.
A base-temperature model would alter thermal accumulation by subtracting a specified threshold before integration. If it used ten degrees as a purely illustrative threshold, the denominator would become twenty times fifteen, or 300 base-adjusted degree-days, and the same numerator with factor 1,000 would give 3.333. The different value arises from a different equation, not a changed fish.
Independent transfer problem: survival changes production
A facility reports 400 fish averaging 50 grams at the start and 350 fish averaging 70 grams at the end of a 30-day interval. Find initial and final live biomass, the mean-weight SGR and the net live biomass gain. State why a report based only on the mean-weight rate would be incomplete. The example deliberately does not assign weights to mortalities or removed animals.
Answer: initial live biomass is 400 times 50 grams, or 20 kilograms. Final live biomass is 350 times 70 grams, or 24.5 kilograms. Net standing biomass gain is 4.5 kilograms. Mean-weight logarithmic SGR is 100 ln(70 divided by 50) divided by 30, approximately 1.122 percent per day. The 40 percent increase in survivor mean weight is not a 40 percent increase in live stock biomass.
The missing explanation is the fate of fifty animals and their mass when lost. Without removals and mortalities, the production account is incomplete. This does not make the SGR calculation false; it clarifies what it does and does not measure. A sound operating record would preserve both the size-growth estimate and a reconciled population ledger.
Common mistakes and their repairs
One frequent mistake is to use the logarithm of final weight minus initial weight instead of the logarithm of the ratio. Logarithms do not distribute over subtraction. The correct expression is ln(final divided by initial), or ln(final) minus ln(initial). Another mistake is dividing by the number of sampling dates rather than elapsed days. Four weekly observations over three weeks define twenty-one days, not four days.
A second family of errors concerns units. TGC based on grams cannot be compared directly with a coefficient calculated from kilograms if the formula factor was not adjusted. An SGR ratio is unitless as long as both endpoint masses use the same unit, but a grams-to-kilograms mismatch creates a thousandfold ratio distortion. Standardize units before calculating, and attach units to every original input field.
A third family involves narrative overreach: calling a growth increase a feed benefit without control groups, identifying disease from poor growth alone, or comparing an early juvenile rate to a market-size rate as though the denominators were biologically equivalent. Repair these by clearly distinguishing measurement, mathematical summary and causal inference. When evidence is missing, write the limitation explicitly.
FAQs: which growth metric is best?
There is no single best metric for every task. Absolute grams per day helps with practical size or harvest planning. Cumulative percent gain is intuitive for describing a fixed interval. Logarithmic SGR is useful for comparing proportional changes over specified durations, subject to body-size and design limitations. TGC offers a particular thermal-normalized framework when its assumptions and formula are appropriate.
Can a negative SGR occur? Yes. Positive initial and final weights with the final value smaller than the initial yield a negative log ratio. Is TGC a mortality-corrected production measure? No. It uses mass endpoints and thermal exposure under a selected equation. Does a higher SGR automatically mean better welfare or profitability? No. Both require additional evidence and different outcome measures.
How many decimal places should be reported? Enough to make meaningful comparisons, but not more than the input quality can support. What happens when samples have different counts? The mean can still be computed, but selection, sampling variance and cohort changes should be reviewed. Why use logarithms? They summarize multiplicative change and allow the same relative growth relationship to be added across consecutive time intervals when the cohort and definitions stay consistent.
Where to go next
Good growth interpretation is a chain of questions rather than a contest between indices. First establish who was measured and when. Next inspect units and definitions. Reproduce the arithmetic, assess sampling error and temperature records, then reconcile survival and biomass. Only after those checks should the reader ask what management response is justified. That order prevents attractive percentages from outrunning the evidence.
For neighboring concepts, consult the feed digestibility and inert-marker guide and the dissolved oxygen guide. They cover separate parts of the biological context. Our How to Translate Easily series index is a wider navigation point, not evidence for the formulas above.
All numerical examples here are teaching calculations created for this article. They are not field measurements or operating recommendations for a particular species. Readers making high-stakes husbandry decisions should use species-appropriate protocols, local water conditions and qualified aquatic animal health advice. The durable skill is to preserve the boundaries of what a rate measures, what a model assumes and what a decision still requires.
Sources and method references
FAO, Recirculating Systems and Re-use of Water in Aquaculture, Chapter 15.
Google Search Central, Creating Helpful, Reliable, People-First Content.