Can Fish Counts Be Trusted? Evidence Shows Survey Method Can Strongly Bias Estimates of Abundance and Biomass

The challenge of counting fish

The challenge of counting mobile animals underwater

Underwater visual censuses are one of the most widely used tools in marine ecology. They are used to estimate fish abundance, compare ecosystems, and assess the effectiveness of marine protected areas.

But these surveys rely on a key assumption: that what is counted represents what is actually present in a fixed area at a fixed moment in time.

In reality, that assumption is often not met.

Many surveys are not truly instantaneous. Fish can enter the survey area after counting has begun, and fast-moving species are more likely to be included than slow or sedentary ones.

This raises a fundamental question: How much do survey methods themselves shape what we think we know about marine populations?

What this study actually does

In “Overestimating Fish Counts by Non-Instantaneous Visual Censuses: Consequences for Population and Community Descriptions” (PLOS ONE), we used a simulation-based approach to test how underwater visual census methods behave when animals differ in movement speed.

We focused on two commonly used survey types:

• belt-transect surveys

• stationary point-count surveys

We then modelled how fish movement, survey time, visibility, and diver speed influence whether animals enter or leave the survey area during the census period.

The goal was to isolate a simple but often overlooked mechanism: what happens when counts are treated as “instantaneous” even when they are not.

What the study shows

Across all simulated scenarios, one pattern was consistent: Mobile animals are more likely to be overcounted in non-instantaneous surveys relative to stationary animals.

This happens because individuals entering the survey area after the census begins are included in the final count, even though they were not present at the start — this does not happen for stationary animals.

The result is a systematic inflation of observed abundance that increases with animal movement speed.

Several key outcomes emerge.

1. Faster fish are more strongly overestimated. Bias increases with swimming speed. Slow-moving or stationary species are relatively unaffected, while highly mobile species—such as many sharks, jacks, and large reef predators—are most strongly overrepresented.

This creates a structural imbalance in survey data, where detectability is not the only factor shaping results—movement itself becomes a source of bias.

2. Survey design determines the magnitude of error. The extent of overestimation depends on how the survey is conducted. Longer survey times, wider transects, lower visibility, and slower diver speeds all increase the probability that animals will enter the sampling area during observation. This means that two studies using similar methods but different field conditions can produce systematically different results, even if underlying populations are identical.

3. Community structure can be distorted. Because large-bodied species tend to be more mobile, the bias is not evenly distributed across ecological groups. This has important implications for how reef communities are described. Mobile predators may appear disproportionately abundant, while slower-moving species contribute less to observed totals than they should. As a result, community composition and trophic structure can be skewed by method alone.

4. Biomass estimates amplify the effect. When abundance data are converted into biomass, the bias increases further. Large, mobile animals contribute disproportionately to total biomass estimates. If those animals are already overrepresented due to movement-related sampling bias, the resulting biomass structure can be substantially distorted. This can influence interpretations of ecosystem structure, including claims of unusually predator-heavy systems or “inverted biomass pyramids.”

Why this matters for marine science

These findings highlight a key issue in ecological monitoring: survey design can directly shape ecological conclusions.

In systems where mobile predators are a focus of conservation concern, even moderate overestimation can influence how ecosystems are interpreted and managed.

This affects:

• baseline population estimates

• comparisons between protected and unprotected areas

• assessments of recovery or decline and

• broader interpretations of ecosystem structure

Importantly, the issue is not that visual censuses are invalid—but that their limitations must be explicitly understood when translating counts into absolute ecological metrics.

The key insight: it is measurement error

This study shows that animal movement is not only a biological trait—it is also a sampling process that interacts directly with survey design. When fish are mobile, and they are counted in a census when they were not there when the census began, their density — and therefore, biomas — is artificially inflated relative to the stationary species.

This means that “abundance” and “biomass” can be improperly represented.

What changes now

These results reinforce a broader principle in marine monitoring: from assuming surveys capture static snapshots
to recognizing that most field methods are time-integrated observations of moving systems. This shift has direct implications for how long-term datasets are interpreted and compared across regions and methods.

It also highlights the importance of clearly reporting survey structure—particularly whether counts are truly instantaneous or not—so that results can be interpreted in the correct methodological context. Where possible, complementary approaches such as tagging, photo identification, or repeated standardized surveys can help ground-truth abundance estimates for highly mobile species.

Frequently asked questions

Are visual censuses still useful? Yes. They remain one of the most efficient, non-destructive, and common tools for assessing communities, especially for relative comparisons.

When do problems arise? When counts are treated as absolute densities without accounting for fish movement and survey timing.

Which species are most affected? Highly mobile species, particularly fast moving, large species.

Does this invalidate past studies? Not necessarily. Many comparative results remain valid, but absolute abundance and biomass estimates may require reinterpretation.

Final thought

Counting fish is hard — measurement itself can reshape our understanding of how marine ecosystems are structured.

Read the full study

Published in PLOS ONE:
Overestimating Fish Counts by Non-Instantaneous Visual Censuses