Fish eDNA: ecological traits remain inferred

Add ecological context without changing the nature of the evidence

Environmental DNA can reveal fish genetic traces in water samples. Once taxa have been identified, their known ecological characteristics can be added from reference sources: habitat, trophic level, migration or maximum body length. This produces a functional view of the detected assemblage, but the trait information comes from outside the water sample.

He and colleagues applied this approach in the Taiwan Strait in a study published in March 2026 in Frontiers in Marine Science. They combined metabarcoding of 33 water samples with traits drawn from FishBase. For environmental managers, the study illustrates how species records can gain ecological context. It also makes the boundary between molecular detection, reference information and ecological inference especially important.

Keep the two information chains visible

Sampling took place in August 2022. The authors compared inshore, midshore, offshore and principal-axis areas of the strait. Water samples were analysed using a mitochondrial 12S marker targeting fish. Taxonomic assignments depended on sequence quality and the reference sequences available for comparison.

The researchers then obtained ecological traits from FishBase. These covered vertical distribution, habitat preference, trophic level, migratory behaviour and maximum body length. The laboratory did not measure these characteristics in the individual animals that contributed DNA to the water. It linked a molecular detection to information already associated with the identified taxon.

Body size makes the distinction easy to understand. A profile dominated by species classified as small describes the detected taxonomic composition in relation to size categories. It does not measure the lengths or ages of fish present at the sampling point. Nor does it independently establish the proportion of juveniles. Reporting it as an observed size distribution would change the meaning of the result.

Species richness and functional diversity answer different questions

Biodiversity can be described through species numbers, composition and the range of characteristics represented. Assemblages with similar species richness can differ in feeding strategies or use of space. Conversely, several species may share the characteristics selected for an analysis.

The study identified 206 species and assigned them to 111 functional entities, each defined by a combination of traits. Of those entities, 78 contained only one species in the inventory. This describes the distribution of detected species under the chosen classification. It does not establish that every ecosystem function has been measured.

Functional redundancy also depends on which traits are included and how finely they are described. Species that sit close together in a selected trait space may still differ in other ecologically important ways. Sharing a few categories therefore does not demonstrate that one species could replace another’s contribution in the actual ecosystem.

For a monitoring programme, this means that the trait list is an analytical choice requiring justification. It should be linked to the question being investigated, with missing information and uncertainty retained rather than hidden behind a single diversity score.

Habitat associations are not direct observations of behaviour

A habitat category in a database describes knowledge about a species. Its appearance in an eDNA profile does not show that each contributing individual occupied that exact habitat at the time and position of sampling. Fish movement and transport of the genetic signal through water complicate that interpretation.

The Taiwan Strait is influenced by currents and coastal environmental gradients. The authors examine relationships between community profiles and environmental variables. These associations can help formulate hypotheses about assemblage distribution. Alone, they cannot establish that a particular pressure caused a community change.

Field records therefore need to retain location, sampling date, depth and relevant conditions beside each result. Environmental measurements provide additional context. Comparing campaigns becomes difficult when the sampling design or habitat coverage changes without being accounted for in the analysis.

A comparable programme in another region would need its own assessment of reference coverage. A taxon reliably identified in one setting may be harder to resolve elsewhere, and uncertain identification carries through into the traits assigned to it.

Read co-occurrence networks as analytical models

The study also constructs networks connecting taxa whose detections vary in a correlated way. Such representations help explore statistical organisation within the dataset. A connection between species is not a direct observation of predation, competition or cooperation.

Two taxa may co-occur because they share favourable conditions, because their signals are transported together or because of other features of sampling and analysis. Networks are therefore useful for exploring patterns and forming hypotheses. They do not independently demonstrate an ecological mechanism.

The same caution applies to stability claims. A functional index or a network structure cannot replace observations of a system responding to disturbance and subsequently recovering. Resilience is a property to investigate, rather than a direct measurement obtained from a sequence inventory.

This distinction does not remove the value of the approach. It identifies which conclusions can guide further checks and which would require independent field evidence before informing a management intervention.

Design the analysis around a management question

Before commissioning a trait analysis, specify the intended comparison. The aim might be to compare habitats, examine replacement among ecological groups or track representation of particular strategies. That purpose should guide trait selection. Producing more indices without a defined decision can make the report less useful.

Traceability needs to cover both taxonomic identification and trait assignment. Uncertain names, missing values, reference sources and database versions should remain accessible. Sensitivity analysis can test whether conclusions change when poorly resolved taxa or incompletely described traits are excluded.

The study reports a regional marine campaign. It does not automatically validate an index for French rivers or every coastal system. Transfer requires suitable taxonomic references, adequate trait coverage and a monitoring design matched to the environment concerned.

Conclusion and Vetofish support

Combining eDNA with ecological traits broadens the description of detected fish assemblages. Its usefulness depends on distinguishing measurements, reference information and hypotheses about ecosystem functioning.

Vetofish can help define these questions and connect analytical outputs with field needs. Our advisory and support service contributes to our environmental work by reviewing monitoring design, interpretation limits and the additional observations needed to support decisions.