Fish assemblages in estuaries are monitored for many reasons: to judge ecological status under water quality legislation, to follow nursery function for commercial species, and to detect arrivals of non-native fish. The conventional tools are seine, fyke and trawl nets. In muddy, tidal water each of these catches a different fraction of the fauna, none can be used over soft mud, mangrove roots or oyster reef without difficulty, and all cause mortality among juveniles that the survey is meant to protect.

Environmental DNA (eDNA) metabarcoding offers a different route. Water is filtered, the DNA retained on the filter is extracted, a short gene region shared by the target group is amplified with general primers and sequenced in bulk, and the sequences are matched to a reference library (Deiner et al., 2017). The approach has given species lists for fishes that match or exceed those from established methods in lakes and rivers (Valentini et al., 2016), and in an urban estuary it tracked the seasonal turnover of the fish fauna (Stoeckle et al., 2017).

Turbid estuaries nevertheless strain the method at several points, and practitioners who follow a clear-water protocol unchanged are likely to obtain small sample volumes, failed amplifications and results that are hard to interpret. This article proposes a procedure designed for these conditions. It is assembled from published methods and from the general guidance on eDNA practice (Goldberg et al., 2016). It is not a report of its application, and no field results are presented.

Three properties of estuarine water matter. The first is the load of suspended particles. Much of the eDNA of fish and other animals occurs in or on particles larger than about one micrometre (Turner et al., 2014), so it is captured by filtration, but the silt, clay and phytoplankton that are captured with it block the filter. In turbid water a fine membrane may pass only a small fraction of the volume routinely filtered in clear water, and the amount of target DNA collected falls accordingly (Williams et al., 2017).

The second is chemistry. Estuarine sediments and the rivers that feed them are rich in humic and fulvic acids, which are co-extracted with DNA and inhibit the polymerase used in amplification. An inhibited reaction yields nothing, and unless the inhibition is recognised the result is read as an absence. Dilution of the extract and passage through a purification column both relieve inhibition, the former at the price of diluting the target (McKee et al., 2015).

The third is movement and persistence. Tidal currents carry DNA away from its source, the salinity and temperature gradients that structure an estuary also affect how fast DNA decays (Barnes and Turner, 2016), and decay rates differ between inshore and offshore waters (Collins et al., 2018). Sediment adds a further complication. Fish DNA is more concentrated and longer lived in bottom sediment than in the water above it (Turner et al., 2015), so resuspension by tides, wind or boat traffic may return old DNA to the water column. Table 1 summarises these problems and the countermeasure proposed for each.

The procedure has eight steps, shown in Figure 1. The first three are preparatory and are done once for an estuary; the remaining five are repeated at every survey. A return loop from the pilot allows the filtration and replication to be revised before routine sampling begins.

A programme adopting the procedure would need to settle several practical matters. Cost is driven mainly by the number of PCR replicates and by sequencing depth, and the pilot gives the information needed to choose both. Field teams need little specialised training, but the laboratory must separate its pre-amplification and post-amplification work physically, as recommended in general guidance (Goldberg et al., 2016). Consistency between years depends on holding the marker, the filter arrangement, the tidal state and the bioinformatic thresholds constant, and on archiving extracts so that samples can be re-run if methods change.

The method complements netting and does not replace it. eDNA gives no information on size, age, condition or reproductive state, which estuarine assessments often require. A sensible arrangement is to use eDNA for broad and frequent coverage of species occurrence and to keep a reduced netting programme for the demographic information that only specimens provide.

The procedure is proposed and has not been validated here. Its components are each supported by published work, but most of that work was done in fresh or clear coastal water, and their combined performance in highly turbid estuaries is untested. A validation would pair eDNA sampling with multi-gear netting at the same stations and tides over at least one annual cycle, and would report the species detected by each method, the detection probabilities estimated for eDNA, and how both vary with turbidity.

Some uncertainties cannot be designed away. The spatial resolution of a detection in a tidal system is coarse, and a species recorded at a station may have been some distance upstream or downstream. The contribution of resuspended sediment to the signal has not been quantified for estuaries. Closely related species, for example among mullets, gobies and some clupeids, may not be separable with a short 12S fragment. Users of the results should be told of these limits in plain terms.

Turbidity does not rule out eDNA metabarcoding as a tool for monitoring estuarine fish, but it changes what a sound protocol looks like. The eight steps proposed here place a filtration pilot, an explicit inhibition test, replicated amplification and occupancy modelling at the centre of the design, and tie sampling to salinity and tide. Whether the procedure delivers reliable species lists in practice remains to be demonstrated by comparison with conventional surveys.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.