Evidence · Fish & aquatic ecology

European eel: a 92% crash — and a fragile recovery

European eel, a Critically Endangered species, crashed in the tidal Thames between Teddington and Battersea: catch per seine haul fell from a mean of 1.22 in 1992–99 to 0.09 in 2008–18, a 92% fall, with every year from 2005 to 2018 below 0.20. But the species did not disappear. 506 eel are recorded across 1989–2025, and since 2019 the series has lifted again — 2025 is the strongest seine year since 1997. Whether that is recovery, survey effort or noise on small numbers is not yet settled, and we say so rather than guess.

EA Ecology & Fish Data Explorer Richmond & Teddington · tidal reach WFD body GB530603911403 RAW-PULL

Corrected 31 August 2026

An earlier version of this page, published 31 July 2026, said that 12 European eel were caught in this reach in 1993 and none after 2019, and was headlined “the eels have gone”. That was wrong. The Environment Agency’s full TraC fish release, which reached the bulk-download directory on 10 August 2026, records 79 eel in 1993 and 506 across 1989–2025, including catches in 2021, 2022, 2023 and 2025.

Two things caused the error. The dataset was not available in usable bulk form when the page was written. And the EA splits eel across three species labels — European eel, European eels > elvers and European elvers — and we had filtered on one. The second mistake was ours alone.

The correction was prompted by the Environment Agency marine ecologist who has advised this programme, who wrote to us on 7 August 2026 to say that he had caught several young eels in kick nets at Ham during the November 2025 Richmond lock drawdown, and that eels of that size would probably swim straight through a normal seine mesh. He was right, and we are grateful. The page now says what the data says.

EA seine & kick-net surveys · tidal reach

The headline figures

The Environment Agency has seine- and kick-netted fish at Richmond and Teddington on the tidal Thames since the early 1990s. The long record is the European eel's near-disappearance — set against a fish community that is otherwise still functioning. These are the numbers the raw survey counts give for the reach.

92%
fall in European eel per seine haul between 1992–99 (mean 1.22) and 2008–18 (mean 0.09). Every one of the fourteen years 2005–2018 sits below 0.20.
Mean of annual catch per seine haul DERIVED from RAW-PULL
506
European eel recorded across the tidal reach in 1989–2025, in 175 records — including 79 in 1993, the largest single year.
EA Ecology & Fish Data Explorer · TraC fish counts RAW-PULL
1.17
eel per seine haul in 2025 — the highest since 1997. 2019, 2021, 2023 and 2025 all sit well above the 2005–18 trough. Whether that is recovery, effort or noise is not yet settled.
Annual catch per seine haul DERIVED from RAW-PULL
0
published biological index values — BMWP, WHPT, AMBI or IQI — for the marine benthic samples, which cover the entire Ham invertebrate series from 2006 to date. Taxa ship without metrics. Update, 14 October 2026: the EA has since generated BMWP and ASPT for Ham on request; see Invertebrates.
TC_BENT_OPEN_DATA_TAXA, reach extract PROVEN GAP

European eel per seine haul, tidal reach, 1989–2025

European eel per seine haul, Teddington Weir to Battersea

Seine netting is the only method in this reach with a consistent unit of effort — 401 of 719 surveys. Raw annual totals are confounded by effort (4 surveys in 1989, 35 in 2014), so catch per haul is the honest measure. Water bodies GB530603911403 (THAMES UPPER) and GB106039023232.

European eel per seine haul in the tidal Thames, Teddington to Battersea, 1989 to 2025 Vertical bar chart of European eel caught per seine haul each survey year, 1989 to 2025. The 1990s run high, peaking at 3.75 in 1997. From 2005 to 2018 every year sits below 0.20, shaded as the trough. From 2019 the series lifts again: 0.40 in 2019, 0.63 in 2021, 0.25 in 2023 and 1.17 in 2025, the highest since 1997. The shape is a steep decline into a long trough followed by a partial and uneven recovery, not a fall to zero. 2005–2018: every year below 0.20 01234 1989: 0.50 eel per seine haul (2 eel, 4 hauls)1992: 1.62 eel per seine haul (21 eel, 13 hauls)1993: 1.26 eel per seine haul (79 eel, 19 hauls)1994: 0.56 eel per seine haul (13 eel, 9 hauls)1995: 0.56 eel per seine haul (10 eel, 9 hauls)1996: 1.43 eel per seine haul (21 eel, 7 hauls)1997: 3.75 eel per seine haul (51 eel, 8 hauls)1998: 0.38 eel per seine haul (8 eel, 8 hauls)1999: 0.25 eel per seine haul (2 eel, 8 hauls)2000: 0.00 eel per seine haul (0 eel, 5 hauls)2001: 0.12 eel per seine haul (3 eel, 8 hauls)2002: 1.00 eel per seine haul (14 eel, 9 hauls)2003: 0.44 eel per seine haul (36 eel, 9 hauls)2004: 0.50 eel per seine haul (14 eel, 4 hauls)2005: 0.13 eel per seine haul (60 eel, 15 hauls)2006: 0.12 eel per seine haul (28 eel, 8 hauls)2007: 0.16 eel per seine haul (12 eel, 19 hauls)2008: 0.06 eel per seine haul (2 eel, 17 hauls)2009: 0.11 eel per seine haul (6 eel, 19 hauls)2010: 0.06 eel per seine haul (28 eel, 18 hauls)2011: 0.12 eel per seine haul (16 eel, 16 hauls)2012: 0.15 eel per seine haul (3 eel, 20 hauls)2013: 0.07 eel per seine haul (7 eel, 14 hauls)2014: 0.10 eel per seine haul (2 eel, 20 hauls)2015: 0.18 eel per seine haul (3 eel, 17 hauls)2016: 0.00 eel per seine haul (0 eel, 8 hauls)2017: 0.06 eel per seine haul (8 eel, 16 hauls)2018: 0.13 eel per seine haul (14 eel, 15 hauls)2019: 0.40 eel per seine haul (10 eel, 15 hauls)2021: 0.62 eel per seine haul (5 eel, 8 hauls)2022: 0.06 eel per seine haul (4 eel, 16 hauls)2023: 0.25 eel per seine haul (4 eel, 8 hauls)2024: 0.00 eel per seine haul (0 eel, 6 hauls)2025: 1.17 eel per seine haul (20 eel, 6 hauls) 2025: 1.17 — highest since 1997 19901995200020052010201520202025 Survey year · European eel per seine haul (401 of the 719 surveys in this reach)
European eel in the tidal reach, Teddington Weir to Battersea (BNG easting 512000–528000, northing 168000–180000): 25 site records, 719 surveys, 2,574 rows, 1989–2025. Eel counts combine the three EA species labels — European eel, European eels > elvers and European elvers. Catch per haul is seine only, the one method here with a consistent unit of effort. 1990, 1991 and 2020 were not surveyed. Provenance: RAW-PULL.
Year Surveys Seine hauls Eel (all methods) Eel per seine haul
19894420.50
19921413211.62
19932919791.26
1994199130.56
1995189100.56
1996137211.43
1997148513.75
199817880.38
199912820.25
20007500.00
200114830.12
2002179141.00
2003239360.44
2004104140.50
20052715600.13
2006158280.12
20073419120.16
2008311720.06
2009331960.11
20103218280.06
20113116160.12
2012332030.15
2013271470.07
2014352020.10
2015311730.18
201615800.00
2017301680.06
20182815140.13
20192715100.40
202113850.62
2022301640.06
202314840.25
202412600.00
2025106201.17

For contrast, freshwater surveys at Molesey (sites 8084/8085, about 6 km above the tidal limit) still catch eels sporadically — a peak of 39 in 2002, then mostly one to three a year — so the disappearance is sharpest in the tidal reach itself. DERIVED from the freshwater fish counts.

The eel aside, the reach's fish community is otherwise still functioning. The 32-year Richmond Seine Net series (98 survey events, 1992–2025; 19,940 fish) is dominated by dace (8,601) and roach (8,055), with flounder (1,354), common goby (622) and sea bass (357), plus scarce smelt and a handful of salmon — an estuarine, diadromous signal, not a dead river. DERIVED from RAW-PULL.

Is the official data good enough / truly open?

The tidal ecology "black spot" — published, but effectively invisible

A common claim about this reach is that the raw tidal fish and ecology data below Teddington (the "TraC" transitional-and-coastal datasets, Mucking → Richmond) is simply not published. Checked against the actual downloads, that claim is largely wrong — and the real failure is more precise, and more interesting, for the question of whether the official evidence base is genuinely usable.

The raw TraC fish counts — including the full 1989–2025 seine series charted above — are published in the EA's bulk download zips, and so are the raw TraC benthic-invertebrate taxa (100+ samples at Isleworth and Richmond). So this is not an availability failure. What is genuinely missing is the connective tissue:

  • No TraC site-summary metadata. The freshwater datasets each ship a site-summary file (how many samples, first and last dates). The TraC phytoplankton and benthic-invertebrate datasets ship none — so sample counts and date ranges are absent, and the Explorer map cannot filter or surface TraC sites at all. A casual user searching the map concludes "there is no data here," when a 32-year fish series and hundreds of benthic samples sit in the bulk zips. A discoverability / usability failure, not an availability one.
  • One genuine availability gap. TraC phytoplankton is not sampled on this reach at all (0 samples), macrophytes have 1 site / 2 samples (2023–24), and diatoms have 1 site / 6 samples ending in 2014. WFD ecological status for these elements therefore rests on very thin or absent local evidence.
Reach data-adequacy — mainstem Thames, Teddington–Richmond, by WFD biological element. Where an element has one site or none, its classification rests on very thin local evidence. Provenance: DERIVED from RAW-PULL (reach_data_adequacy_summary.csv).
Biological element Mainstem sites Surveys / samples Span
Fish (freshwater / non-tidal)10571995–2025
Fish (TraC / tidal)61421989–2025
Invertebrates (freshwater)61731983–2026
Invertebrates (TraC benthic)22192005–2026
Macrophytes122023–2024
Diatoms (phytobenthos)162007–2014
Phytoplankton (TraC)00none

Highlighted rows are the thin ones: three of the seven WFD biological elements on this reach rest on a single site or no local sampling at all. Fish and invertebrates, by contrast, have long, genuinely useful records — they are just hard to find for the tidal reach.

How to read this

Caveats a regulator would press — stated up front

  • Small numbers. The largest single-year eel total in 37 years is 79, in 1993; several years are between 0 and 5 fish. Confidence intervals are wide. Do not read year-on-year percentage changes off this series — the periods are the story, not the individual years.
  • Effort can only be partly controlled. SURVEY_LENGTH, SURVEY_WIDTH and SURVEY_AREA are populated in just 2.3% of rows for this reach, so a true catch-per-unit-area cannot be computed for the tidal Thames at all. Our figure is per seine haul, which assumes hauls are comparable; the EA publishes nothing that lets us verify that assumption.
  • Gear matters, and we do not mix it. Four methods appear in the reach — seine netting (1,925 rows), 1.5 m beam trawl (447), kick netting (198) and 2.0 m beam trawl (4). Species composition differs sharply by method, so the chart uses seine only. Totals in the table are all methods and are not comparable between years.
  • An empty net is not an empty river. The EA marine ecologist advising this programme wrote on 7 August 2026 that while kick-netting at Ham during the November 2025 Richmond lock drawdown his team caught several young eels incidentally, and that eels of that size “would probably swim straight through a normal seine net mesh.” Seine surveys under-detect small eel. Absence of catch is not absence of fish.
  • The recovery since 2019 is not established. Four years above the trough, on small numbers and variable effort, is a signal worth reporting and not a trend worth asserting. We are seeking expert review of it rather than publishing a conclusion.
  • The eel decline is international, not local. European eel recruitment is down more than 90% Europe-wide and the species is IUCN Critically Endangered. This reach mirrors a continental decline. It is a wildlife-harm signal; on its own it does not prove a local cause or attribute anything to any one discharger.
  • “TraC data is unpublished” is no longer true. As of the 6–10 August 2026 release the TraC fish data is published in full, 1989 to 2025, species-level, downloadable. The remaining gaps are narrower and we state them precisely: no published metrics file for marine benthic invertebrates (taxa only, for the entire Ham series); no derivable catch-per-unit-area; no site names in any Biosys dataset, only grid references; and third-party surveys such as ZSL’s are explicitly excluded from the NFPD dataset.
  • Tidal biology is otherwise healthy. Under estuarine WFD tools the reach’s fish and invertebrates score Good; THAMES UPPER’s Moderate status is driven by chemistry, specific pollutants and hydrology, not a dead biological community. The eel is one struggling species within a still-working assemblage.

Sources & provenance

  • RAW-PULL  EA Ecology & Fish Data Explorer — bulk downloads https://environment.data.gov.uk/ecology/explorer/downloads/ (documentation /ecology/explorer/how-to-use/). National zips downloaded 2026-07-31 (files dated 2026-07-05/06), filtered to the reach bounding box (lat 51.40–51.48, lon −0.35 to −0.28; OSGB→WGS84 via pyproj). Open Government Licence; England only.
  • Tidal sites: Richmond Seine Net SITE_ID 20562 (98 events, 1992–2025), Richmond Kick Sample 20560 (37 events), Teddington Weir Kick 20625; WFD body GB530603911403. Freshwater contrast: Molesey 8084/8085, ~6 km above Teddington.
  • Working files: /tmp/thames_datastore/fish_ecology/reach_TRAC_fish_counts.csv, reach_FW_fish_counts.csv, reach_data_adequacy_summary.csv, reach_sites_TRAC.csv, and the domain NOTE.md.
  • Conservation status: European eel (Anguilla anguilla) — IUCN Critically Endangered; >90% fall in European recruitment since the 1980s. PUBLISHED