Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 50 of the 51 references it cites were found in a registry or an index; none of the 111 sources its claims cite is checked for support yet.

Modelling remote barrier detection to achieve free-flowing river targets

Open access
Parks, Millicent V, Garcia de Leaniz, Carlos, Jones, Peter E, Jones, Josh · 2024 · Environmental Research Letters

Abstract

Fragmentation caused by artificial barriers is one of the main stressors of rivers worldwide. However, many barrier inventories only record large barriers, which underestimates barrier numbers, and hence fragmentation. Corrected barrier numbers can be obtained via river walkovers, but these are costly and time consuming. We assessed the performance of remote sensing as an alternative to river walkovers for barrier discovery by comparing the number and location of barriers detected in the field with those detected using Google Earth imagery. Only 56% of known barriers could be detected remotely, but machine learning models predicted the likelihood of remote detection with 62%–65% accuracy. Barriers located downstream were twice as likely to be detected remotely than those in the headwaters, the probability of detection diminishing by 3%–4% for every decrease in Strahler stream order and for every 10 km increase in distance from the river mouth. Barriers located in forested reaches were 35% less likely to be detected than those in open reaches. Observer skills also affected the ability to locate barriers remotely and detection rate varied by 11% between experienced and less experienced observers, suggesting that training might improve barrier detection. Our findings have implications for estimates of river fragmentation because they show that the most under-represented structures in barrier inventories, i.e. small barriers located in forested headwaters, are unlikely to be detected remotely. Although remote sensing cannot fully replace ‘boots on the ground’ field surveys for filling barrier data gaps, it can reduce the field work necessary to improve barrier inventories and help inform optimal strategies for barrier removal under data-poor scenarios.

Overview

CoChat Summary

TL;DR

Remote sensing via Google Earth detected only 56% of known river barriers, and machine-learning models predicted remote detection with 62%–65% accuracy. Detection was strongly biased toward downstream, open, larger-order reaches, meaning small barriers in forested headwaters are most likely to remain missing from inventories.

Key findings

  • Across both datasets, 56% of known barriers were detected remotely. In the exhaustive River Afan walkover, observers detected on average 50% of barriers (47%–55% by observer), and 26% were not detected by any of the three observers; 44% of barriers could not be detected remotely, mostly small structures <2 m in height.
  • Machine-learning models predicted whether a barrier could be detected remotely with 62%–65% accuracy: logistic regression 62% (95% CI 57%–67%), random forest 57% (95% CI 48%–65%), and adaptive boosting 65% (95% CI 60%–69%). Logistic model predictive power was low (R² Tjur = 0.09, P < 0.001).
  • Detection declined upstream and in forest: barriers near the river mouth had P-detection = 0.70 versus 0.33 in headwaters, with detection decreasing by ~3% per 10 km upstream; detection fell from P = 0.90 in sixth-order reaches to P = 0.64 in first-order streams, diminishing by ~4% per decrease in Strahler stream order. Barriers in forested reaches were 35% less likely to be detected than those in open reaches (P = 0.45 vs 0.64; odds ratio = 0.46, P = 0.005).
  • Barrier type and survey type mattered: culverts (P = 0.63) and fords (P = 0.48) were more detectable than ramps (P = 0.33) or weirs (P = 0.34; odds ratio = 0.28, P < 0.001), though river-road crossings had lower specificity (18%) than other barriers (92%). Barriers found in exhaustive walkovers were less detectable remotely (P = 0.64) than those from partial surveys (P = 0.77; odds ratio = 1.90, P = 0.049), and detection was overestimated by ~18% when ground-truthing used partial rather than complete surveys. Observer skill varied detection by 11% between experienced and less experienced observers.

Methods

  • Ground-truthing used two datasets: an exhaustive walkover of the entire River Afan catchment in Wales (171 km total river length; n = 279 barriers recorded; five dams excluded and no sluices present) and a sample dataset of 208 barriers from partial walkovers of principal salmon rivers in Wales during 2018–2021. Three observers independently scanned Google Earth at 500 m eye altitude, zooming to 100–300 m to classify barrier type; coordinates were matched to field barriers using a 10 m buffer and a 100 m match radius.
  • Detection performance was assessed with detection rate, sensitivity, specificity, and accuracy with 95% CIs. Repeatability among three raters was measured with the S-statistic (S = 0.66, 95% CI 0.59–0.72, P < 0.001).
  • Remote detection was modelled for 468 barriers with complete data (n = 260 full survey, n = 208 partial surveys) using binary logistic regression with barrier type, survey type, forest cover, Strahler stream order, barrier height class, altitude, and distance to mouth as predictors. All 128 main-effect models were compared via AICc using `glmulti`; results were complemented with random forest and adaptive boosted trees. Forested reaches were defined as barrier locations with 5 or more trees within a 10 m radius of either bank.

Limitations

  • The paper states that Google Earth images are typically 1–3 years old and could miss new barriers, and that resolution and level of detail differ between locations, affecting detection probability. Observer skill also affected detection, with an 11% difference between experienced and less experienced observers.
  • The authors state that remote sensing cannot fully replace “boots on the ground” river walkovers for filling barrier data gaps. LiDAR could help detect barriers through canopy only if barriers have distinctive surface roughness or are at least 2 m high, which would miss more than 68% of instream barriers in Europe; improving image resolution alone may not be enough in heavily forested streams, which made up 75% of barriers in this study.
  • Evident from the text: the logistic model had low predictive power (R² Tjur = 0.09), and the study excluded five dams because there were too few for analysis, so dam-specific detection could not be assessed. Partial river surveys overestimated remote detection by ~18% relative to the complete walkover, indicating that ground-truthing based on incomplete surveys can bias detection estimates.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1088/1748-9326/ad6460

Cites

50 of 50 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Image 1
CC-BY · From the PDF
Figure 1 Location of the four main barrier types (culvert, ford, ramp, weir. N = 260) found in the complete walkover of the River Afan catchment showing whether they were detected (closed fill) or undetected (open fill) in Google Earth.
Figure 1 Location of the four main barrier types (culvert, ford, ramp, weir. N = 260) found in the complete walkover of the River Afan catchment showing whether they were detected (closed fill) or undetected (open fill) in Google Earth.CC-BY · From the PDF
Figure 2 Performance metrics (%) of remote detection of barriers at river-road crossings (culverts and fords) and other barriers (ramps and weirs) found in an exhaustive river walkover. The results show the means ( ± 95 CI) of three raters working independently.
Figure 2 Performance metrics (%) of remote detection of barriers at river-road crossings (culverts and fords) and other barriers (ramps and weirs) found in an exhaustive river walkover. The results show the means ( ± 95 CI) of three raters working independently.CC-BY · From the PDF
Image 4
CC-BY · From the PDF
Figure 4 Parameter estimates ( ± 95 CI) of five significant predictors of remote barrier detection using logistic regression.
Figure 4 Parameter estimates ( ± 95 CI) of five significant predictors of remote barrier detection using logistic regression.CC-BY · From the PDF
Figure 5 Odds ratios of five significant predictors of remote barrier detection using logistic regression compared to the reference values for barrier type (culvert), survey type (exhaustive), forest cover (0 = open), and stream order (first order). Predictors that increase the probability of barrier detection compared to the reference values are shown in blue, and those that decrease it are shown in red.
Figure 5 Odds ratios of five significant predictors of remote barrier detection using logistic regression compared to the reference values for barrier type (culvert), survey type (exhaustive), forest cover (0 = open), and stream order (first order). Predictors that increase the probability of barrier detection compared to the reference values are shown in blue, and those that decrease it are shown in red.CC-BY · From the PDF

Verification

Claims the paper makes

0 Valid · 0 Not valid · 59 Not checked · 61 Own results and interpretations

60 more claims

Create a free account to see all 120 claims and to verify the 96 references still open.

See all claims — free account
  • It may state two separate claims as one. 43
  • The quote only partly states this claim. 2
  • Probably not a substantial claim. 17

Create a free account to see these claims.

Citations

Cites

Show all 51 references

Research works printed but not linked: 1

  1. Zhu, 2010, Sensitivity, specificity, accuracy, associated confidence interval and ROC analysis with practical SAS implementations, vol 19, p 67

Cited by

See all 7

Metadata

Authors
Parks, M. V., Garcia de Leaniz, C., Jones, P. E., & Jones, J.
Affiliations
  • Swansea University (GB)
  • Natural Resources Wales (GB)
Venue
Environmental Research Letters
Published
2024-08-01
Volume
19
Issue
8
Pages
084055
ISSN
1748-9326 (ISSN-L)
OpenAlex
W4400720549
Cited by
7
References
51
Citations by year
2026: 4 · 2025: 3
FWCI
1.34
Type
journal article
Language
en
Open access
Licence
Locations
  • Environmental Research Letters (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • European Commission Horizon 2020 Framework Programme
  • AMBER (Adaptive Management of Barriers in European Rivers) Project (Grant Agreement Number 689682)
Topics
  • Physical Sciences › Environmental Science › Nature and Landscape Conservation › Fish Ecology and Management Studies
  • Physical Sciences › Earth and Planetary Sciences › Earth-Surface Processes › Hydrology and Sediment Transport Processes
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Remote-Sensing Image Classification
Keywords
river barriers, river fragmentation, Google Earth imagery, machine learning detection, Strahler stream order, barrier removal, free-flowing rivers

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-18, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-18 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Modelling remote barrier detection to achieve free-flowing river targets. https://app.cochat.ai/papers/10.1088/1748-9326/ad6460
The paper
Parks, M. V., Garcia de Leaniz, C., Jones, P. E., & Jones, J. (2024). Modelling remote barrier detection to achieve free-flowing river targets. Environmental Research Letters. https://doi.org/10.1088/1748-9326/ad6460

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.