CoChat

The DOI matches its Crossref record; 47 of the 50 references it cites were found in a registry or an index; none of the 16 sources its claims cite is checked for support yet.

Sampling Strategies to Improve Passive Optical Remote Sensing of River Bathymetry

Open access

Abstract

Passive optical remote sensing of river bathymetry involves establishing a relation between depth and reflectance that can be applied throughout an image to produce a depth map. Building upon the Optimal Band Ratio Analysis (OBRA) framework, we introduce sampling strategies for constructing calibration data sets that lead to strong relationships between an image-derived quantity and depth across a range of depths. Progressively excluding observations that exceed a series of cutoff depths from the calibration process improved the accuracy of depth estimates and allowed the maximum detectable depth (dmax) to be inferred directly from an image. Depth retrieval in two distinct rivers also was enhanced by a stratified version of OBRA that partitions field measurements into a series of depth bins to avoid biases associated with under-representation of shallow areas in typical field data sets. In the shallower, clearer of the two rivers, including the deepest field observations in the calibration data set did not compromise depth retrieval accuracy, suggesting that dmax was not exceeded and the reach could be mapped without gaps. Conversely, in the deeper and more turbid stream, progressive truncation of input depths yielded a plausible estimate of dmax consistent with theoretical calculations based on field measurements of light attenuation by the water column. This result implied that the entire channel, including pools, could not be mapped remotely. However, truncation improved the accuracy of depth estimates in areas shallower than dmax, which comprise the majority of the channel and are of primary interest for many habitat-oriented applications.

Overview

CoChat Summary

TL;DR

This study introduces two sampling strategies—OPTID and SOBRA—for calibrating passive optical depth retrieval from river imagery, building on Optimal Band Ratio Analysis (OBRA). Tests on hyperspectral images and field spectra from the Deschutes and Snake Rivers show that truncating deep observations can improve depth estimates and infer maximum detectable depth, while depth-stratified sampling can modestly improve calibration strength.

Key findings

  • OPTID improved OBRA calibration for the Deschutes image, with OBRA R² reaching a maximum at a cutoff depth of 3.57 m and then decreasing; this yielded an image-based estimate of maximum detectable depth d_max = 3.57 m. For the Snake image, OBRA R² plateaued at about 1.8 m and did not decrease for greater cutoff depths, implying d_max was not exceeded.
  • SOBRA increased OBRA R² relative to standard OBRA: Deschutes image 0.59 to 0.64; Deschutes field spectra 0.79 to 0.86; Snake image 0.79 to 0.80; Snake field spectra 0.85 to 0.88.
  • Irradiance-profile calculations gave d_max(λ₁) for the Deschutes at λ₁ = 519 nm of 2.78 m for ΔL_N/L_B = 0.1 and 5.55 m for 0.01; for Deschutes field spectra, 3.56 m and 7.12 m. For the Snake at λ₁ = 583 nm, d_max(λ₁) was 8.99 m and 17.98 m for ΔL_N/L_B = 0.1 and 0.01.
  • For the Deschutes image, truncating the calibration data set increased OBRA R² and reduced the minimum predicted depth from 0.73 m to 0.25 m; approximately 10% of the channel was deeper than the inferred 3.57 m d_max.

Methods

  • Empirical calibration/validation used field spectra and CASI hyperspectral images from two rivers: the Deschutes River, Oregon (3.67 km reach, depths up to 8.87 m, multibeam echo sounder survey), and the Snake River, Wyoming (1.2 km reach, mean depth 0.98 m, hybrid wading/ADCP survey).
  • OBRA regressed X = ln[R(λ₁)/R(λ₂)] against depth d with a quadratic relation. OPTID iteratively truncated field depths in 5 cm increments from the maximum depth down to approximately 0.5 m; SOBRA binned depths into 10 bins with equal sample sizes, with the deepest bin extending from the 95th percentile to the maximum depth.
  • Validation used observed vs. predicted regressions on a 50% random sample withheld from calibration; OPTID-based d_max estimates were compared with d_max(λ) calculated from measured irradiance profiles using K_d(λ) and assumed ΔL_N(λ)/L_B(λ) values.

Limitations

  • Stated: OBRA is empirical and requires ground-based depth measurements for calibration, which is an important limitation. The study holds the curve-fitting method (regression) and functional form (quadratic) constant, so sampling is only one aspect of calibration; future work should evaluate other statistical approaches and equations.
  • Stated: OPTID-based d_max was unambiguous only for the Deschutes image spectra; for the Snake image and field spectra from both sites, no clear maximum was evident. Calculating d_max from irradiance profiles requires assuming ΔL_N(λ)/L_B(λ), and how to specify this parameter a priori is not obvious.
  • Evident from the text: results are based on two rivers and one hyperspectral sensor (CASI) plus field spectra; field spectra had fewer observations, especially on the Deschutes, and low R² values at shallow cutoffs were partly attributed to small sample sizes.

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

Integrity signals

Registry

Crossref · 10.3390/rs10060935

Cites

47 of 47 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 Example of Optimal Band Ratio Analysis (OBRA)-based depth retrieval from the Deschutes River using a quadratic relation between the image-derived quantity X and flow depth d . ( a ) calibration plot showing the non-monotonic relationship between X and d due to the X 2 term. The X vs. d equation represented by the red line reaches a minimum at d = 0.73 m; ( b ) depth map produced from a hyperspectral image based on the quadratic X vs. d relation shown in ( a ). Note the depths along the channel margins do not go to to zero but rather the minimum estimated depth of 0.73 m. Flow is from left to right; ( c ) observed vs. predicted (OP) regression used to validate depth estimates, showing the absence of depth estimates less than 0.73 m. The best-fit regression line is shown in red. d f refers to the field-surveyed depth, di to the image-derived depth estimate, n is the sample size, and s R is the standard error of the regression in units of m.
Figure 1 Example of Optimal Band Ratio Analysis (OBRA)-based depth retrieval from the Deschutes River using a quadratic relation between the image-derived quantity X and flow depth d . ( a ) calibration plot showing the non-monotonic relationship between X and d due to the X 2 term. The X vs. d equation represented by the red line reaches a minimum at d = 0.73 m; ( b ) depth map produced from a hyperspectral image based on the quadratic X vs. d relation shown in ( a ). Note the depths along the channel margins do not go to to zero but rather the minimum estimated depth of 0.73 m. Flow is from left to right; ( c ) observed vs. predicted (OP) regression used to validate depth estimates, showing the absence of depth estimates less than 0.73 m. The best-fit regression line is shown in red. d f refers to the field-surveyed depth, di to the image-derived depth estimate, n is the sample size, and s R is the standard error of the regression in units of m.CC-BY · From the PDF
Image 2
CC-BY · From the PDF
Image 3
CC-BY · From the PDF
Figure 3 Histograms of the original field data set, the binned data set, and the final calibration data set used for Stratified Optimal Band Ratio Analysis (SOBRA) for each study site. The binned histograms used for SOBRA consist of 10 bins, each with an equal number of observations sampled at random from each depth bin. The deepest bin extends from the 95th percentile of the original depth distribution to the maximum depth in the data set.
Figure 3 Histograms of the original field data set, the binned data set, and the final calibration data set used for Stratified Optimal Band Ratio Analysis (SOBRA) for each study site. The binned histograms used for SOBRA consist of 10 bins, each with an equal number of observations sampled at random from each depth bin. The deepest bin extends from the 95th percentile of the original depth distribution to the maximum depth in the data set.CC-BY · From the PDF
Figure 4 Optimal Band Ratio Analysis (OBRA) matrix based on spectra extracted from the Deschutes River hyperspectral image for the original field data from the multibeam echo sounder survey, followed by OBRA matrices for a series of cutoff depths from OBRA of Progressively Truncated Input Depths (OPTID).
Figure 4 Optimal Band Ratio Analysis (OBRA) matrix based on spectra extracted from the Deschutes River hyperspectral image for the original field data from the multibeam echo sounder survey, followed by OBRA matrices for a series of cutoff depths from OBRA of Progressively Truncated Input Depths (OPTID).CC-BY · From the PDF
Figure 5 Optimal Band Ratio Analysis (OBRA) matrix based on spectra extracted from the Snake River hyperspectral image for the original field data from the hybrid wading/acoustic Doppler current profiler (ADCP) survey, followed by OBRA matrices for a series of cutoff depths from OBRA of Progressively Truncated Input Depths (OPTID).
Figure 5 Optimal Band Ratio Analysis (OBRA) matrix based on spectra extracted from the Snake River hyperspectral image for the original field data from the hybrid wading/acoustic Doppler current profiler (ADCP) survey, followed by OBRA matrices for a series of cutoff depths from OBRA of Progressively Truncated Input Depths (OPTID).CC-BY · From the PDF
Figure 6 Summary of Optimal Band Ratio Analysis (OBRA) of Progressively Truncated Input Depths (OPTID) results for image spectra from the ( a ) Deschutes and ( b ) Snake Rivers. The numerator and denominator wavelengths identified by OBRA are plotted for each cutoff depth.
Figure 6 Summary of Optimal Band Ratio Analysis (OBRA) of Progressively Truncated Input Depths (OPTID) results for image spectra from the ( a ) Deschutes and ( b ) Snake Rivers. The numerator and denominator wavelengths identified by OBRA are plotted for each cutoff depth.CC-BY · From the PDF
Figure 7 Summary of Optimal Band Ratio Analysis (OBRA) of Progressively Truncated Input Depths (OPTID) for each study site and data type. Both the OBRA R 2 for calibration and the observed vs. predicted (OP) R 2 for validation are plotted in each panel as blue and red lines, respectively.
Figure 7 Summary of Optimal Band Ratio Analysis (OBRA) of Progressively Truncated Input Depths (OPTID) for each study site and data type. Both the OBRA R 2 for calibration and the observed vs. predicted (OP) R 2 for validation are plotted in each panel as blue and red lines, respectively.CC-BY · From the PDF
Figure 8 Maximum detectable depths calculated from measured irradiance profiles for two different values of D LN / LB for the ( a ) Deschutes and ( b ) Snake Rivers. Also plotted for each site are dmax values based on the irradiance profiles for the numerator ( l 1 ) and denominator ( l 2 ) wavelengths identified via OPTID of hyperspectral images and field spectra. For comparison, the dmax inferred from OPTID of the Deschutes River image is represented by the dashed line in ( a ).
Figure 8 Maximum detectable depths calculated from measured irradiance profiles for two different values of D LN / LB for the ( a ) Deschutes and ( b ) Snake Rivers. Also plotted for each site are dmax values based on the irradiance profiles for the numerator ( l 1 ) and denominator ( l 2 ) wavelengths identified via OPTID of hyperspectral images and field spectra. For comparison, the dmax inferred from OPTID of the Deschutes River image is represented by the dashed line in ( a ).CC-BY · From the PDF
Figure 9 ( a ) bathymetric map derived from a hyperspectral image of the Deschutes River based on the Optimal Band Ratio Analysis (OBRA) relation for the depth increment inferred to correspond to the sensor's maximum detectable depth dmax ; ( b ) zooms in on the portion of the channel indicated by the rectangle in ( a ). Pixels inferred to have depths in excess of dmax have been masked and appear as blank (white) areas in these maps; ( c ) calibration scatter plot relating the image-derived quantity X to depth d .
Figure 9 ( a ) bathymetric map derived from a hyperspectral image of the Deschutes River based on the Optimal Band Ratio Analysis (OBRA) relation for the depth increment inferred to correspond to the sensor's maximum detectable depth dmax ; ( b ) zooms in on the portion of the channel indicated by the rectangle in ( a ). Pixels inferred to have depths in excess of dmax have been masked and appear as blank (white) areas in these maps; ( c ) calibration scatter plot relating the image-derived quantity X to depth d .CC-BY · From the PDF

Verification

Claims the paper makes

0 Valid · 2 Not valid · 12 Not checked · 136 Own results and interpretations

96 more claims

Create a free account to see all 150 claims and to verify the 16 references still open.

See all claims — free account
  • It may state two separate claims as one. 38
  • Probably not a substantial claim. 15
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Citations

Cites

Show all 47 references

Research works printed but not linked: 1

  1. Lyon, 1992, Airborne multispectral scanner data for evaluating bottom sediment types and water depths of the St. Marys River, Michigan, Photogramm. Eng. Remote Sens., 58, 951

Cited by

See all 38

Metadata

Authors
Legleiter, C., Overstreet, B., & Kinzel, P.
Affiliations
  • University of Wyoming (US)
  • United States Geological Survey (US)
Corresponding author
Legleiter, Carl
Venue
Remote Sensing
Publisher
MDPI AG
Published
2018-06-13
Volume
10
Issue
6
Pages
935
ISSN
2072-4292 (ISSN-L)
OpenAlex
W2807808709
Cited by
38
References
50
Citations by year
2026: 5 · 2025: 4 · 2024: 10 · 2023: 6 · 2022: 4 · 2021: 5 · 2020: 1 · 2019: 1 · 2018: 2
FWCI
1.73
Type
journal article
Language
en
Open access
Locations
  • Remote Sensing (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Office of Naval Research (N000141010873)
  • University of Wyoming - National Park Service Research Center (1001576E)
Topics
  • Physical Sciences › Earth and Planetary Sciences › Earth-Surface Processes › Hydrology and Sediment Transport Processes
  • Life Sciences › Agricultural and Biological Sciences › Forestry › Remote Sensing and LiDAR Applications
  • Physical Sciences › Earth and Planetary Sciences › Earth-Surface Processes › Coastal and Marine Dynamics
Keywords
river bathymetry, passive optical remote sensing, sampling strategy, light attenuation, turbidity

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

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.