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

The DOI matches its Crossref record; 52 of the 52 references it cites were found in a registry or an index; none of the 2 sources its claims cite is checked for support yet.

A multi-sensor framework for comparing and integrating optical, C-band and L-band SAR inundation products in tropical floodplain of Australia

Open access

Abstract

Accurate inundated landscape mapping is essential for understanding hydrological connectivity, wetland dynamics and floodplain processes in tropical environments. However, optical and Synthetic Aperture Radar (SAR) sensors often produce inconsistent inundation estimates due to differences in sensor physics, wavelength sensitivity, vegetation and cloud penetration, and analytical methods. These challenges are amplified in wet–dry tropical systems, where cloud cover, heterogeneous vegetation and seasonal flooding complicate water detection. This study develops a unified multi-sensor framework for comparing and integrating inundation products derived from Sentinel-2 optical imagery, Sentinel-1 C-band and PALSAR-2 L-band SAR in the Daly River floodplain system in northern Australia. Inundated landscape maps were generated using supervised Random Forest classification and unsupervised multi-Otsu thresholding, followed by multi-sensor fusion analysis. Products were developed and validated using wet- and dry-season field observations and Indigenous Knowledges. Results showed strong seasonal and sensor-specific differences. Sentinel-2 provided the most balanced representation of dry-season floodplain vegetation and inundation, while Sentinel-1 achieved high wet-season overall accuracy (>0.95) under cloud-affected conditions. PALSAR-2 improved the detection of inundated vegetation and structurally complex environments, reaching an overall accuracy of 0.97. Multi-sensor evidential reasoning (fusion) improved spatial coherence and class delineation, but performance varied across sensor combinations. These findings show that optical, C-band SAR and L-band SAR inundation products are not directly interchangeable in tropical wetlands. Instead, they performed separately for each landscape. The proposed framework supports the comparison and integration of multi-sensor inundation products to enhance wetland conservation initiatives, wetland monitoring, hydrological assessment, and environmental management in data-limited tropical regions globally.

Overview

CoChat Summary

TL;DR

This study develops a multi-sensor framework to compare and integrate Sentinel-2 optical, Sentinel-1 C-band SAR, and PALSAR-2 L-band SAR inundation products in the Daly River floodplain, northern Australia, using Random Forest, multi-Otsu, and Dempster-Shafer fusion validated with field observations and Indigenous Knowledges. It finds that no single sensor performs consistently across seasons and landscape classes, with sensor fusion improving spatial coherence but varying by sensor combination.

Key findings

  • Sentinel-2 provided the most balanced dry-season representation of floodplain vegetation and inundation; Sentinel-1 achieved wet-season overall accuracy >0.95 under cloud-affected conditions; PALSAR-2 improved detection of inundated vegetation and structurally complex environments, reaching an overall accuracy of 0.97. Multi-sensor evidential reasoning (fusion) improved spatial coherence and class delineation, but performance varied across sensor combinations.
  • In 2024 dry-season classifications, Sentinel-2 Random Forest had OA 0.96 (best overall), Sentinel-1 + Sentinel-2 Random Forest 0.95, Sentinel-1 multi-Otsu 0.94, Sentinel-1 + Sentinel-2 multi-Otsu 0.95, and Sentinel-1 Random Forest 0.44 (lowest).
  • In the 2026 wet season, Sentinel-1 Random Forest OA was 0.98 and Sentinel-1 multi-Otsu OA was 0.92; RF mapped grass-sedge swamp as the largest class (~720 km²) and ephemerally inundated woodland ~150 km², while multi-Otsu mapped floodplain as the largest class (~890 km²) and open water ~135 km².
  • For 2020, dry-season fused P2+S2+S1, P2+S2, and S2 had OA ≥0.93, while fused S1+S2 (0.22), P2 (0.19), and fused P2+S1 (0.19) were weak; wet-season fused P2+S1 had OA 0.99, P2-only 0.94, S1 0.82, and multi-Otsu fused/S2-only products were poor (OA = 0.43, range 0.42–0.48).

Methods

  • Study area: downstream Daly River catchment, Northern Territory, Australia, approximately 110 km inland and about 3,585 km², with tropical monsoon wet and dry seasons. Data: Sentinel-1 C-band SAR (VV/VH), PALSAR-2 L-band SAR (HH/HV), and Sentinel-2 optical (RGB, NIR, SWIR) from 2020, 2024, and 2026; maximum temporal difference among sensor observations was four days.
  • Reference data: 605 dry-season field samples (September 2024) and 611 wet-season samples (February 2026); Traditional Ecological Knowledges from the Nauiyu community and Malak Malak Rangers informed study design, fieldwork planning, site selection, and interpretation, and supported collection of 587 GPS reference samples during the dry and wet seasons of 2024 and 2026; Malak Malak Rangers reviewed and verified classified wetland maps, including the 2020 classification.
  • Classification and validation: six LULC/inundation classes; supervised Random Forest (500 trees, 70/30 training-validation split) and unsupervised multi-Otsu thresholding; Dempster-Shafer Theory fusion with equal sensor reliability weights (singleton mass 0.90, ignorance 0.10); validation via confusion matrices, OA, Kappa, PA, and UA, using 30% validation subsets (183 wet 2026, 182 dry 2024), 605 independent samples for 2024 dry multi-Otsu, and 595 reference samples for 2020 P2/S1 agreement (232 for S2 wet 2020, 595 for S2 dry 2020).

Limitations

  • Stated: unsupervised multi-Otsu had limited capacity to classify inundated vegetation, especially densely forested swamp, where canopy cover and spectral mixing obscure surface water; flooding beneath vegetation was not reliably detected by threshold-based methods alone. Stated: sensor acquisition dates differ by up to four days, creating potential temporal mismatch; 2020 lacked wet- and dry-season validation data, increasing uncertainty and limiting interannual comparison.
  • Stated: S1 and S2 were resampled from 10 m to 25 m to match P2, so narrow channels (<25 m) or vegetation-obscured channels could not be mapped reliably due to mixed-pixel effects, and fusion could not overcome this. Stated: residual cloud contamination in wet-season S2 reduced valid optical observations and introduced uncertainty.
  • Stated: equal DST reliability weights and tie-breaking by retaining the first input made tied outcomes dependent on sensor order and potentially favored SAR-derived classifications; constant weights may underweight the best-performing sensor in a season or overstate a sensor with reduced seasonal accuracy. Evident from the text: single-day Sentinel-1 scenes produced data gaps in some maps, and PALSAR-2 availability was limited to 2020, constraining comparisons.

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

Integrity signals

Registry

Crossref · 10.1080/15481603.2026.2756795

Cites

52 of 52 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Image 1
CC-BY · From the PDF
Image 2
CC-BY · From the PDF
Figure 1 Downstream area of the Daly River catchment as study site; the background image is ESRI Basemap 2026 (Service Layer Credits: Earthstar Geographics ).
Figure 1 Downstream area of the Daly River catchment as study site; the background image is ESRI Basemap 2026 (Service Layer Credits: Earthstar Geographics ).CC-BY · From the PDF
Figure 2 Overview of the inundated landscape mapping workflow for multi-sensor (i.e. Sentinel-1, Sentinel-2, and PALSAR-2) data.
Figure 2 Overview of the inundated landscape mapping workflow for multi-sensor (i.e. Sentinel-1, Sentinel-2, and PALSAR-2) data.CC-BY · From the PDF
Image 5
CC-BY · From the PDF
Image 6
CC-BY · From the PDF
Image 7
CC-BY · From the PDF
Image 8
CC-BY · From the PDF
Image 9
CC-BY · From the PDF
Image 10
CC-BY · From the PDF
Figure 3 Classification maps of SAR (Sentinel-1) and optical (Sentinel-2) single and fused sensors for the dry season, September 2024, showing five inundated classes: open water, grass-sedge swamp, forested swamp, floodplain, and floodplain woodland, and one ephemerally inundated woodland class. (a) Unsupervised multi-Otsu classification-derived map using S1, (b) S2, (c) fused S1 and S2; (d) supervised Random Forest (RF) classification-derived map using S1, (e) S2, and (f) fused S1 and S2. For (a, c, d, f), the S1 single-scene and fused S1 maps do not provide full study-area coverage because they rely on a single-day scene, leading to data gaps.
Figure 3 Classification maps of SAR (Sentinel-1) and optical (Sentinel-2) single and fused sensors for the dry season, September 2024, showing five inundated classes: open water, grass-sedge swamp, forested swamp, floodplain, and floodplain woodland, and one ephemerally inundated woodland class. (a) Unsupervised multi-Otsu classification-derived map using S1, (b) S2, (c) fused S1 and S2; (d) supervised Random Forest (RF) classification-derived map using S1, (e) S2, and (f) fused S1 and S2. For (a, c, d, f), the S1 single-scene and fused S1 maps do not provide full study-area coverage because they rely on a single-day scene, leading to data gaps.CC-BY · From the PDF
Figure 4 Comparison of Sentinel-1 (S1) derived (a) multi-Otsu thresholding and (b) Random Forest (RF) classifications for the 2026 wet season, showing five inundation classes (open water, grass-sedge swamp, forested swamp, floodplain, and floodplain woodland) and one ephemerally inundated woodland class. For (a) and (b), the S1 single-scenes do not provide full study-area coverage because they use a single-day scene, resulting in some data gaps.
Figure 4 Comparison of Sentinel-1 (S1) derived (a) multi-Otsu thresholding and (b) Random Forest (RF) classifications for the 2026 wet season, showing five inundation classes (open water, grass-sedge swamp, forested swamp, floodplain, and floodplain woodland) and one ephemerally inundated woodland class. For (a) and (b), the S1 single-scenes do not provide full study-area coverage because they use a single-day scene, resulting in some data gaps.CC-BY · From the PDF
Figure 5 Comparison of the class distribution of the inundated landscape derived from three individual sensors. (a) Sentinel-1, (b) PALSAR-2, (c) Sentinel-2, (d) fused PALSAR-2 and Sentinel-1, (e) fused PALSAR-2 and Sentinel-2, (f) used P2, Sentinel-2 and Sentinel-1, and (g) Sentinel-2 and Sentinel-1. Inundated landscape mapping outputs have been produced across six dry season inundation classes in the downstream reach of the catchment. For (a), (d), (f), (g), the Sentinel-1 single-scene and fused Sentinel-1 maps do not provide full study-area coverage because they rely on a single-day scene, leading to data gaps.
Figure 5 Comparison of the class distribution of the inundated landscape derived from three individual sensors. (a) Sentinel-1, (b) PALSAR-2, (c) Sentinel-2, (d) fused PALSAR-2 and Sentinel-1, (e) fused PALSAR-2 and Sentinel-2, (f) used P2, Sentinel-2 and Sentinel-1, and (g) Sentinel-2 and Sentinel-1. Inundated landscape mapping outputs have been produced across six dry season inundation classes in the downstream reach of the catchment. For (a), (d), (f), (g), the Sentinel-1 single-scene and fused Sentinel-1 maps do not provide full study-area coverage because they rely on a single-day scene, leading to data gaps.CC-BY · From the PDF
Figure 6 Comparison of the class distribution of the inundated landscape derived from three individual sensors. (a) Sentinel-1, (b) PALSAR-2, (c) Sentinel-2, (d) fused PALSAR-2 and Sentinel-1, (e) fused PALSAR-2 and Sentinel-2, (f) used PALSAR-2, Sentinel-2 and Sentinel-1, and (g) Sentinel-2 and Sentinel-1. Inundated landscape mapping outputs have been produced across six inundation classes in the wet season in the downstream reach of the catchment. For (a, d, f, g), the Sentinel-1 single-scene and fused Sentinel-1 maps do not provide full study-area coverage because they use a single-day scene; (c), (e), (f), (g) have cloud contamination, resulting in some data gaps.
Figure 6 Comparison of the class distribution of the inundated landscape derived from three individual sensors. (a) Sentinel-1, (b) PALSAR-2, (c) Sentinel-2, (d) fused PALSAR-2 and Sentinel-1, (e) fused PALSAR-2 and Sentinel-2, (f) used PALSAR-2, Sentinel-2 and Sentinel-1, and (g) Sentinel-2 and Sentinel-1. Inundated landscape mapping outputs have been produced across six inundation classes in the wet season in the downstream reach of the catchment. For (a, d, f, g), the Sentinel-1 single-scene and fused Sentinel-1 maps do not provide full study-area coverage because they use a single-day scene; (c), (e), (f), (g) have cloud contamination, resulting in some data gaps.CC-BY · From the PDF
Image 15
CC-BY · From the PDF

Verification

Claims the paper makes

0 Valid · 0 Not valid · 2 Not checked · 78 Own results and interpretations

37 more claims

Create a free account to see all 80 claims and to verify the 1 reference still open.

See all claims — free account
  • The quote only partly states this claim. 2
  • It may state two separate claims as one. 32
  • Probably not a substantial claim. 7

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 52 references

Cited by

Still being checked.

Metadata

Authors
Mollick, P., Catry, T., Hostache, R., de Mello, K., Irvine, D. J., Garcia, E. A., Rangers, M. M., & Crabbe, R. A.
Affiliations
  • Université de Montpellier (FR)
  • Institute of Technology of Cambodia (KH)
  • Commonwealth Scientific and Industrial Research Organisation (AU)
  • Charles Darwin University (AU)
  • Research Institute for the Environment and Livelihoods, Charles Darwin University
  • Malak Malak Ranger Group, Northern Land Council
Corresponding author
Mollick, Porni
Venue
GIScience & Remote Sensing
Published
2026-11-11
Volume
63
Issue
1
ISSN
  • 1548-1603 (ISSN-L)
  • 1943-7226
OpenAlex
W7165730296
Cited by
0
References
52
FWCI
0.00
Type
journal article
Language
en
Open access
Licence
Locations
  • GIScience & Remote Sensing (publishedVersion, cc-by)
  • SSRN Electronic Journal (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • Charles Darwin University (Cycas Scholarship)
  • Cooperative Research Centre for Developing Northern Australia (AT.7.2223014)
  • Australian Government Research Training Program (doi.org/10.82133/C42F-K220)
Topics
  • Physical Sciences › Environmental Science › Global and Planetary Change › Flood Risk Assessment and Management
  • Physical Sciences › Earth and Planetary Sciences › Geophysics › Synthetic Aperture Radar (SAR) Applications and Techniques
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Remote-Sensing Image Classification
Keywords
flood inundation mapping, sensor fusion, C-band SAR, L-band SAR, Sentinel-2, Sentinel-1, PALSAR-2, tropical wetlands, floodplain vegetation, seasonal flooding, cloud cover, evidential reasoning, random forest classification, indigenous knowledge, wetland conservation

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 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-19 (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: A multi-sensor framework for comparing and integrating optical, C-band and L-band SAR inundation products in tropical floodplain of Australia. https://app.cochat.ai/papers/10.1080/15481603.2026.2756795
The paper
Mollick, P., Catry, T., Hostache, R., de Mello, K., Irvine, D. J., Garcia, E. A., Rangers, M. M., & Crabbe, R. A. (2026). A multi-sensor framework for comparing and integrating optical, C-band and L-band SAR inundation products in tropical floodplain of Australia. GIScience & Remote Sensing. https://doi.org/10.1080/15481603.2026.2756795

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.