CoChat

Towards global flood mapping onboard low cost satellites with machine learning

Open access

The DOI matches its Crossref record; 60 of the 60 references it cites were found in a registry or an index.

Abstract

Spaceborne Earth observation is a key technology for flood response, offering valuable information to decision makers on the ground. Very large constellations of small, nano satellites— ’CubeSats’ are a promising solution to reduce revisit time in disaster areas from days to hours. However, data transmission to ground receivers is limited by constraints on power and bandwidth of CubeSats. Onboard processing offers a solution to decrease the amount of data to transmit by reducing large sensor images to smaller data products. The ESA’s recent PhiSat-1 mission aims to facilitate the demonstration of this concept, providing the hardware capability to perform onboard processing by including a power-constrained machine learning accelerator and the software to run custom applications. This work demonstrates a flood segmentation algorithm that produces flood masks to be transmitted instead of the raw images, while running efficiently on the accelerator aboard the PhiSat-1. Our models are trained on WorldFloods: a newly compiled dataset of 119 globally verified flooding events from disaster response organizations, which we make available in a common format. We test the system on independent locations, demonstrating that it produces fast and accurate segmentation masks on the hardware accelerator, acting as a proof of concept for this approach.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1038/s41598-021-86650-z

Cites

60 of 60 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 An example of a data product from the Copernicus EMS catalogue (activation EMSR312), in this case a map showing flood extent over the city of Vigan in the North West of Luzon island in the Philippines in September 2018. A blue water mask (here generated using an automatic method from a RADARSAT-2 image) is overlaid on top of a Sentinel-2 image, showing the extent of flooding. Sentinel 2 imagery and Copernicus EMS mapping products are provided as public domain. Base image and reference labels mask are included in the WorldFloods database and code for plotting this images may be found in our repository6.
Figure 1 An example of a data product from the Copernicus EMS catalogue (activation EMSR312), in this case a map showing flood extent over the city of Vigan in the North West of Luzon island in the Philippines in September 2018. A blue water mask (here generated using an automatic method from a RADARSAT-2 image) is overlaid on top of a Sentinel-2 image, showing the extent of flooding. Sentinel 2 imagery and Copernicus EMS mapping products are provided as public domain. Base image and reference labels mask are included in the WorldFloods database and code for plotting this images may be found in our repository6.CC BY · PubMed Central
Figure 2 Spatial resolution and spectral response of Sentinel-2 and HyperScout-2 sensors.
Figure 2 Spatial resolution and spectral response of Sentinel-2 and HyperScout-2 sensors.CC BY · PubMed Central
Figure 3 Locations of flood events contained in WorldFloods. Blue, orange and green areas denote Copernicus EMS, UNOSAT and GloFIMR data, respectively. Red circles denote test regions. Basemap credit: http://www.simplemaps.com.
Figure 3 Locations of flood events contained in WorldFloods. Blue, orange and green areas denote Copernicus EMS, UNOSAT and GloFIMR data, respectively. Red circles denote test regions. Basemap credit: http://www.simplemaps.com.CC BY · PubMed Central
Figure 4 (a) Sentinel 2 RGB bands and (b) associated labelled map (land/brown, water/blue, cloud/white) over Porto Palo (Sicily) derived from Copernicus EMS 333 activation. Cloud mask obtained automatically with s2cloudless49. Base image and reference labels are included in the WorldFloods database and code for plotting this images may be found in our repository6.
Figure 4 (a) Sentinel 2 RGB bands and (b) associated labelled map (land/brown, water/blue, cloud/white) over Porto Palo (Sicily) derived from Copernicus EMS 333 activation. Cloud mask obtained automatically with s2cloudless49. Base image and reference labels are included in the WorldFloods database and code for plotting this images may be found in our repository6.CC BY · PubMed Central
Figure 5 (a) Satellite used to derive each flood map in the WorldFloods data set. ‘Other’ satellites (all optical): GeoEye-1, PlanetScope, Earth Observing (EO)-1. (b) Difference in days between the flood map and the Sentinel-2 image (Sentinel-2 image is always posterior in time to the flood map).
Figure 5 (a) Satellite used to derive each flood map in the WorldFloods data set. ‘Other’ satellites (all optical): GeoEye-1, PlanetScope, Earth Observing (EO)-1. (b) Difference in days between the flood map and the Sentinel-2 image (Sentinel-2 image is always posterior in time to the flood map).CC BY · PubMed Central
Figure 6 Overview of the model training pipeline used in this work. Note that WorldFloods provides images from S2, but reference flood extent maps may have been labelled from other sources, such as radar satellites.
Figure 6 Overview of the model training pipeline used in this work. Note that WorldFloods provides images from S2, but reference flood extent maps may have been labelled from other sources, such as radar satellites.CC BY · PubMed Central
Figure 7 Precision–recall curves of different models trained on (a), the Sentinel-2 original resolution (10 m) and (b), in the degraded resolution of HyperScout-2 (80 m). In gray 95% recall threshold.
Figure 7 Precision–recall curves of different models trained on (a), the Sentinel-2 original resolution (10 m) and (b), in the degraded resolution of HyperScout-2 (80 m). In gray 95% recall threshold.CC BY · PubMed Central
Figure 8 Performance of models trained with all WorldFloods flood maps compared with models trained only with flood maps derived from Sentinel-2.
Figure 8 Performance of models trained with all WorldFloods flood maps compared with models trained only with flood maps derived from Sentinel-2.CC BY · PubMed Central
Figure 9 Segmentation results of degraded models (SCNN 80 m) run on Myriad 2 device. Sentinel 2 imagery and Copernicus EMS mapping products are provided as public domain. Base images and reference labels are included in the WorldFloods database and code for plotting these images may be found in our repository6. Colours are as follows: brown/land, blue/water, white/cloud.
Figure 9 Segmentation results of degraded models (SCNN 80 m) run on Myriad 2 device. Sentinel 2 imagery and Copernicus EMS mapping products are provided as public domain. Base images and reference labels are included in the WorldFloods database and code for plotting these images may be found in our repository6. Colours are as follows: brown/land, blue/water, white/cloud.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

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

The self-check needs the paper’s full text.

Citations

Cites

Show all 60 references

Research works printed but not linked: 1

  1. Manzillo, P. F. et al. Hyperspectral imaging for real time land and vegetation inspection. In The 4S Symposium (2017).

Cited by

See all 202

Metadata

Authors
Mateo-Garcia, G., Veitch-Michaelis, J., Smith, L., Oprea, S. V., Schumann, G., Gal, Y., Baydin, A. G., & Backes, D.
Affiliations
  • Liverpool John Moores University (GB)
  • University of Oxford (GB)
  • University of Edinburgh (GB)
  • University of Bristol (GB)
  • University of Luxembourg (LU)
Venue
Scientific Reports
Published
2021-03-31
Volume
11
Issue
1
Pages
7249-7249
ISSN
2045-2322 (ISSN-L)
OpenAlex
W3140841781
Cited by
202
References
56
Citations by year
2026: 23 · 2025: 41 · 2024: 53 · 2023: 51 · 2022: 26 · 2021: 8
FWCI
14.14
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Open Repository and Bibliography (University of Luxembourg) (submittedVersion, cc-by)
  • Edinburgh Research Explorer (University of Edinburgh) (submittedVersion, cc-by)
Funders
  • Ministerio de Ciencia e Innovación (TEC2016-77741-R)
  • Science and Technology Facilities Council (ST/R002673/1)
  • Engineering and Physical Sciences Research Council (EP/L016427/1, EP/N019474/1)
Topics
  • Physical Sciences › Environmental Science › Global and Planetary Change › Flood Risk Assessment and Management
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Remote-Sensing Image Classification
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
Keywords
flood mapping, CubeSat, on-board processing, machine learning accelerators, flood segmentation, disaster response, low-cost satellites, data reduction

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.

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: Towards global flood mapping onboard low cost satellites with machine learning. https://app.cochat.ai/papers/10.1038/s41598-021-86650-z
The paper
Mateo-Garcia, G., Veitch-Michaelis, J., Smith, L., Oprea, S. V., Schumann, G., Gal, Y., Baydin, A. G., & Backes, D. (2021). Towards global flood mapping onboard low cost satellites with machine learning. Scientific Reports. https://doi.org/10.1038/s41598-021-86650-z

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.