Map Floodwater Radar Imagery using Machine Learning Algorithms

Flooding is a widespread and costly natural disaster around the world. Accurately assessing the extent of flooding in near real-time is crucial for governments and humanitarian organizations. This information strengthens early warning systems, evaluates risks, and guides effective relief efforts. Therefore, precise flood mapping is essential for saving lives through improved early warning systems and targeted emergency responses. In this study, radar imagery available on the Planetary Computer Data was utilized to train a U-Net model specifically designed to label flood-affected pixels in an image from a flood event. Different blocks of the U-Net encoder architecture were fine-tuned to identify the most efficient fine-tuned model, and their results were compared. As a result, the model with blocks 1 and 2 being fine-tuned demonstrated the highest Intersection over Union (IoU) score of 78.904%, an increase of 8.663% over the baseline methods. ABSTRAK: Banjir merupakan bencana alam yang meluas dan mahal di seluruh dunia. Penilaian yang tepat terhadap skala banjir secara hampir masa nyata adalah penting bagi kerajaan dan organisasi kemanusiaan. Maklumat ini memperkukuhkan sistem amaran awal, menilai risiko, dan membimbing usaha bantuan yang lebih berkesan. Oleh itu, pemetaan banjir yang tepat adalah penting untuk menyelamatkan nyawa melalui sistem amaran awal yang lebih baik dan respons kecemasan yang disasarkan. Dalam kajian ini, imej radar yang tersedia pada Planetary Computer Data digunakan untuk melatih model U-Net yang direka khas untuk melabelkan piksel yang terjejas oleh banjir dalam imej daripada kejadian banjir. Bagi mengenal pasti model ditala-halus yang paling cekap, blok-blok berlainan dalam arkitektur pengekod U-Net telah ditala-halus, dan hasilnya dibandingkan. Hasilnya, model dengan blok 1 dan 2 yang ditala-halus menunjukkan skor Intersection over Union (IoU) tertinggi sebanyak 78.904%, iaitu peningkatan sebanyak 8.663% berbanding kaedah asas.

Authors

Institutions

Publication Details

Journal
IIUM Engineering Journal
Published
2025-01-10
DOI
https://doi.org/10.31436/iiumej.v26i1.3157
Citations
1
Primary Topic
Flood Risk Assessment and Management
Type
article
Field-Weighted Citation Impact
1.11
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Map Floodwater Radar Imagery using Machine Learning Algorithms

Thanh-Nghi Doan, Duc-Ngoc Le-Thi
1 citations
IIUM Engineering Journal
Flood Risk Assessment and Management
1.11
article

Map Floodwater Radar Imagery using Machine Learning Algorithms

Thanh-Nghi Doan, Duc-Ngoc Le-Thi
article en
1 citations

Abstract

Flooding is a widespread and costly natural disaster around the world. Accurately assessing the extent of flooding in near real-time is crucial for governments and humanitarian organizations. This information strengthens early warning systems, evaluates risks, and guides effective relief efforts. Therefore, precise flood mapping is essential for saving lives through improved early warning systems and targeted emergency responses. In this study, radar imagery available on the Planetary Computer Data was utilized to train a U-Net model specifically designed to label flood-affected pixels in an image from a flood event. Different blocks of the U-Net encoder architecture were fine-tuned to identify the most efficient fine-tuned model, and their results were compared. As a result, the model with blocks 1 and 2 being fine-tuned demonstrated the highest Intersection over Union (IoU) score of 78.904%, an increase of 8.663% over the baseline methods. ABSTRAK: Banjir merupakan bencana alam yang meluas dan mahal di seluruh dunia. Penilaian yang tepat terhadap skala banjir secara hampir masa nyata adalah penting bagi kerajaan dan organisasi kemanusiaan. Maklumat ini memperkukuhkan sistem amaran awal, menilai risiko, dan membimbing usaha bantuan yang lebih berkesan. Oleh itu, pemetaan banjir yang tepat adalah penting untuk menyelamatkan nyawa melalui sistem amaran awal yang lebih baik dan respons kecemasan yang disasarkan. Dalam kajian ini, imej radar yang tersedia pada Planetary Computer Data digunakan untuk melatih model U-Net yang direka khas untuk melabelkan piksel yang terjejas oleh banjir dalam imej daripada kejadian banjir. Bagi mengenal pasti model ditala-halus yang paling cekap, blok-blok berlainan dalam arkitektur pengekod U-Net telah ditala-halus, dan hasilnya dibandingkan. Hasilnya, model dengan blok 1 dan 2 yang ditala-halus menunjukkan skor Intersection over Union (IoU) tertinggi sebanyak 78.904%, iaitu peningkatan sebanyak 8.663% berbanding kaedah asas.

IIUM Engineering JournalVol. 26(1)
An Giang University (VN)
Climate action
Openalex Percentile: Top 28%
Flood Risk Assessment and Management
1.11
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.