A Novel Dataset for Flood Detection Robust to Seasonal Changes in Satellite Imagery

2025-07-30

AI Summary

Existing remote sensing flood segmentation datasets lack seasonal robustness and multi-temporal annotations, hindering generalization across varying environmental conditions. Method: We introduce FloodSeg-MW, the first benchmark dataset for satellite-based flood semantic segmentation, covering ten sites across five U.S. Midwest states during the 2019 flood events; each site includes ten co-registered, high-resolution optical images enabling seasonal variation modeling and temporal generalization assessment. Contribution/Results: Comprehensive evaluation on U-Net, SegFormer, and SiamUnet reveals moderate performance (mIoU ≈ 58%), underscoring the need for advanced temporal modeling and multimodal fusion. Ablation studies using sliding temporal windows confirm significant gains from time-series features in flood identification. FloodSeg-MW will be publicly released, addressing the critical gap in seasonally adaptive flood segmentation data and advancing the practical deployment of remote sensing semantic segmentation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Time-Series/Data Streams
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications
• Web Mining and Content Analysis: Web data integration and cleaning
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

This study introduces a novel dataset for segmenting flooded areas in satellite images. After reviewing 77 existing benchmarks utilizing satellite imagery, we identified a shortage of suitable datasets for this specific task. To fill this gap, we collected satellite imagery of the 2019 Midwestern USA floods from Planet Explorer by Planet Labs (Image c{opyright} 2024 Planet Labs PBC). The dataset consists of 10 satellite images per location, each containing both flooded and non-flooded areas. We selected ten locations from each of the five states: Iowa, Kansas, Montana, Nebraska, and South Dakota. The dataset ensures uniform resolution and resizing during data processing. For evaluating semantic segmentation performance, we tested state-of-the-art models in computer vision and remote sensing on our dataset. Additionally, we conducted an ablation study varying window sizes to capture temporal characteristics. Overall, the models demonstrated modest results, suggesting a requirement for future multimodal and temporal learning strategies. The dataset will be publicly available on <https://github.com/youngsunjang/SDSU_MidWest_Flood_2019>.

Problem

• Lack of suitable flood detection datasets in satellite imagery
• Need for robust flood segmentation across seasonal changes
• Evaluating models on new Midwest USA flood dataset

Innovation

• Novel flood detection dataset from satellite imagery
• Uniform resolution and resizing in data processing
• Tested state-of-the-art semantic segmentation models