MFogHub: Bridging Multi-Regional and Multi-Satellite Data for Global Marine Fog Detection and Forecasting

TL;DR

MFogHub tackles the lack of open, global marine fog datasets by providing a multi-regional, multi-satellite benchmark with 68,000 samples across 15 regions and 6 geostationary satellites, structured as cube-streams to support flexible, spatio-temporal analysis. The paper benchmarks 16 detection/forecasting baselines, analyzes regional and satellite generalization, and investigates spectral-band sensitivity and class imbalance effects, revealing notable generalization gaps and the value of multi-regional training. Key contributions include the first large-scale cross-region dataset, a cube-stream organization, and comprehensive experiments that quantify cross-regional and cross-satellite performance variations. By making the dataset and code publicly available, MFogHub aims to advance practical marine fog monitoring and the scientific understanding of fog dynamics on a global scale.

Abstract

Deep learning approaches for marine fog detection and forecasting have outperformed traditional methods, demonstrating significant scientific and practical importance. However, the limited availability of open-source datasets remains a major challenge. Existing datasets, often focused on a single region or satellite, restrict the ability to evaluate model performance across diverse conditions and hinder the exploration of intrinsic marine fog characteristics. To address these limitations, we introduce \textbf{MFogHub}, the first multi-regional and multi-satellite dataset to integrate annotated marine fog observations from 15 coastal fog-prone regions and six geostationary satellites, comprising over 68,000 high-resolution samples. By encompassing diverse regions and satellite perspectives, MFogHub facilitates rigorous evaluation of both detection and forecasting methods under varying conditions. Extensive experiments with 16 baseline models demonstrate that MFogHub can reveal generalization fluctuations due to regional and satellite discrepancy, while also serving as a valuable resource for the development of targeted and scalable fog prediction techniques. Through MFogHub, we aim to advance both the practical monitoring and scientific understanding of marine fog dynamics on a global scale. The dataset and code are at \href{https://github.com/kaka0910/MFogHub}{https://github.com/kaka0910/MFogHub}.

Figures (9)

• Figure 1: Overview of MFogHub. Right: MFogHub collects data from 15 marine fog-prone regions worldwide, captured by 6 geostationary satellites. Middle: Data for each region-satellite pair is organized in a cube-stream structure with dimensions of "timestamp-spectral band-latitude-longitude." MFogHub includes 21 cube-streams in total, each with corresponding masks, supporting both detection and forecasting tasks. Left: MFogHub enables unique evaluations of model generalization across multiple regions and satellite.
• Figure 2: Visualization of marine fog occurrence frequency based on 9.5 million ICOADS observations (2015-2024) for identifying costal fog-prone regions worldwide.
• Figure 3: Illustration of data proportion and flow across various regions and satellites for marine fog detection and forecasting in the MFogHub dataset.
• Figure 4: Spatial distribution and intensity variations across B.C., C.C. and G.A. sub-datasets using marine fog detection labels.
• Figure 5: Visualization of pixel histogram distributions and band information (Index, Central Wavelength (CW), and Spatial Resolution (SR)) across multiple bands for FY4A and H8/9 satellites.