A New Application of Unsupervised Learning to Nighttime Sea Fog Detection

Abstract

This paper presents a nighttime sea fog detection algorithm incorporating unsupervised learning technique. The algorithm is based on data sets that combine brightness temperatures from the 3.7 μm and 10.8 μm channels of the meteorological imager (MI) onboard the Communication, Ocean and Meteorological Satellite (COMS), with sea surface temperature from the Operational Sea Surface Temperature and Sea Ice Analysis (OSTIA). Previous algorithms generally employed threshold values including the brightness temperature difference between the near infrared and infrared. The threshold values were previously determined from climatological analysis or model simulation. Although this method using predetermined thresholds is very simple and effective in detecting low cloud, it has difficulty in distinguishing fog from stratus because they share similar characteristics of particle size and altitude. In order to improve this, the unsupervised learning approach, which allows a more effective interpretation from the insufficient information, has been utilized. The unsupervised learning method employed in this paper is the expectation–maximization (EM) algorithm that is widely used in incomplete data problems. It identifies distinguishing features of the data by organizing and optimizing the data. This allows for the application of optimal threshold values for fog detection by considering the characteristics of a specific domain. The algorithm has been evaluated using the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) vertical profile products, which showed promising results within a local domain with probability of detection (POD) of 0.753 and critical success index (CSI) of 0.477, respectively.

Fog Season Risk Assessment for Maritime Transportation Systems Exploiting Himawari-8 Data: A Case Study in Bohai Sea, China

“…The MLP sea fog identification method was applied using Python programming, and the algorithm was verified using the CALIPSO-VFM classification points; the results it produced were then compared with those obtained with the NDSI threshold method proposed by [11]. The accuracy evaluation indices used in this paper are based on the work of [7], including the probability of detection (POD), probability of false detection (PFD), probability of missing detection (PMD), and equitable threat score (ETS). The calculation formulas of the four indicators are respectively shown in Equations ( 21)-( 24):…”

“…For instance, the multiband threshold algorithm [5] and U-Net deep learning method [6] were used for sea fog detection based on Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data. An unsupervised learning algorithm was used for sea fog detection based on Communication, Ocean, and Meteorological Satellite (COMS) data [7], while a dual-satellite method was proposed by [8], which combined Himawari-8 and FY-4A satellite data to detect sea fog at dawn. The authors of [9] proposed a decision tree approach that combined Himawari-8 and Geostationary Ocean Color Imager (GOCI) satellite data to detect sea fog.…”

“…where R i represents the navigational risk value in grid i, and x i,z represents the standardized value of criterion z in grid i. The calculated navigational risk values are standardized using Equation (7). After standardization, the final navigational risk assessment values are obtained, which are divided into four levels: [0, 0.25), [0.25, 0.5), [0.5, 0.75), and [0.75, 1], representing "safe", "a little safer", "a little dangerous", and "dangerous", respectively.…”

A scSE-LinkNet Deep Learning Model for Daytime Sea Fog Detection

“…Additionally, a number of studies have taken a different approach for sea fog detection using machine learning. This includes methods such as the expectation maximization algorithm (EM) [21] and decision tree (DT) [22] to precisely differentiate between stratus and sea fog. Although the introduction of machine learning further clarifies the stratus and sea fog boundaries, the process is more cumbersome because of the transformation and visualization of the detection results.…”

Probability Index of Low Stratus and Fog at Dawn using Dual Geostationary Satellite Observations from COMS and FY-2D near the Korean Peninsula

“…Kim et al [20], and Shin and Kim [21] used satellite observations (COMS or Himawari-8) for nighttime fog detection in the region, in conjunction with additional land and sea surface temperature information. Although the BTD 3.7-11 threshold is expected to be more useful for night observations rather than dawn and dusk, the detection in their studies [20,21] was inaccurate (false alarm ratio; FAR = 0.43 − 0.56) and requires additional information (e.g., multiple satellite observations, more independent channels) for improvement. The inaccuracy may also be due to uncertainties caused by using land/sea surface temperatures as low boundary conditions, and the presence of higher clouds [22].…”