Search

Ocean fog, a meteorological phenomenon characterized by reduced visibility due to tiny water droplets or ice particles, poses significant safety risks for maritime activities and coastal regions. Accurate prediction of ocean fog is crucial but challenging due to its complex formation mechanisms and variability. This study proposes an advanced ocean fog prediction model for the Yellow Sea region, leveraging satellite-based detection and high-performance data-driven methods. We used Himawari-8 satellite data to obtain a lot of spatiotemporal ocean fog references and employed AutoML to integrate numerical weather prediction (NWP) outputs and sea surface temperature (SST)-related variables. The model demonstrated superior performance compared to traditional NWP-based methods, achieving high performance in both quantitative—probability of detection of 81.6%, false alarm ratio of 24.4%, f1 score of 75%, and proportion correct of 79.8%—and qualitative evaluations for 1 to 6 h lead times. Key contributing variables included relative humidity, accumulated shortwave radiation, and atmospheric pressure, indicating the importance of integrating diverse data sources. The study emphasizes the potential of using satellite-derived data to improve ocean fog prediction, while also addressing the challenges of overfitting and the need for more comprehensive reference data.

Ocean–fog is a type of fog that forms over the ocean and has a visibility of less than 1 km. Ocean–fog frequently causes incidents over oceanic and coastal regions; ocean–fog detection is required regardless of the time of day. Ocean–fog has distinct thermo-optical properties, and spatially and temporally extensive ocean–fog detection methods based on geostationary satellites are typically employed. Infrared (IR) channels of Himawari-8 were used to construct three machine-learning models for the continuous detection of ocean–fog. In contrast, visible channels are valid only during the daytime. As control models, we used fog products from the National Meteorological Satellite Center (NMSC) and machine-learning models trained by adding a visible channel. The extreme gradient boosting model utilizing IR channels corrected ocean–fog perfectly day and night, with the highest F1 score of 97.93% and a proportion correct (PC) of 98.59% throughout the day. In contrast, the NMSC product had a probability of detection of 87.14%, an F1 score of 93.13%, and a PC of 71.9%. As demonstrated by the qualitative evaluation, the NMSC product overestimates clouds over small and coarsely textured ocean–fog regions. In contrast, the proposed model distinguishes between ocean–fog, clear skies, and clouds at the pixel scale. The Shapley additive explanation analysis demonstrated that the difference between channels 14 and 7 was very useful for ocean–fog detection at night, and its extremely low values contributed significantly to distinguishing nonfog during the daytime. Channel 15, affected by water vapor absorption, contributed most to ocean–fog detection among atmospheric window channels. The research findings can be used to improve operational ocean–fog detection and forecasting.

Near-surface relative humidity (RHns) is an essential meteorological parameter for water, carbon, and climate studies. However, spatially continuous RHns estimation is difficult due to the spatial discontinuity of in situ observations and the cloud contamination of satellite-based data. This article proposed machine learning-based models to estimate spatially continuous daily RHns at 1 km resolution over Japan and South Korea under all sky conditions and examined the spatiotemporal patterns of RHns. All sky estimation of RHns using machine learning has been rarely conducted, and it can be an alternative to the currently available RHns data mostly from numerical models, which have relatively low spatial resolution. We combined two schemes for clear sky conditions (scheme A, which uses satellite and reanalysis data) and cloudy sky conditions (scheme B, which uses reanalysis data solely). The relatively small numbers of data in extremely low and high RHns conditions (i.e., 70%, respectively) were augmented by applying an oversampling method to avoid biased training. The machine learning models based on random forest (RF) and XGBoost were trained and validated using 94 in situ observation sites from meteorological administrations of both countries from 2012 to 2017. The results showed that XGBoost produced slightly better performance than RF, and the spatially continuous RHns model combined based on XGBoost yielded the coefficient of determination of 0.72 and a root-mean-square error of 10.61%. Spatiotemporal patterns of the estimated RHns agreed with in situ observations, reflecting the effect of topography on RHns. We expect that the proposed RHns model could be used in various environmental studies that require RHns under all sky conditions as input data.

Tropical cyclones (TCs) are destructive natural disasters. Accurate prediction and monitoring are important to mitigate the effects of natural disasters. Although remarkable efforts have been made to understand TCs, operational monitoring information still depends on the experience and knowledge of forecasters. In this study, a fully automated geostationary-satellite-based TC center estimation approach is proposed. The proposed approach consists of two improved methods: the setting of regions of interest (ROI) using a score matrix (SCM) and a TC center determination method using an enhanced logarithmic spiral band (LSB) and SCM. The former enables prescreening of the regions that may be misidentified as TC centers during the ROI setting step, and the latter contributes to the determination of an accurate TC center, considering the size and length of the TC rainband in relation to its intensity. Two schemes, schemes A and B, were examined depending on whether the forecasting data or real-time observations were used to determine the initial guess of the TC centers. For each scheme, two models were evaluated to discern whether SCM was combined with LSB for TC center determination. The results were investigated based on TC intensity and phase to determine the impact of TC structural characteristics on TC center determination. While both proposed models improved the detection performance over the existing approach, the best-performing model (i.e., LSB combined with SCM) achieved skill scores (SSs) of +17.4% and +20.8% for the two schemes. In particular, the model resulted in a significant improvement for strong TCs (categories 4 and 5), with SSs of +47.8% and +72.8% and +41.2% and +72.3% for schemes A and B, respectively. The research findings provide an improved understanding of the intensity- and phase-wise spatial characteristics of TCs, which contributes to objective TC center estimation.

For a long time, researchers have tried to find a way to analyze tropical cyclone (TC) intensity in real-time. Since there is no standardized method for estimating TC intensity and the most widely used method is a manual algorithm using satellite-based cloud images, there is a bias that varies depending on the TC center and shape. In this study, we adopted convolutional neural networks (CNNs) which are part of a state-of-art approach that analyzes image patterns to estimate TC intensity by mimicking human cloud pattern recognition. Both two dimensional-CNN (2D-CNN) and three-dimensional-CNN (3D-CNN) were used to analyze the relationship between multi-spectral geostationary satellite images and TC intensity. Our best-optimized model produced a root mean squared error (RMSE) of 8.32 kts, resulting in better performance (~35%) than the existing model using the CNN-based approach with a single channel image. Moreover, we analyzed the characteristics of multi-spectral satellite-based TC images according to intensity using a heat map, which is one of the visualization means of CNNs. It shows that the stronger the intensity of the TC, the greater the influence of the TC center in the lower atmosphere. This is consistent with the results from the existing TC initialization method with numerical simulations based on dynamical TC models. Our study suggests the possibility that a deep learning approach can be used to interpret the behavior characteristics of TCs.

This paper proposes a complete framework of a machine learning-based model that detects convective initiation (CI) from geostationary meteorological satellite data. The suggested framework consists of three main processes: (1) An automated sampling tool; (2) machine learning-based CI detection modelling; (3) repeated model tuning through validation. In this study, the automated sampling tool was able to track the CI objects iteratively, even without ancillary data such as an atmospheric motion vector (AMV). The collected samples were used to train the machine learning model for CI detection. Random forest (RF) was used to classify the CI and non-CI. To enhance the advantages of the machine learning approach, we adopted model tuning to iteratively update the training dataset from each validation result by adding hits and misses to the CI samples, and false alarms and correct negatives to the non-CI samples. Using 12 interest fields from the Himawari-8 Advanced Himawari Imager (AHI) over the Korean Peninsula, this simple and intuitive tuning process increased the overall probability of detection (POD) from 0.79 to 0.82 and decreased the overall false alarm rate (FAR) from 0.46 to 0.37 with around 40 min of the lead-time. Amongst the 12 interest fields, T b (11.2) µm was identified as the most significant predictor in the RF model, followed by T b (8.6—11.2) µm, and T b (6.2−7.3) µm. The effect of model tuning on the CI detection performance was also analyzed using spatiotemporal validation maps. By automatically collecting and updating the machine learning training dataset, the suggested framework is expected to help the maintenance of the CI detection model from an operational perspective.

Even though deicing or airframe coating technologies continue to develop, aircraft icing is still one of the critical threats to aviation. While the detection of potential icing clouds has been conducted using geostationary satellite data in the US and Europe, there is not yet a robust model that detects potential icing areas in East Asia. In this study, we proposed machine-learning-based icing detection models using data from two geostationary satellites—the Communication, Ocean, and Meteorological Satellite (COMS) Meteorological Imager (MI) and the Himawari-8 Advanced Himawari Imager (AHI)—over Northeast Asia. Two machine learning techniques—random forest (RF) and multinomial log-linear (MLL) models—were evaluated with quality-controlled pilot reports (PIREPs) as the reference data. The machine-learning-based models were compared to the existing models through five-fold cross-validation. The RF model for COMS MI produced the best performance, resulting in a mean probability of detection (POD) of 81.8%, a mean overall accuracy (OA) of 82.1%, and mean true skill statistics (TSS) of 64.0%. One of the existing models, flight icing threat (FIT), produced relatively poor performance, providing a mean POD of 36.4%, a mean OA of 61.0, and a mean TSS of 9.7%. The Himawari-8 based models also produced performance comparable to the COMS models. However, it should be noted that very limited PIREP reference data were available especially for the Himawari-8 models, which requires further evaluation in the future with more reference data. The spatio-temporal patterns of the icing areas detected using the developed models were also visually examined using time-series satellite data.

Melt ponds, a common feature on Arctic sea ice, absorb most of the incoming solar radiation and have a large effect on the melting rate of sea ice, which significantly influences climate change. Therefore, it is very important to monitor melt ponds in order to better understand the sea ice-climate interaction. In this study, melt pond retrieval models were developed using the TerraSAR-X dual-polarization synthetic aperture radar (SAR) data with mid-incidence angle obtained in a summer multiyear sea ice area in the Chukchi Sea, the Western Arctic, based on two rule-based machine learning approaches—decision trees (DT) and random forest (RF)—in order to derive melt pond statistics at high spatial resolution and to identify key polarimetric parameters for melt pond detection. Melt ponds, sea ice and open water were delineated from the airborne SAR images (0.3-m resolution), which were used as a reference dataset. A total of eight polarimetric parameters (HH and VV backscattering coefficients, co-polarization ratio, co-polarization phase difference, co-polarization correlation coefficient, alpha angle, entropy and anisotropy) were derived from the TerraSAR-X dual-polarization data and then used as input variables for the machine learning models. The DT and RF models could not effectively discriminate melt ponds from open water when using only the polarimetric parameters. This is because melt ponds showed similar polarimetric signatures to open water. The average and standard deviation of the polarimetric parameters based on a 15 × 15 pixel window were supplemented to the input variables in order to consider the difference between the spatial texture of melt ponds and open water. Both the DT and RF models using the polarimetric parameters and their texture features produced improved performance for the retrieval of melt ponds, and RF was superior to DT. The HH backscattering coefficient was identified as the variable contributing the most, and its spatial standard deviation was the next most contributing one to the classification of open water, sea ice and melt ponds in the RF model. The average of the co-polarization phase difference and the alpha angle in a mid-incidence angle were also identified as the important variables in the RF model. The melt pond fraction and sea ice concentration retrieved from the RF-derived melt pond map showed root mean square deviations of 2.4% and 4.9%, respectively, compared to those from the reference melt pond maps. This indicates that there is potential to accurately monitor melt ponds on multiyear sea ice in the summer season at a local scale using high-resolution dual-polarization SAR data.

Melt ponds, a common feature on Arctic sea ice, absorb most of the incoming solar radiation and have a large effect on the melting rate of sea ice, which significantly influences climate change. Therefore, it is very important to monitor melt ponds in order to better understand the sea ice-climate interaction. In this study, melt pond retrieval models were developed using the TerraSAR-X dual-polarization synthetic aperture radar (SAR) data with mid-incidence angle obtained in a summer multiyear sea ice area in the Chukchi Sea, the Western Arctic, based on two rule-based machine learning approaches--decision trees (DT) and random forest (RF)--in order to derive melt pond statistics at high spatial resolution and to identify key polarimetric parameters for melt pond detection. Melt ponds, sea ice and open water were delineated from the airborne SAR images (0.3-m resolution), which were used as a reference dataset. A total of eight polarimetric parameters (HH and W backscattering coefficients, co-polarization ratio, co-polarization phase difference, co-polarization correlation coefficient, alpha angle, entropy and anisotropy) were derived from the TerraSAR-X dual-polarization data and then used as input variables for the machine learning models. The DT and RF models could not effectively discriminate melt ponds from open water when using only the polarimetric parameters. This is because melt ponds showed similar polarimetric signatures to open water. The average and standard deviation of the polarimetric parameters based on a 15 x 15 pixel window were supplemented to the input variables in order to consider the difference between the spatial texture of melt ponds and open water. Both the DT and RF models using the polarimetric parameters and their texture features produced improved performance for the retrieval of melt ponds, and RF was superior to DT. The HH backscattering coefficient was identified as the variable contributing the most, and its spatial standard deviation was the next most contributing one to the classification of open water, sea ice and melt ponds in the RF model. The average of the co-polarization phase difference and the alpha angle in a mid-incidence angle were also identified as the important variables in the RF model. The melt pond fraction and sea ice concentration retrieved from the RF-derived melt pond map showed root mean square deviations of 2.4% and 4.9%, respectively, compared to those from the reference melt pond maps. This indicates that there is potential to accurately monitor melt ponds on multiyear sea ice in the summer season at a local scale using high-resolution dual-polarization SAR data. [ABSTRACT FROM AUTHOR]