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.

Overshooting convective cloud Top (OT)-accompanied clouds can cause severe weather conditions, such as lightning, strong winds, and heavy rainfall. The distribution and behavior of OTs can affect regional and global climate systems. In this paper, we propose a new approach for OT detection by using machine learning methods with multiple infrared images and their derived features. Himawari-8 satellite images were used as the main input data, and binary detection (OT or nonOT) with class probability was the output of the machine learning models. Three machine learning techniques-random forest (RF), extremely randomized trees (ERT), and logistic regression (LR)-were used to develop OT classification models to distinguish OT from non-OT. The hindcast validation over the Southeast Asia and West Pacific regions showed that RF performed best, resulting in a mean probabilities of detection (POD) of 77.06% and a mean false alarm ratio (FAR) of 36.13%. Brightness temperature at 11.2 µm (Tb11) and its standard deviation (STD) in a 3 × 3 window size were identified as the most contributing variables for discriminating OT and nonOT classes. The proposed machine learning-based OT detection algorithms produced promising results comparable to or even better than the existing approaches, which are the infrared window (IRW)-texture and water vapor (WV) minus IRW brightness temperature difference (BTD) methods.

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.

In this research, climate classification maps over the Korean Peninsula at 1 km resolution were generated using the satellite-based climatic variables of monthly temperature and precipitation based on machine learning approaches. Random forest (RF), artificial neural networks (ANN), k-nearest neighbor (KNN), logistic regression (LR), and support vector machines (SVM) were used to develop models. Training and validation of these models were conducted using in-situ observations from the Korea Meteorological Administration (KMA) from 2001 to 2016. The rule of the traditional Köppen-Geiger (K-G) climate classification was used to classify climate regions. The input variables were land surface temperature (LST) of the Moderate Resolution Imaging Spectroradiometer (MODIS), monthly precipitation data from the Tropical Rainfall Measuring Mission (TRMM) 3B43 product, and the Digital Elevation Map (DEM) from the Shuttle Radar Topography Mission (SRTM). The overall accuracy (OA) based on validation data from 2001 to 2016 for all models was high over 95%. DEM and minimum winter temperature were two distinct variables over the study area with particularly high relative importance. ANN produced more realistic spatial distribution of the classified climates despite having a slightly lower OA than the others. The accuracy of the models using high altitudinal in-situ data of the Mountain Meteorology Observation System (MMOS) was also assessed. Although the data length of the MMOS data was relatively short (2013 to 2017), it proved that the snowy, dry and cold winter and cool summer class (Dwc) is widely located in the eastern coastal region of South Korea. Temporal shifting of climate was examined through a comparison of climate maps produced by period: from 1950 to 2000, from 1983 to 2000, and from 2001 to 2013. A shrinking trend of snow classes (D) over the Korean Peninsula was clearly observed from the ANN-based climate classification results. Shifting trends of climate with the decrease/increase of snow (D)/temperate (C) classes were clearly shown in the maps produced using the proposed approaches, consistent with the results from the reanalysis data of the Climatic Research Unit (CRU) and Global Precipitation Climatology Centre (GPCC).