Abstract This paper proposes a novel GAN framework with self-clustering approach for precipitation nowcasting (ClusterCast). Previous studies have primarily captured the motion vector using only single latent space, making models difficult to adapt disparate space-time distribution of precipitation. Environmental factors (e.g., regional characteristics and scale) an impact on systems can cause non-stationary distribution. To tackle this problem, our key idea is train generator network...

Accurate short-term precipitation forecast is of social and economic significance for preventing severe weather damage. Deep learning has been rapidly adopted in nowcasting based on radar, which plays a key role dangerous conditions such as torrential rainfall. However, the limited observation range radar imposes constraints shorter lead times. Securing sufficient time timely flood warnings emergency responses crucial. Here, we propose novel GAN-based framework that combines satellite data...

Summary This study aims to recommend metadata for building a high ranking in Search Engine Result Page (SERP) by considering Optimizations (SEO). For online marketing, it is important place their websites on the top rank result of search engines. However, on‐page techniques traditional SEO do not have logical foundation select metadata. Metadata an element prioritize when engine indexing user queries. Thereby, this proposes method recommending metadata, which consists two steps: i) combining...

Deep learning-based time series forecasting has dominated the short-term precipitation field with help of its ability to estimate motion flow in high-resolution datasets. The growing interest nowcasting offers substantial opportunities for advancement current technologies. Nevertheless, there been a scarcity in-depth surveys using deep learning. Thus, this paper systemically reviews recent progress models. Specifically, we investigate following key points within background components,...

This paper presents a system, namely, the abnormal-weather monitoring and curation service (AWMC), which provides people with better understanding of abnormal weather conditions. The can analyze set multivariate datasets (i.e., 7 meteorological from 18 cities in Korea) show (i) dates are mostly certain city, (ii) on date. In particular, dynamic graph-embedding-based anomaly detection method was employed to measure scores. We implemented conducted evaluations. Regarding results weather, AWMC...

Numerical weather prediction (NWP) models are fundamental in meteorology for simulating and forecasting the behavior of various atmospheric variables. The accuracy precipitation forecasts acquisition sufficient lead time crucial preventing hazardous events. However, performance NWP is limited by nonlinear unpredictable patterns extreme phenomena driven temporal dynamics. In this regard, we propose a \textbf{S}elf-\textbf{S}upervised \textbf{L}earning with \textbf{P}robabilistic...

Social event detection is regarded as an useful tool for understanding our society and more importantly providing people with various smart city services. However, given a large number of social big data, it hard to find out the meaningful patterns traditional statistical analysis. The aim this concept paper present how generate spatiotemporal constraints analyzing data. Particularly, To discover hidden event, spatio-temporal constraint generation process consists three methods, which are i)...

Deep learning has been rapidly adopted in short-term precipitation prediction, such as simulating movement and predicting extreme weather events. Recently, generative adversarial neural networks (GANs) have shown to be effective at dealing with field smoothing increasing lead time. Several studies (Jing et al., 2019; Ravuri 2021) demonstrated the potential of GAN by solving spatial problems demonstrating reliable predictive performance. However, despite promising results from GANs,...

Physical parameterization is one of the major components Numerical Weather Prediction system. In Korean Integrated Model (KIM), physical parameterizations account for about 30 % total computation time. There are many studies developing neural network based emulators to replace and accelerate physics parameterization. this study, we develop a planetary boundary layer(PBL) emulator which on Shin-Hong (Hong et al., 2006, 2010; Shin Hong, 2013, 2015) scheme that computes parameterized effects...

This paper proposes a novel generative method based on self-supervised learning (SSL) and GAN. The key idea is to train generator network predict precipitation sequences by subnetwork that automatically labels types from model. training process consists of (i) clustering the condition features radar observations using sub-network with convolutional layers in (ii) generalizing diverse according clustered labels, enabling various latent representations. Additionally, we attempt an ensemble...

This study utilized a convolutional neural network (CNN) architecture based on Convlstm and Trajgru models with CBAM method to predict rainfall. The radar satellite data collected preprocessed during summer seasons from 2019 2022 were used along additional auxiliary including longitude, latitude, terrain information perform deep learning-based rainfall prediction.CBAM is focus attention specific spatial channel relationships. are architectures designed process spatiotemporal data.Assuming...

Obtaining a sufficient forecast lead time for local precipitation is essential in preventing hazardous weather events. Global warming-induced climate change increases the challenge of accurately predicting severe events, such as heavy rainfall. In this paper, we propose deep learning-based post-processor numerical prediction (NWP) models. The consists (i) employing self-supervised pre-training, where parameters encoder are pre-trained on reconstruction masked variables atmospheric physics...

<p>In order to capture spatio-temporal characteristics of precipitation process in machine learning context, many studies applied convolutional and recurrent neural networks. Many state-of-the-art approaches focused on a single latent representation the quantitative forecast (QPF). To describe reflectivity echoes with variable may be an overly restrictive assumption, impeding effective features. Therefore, we propose conditioned forecasting model based self-supervised (SSL),...