• Journal of Information Science Theory and Practice
• /
• v.10 no.spc
• /
• pp.135-142
• /
• 2022

Recently, with the development of data processing technology and the increase of computational power, methods to solving social problems using Artificial Intelligence (AI) are in the spotlight, and AI technologies are replacing and supplementing existing traditional methods in various fields. Meanwhile in Korea, heavy rain is one of the representative factors of natural disasters that cause enormous economic damage and casualties every year. Accurate prediction of heavy rainfall over the Korean peninsula is very difficult due to its geographical features, located between the Eurasian continent and the Pacific Ocean at mid-latitude, and the influence of the summer monsoon. In order to deal with such problems, the Korea Meteorological Administration operates various state-of-the-art observation equipment and a newly developed global atmospheric model system. Nevertheless, for precipitation nowcasting, the use of a separate system based on the extrapolation method is required due to the intrinsic characteristics associated with the operation of numerical weather prediction models. The predictability of existing precipitation nowcasting is reliable in the early stage of forecasting but decreases sharply as forecast lead time increases. At this point, AI technologies to deal with spatio-temporal features of data are expected to greatly contribute to overcoming the limitations of existing precipitation nowcasting systems. Thus, in this project the dataset required to develop, train, and verify deep learning-based precipitation nowcasting models has been constructed in a regularized form. The dataset not only provides various variables obtained from multiple sources, but also coincides with each other in spatio-temporal specifications.

• Atmosphere
• /
• v.28 no.4
• /
• pp.443-455
• /
• 2018

Nowcasting algorithms using weather radar data are mostly based on extrapolating the radar echoes. We estimate the echo motion vectors that are used to extrapolate the echo properly. Therefore, understanding the general characteristics of these motion vectors is important to improve the performance of nowcasting. General characteristics of radar-based motions are analyzed for warm season precipitation over Jeju region. Three-year summer season data (June~August, 2011~2013) from two radars (GSN, SSP) in Jeju are used to obtain echo motion vectors that are retrieved by Variational Echo Tracking (VET) method which is widely used in nowcasting. The highest frequency occurs in precipitation motion toward east-northeast with the speed of $15{\sim}16m\;s^{-1}$ during the warm season. Precipitation system moves faster and eastward in June-July while it moves slower and northeastward in August. The maximum frequency of speed appears in $10{\sim}20m\;s^{-1}$ and $5{\sim}10m\;s^{-1}$ in June~July and August respectively while average speed is about $14{\sim}15m\;s^{-1}$ in June~July and $8m\;s^{-1}$ in August. In addition, the direction of precipitation motion is highly variable in time in August. The speed of motion in Lee side of the island is smaller than that of the windward side.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.123-123
• /
• 2020

The present study is aimed to correcting radar-based mean areal precipitation forecasts to improve urban flood predictions and uncertainty analysis of water levels contributed at each stage in the process. For this reason, a long short-term memory (LSTM) network is used to reproduce three-hour mean areal precipitation (MAP) forecasts from the quantitative precipitation forecasts (QPFs) of the McGill Algorithm for Precipitation nowcasting by Lagrangian Extrapolation (MAPLE). The Gangnam urban catchment located in Seoul, South Korea, was selected as a case study for the purpose. A database was established based on 24 heavy rainfall events, 22 grid points from the MAPLE system and the observed MAP values estimated from five ground rain gauges of KMA Automatic Weather System. The corrected MAP forecasts were input into the developed coupled 1D/2D model to predict water levels and relevant inundation areas. The results indicate the viability of the proposed framework for generating three-hour MAP forecasts and urban flooding predictions. For the analysis uncertainty contributions of the source related to the process, the Bayesian Markov Chain Monte Carlo (MCMC) using delayed rejection and adaptive metropolis algorithm is applied. For this purpose, the uncertainty contributions of the stages such as QPE input, QPF MAP source LSTM-corrected source, and MAP input and the coupled model is discussed.

• Journal of Korea Water Resources Association
• /
• v.46 no.8
• /
• pp.857-871
• /
• 2013

The interest in rainfall observation and forecasting using remote sensing method like RADAR (Radio Detection and Ranging) and satellite image is increased according to increased damage by rapid weather change like regional torrential rain and flash flood. In this study, the basin runoff was calculated using adaptive neuro-fuzzy technique, one of the data driven model and MAPLE (McGill Algorithm for Precipitation Nowcasting by Lagrangian Extrapolation) forecasted precipitation data as one of the input variables. The flood estimation method using neuro-fuzzy technique and RADAR forecasted precipitation data was evaluated. Six rainfall events occurred at flood season in 2010 and 2011 in Chungju Reservoir basin were used for the input data. The flood estimation results according to the rainfall data used as training, checking and testing data in the model setup process were compared. The 15 models were composed of combination of the input variables and the results according to change of clustering methods were compared and analysed. From this study was that using the relatively larger clustering radius and the biggest flood ever happened for training data showed the better flood estimation. The model using MAPLE forecasted precipitation data showed relatively better result at inflow estimation Chungju Reservoir.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.32 no.5B
• /
• pp.267-272
• /
• 2012

To reduce the flood risk caused by unexpected heavy rainfall, many prediction methods for flood have been developed. A major constituent of flood prediction is an accurate rainfall estimation which is an input of hydrologic models. In this study, a regional-scale weather model which can provide relatively longer lead time for flood mitigation compared to the Nowcasting based on radar system will be introduced and applied to the Chongmi river basin located in central part of South Korea. The duration of application of a regional weather model is from July 11 to July 23 in 2006. The estimated rainfall amounts were compared with observations from rain gauges (Sangkeuk, Samjook, and Sulsung). For this rainfall event at Chongmi river basin, Thomson and Kain-Frisch Schemes for microphysics and cumulus parameterization, respectively, were selected as optimal physical conditions to present rainfall fall amount in terms of Mean Absolute Relative Errors (MARE>0.45).

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2021.05a
• /
• pp.43-45
• /
• 2021

The short-term quantitative precipitation prediction (QPF) system is important socially and economically to prevent damage from severe weather. Recently, many studies for short-term QPF model applying the Deep Neural Network (DNN) has been conducted. These studies require the sophisticated pre-processing because the mistreatment of various and vast meteorological data sets leads to lower performance of QPF. Especially, for more accurate prediction of the non-linear trends in precipitation, the dataset needs to be carefully handled based on the physical and dynamical understands the data. Thereby, this paper proposes the following approaches: i) refining and combining major factors (weather radar, terrain, air temperature, and so on) related to precipitation development in order to construct training data for pattern analysis of precipitation; ii) producing predicted precipitation fields based on Convolutional with ConvLSTM. The proposed algorithm was evaluated by rainfall events in 2020. It is outperformed in the magnitude and strength of precipitation, and clearly predicted non-linear pattern of precipitation. The algorithm can be useful as a forecasting tool for preventing severe weather.

• Atmosphere
• /
• v.16 no.3
• /
• pp.247-257
• /
• 2006

The Yeong-dong heavy snowfall forecast supporting system has been developed during the last several years. In order to construct the conceptual model, we have examined the characteristics of heavy snowfalls in the Yeong-dong region classified into three precipitation patterns. This system is divided into two parts: forecast and observation. The main purpose of the forecast part is to produce value-added data and to display the geography based features reprocessing the numerical model results associated with a heavy snowfall. The forecast part consists of four submenus: synoptic fields, regional fields, precipitation and snowfall, and verification. Each offers guidance tips and data related with the prediction of heavy snowfalls, which helps weather forecasters understand better their meteorological conditions. The observation portion shows data of wind profiler and snow monitoring for application to nowcasting. The heavy snowfall forecast supporting system was applied and tested to the heavy snowfall event on 28 February 2006. In the beginning stage, this event showed the characteristics of warm precipitation pattern in the wind and surface pressure fields. However, we expected later on the weak warm precipitation pattern because the center of low pressure passing through the Straits of Korea was becoming weak. It was appeared that Gangwon Short Range Prediction System simulated a small amount of precipitation in the Yeong-dong region and this result generally agrees with the observations.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.100-100
• /
• 2023

단기 강우 예측에는 주로 물리과정 기반 수치예보모델(NWPs, Numerical Prediction Models) 과 레이더 기반 확률론적 방법이 사용되어 왔으며, 최근에는 머신러닝을 이용한 레이더 기반 강우예측 모델이 단기 강우 예측에 뛰어난 성능을 보이는 것을 확인하여 관련 연구가 활발히 진행되고 있다. 하지만 머신러닝 기반 모델은 예측 선행시간 증가 시 성능이 크게 저하되며, 또한 대기의 물리적 과정을 고려하지 않는 Black-box 모델이라는 한계점이 존재한다. 본 연구에서는 이러한 한계를 극복하기 위해 머신러닝 기반 blending 기법을 통해 물리과정 기반 수치예보모델인 Weather Research and Forecasting (WRF)와 최신 머신러닝 기법 (cGAN, conditional Generative Adversarial Network) 기반 모델을 결합한 Hybrid 강우예측모델을 개발하고자 하였다. cGAN 기반 모델 개발을 위해 1시간 단위 1km 공간해상도의 레이더 반사도, WRF 모델로부터 산출된 기상 자료(온도, 풍속 등), 유역관련 정보(DEM, 토지피복 등)를 입력 자료로 사용하여 모델을 학습하였으며, 모델을 통해 물리 정보 및 머신러닝 기반 강우 예측을 생성하였다. 이렇게 생성된cGAN 기반 모델 결과와 WRF 예측 결과를 결합하는 머신러닝 기반 blending 기법을 통해Hybrid 강우예측 결과를 최종적으로 도출하였다. 본 연구에서는 Hybrid 강우예측 모델의 성능을 평가하기 위해 수도권 및 안동댐 유역에서 발생한 호우 사례를 기반으로 최대 선행시간 6시간까지 모델 예측 결과를 분석하였다. 이를 통해 물리과정 기반 모델과 머신러닝 기반 모델을 결합하는 Hybrid 기법을 적용하여 높은 정확도와 신뢰도를 가지는 고해상도 강수 예측 자료를 생성할 수 있음을 확인하였다.

• Journal of Korea Water Resources Association
• /
• v.50 no.5
• /
• pp.335-346
• /
• 2017

This study generated the radar-based forecasted rainfall using spatial-scale decomposition method (SCDM) and evaluated the hydrological applicability with forecasted rainfall by KMA (MAPLE, KONOS) in terms of urban flood forecasting. SCDM is to separate the small-scale field (convective cell) and large-scale field (straitform cell) from radar rainfield. And each separated field is forecasted by translation model and storm tracker nowcasting model for improvement of QPF accuracy. As the evaluated results of various QPF for three rainfall events in Seoul and Metropolitan area, proposed method showed better prediction accuracy than MAPLE and KONOS considering the simplicity of the methodology. In addition, this study assessed the urban hydrological applicability for Gangnam basin. As the results, KONOS simulated the peak of water depth more accurately than MAPLE and SCDM, however cannot simulated the timeseries pattern of water depth. In the case of SCDM, the quantitative error was larger than observed water depth, but the simulated pattern was similar to observation. The SCDM will be useful information for flood forecasting if quantitative accuracy is improved through the adjustment technique and blending with NWP.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2012.05a
• /
• pp.984-984
• /
• 2012

지구온난화와 기후변화의 영향으로 전 지구적으로 이상홍수, 이상가뭄, 한파와 같은 이상기상 현상이 빈번하게 발생하고 있다. 국내에서는 2010년 추석 광화문 침수사태와 2011년 우면산 산사태와 같은 국지성 집중호우로 인한 인적 물적 피해가 속출하고 있다. 전통적으로 시기나 양적인 측면에서 대부분 장마기간에 국한되었던 강우집중현상이 과거와 달리 특정기간에 상관없이 발생하고 단기성, 국지성을 지닌 호우의 발생빈도가 높아지는 등 국내 강우의 특성이 변하고 있다. 이러한 변화에 대응하기 위해서 강우예측과 유출량예측의 정확도를 높이기 위한 시도가 다양하게 이루어지고 있다. 강우예측의 정확성을 높이기 위해 기상청에서는 단기예보를 목적으로 전지구 통합모델과 지역 통합모델을 연계한 동네예보를 수행하고 있으며, 초단기 예보를 위한 목적으로 VSRF, SCAN, VDRAS, MAPLE 등의 예보를 수행하고 있다. 홍수량 예측에서는 일반적으로 사용하고 있는 물리적 기반의 모형에 레이더강우와 같은 격자형 강우자료를 사용하여 정확성을 높이거나, 기존의 집중형 모형을 분포형 모형으로 대체하기 위한 연구 등이 이루어지고 있으며, 모형 구축이 간편하고 예측 정확도가 우수하다는 장점으로 인해 신경회로망이나 퍼지추론기법 등을 사용한 연구도 지속적으로 이루어지고 있다. 본 연구에서는 수자원분야에 산재한 불확실성을 적극적으로 인정하고 수학적으로 해석하기 위한 이론인 퍼지이론에 신경망 이론을 도입한 neuro-fuzzy 기법을 사용하여 홍수량을 예측하였다. 모형의 입력자료로는 관측된 강우자료와 유출량자료 및 기상청에서 제공하는 MAPLE(McGill Algorithm for Precipitation Nowcasting by Lagrangian Extrapolation) 강우예측자료를 사용하여 적용성을 평가해보았다. 모형의 적용성을 평가하기 위해 시험유역을 충주댐 상류 유역으로 선정하였으며, 2010년 2011년 홍수기의 충주댐 유입량을 예측하였다. 모형의 입력자료를 변경하여 입력자료의 변화에 따른 결과를 비교하였고, clustering 반경의 변화에 따른 정확도를 비교하였다. 모형의 정확도는 평균제곱근오차와 첨두수위오차를 통해 비교하였으며, 비교결과 전반적으로 lead time이 길어질수록 MAPLE 사용 시 예측 정확도가 우수하였고, clustering 반경은 0.5일 때 가장 우수한 결과를 보였다.