• Journal of Korea Water Resources Association
• /
• v.56 no.8
• /
• pp.471-484
• /
• 2023

Deep learning models based on generative adversarial neural networks are specialized in generating new information based on learned information. The deep generative models (DGMR) model developed by Google DeepMind is an generative adversarial neural network model that generates predictive radar images by learning complex patterns and relationships in large-scale radar image data. In this study, the DGMR model was trained using radar rainfall observation data from the Ministry of Environment, and rainfall prediction was performed using an generative adversarial neural network for a heavy rainfall case in August 2021, and the accuracy was compared with existing prediction techniques. The DGMR generally resembled the observed rainfall in terms of rainfall distribution in the first 60 minutes, but tended to predict a continuous development of rainfall in cases where strong rainfall occurred over the entire area. Statistical evaluation also showed that the DGMR method is an effective rainfall prediction method compared to other methods, with a critical success index of 0.57 to 0.79 and a mean absolute error of 0.57 to 1.36 mm in 1 hour advance prediction. However, the lack of diversity in the generated results sometimes reduces the prediction accuracy, so it is necessary to improve the diversity and to supplement it with rainfall data predicted by a physics-based numerical forecast model to improve the accuracy of the forecast for more than 2 hours in advance.

• Journal of Korea Water Resources Association
• /
• v.38 no.1
• /
• pp.11-23
• /
• 2005

The objective of this study is to develop a short-term precipitation-streamflow coupling method for real-time river flow forecast. The coupled method is based on the RDAPS model for precipitation and atmospheric simulation and the SFM model for streamflow simulation. The selected study area is the 2,703-km$^2$ Soyang River basin with outlet at Soyang dam site. The rainfall-runoff event from 18 to 24 July 2003 is selected for the performance test of predicted precipitation and streamflow. It can be seen that the simulated basin-scale precipitation from the RDAPS can be useable as an input for SFM hydrologic model. Short-term hydrometeorological simulations using the RDAPS and SFM model were well captured important hydrometeorological characteristics in this study area. It is concluded that atmospheric precipitation forecast would be useful for streamflow forecast.

• Journal of Korea Water Resources Association
• /
• v.49 no.4
• /
• pp.327-333
• /
• 2016

Due to climate changes accelerated by global warming, South Korea has experienced regional climate variations as well as increasing severities and frequencies of extreme weather. The precipitation in South Korea during the summer season in 2013 was concentrated mainly in the central region; the maximum number of rainy days were recorded in the central region while the southern region had the minimum number of rainy days. As a result, much attention has been paid to the importance of flood control due to damage caused by spatiotemporal intensive rainfalls. In this study, forecast rainfall data was used for rapid responses to prevent disasters during flood seasons. For this purpose, the applicability of numerical weather forecast data was analyzed using the ground observation rainfall and inflow rate. Correlation coefficient, maximum rainfall intensity percent error and total rainfall percent error were used for the quantitative comparison of ground observation rainfall data. In addition, correlation coefficient, Nash-Sutcliffe efficiency coefficient, and standardized RMSE were used for the quantitative comparison of inflow rate. As a result of the simulation, the correlation coefficient up to six hours was 0.7 or higher, indicating a high correlation. Furthermore, the Nash-Sutcliffe efficiency coefficient was positive until six hours, confirming the applicability of forecast rainfall.

• Journal of Korea Water Resources Association
• /
• v.57 no.5
• /
• pp.333-346
• /
• 2024

High-resolution medium-range streamflow prediction is crucial for sustainable water quality and aquatic ecosystem management. For reliable medium-range streamflow predictions, it is necessary to understand the characteristics of forcings and to effectively utilize weather forecast data with low spatio-temporal resolutions. In this study, we presented a comparative analysis of medium-range streamflow predictions using the distributed hydrological model, WRF-Hydro, and the numerical weather forecast Global Data Assimilation and Prediction System (GDAPS) in the Geumho River basin, Korea. Multiple forcings, ground observations (AWS&ASOS), numerical weather forecast (GDAPS), and Global Land Data Assimilation System (GLDAS), were ingested to investigate the performance of streamflow predictions with highresolution WRF-Hydro configuration. In terms of the mean areal accumulated rainfall, GDAPS was overestimated by 36% to 234%, and GLDAS reanalysis data were overestimated by 80% to 153% compared to AWS&ASOS. The performance of streamflow predictions using AWS&ASOS resulted in KGE and NSE values of 0.6 or higher at the Kangchang station. Meanwhile, GDAPS-based streamflow predictions showed high variability, with KGE values ranging from 0.871 to -0.131 depending on the rainfall events. Although the peak flow error of GDAPS was larger or similar to that of GLDAS, the peak flow timing error of GDAPS was smaller than that of GLDAS. The average timing errors of AWS&ASOS, GDAPS, and GLDAS were 3.7 hours, 8.4 hours, and 70.1 hours, respectively. Medium-range streamflow predictions using GDAPS and high-resolution WRF-Hydro may provide useful information for water resources management especially in terms of occurrence and timing of peak flow albeit high uncertainty in flood magnitude.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2009.05a
• /
• pp.1141-1144
• /
• 2009

인공신경망(artificial neural network) 기법은 인간의 두뇌 신경세포의 활동을 모형화한 것으로 오랜 시간동안 발전해 왔으며 여러 분야에서 활용되고 있고 수문분야에서도 인공신경망을 이용한 연구가 활발히 진행되어 왔다. RDAPS와 같은 단기수치예보 자료는 강우의 유무 판단과 같은 정성적인 분석에서 비교적 정확도가 높지만 정확한 강우량의 추정과 같은 정량적인 부분에서는 정확도가 매우 낮으므로 인공신경망 기법과 같은 후처리 기법을 통해서 정확도를 높이게 된다. 인공신경망 기법을 수행할 때, 가장 중요한 것은 입력변수선택(input variable selection)으로 입력 변수의 적절한 선택이 결과값에 큰 영향을 주게 된다. 본 연구에서는 mutual information을 입력 변수 선택 기법으로 채택하여, 인공신경망의 입력변수 선정의 정확도를 알아보고자 한다. Mutual information은 주어진 자료의 엔트로피값을 이용하여 변수들 간의 독립과 종속의 관계를 나타내는 기법으로서, MI값은 '0'에서 '1'의 값을 가지며 '0'에 가까울수록 변수들 간의 관계가 독립적이고 '1'에 가까울수록 종속적인 관계를 나타낸다. 인공신경망의 입력변수선정에 대한 mutual information의 정확도를 알아보기 위해, 기존 입력변수선택 기법과 mutual information을 이용했을 경우의 인공신경망의 처리능력, 정확도를 비교 검토하였다.

• Journal of Korea Water Resources Association
• /
• v.41 no.9
• /
• pp.895-905
• /
• 2008

To analyze hydrologic processes in a watershed requires both various geographical data and hydrological time series data. Recently, not only geographical data such as DEM(Digital Elevation Model) and hydrologic thematic map but also hydrological time series from numerical weather prediction and rainfall radar have been provided as grid data, and there are studies on hydrologic analysis using these grid data. In this study, GRM(Grid based Rainfall-runoff Model) which is physically-based distributed rainfall-runoff model has been developed to simulate short term rainfall-runoff process effectively using these grid data. Kinematic wave equation is used to simulate overland flow and channel flow, and Green-Ampt model is used to simulate infiltration process. Governing equation is discretized by finite volume method. TDMA(TriDiagonal Matrix Algorithm) is applied to solve systems of linear equations, and Newton-Raphson iteration method is applied to solve non-linear term. Developed model was applied to simplified hypothetical watersheds to examine model reasonability with the results from $Vflo^{TM}$. It was applied to Wicheon watershed for verification, and the applicability to real site was examined, and simulation results showed good agreement with measured hydrographs.

• Korean Journal of Remote Sensing
• /
• v.29 no.4
• /
• pp.423-441
• /
• 2013

On 21 September 2010, one of Chuseok holidays in Korea, localized heavy rainfalls occurred over the midwestern region of the Korean peninsula. In this study MTSAT-2 infrared and water vapor channel imagery are examined to find out some features which are obvious in each stage of the life cycle of convective cell for this heavy rain event. Also the kinematic and thermodynamic features probably associated with them are investigated. The first clouds related with the Chuseok heavy rain are detected as low-level multicell cloud (brightness temperature: $-15{\sim}0^{\circ}C$) in the middle of the Yellow sea at 1630~1900 UTC on 20 Sept., which are probably associated with the convergence at 1000 hPa. Convective cells are initiated in the vicinity of Shantung peninsula at 1933 UTC 20, which have developed around the edge of the dark region in water vapor images. At two times of 0033 and 0433 UTC 21 the merging of two convective cells happens near midwestern coast of the peninsula and then they have developed rapidly. From 0430 to 1000 UTC 21, key features of convective cell include repeated formation of secondary cell, slow horizontal cloud motion, persistence of lower brightness temperature ($-75{\sim}-65^{\circ}C$), and relatively small cloud size (${\leq}-50^{\circ}C$) of about $30,000km^2$. Radar analysis showed that this heavy rain is featured by a narrow line-shaped rainband with locally heavy rainrate (${\geq}50$ mm/hr), which is located in the south-western edge of the convective cell. However there are no distinct features in the associated synoptic-scale dynamic forcing. After 1000 UTC 21 the convective cell grows up quickly in cloud size and then is dissipated. These satellite features may be employed for very short range forecast and nowcasting of mesoscale heavy rain system.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2008.05a
• /
• pp.2179-2183
• /
• 2008

지금까지 분포형 모형 개발에 대한 많은 노력이 있음에도 불구하고 여러 제약사항들에 의해 잠재력을 보여주는 정도로 활용되어 왔으나, 최근 급속도로 발전하는 컴퓨터의 계산능력, DEM 등 디지털정보의 구축이 진행되어 오고 있고, GIS 및 인공위성 영상기법의 발달로 공간적인 비균질성을 고려하여 유출과정에서 운동역학적인 이론을 기반으로 물의 흐름을 수리학적으로 추적해 나가는 물리적기반의 분포형 유출모형의 활용도가 높아지고 있다. 본 모형개발에 있어 이론적 배경이 된 모형은 1998년부터 일본 교토대학 방재연구소 코지리 연구실에서 개발 중인 Hydro-BEAM으로 유역 물순환의 건전성을 평가하기 위하여 장기간의 유역 내 유량, 수질을 시계열 및 공간적으로 파악하여 유역의 영향평가를 위해 개발된 물리적 기반의 격자구조를 가진 분포형 장기유출 모형이다. 유출량 계산은 유역내 수평 유출량산정 모듈로서 평면 분포형의 격자형을, 연직 분포형으로는 $A{\sim}B$층의 수평유출량은 하천으로 유입하고, C층은 하천유량에 영향을 미치지 않는 지하수층으로 가정하는 다층모형을 이용해서 A층, 지표 및 하도흐름은 운동파 법(kinematic wave)으로, $B{\sim}C$층의 유출량은 선형저류법으로 계산하는 모형이다. 본 연구에서는 격자흐름방향을 4방향에서 8방향으로 개선하였고, 모형의 각종 수문매개변수들을 GIS와 연계하여 직접 입력할 수 있도록 하였으며, 물리적기반의 침투과정을 모의할 수 있도록 Green & Ampt모듈을 추가하고, 향후 레이더 강우 및 수치예보강우의 홍수유출예측을 염두에 두고 격자강우량을 활용할 수 있도록 하는 등 홍수유출해석을 위한 분포형 강우-유출모형으로 개선 하였고, 이를 남강댐유역에 적용해 봄으로써 모형의 적용성을 검토해 보고자 하였다. 홍수기동안의 지표흐름과 지표하 흐름의 시간적 변화와 공간적 분포를 모의할 수 있었으며, 전처리과정으로서 ArcGIS 혹은 ArcView등의 GIS 프로그램을 이용하여 모형에 필요한 ASCII형태의 입력 매개 변수 자료들을 가공하였다. 또한 후처리과정으로서 모형의 수행결과인 유역내의 유출량 분포 등을 GIS상에서 나타낼 수 있도록 ASCII형태로 출력하도록 구성하였다. 남강댐유역을 대상으로 유역을 500m의 정방형 격자로 분할하고 수계망을 통하여 유역 출구까지 운동파이론에 의해 추적 계산하였으며, 수문곡선 비교결과 재현성 높은 결과를 보여주었다. 모형의 정확성 및 실용성에 대한 보다 정확한 평가를 위해서는 향후 다양한 강우 사상 혹은 다양한 크기의 유역에 대한 유출량의 재현성 및 매개변수 등에 검증이 이루어져야 할 것이다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.19-19
• /
• 2023

최근 기후변화에 의해 단기간에 많은 양의 집중호우가 발생하여 도시지역의 침수 피해가 증가하고 있다. 이에 도시지역의 홍수 피해 해결을 위해 도심지 홍수 발생 시 홍수정도 및 상황을 파악할 수 있는 장비가 개발되었으나, 실용화 단계까지는 진행이 미흡한 상황이다. 또한 기존 도시지역 홍수 현상 및 원인을 분석하기 위해 수치모형을 활용하고 있으나, 우수관망의 노후화 및 초기 강우패턴 적용에 대한 정확한 해석결과의 어려워 활용성이 낮다. 또한 홍수정도와 발생상황 인지를 위한 계측 장비의 개발 연구는 지속적으로 진행되고 있으나, 계측 장비의 높은 가격으로 전국적으로 설치 할 수 없는 상황으로 이를 대응하기 위한 별도의 방안 마련이 필요한 실정이다. 이를 위해 본 과제에서는 고성능·저비용 계측센서를 개발하여 실용화 가능성을 높이고, 전국에 산재되어있는 CCTV(교통상황, 방법용 등)의 영상을 활용한 침수상황 인지 기술 개발, 계측 데이터와 모니터링 데이터의 활용을 위한 빅데이터 개방 플랫폼을 구축하여, 상습 침수지역에 대해 실시간 감시가 가능한 계측 시스템의 정형 데이터와 CCTV 및 영상 등 모니터링 장비의 비정형 데이터의 분석 기술을 결합한 새로운 도심지 홍수 감시 기술의 개발을 목표로 한다. 이를 위해 본 연구 1차년도에 지표면 침수심 계측센서와 우수관망 월류심 계측센서를 개발하였으며, 2차년도에는개발된 계측센서의 현장실증을 통해 데이터를 수집한다. 수집된 계측센서 데이터와 비정형(CCTV 영상) 데이터의 AI학습을 통해 분석된 침수심, 침수범위, 침수면적 데이터는 도심지 홍수 정보 프로그램을 통해 표출되며, 최종적으로는 현장 상황을 쉽게 파악 가능한 3D 레이어의 형식으로 표출하고자 한다. 추후 도심지 홍수 정보 프로그램을 통해 표출되는 3D 레이어는 환경부가 추진하는 DT(Digital Twin) 연계 인공지능(AI) 홍수예보 사업과의 연계 시 도심지 홍수 지도 구축을 위한 자료로 활용될 수 있을 것으로 판단된다.

• Atmosphere
• /
• v.33 no.2
• /
• pp.275-295
• /
• 2023

In Korea, there are four institutions related to atmospheric science: the university's atmospheric science-related department, the Korea Meteorological Administration (KMA), the ROK Air Force Weather Group, and the Meteorological Industry Association. These four institutions have developed while maintaining a deep cooperative relationship with the Korea Meteorological Society (KMS) for the past 60 years. At the university, 6,986 bachelors, 1,595 masters, and 505 doctors, who are experts in meteorology and climate, have been accredited by 2022 at 7 universities related to atmospheric science. The KMA is carrying out national meteorological tasks to protect people's lives and property and foster the meteorological industry. The ROK Air Force Weather Group is in charge of military meteorological work, and is building an artificial intelligence and space weather support system through cooperation with universities, the KMA, and the KMS. Although the Meteorological Industry Association has a short history, its members, sales, and the number of employees are steadily increasing. The KMS greatly contributed to raising the national meteorological service to the level of advanced countries by supporting the development of universities, the KMA, the Air Force Meteorological Agency, and the Meteorological Industry Association.