• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2021년도 학술발표회
• /
• pp.140-140
• /
• 2021

Urban flood management is a crucial and challenging task, particularly in developed cities. Therefore, accurate prediction of urban flooding under heavy precipitation is critically important to address such a challenge. In recent years, machine learning techniques have received considerable attention for their strong learning ability and suitability for modeling complex and nonlinear hydrological processes. Moreover, a survey of the published literature finds that hybrid computational intelligent methods using nature-inspired algorithms have been increasingly employed to predict or simulate the streamflow with high reliability. The present study is aimed to propose a novel approach, an ensemble tree, Bayesian Additive Regression Trees (BART) model incorporating a nature-inspired algorithm to predict hourly multi-step ahead streamflow. For this reason, a hybrid intelligent model was developed, namely GA-BART, containing BART model integrating with Genetic algorithm (GA). The Jungrang urban basin located in Seoul, South Korea, was selected as a case study for the purpose. A database was established based on 39 heavy rainfall events during 2003 and 2020 that collected from the rain gauges and monitoring stations system in the basin. For the goal of this study, the different step ahead models will be developed based in the methods, including 1-hour, 2-hour, 3-hour, 4-hour, 5-hour, and 6-hour step ahead streamflow predictions. In addition, the comparison of the hybrid BART model with a baseline model such as super vector regression models is examined in this study. It is expected that the hybrid BART model has a robust performance and can be an optional choice in streamflow forecasting for urban basins.

• 한국농공학회논문집
• /
• 제57권5호
• /
• pp.1-12
• /
• 2015

The study aimed to project inflows and demmands for the agricultural reservoir watersheds in South Korea considering a variety of regional characteristics and the uncertainty of future climate information. The study bias-corrected and spatially downscaled retrospective daily Global Climate Model (GCM) outputs under Representative Concentration Pathways (RCP) 4.5 and 8.5 emission scenarios using non-parametric quantile mapping method to force Soil and Water Assessment Tool (SWAT) model. Using the historical simulation, the skills of un-calibrated SWAT model (without calibration process) was evaluated for 5 reservoir watersheds (selected as well-monitored representatives). The study then, evaluated the performance of 9 GCMs in reproducing historical upstream inflow and irrigation demand at the five representative reservoirs. Finally future inflows and demands for 58 watersheds were projected using 9 GCMs projections under the two RCP scenarios. We demonstrated that (1) un-calibrated SWAT model is likely applicable to agricultural watershed, (2) the uncertainty of future climate information from different GCMs is significant, (3) multi-model ensemble (MME) shows comparatively resonable skills in reproducing water balances over the study area. The results of projection under the RCP 4.5 and RCP 8.5 scenario generally showed the increase of inflow by 9.4% and 10.8% and demand by 1.4% and 1.7%, respectively. More importantly, the results for different seasons and reservoirs varied considerably in the impacts of climate change.

• 대기
• /
• 제31권5호
• /
• pp.637-656
• /
• 2021

The United States has been known as the world's major producer of crops such as wheat, corn, and soybeans. Therefore, using meteorological long-term forecast data to project reliable crop yields in the United States is important for planning domestic food policies. The current study is part of an effort to improve the seasonal predictability of regional-scale precipitation across the United States for estimating crop production in the country. For the purpose, a dynamic downscaling method using Weather Research and Forecasting (WRF) model is utilized. The WRF simulation covers the crop-growing period (March to October) during 2000-2020. The initial and lateral boundary conditions of WRF are derived from the Pusan National University Coupled General Circulation Model (PNU CGCM), a participant model of Asia-Pacific Economic Cooperation Climate Center (APCC) Long-Term Multi-Model Ensemble Prediction System. For bias correction of downscaled daily precipitation, empirical quantile mapping (EQM) is applied. The downscaled data set without and with correction are called WRF_UC and WRF_C, respectively. In terms of mean precipitation, the EQM effectively reduces the wet biases over most of the United States and improves the spatial correlation coefficient with observation. The daily precipitation of WRF_C shows the better performance in terms of frequency and extreme precipitation intensity compared to WRF_UC. In addition, WRF_C shows a more reasonable performance in predicting drought frequency according to intensity than WRF_UC.

• 대기
• /
• 제25권2호
• /
• pp.295-308
• /
• 2015

Based on the CMIP5 historical simulation datasets, we assessed the performance of state-of-the-art climate models in respect to the relationship between interannual variabilities of the North Pacific synoptic eddy (NPSE) and East Asian winter monsoon (EAWM). Observation (ERA-Interim) shows a high negative correlation (-0.73) between the interannual variabilities of East Asian winter monsoon (EAWM) intensity and North Pacific synoptic eddy (NPSE) activity during the period of 1979~2005. Namely, a stronger (weaker) EAWM is related to a weaker (stronger) synoptic eddy activities over the North Pacific. This strong reverse relationship can be well explained by latitudinal distributions of the surface temperature anomalies over East Asian continent, which leads the variation of local baroclinicity and significantly weakens the baroclinic wave activities over the northern latitudes of $40^{\circ}N$. This feature is supported by the distribution of the meridional heat flux (${\overline{{\nu}^{\prime}{\theta}^{\prime}}}$) anomalies, which have negative (positive) values along the latitudes $40{\sim}50^{\circ}N$ for strong(weak) EAWM years. In this study, the historical simulations by 11 CMIP5 climate models (BCC-CSM1.1, CanESM2, GFDL-ESM2G, GFDL-ESM2M, HadGEM2-AO, HadGEM2-CC, IPSL-CM5A-LR, MPI-ESM-LR, MPI-ESM-MR, MRI-CGCM3, and NorESM1-M) are analyzed for DJF of 1979~2005. Correlation coefficient between the two phenomena is -0.59, which is comparable to that of observation. Model-to-model variation in this relationship is relatively large as the range of correlation coefficient is between -0.76 (HadGEM2-CC and HadGEM2-AO) and -0.33 (MRI-CGCM3). But, these reverse relationships are shown in all models without any exception. We found that the multi-model ensemble is qualitatively similar to the observation in reasoning (that is, latitudinal distribution of surface temperature anomalies, variation of local baroclinicity and meridional heat flux by synoptic eddies) of the reverse relationship. However, the uncertainty for weak EAWM is much larger than strong EAWM. In conclusion, we suggest that CMIP5 models as an ensemble have a good performance in the simulation of EAWM, NPSE, and their relationship.

• 한국지하수토양환경학회지:지하수토양환경
• /
• 제18권3호
• /
• pp.119-133
• /
• 2013

In the present study, several groundwater vulnerability assessment methods were applied to an agricultural area of Gumma in Korea. For the groundwater intrinsic vulnerability assessment, the performance of DRASTIC, SINTACS and GOD models was compared and an ensemble approach was suggested. M-DRASTIC and multi-linear regression (MLR) models were applied for the groundwater specific vulnerability assessment to nitrate of the study site. The correlation coefficient between the nitrate concentration and M-DRASTIC index was as low as 0.24. The result of the MLR model showed that the correlation coefficient is 0.62 and the areal extents of livestock farming and upland field are most influential factors for the nitrate contamination of groundwater in the study site.

• 한국수자원학회:학술대회논문집
• /
• 한국수자원학회 2015년도 학술발표회
• /
• pp.20-20
• /
• 2015

최근 기후변화의 영향으로 전 세계적으로 홍수와 가뭄의 발생빈도가 증가하고 있다. 특히, 가뭄은 우리나라에서 겨울과 봄철을 중심으로 매년 발생되고 있다. 가뭄의 정확한 발생을 판단하기는 어려우나, 가뭄이 발생되면 그 진행속도는 홍수보다 느리기 때문에 초기에 가뭄의 발생가능성을 예측한다면 가뭄에 대한 피해를 줄일 수 있다. 따라서 최근 가뭄 예측에 대한 다양한 연구가 이루어지고 있다. 본 연구에서는 가뭄발생의 불확실성을 내포하기 위하여 Bayesian Network (BN) 모형과 SPI의 자기상관성을 바탕으로 가까운 미래의 가뭄 발생확률을 예측하는 방법을 제안하였다. BN은 변수들 간의 인과관계를 확률적으로 나타낼 수 있는 네트워크 모형으로, 자연현상에 대한 위험도 분석 및 의학 분야에서 질병추정을 위한 모형으로 활용되고 있다. 본 연구에서는 가까운 미래의 가뭄 예측을 위하여 APEC 기후센터(APEC Climate Center, APCC)에서 제공하는 다중모형앙상블(Multi-model Ensemble, MME) 강우예측 결과로 도출한 미래 SPI 및 과거 강우량 자료로 구축한 SPI를 부모노드로, 예측 SPI를 자식노드로 BN을 구축하였다. BN의 각각의 노드를 Gaussian 확률분포모형으로 가정한 뒤, Likelihood weighting 방법으로 주변사후분포확률(Marginal posterior distribution)을 추정하여 미래의 SPI의 발생확률을 계산하였다. 2008년부터 2013년의 BN 가뭄 예측값과 MME 강우예측 결과로 도출한 SPI를 실제 관측 강우량으로 산정한 SPI와 비교하였으며, BN이 실제 관측결과에 가까운 결과가 도출되었다. 본 연구에서는 BN을 활용하여 가까운 미래의 가뭄 발생가능성을 확률적으로 나타낼 수 있는 방법을 제시하였으며, 그 결과 가뭄상태별 가뭄 발생확률이 산정되었다.

• 한국기후변화학회지
• /
• 제9권1호
• /
• pp.31-45
• /
• 2018

This study implemented the prediction of drought properties (number of drought events, intensity, duration) using the user-oriented systematical procedures of downscaling climate change scenarios based the multiple global climate models (GCMs), AIMS (APCC Integrated Modeling Solution) program. The drought properties were defined and estimated with Effective Drought Index (EDI). The optimal 10 models among 29 GCMs were selected, by the estimation of the spatial and temporal reproducibility about the five climate change indices related with precipitation. In addition, Simple Quantile Mapping (SQM) as the downscaling technique is much better in describing the observed precipitation events than Spatial Disaggregation Quantile Delta Mapping (SDQDM). Even though the procedure was systematically applied, there are still limitations in describing the observed spatial precipitation properties well due to the offset of spatial variability in multi-model ensemble (MME) analysis. As a result, the farther into the future, the duration and the number of drought generation will be decreased, while the intensity of drought will be increased. Regionally, the drought at the central regions of the Korean Peninsula is expected to be mitigated, while that at the southern regions are expected to be severe.

• 한국공간구조학회논문집
• /
• 제23권3호
• /
• pp.87-94
• /
• 2023

Chloride is one of the most common threats to reinforced concrete (RC) durability. Alkaline environment of concrete makes a passive layer on the surface of reinforcement bars that prevents the bar from corrosion. However, when the chloride concentration amount at the reinforcement bar reaches a certain level, deterioration of the passive protection layer occurs, causing corrosion and ultimately reducing the structure's safety and durability. Therefore, understanding the chloride diffusion and its prediction are important to evaluate the safety and durability of RC structure. In this study, the chloride diffusion coefficient is predicted by machine learning techniques. Various machine learning techniques such as multiple linear regression, decision tree, random forest, support vector machine, artificial neural networks, extreme gradient boosting annd k-nearest neighbor were used and accuracy of there models were compared. In order to evaluate the accuracy, root mean square error (RMSE), mean square error (MSE), mean absolute error (MAE) and coefficient of determination (R2) were used as prediction performance indices. The k-fold cross-validation procedure was used to estimate the performance of machine learning models when making predictions on data not used during training. Grid search was applied to hyperparameter optimization. It has been shown from numerical simulation that ensemble learning methods such as random forest and extreme gradient boosting successfully predicted the chloride diffusion coefficient and artificial neural networks also provided accurate result.

• 한국농림기상학회지
• /
• 제18권1호
• /
• pp.42-54
• /
• 2016

최근 국내에서 재배면적이 증가하고 있는 골드키위 해금의 개화시기를 예측할 수 있는 휴면시계모형의 모수를 추정하고 해금 주산지에서 미래 기후변화에 의한 개화시기의 변화와 불확실성을 전망하고자 본 연구를 수행하였다. 해금 개화시기 예측을 위한 휴면시계모형의 모수는 $6.3^{\circ}C$(base temperature, $T_b$), 102.5(chill requirement, $R_c$), 575(heat requirement, $R_h$)로 추정되었다. 2가지 방법으로 추정된 모수를 검증하였는데, 4개 표준기상관측소의 3년 동안(2013-2015)의 기상자료로부터 해금의 개화시기를 예측하고 25개 해금 노지 재배농가에서 수집된 2년 동안(2014-2015)의 관측 개화일과 비교한 결과 5.2일의 추정오차를 보였다. 또한 격자형 기후표면에 의해 계산된 격자형 개화시기 표면으로부터 25개 해금 노지 재배농가가 위치한 격자들의 예상 개화시기를 추출하여 비교한 결과, 3.4일의 추정오차를 보였다. 이 모수를 2021-2040년 동안의 6개 GCMs의 미래 기후변화 시나리오와 결합하여 해금의 미래 개화시기를 예측하였다. 전남 키위 주산지역에서 가장 빠른 개화시기는 4월 21일(111일), 가장 늦은 개화시기는 6월 2일(153일)로 나타났다. 6개 개별 GCM 중에서 RCP 4.5의 CanESM2과 GFDL-ESM2G, RCP 8.5의 HadGEM2-AO에서 20년 후 전남 키위 주산지역에서 해금의 개화시기는 현재보다 2-3일 단축될 뿐 현재와 큰 차이가 발생하지 않는 것으로 전망되었다. 그러나 RCP 4.5와 RCP 8.5의 6개 GCMs의 평균 미래 개화시기에서 현재보다 10일 이상 단축되고 현재와 같은 개화시기는 전북 및 충남 해안지역 등 북쪽으로 약 150km 이상까지 확대될 수 있는 것으로 전망되었다. 본 연구의 예비 결과는 국내 육종 과수의 생장발육 및 개화시기 예측 등을 위한 생물계절 연구와 기후변화에 대한 영향평가 개선에 기여할 수 있을 것으로 기대한다.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제12권11호
• /
• pp.471-480
• /
• 2023

국내 건설수주 규모는 2013년 91.3조원에서 2021년 총 212조원으로 특히 민간부문에서 크게 성장하였다. 국내외 시장 규모가 성장하면서, EPC(Engineering, Procurement, Construction) 프로젝트의 규모와 복잡성이 더욱 증가되고, 이에 프로젝트 관리 및 ITB(Invitation to Bid) 문서의 위험 관리가 중요한 이슈가 되고 있다. EPC 프로젝트 발주 이후 입찰 절차에서 실제 건설 회사에게 부여되는 대응 시간은 한정적일 뿐만 아니라, 인력 및 비용의 문제로 ITB 문서 계약 조항의 모든 리스크를 검토하는데 매우 어려움이 있다. 기존 연구에서는 이와 같은 문제를 해결하고자 EPC 계약 문서의 위험 조항을 범주화하고, 이를 AI 기반으로 탐지하려는 시도가 있었으나, 이는 레이블링 데이터 활용의 한계와 클래스 불균형과 같은 데이터 측면의 문제로 실무에서 활용할 수 있는 수준의 지원 시스템으로 활용하기 어려운 상황이다. 따라서 본 연구는 기존 연구와 같이 위험 조항 자체를 정의하고 분류하는 것이 아니라, FIDIC Yellow 2017(국제 컨설팅엔지니어링 연맹 표준 계약 조건) 기준 계약 조항을 세부적으로 분류할 수 있는 AI 모델을 개발하고자 한다. 프로젝트의 규모, 유형에 따라서 세부적으로 검토해야 하는 계약 조항이 다를 수 있기 때문에 이와 같은 다중 텍스트 분류 기능이 필요하다. 본 연구는 다중 텍스트 분류 모델의 성능 고도화를 위해서 최근 텍스트 데이터의 컨텍스트를 효율적으로 학습할 수 있는 ELECTRA PLM(Pre-trained Language Model)을 사전학습 단계부터 개발하고, 해당 모델의 성능을 검증하기 위해서 총 4단계 실험을 진행했다. 실험 결과, 자체 개발한 ITB-ELECTRA 모델 및 Legal-BERT의 앙상블 버전이 57개 계약 조항 분류에서 가중 평균 F1-Score 기준 76%로 가장 우수한 성능을 달성했다.