물리 기반 순간단위도 및 딥러닝 기반의 하이브리드 홍수 예측 기법

정민엽 서울시립대학교 일반대학원 2023 국내박사

Various rainfall-runoff models have been developed and used to predict floods in watersheds. Conceptual rainfall-runoff models have been widely used in practice due to their high reliability and applicability. However, these models simplify the rainfall-runoff process, making it difficult to incorporate the physical processes into runoff predictions. Additionally, the models have various parameter uncertainties, requiring calibration using a sufficient amount of observational data. On the other hand, physically-based rainfall-runoff models physically consider the flow within the watershed, providing relatively accurate results. However, their high computational cost and numerical instability limit their practical application. In this study, we developed a flood prediction model that integrates a physically-based dynamic wave model, an instantaneous unit hydrograph (IUH) model, and a deep learning model to achieve both accuracy and efficiency. We derived and interpolated the IUH based on the dynamic wave simulation and the power-law relationship between peak characteristics of IUH and rainfall excess intensity. In addition, we estimated rainfall losses using the Green-Ampt model and the NRCS-CN model. To estimate the parameters of each rainfall loss model, we incorporated real-time GLDAS-Noah land surface modeling data and rainfall time series data into the LSTM model, a deep learning model. Finally, we predicted the flood hydrograph of the watershed through convolutional integration. We applied our flood prediction model to both ideal and real watersheds. The LSTM model for rainfall loss estimation predicted the effective rainfall with high accuracy, achieving a Nash-Sutcliffe efficiency and correlation coefficient of over 0.9 and a normalized mean absolute error and normalized root mean square error below 0.15. In the flood prediction results, we obtained high accuracy with Nash-Sutcliffe efficiency between 0.55 and 0.9 and coefficients of determination between 0.67 and 0.95. Our physically-based IUH can be derived using only the topography and Manning's roughness coefficient of the watershed, without the need for parameter calibration using observed data. This makes it advantageous for application to ungauged watersheds or watersheds with limited observational data. Additionally, the derived IUHs account for nonlinear rainfall-runoff relationship, as it is used as a function of rainfall excess intensity for a single watershed. In addition, we efficiently interpolated the IUH for arbitrary rainfall excess intensity using a power-law-based technique. Furthermore, the derivation of the flood hydrograph through convolutional integration enables instantaneous and efficient flood prediction. By incorporating deep learning models and real-time watershed data for rainfall loss estimation, our flood prediction technique can reflect the moisture conditions of the watershed during flood events and consider various influencing factors related to rainfall loss. This approach reduces uncertainties in parameter estimation for rainfall losses and minimizes the need for subjective judgments. The flood prediction model in this study is expected to provide several advantages for flood damage prevention and the operation of water control structures. Except for the initial construction processes of optimizing LSTM and deriving IUHs, flood prediction can be completed rapidly within minutes, making it highly efficient. Moreover, by considering real-time moisture conditions in the watershed and physically reflecting the rainfall-runoff process, the model ensures the accuracy of flood prediction. 유역에서의 홍수를 예측하기 위한 다양한 강우-유출 모형들이 개발되어 사용되어왔다. 개념적 강우-유출 모형들은 신뢰성과 적용성이 높아 실무에서 널리 활용되어왔으나, 강우-유출 과정을 단순화하여 고려하므로 유출예측에 강수의 흐름을 물리적으로 고려하기 힘들다는 한계가 있다. 또한 모형의 매개변수에 여러 불확실성이 존재하므로 충분한 양의 관측 자료를 사용한 보정 작업이 필요하다. 물리적 강우-유출 모형들은 유역 내 흐름을 고려하므로 유출예측 결과가 비교적 물리적으로 정확하다는 장점이 있지만, 높은 계산 비용 및 수치적 불안정성으로 인하여 실무에의 적용이 힘들다는 한계가 있다. 본 연구에서는 물리기반의 동역학파 모형과 순간단위도 모형, 그리고 딥러닝 모형을 결합함으로써 정확도와 효율성을 모두 확보할 수 있는 홍수예측 기법을 개발하였다. 동역학파 시뮬레이션 및 순간단위도 첨두 특성과 유효강우강도 간의 멱함수관계를 이용하여 물리기반의 순간단위도를 유도 및 보간하였다. 또한, Green-Ampt 모형과 NRCS-CN 모형을 적용하여 강우 손실을 산정하였으며, 각 모형의 매개변수를 추정하기 위하여 실시간으로 제공되는 GLDAS-Noah 지표면 모델링 자료 및 강우 시계열 자료를 딥러닝 모형인 LSTM 모형을 통해 고려하였다. 최종적으로, 유효강우주상도와 순간단위도를 활용한 회선적분을 통해 순간적으로 유역의 홍수수문곡선을 예측하였다. 본 연구의 홍수예측기법을 가상유역 및 실제유역에 적용해보았다. 강우손실 산정을 위한 LSTM 모형은 NSE (Nash-Sutcliffe efficiency)와 CR (correlation coefficient) 0.9 이상, NMAE (Normalized Mean absolute error)와 NRMSE (Normalized root mean square error) 0.15 이하의 높은 정확도로 유효우량을 예측하였으며, 최종적인 홍수 예측 결과에서는 NSE 0.55이상 0.9이하, R^2 (Coefficient of determination) 0.67이상 0.95이하의 높은 정확도가 확인되었다. 본 연구의 물리기반 순간단위도는 관측수문자료를 이용한 매개변수 보정작업 없이도 유역의 지형과 조도계수 만을 사용하여 유도될 수 있으므로 미계측 유역이나 관측 자료가 부족한 유역에 대한 적용이 유리하다. 또한, 유도되는 순간단위도는 하나의 유역에서도 유효강우강도의 함수로 사용되므로 홍수 예측에 비선형적 강우-유출 관계를 고려할 수 있었으며, 멱함수 기반의 순간단위도 보간 방법을 통하여 임의의 유효 강우강도에 대한 순간단위도를 효율적으로 보간할 수 있었다. 그리고 최종적인 홍수수문곡선 유도는 순간단위도의 회선적분을 통하여 이루어지므로 순간적, 효율적인 홍수 예측이 가능하였다. 강우 손실의 산정에 딥러닝 모형 및 실시간 유역 데이터를 활용함으로써 홍수 사상 발생 시의 유역의 습윤 조건을 반영할 수 있었으며, 강우 손실과 관련된 여러 영향인자들을 고려할 수 있었다. 또한 이를 통하여 강우 손실 매개변수 산정의 불확실성 및 주관적 판단의 필요성을 줄일 수 있었다. 본 연구에서 개발된 홍수 예측 모형은 홍수 피해 방지 및 수공구조물의 운영 등에 여러 가지 이점을 제공할 것으로 기대된다. 모형의 초기 구축과정인 LSTM의 최적화 및 순간단위도 유도를 위한 동역학파 시뮬레이션을 제외한다면, 홍수 예측은 수 분 내에 빠르게 완료되므로 효율적이다. 또한, 유역의 실시간 습윤 조건을 반영하면서도 강우-유출 과정을 물리적으로 반영하므로 홍수 예측의 정확도를 확보할 수 있다.

분포형모델을 이용한 홍수유출예측 : 앙상블 칼만필터 적용을 중심으로

박효길 영남대학교 대학원 2009 국내석사

In this study, the distributed model is applied by ensemble kalman filter using the data of real time runoff and the parameter of model to meet accurately the flood flow forecasting. Also reducing uncertainty and updating measurements with errors are applied in nonlinear systems in wicheon basin, which is the typical IHP basin. The results of study are as follows. 1) The result of the percent error for peak flow in each flood event, the ensemble 100 is average mark of 1.21% and the ensemble 80 is average mark of -1.93%. The ensemble 100 and 80 are the lowest in the others. And the percent error for peak flow is more than 3% in the numble of ensemble decreased. Also the result of the error for peak time, four of flood events is accurate and three of flood events is less than 1hr. 2) The result of the coefficient of model efficiency in each flood event, the ensemble 100 is the highest average mark of 0.96 and the coefficient of model efficiency is low in the numble of ensemble decreased. It gives its confidence to the numble of more than ensemble 100(Evesen, 1994 / Wen&Chen, 2005). 3) Henceforth, the sensitivity analysis of the model and the decision making of the optimization will be studied for the analysis of uncertainty about the generation of initial ensemble, the variance of the measurements of the rainfall and the runoff and the variance of initial runoff to apply accurate real time forecasting. Consequently, the flood forecasting by using distributed models applied the ensemble kalman filter is comparatively accurate in the domestic flood forecasting. 본 연구에서는 홍수유출예측의 정확성을 높이기 위해 실시간으로 관측되는 유출량자료와 모델의 매개변수를 이용하여 분포형 매개변수 모형(Distributed-Parameter Model)인 분포형 모델을 대상으로 앙상블 칼만필터(Ensemble Kalman Filter, EnKF) 이론을 적용하였다. 또한 비선형시스템에서 오차를 포함한 반응을 실시간으로 처리하여 불확실성을 정량적으로 감소시켜 홍수유출을 예측하기 위하여 IHP대표유역인 위천유역을 대상으로 지금까지 연구한 결과를 요약하면 다음과 같다. 1) 각 호우사상별 첨두유량오차율을 분석한 결과 앙상블 개수가 100개, 80개일 때 각각 평균이 1.21%, -1.93%로 가장 낮게 나왔으며, 앙상블 개수가 적어질수록 약 3%가 넘는 것으로 나타났다. 또한 첨두시간오차는 8개 호우사상에서 4개 호우사상이 일치하였으며, 3개의 호우사상이 1hr이하로 예측을 실시하는 것으로 나타났다. 2) 각 호우사상에 대한 앙상블 개수별 모델 효율성 계수(Nash&Sutcliffe, 1970)를 비교한 결과, 앙상블 개수가 100개일 때 평균 0.96으로 가장 높았고 개수가 작아질수록 효율성 계수는 낮아지는 것으로 나타난 것으로 나타났으며, 앙상블 개수가 100개 이상일 때 신뢰할만한 결과를 얻을 수 있는 것으로 판단된다(Evesen, 1994 / Wen&Chen, 2005). 3) 실시간 홍수유출예측을 불확실성에 대한 고려없이는 앞으로 발생하게 되는 흐름상태에 대한 예측이 불가능할 뿐만 아니라 매 시간마다 예측을 통한 홍수모의결과와 실측자료간에 오차가 발생하게 되는데, 모형의 수행을 통한 오차의 제거가 없을 경우 오차의 누적으로 인한 모형의 정확성은 더욱 낮아지게 될 것이다. 따라서 실시간 홍수예측을 위해 사용되는 홍수유출모형의 정확도 향상을 위해서 앙상블 칼만필터를 이용하여 오차를 포함한 반응을 실시간으로 처리하여 불확실성을 정량적으로 감소시켜 더욱 정밀한 홍수유출예측이 가능할 것으로 판단된다. 4) 향후 모형에 대한 더욱 정교한 실시간 기법의 적용을 위해 초기앙상블의 형성, 관측된 강우자료와 유출량자료의 분산, 초기 유량의 분산에 관한 불확실도 해석을 통한 모형의 민감도 분석과 최적값 결정을 위한 구체적인 연구가 필요한 것으로 나타났다. 따라서 지금까지의 연구결과에서 앙상블 칼만필터를 이용한 홍수유출예측은 국내 홍수예경보에 이용할 경우 비교적 정밀한 홍수유출을 예측할 수 있을 것으로 판단된다.

정선유역의 확률론적 홍수예측을 위한 강우 불확실성 분석

이경태 서울대학교 대학원 2012 국내석사

기존의 홍수예측은 확정적인 단일 예측 값만 제공함으로써 신속하고 편리하였지만 이에 대한 불확실성이 클 경우 예상치 못한 인적 물적 피해를 가져올 수 있다. 이처럼 확률론적 홍수예측의 필요성이 대두되면서 미국이나 유럽 등의 선진국에서는 NWSRFS(National Weather Service River Forecast System)와 EFFS(European Flood Forecasting System)와 같이 확률론적 홍수예측에 대한 연구 및 기술개발이 이미 활발하게 진행되고 있으며, 실무에도 적용된 사례가 많다. 홍수예측의 확률론적 접근에 있어서는 많은 불확실성들이 내포되어 있으므로 예측시스템에서 생성된 앙상블 유량예측의 결과의 정확성 분석과 올바른 불확실성 정보의 제공이 필요하다. 국내에서는 중장기의 확률론적 유량예측의 연구는 활발하게 이루어지지만 단기 예측에 관한 연구는 미흡하다. 또한 국외에서도 확률홍수예측에 대한 불확실성 분석에 관한 연구가 전무한 현실이다. 본 연구는 홍수예측에서 수문 모형의 적용에서 오는 불확실성은 없다고 가정한 후, 확률론적 홍수예측 방법을 국내에 적용시켜서 단기예측을 중심으로 기상청의 기상예보 시스템인 MAPLE(McGill Algorithm for Predictation nowcasting by Lagrangian Extrapolation), KLAPS(Korea Local Analysis and Prediction System) 그리고 UM(Unified Model)의 예측장 결과 중 강우데이터를 사용하였다. 기상예측 시스템들은 서로 다른 예측값 제공 시간간격과 공간 스케일을 갖고 있다. 따라서 확정적으로 예측되는 강우자료를 이용하여 확률적 예측을 위한 앙상블 생성기법인 시간지연 앙상블(Time lagged ensemble) 방법과 다중모형 앙상블(Multi model ensemble) 방법을 적용시켜서 생성된 기상 앙상블을 현재 국토해양부 홍수통제소에서 사용하고 있는 강우 유출모형인 저류함수 모형의 입력 자료로 사용하여 남한강 상류에 위치한 정선유역에 적용시켰다. 우선 제공되는 확정적 기상예측으로부터 생성된 기상 앙상블들의 불확실성을 분석하였다. 일반적 기대와 같이 선행시간이 길어질수록 정확성이 점차 낮아지는 것을 확인하였고, 평균적으로 이수기의 불확실성 범위가 홍수기 보다 약 55 % 증가하는 것으로 나타났다. 또한 기상예측에 있어서 계통적 오차 보다는 무작위 오차에 크게 영향을 받는 것으로 확인되었다. 다음은 생성된 유량앙상블들로부터 선행시간별로 총 유량과 첨두 유량 예측에 있어서의 정확성과 불확실성을 분석하였다. Reliability와 resolution을 정량적으로 계산하였으며 불확실성의 범위를 나타내는 resolution의 영향에 따라 예보의 정확성을 판단하는 BS(Brier Score)와 RPS(Ranked Probability Score)의 값에 크게 영향을 미치는 것으로 나타났다. 선행시간이 증가함에 따라 MAPLE과 KLAPS에서는 평균적으로 매 1 시간당 51 %, UM에서는 매 3시간당 55 %씩 불확실성의 범위가 증가한다. 전반적으로 UM의 예측보다 MAPLE과 KLAPS의 예측의 정확성이 더 높은 결과를 얻었다. 서로 다른 예측 공간규모에 따른 불확실성 분석을 통해 UM의 첨두유량 예측의 불확실성 범위가 MAPLE과 KLAPS의 값보다 더 크게 나타났다. 이는 정선유역에 포함되는 UM의 격자수가 다른 두 시스템들보다 작아서 발생하는 공간적 불확실성이라고 할 수 있다. 반면에 예측된 첨두유량 값들을 비교하였을 때에는 UM이 MAPLE과 KLAPS에 비해 약 7.2 % 정도 과소추정 되었다. 마지막으로 지금까지 단기예측에 중점을 둔 시간(hourly) 단위의 불확실성 특성들과 기존의 중장기예측인 월 단위(monthly) 확률적 예측의 결과들과 비교하였을 때, 예측의 범위가 시간단위로 상당히 짧은 단기 확률적 예측에 있어서도 무시할 수 없는 불확실성이 존재함을 보여주고 있다. 따라서 기상예측에서 오는 불확실성 뿐 만 아니라 그 밖에 초기조건, 수문모형, 매개변수에서 오는 기타 불확실성의 추가적인 연구가 필요함을 보여준다. The existing flood forecasting system only provides a single deterministic value of prediction which is fast and convenient. However, there would be huge unexpected property damages and loss of life when a large uncertainty does occur. Due to the increased global interest in probabilistic flood forecasting, U.S. and European countries have already begun actively studying and developing probabilistic flood forecasting systems such as NWSRFS (National Weather Service River Forecast System) and EFFS (European Flood Forecasting System). There are many cases in which these systems have already been applied in practice. Since there are many uncertainties in probabilistic forecasting, the accuracy and uncertainty analyses should be performed and examined closely. There are many previous studies and reports regarding medium- and long-range probabilistic forecasting but there are insufficient research data regarding short-range forecasting in Korea. I assumed that there are no uncertainties from hydrologic model simulation. Uncertainty only came from the rainfall generating ensembles with the different deterministic weather forecasting systems such as MAPLE (McGill Algorithm for Prediction nowcasting Lagrangian Extrapolation), KLAPS (Korea Local Analysis and Prediction System), UM (Unified Model), all of which have different prediction periods and spatial grids using time lagged and multi model ensemble generation techniques. I placed them into Storage function method as meteorological inputs to generate streamflow ensembles. The results from the analysis of uncertainty in the weather predictions show that there are inverse relations between increasing lead time and accuracy. Also, more random errors exist than systematic errors in weather prediction. UM predictions are less accurate than MAPLE and KLAPS predictions for both total discharge and peak discharge by quantitatively measuring Brier Score and Ranked Probability Score. The study results show that not only monthly ensemble prediction of streamflow but also hourly ensemble prediction of streamflow need to be considered since there exist remarkable uncertainties in both short- and long-range probabilistic predictions.

중소하천에서의 실시간 홍수예측 및 범람해석

최승용 경북대학교 대학원 2011 국내박사

Since the frequency and scale of damage from the torrential rain increases recently due to climate change and global warming, the significance of flood forecasting and warning becomes important in medium and small streams as well as large rivers. Real time flood forecasting is important in a flood-prone region within medium and small sized watershed, for the issue of flood warnings in order to allow ample leadtime for the evacuation of populations and protection of primary facilities endangered by imminent rising water levels. In the modeling the relations with a rainfall and runoff by using the existing hydrological model, it implies lots of non-linear and uncertain elements. Through watershed flood routing, reservoir routing and channel routing for the estimation of runoff, many diverse errors can be occurred and accumulated, so that the results of rainfall-runoff analysis may contain the uncertainties in the existing flood forecasting and warning system. The purpose of this study is develop the data driven model for the real-time flood forecasting and inundation analysis that are improved the problems of the existing hydrological model. Neuro-Fuzzy model which use the fuzzy inference theory with neural network, that can forecast flood only by using the rainfall and flood level and discharge data without using lots of physical data that are necessary in existing hydrological rainfall-runoff model is used. First of all, the various combinations of input data were constructed by using rainfall and water level to select optimal combinations of input data for applying Neuro-Fuzzy model. The optimal combination of input data that composed of rainfall data at t time, rainfall data at t+1 time, water level data at t time, water level data at t+1 time, water level data at t+2 time was selected when current time is t time. Based on optimal combination of input data, training and checking events are selected by using various flood events for applying Neuro-Fuzzy model in Jungrangcheon. Neuro-Fuzzy system was constructed by estimating the number and parameters of membership function for each leadtime of selected training and checking events. In addition, DEM based two-dimensional inundation model that can perform effectively real-time inundation analysis in protected lowland has been developed. The suggested Neuro-Fuzzy method was applied to Jungrangcheon. For the verification of the model, the comparisons between forecasting floods and observation data are presented. The comparison results of Neuro-Fuzzy model in Jungrangcheon with the observed data show that the real-time flood forecasting can be simulated successfully without large errors. In case of the 180 minutes leadtime, a root mean square error turned out to be 0.174m, the correlation coefficient to be 0.887 and in other statistical indicators such as mean absolute deviation and Nash-Sutcliffe efficient coefficient, forecasting results are much improved in terms of flood level and discharge. Accordingly, three hours leadtime with high accuracy can be obtained for real-time flood forecasting. The forecasted floods by applying Neuro-Fuzzy model are used as boundary conditions to simulate the real-time one-dimensional hydraulic analysis for jungrangcheon. As a result of verification, the simulated results by using forecasted floods are in good agreements with the simulated results which use actual observation floods as boundary conditions. In order to estimate for calibration and verification of real-time two-dimensional inundation forecasting model, DEM based inundation model and level-pool model developed in this study are applied to a flood event in 1998 year in Jungrangcheon. The results of two models were compared with the actually investigated inundation area. From the comparisons , it revealed that results of DEM-based model are better accurate and improved than those of level-pool model. The computational time for each used grid size is reviewed in terms of real-time inundation forecasting. As a result, the computational time of DEM-based inundation model used 20m×20m DEM data was only 125 seconds. Therefore DEM-based inundation model developed in this study will be able to obtain sufficient leadtime with high accuracy for real-time inundation forecasting within 3 minutes. The methodologies presented in this study provide an advanced integrated system of real-time flood forecasting and inundation analysis which can be applicable for flood prevention and mitigation in medium and small streams. This study can greatly contribute to the construction of a high accuracy flood information system that secure leadtime in medium and small streams.

水門資料 品質管理에 따른 洪水豫測正確度의 影響 評價와 改善

김휘린 高麗大學校 大學院 2014 국내박사

This study examined the effect of the flood forecasting accuracy based on hydrological data quality control. Initially, several improvement methods suggested for the hydrological data quality based on the investigation of the status and problems of hydrological survey facilitation, data transmission and process. All of the rainfall and water level stations of the Han River were chosen and a 10-year data was collected to analyze the effect of the improvement methods. Each station data were analyzed, during the flood season, monthly and yearly. Consequently, hydrological data quality improved significantly after applying the suggested methods. Moreover, there is a direct correlation between estimation results of the basin or precipitation and data quality control of each rainfall station by analysis of quantitative traits. The Storage Function and FLDWAV model are applied to simulate the flood events which occurred in the Han River basin. The parameter calibration are minimized to evaluate the effect of hydrological data quality control. The case studies are divided before and after the establishment of hydrological data quality control improvement. This study carefully analyzes the forecasting results of rainfall data quality control between raw data and calibration data respectively. Furthermore, the accuracy of flood forecasting based on water level data quality control compares with upstream and downstream water level station. This research provides the ways of improving hydrological data quality control such as the installation of dual measurement and upgrading the transmission system. In addition, those contribute not only to the good quality of data production but also accuracy of model performance. Therefore, it is necessary to obtain a high quality of hydrological data for accurate flood forecasting result instead of concentrating on forecast model’s development and parameter calibration.

낙동강 유역의 실시간 홍수예측 시스템 구축

노홍식 경북대학교 대학원 2013 국내석사

Recently, the damage on people and properties are rapidly increasing by flood disaster such as inundation due to typhoons and localized heavy rain affected by abnormal climate. In domestic cases, large-scaled flood damage habitually occurred by seasonal concentrated phenomena of precipitation due to typhoon lusa in 2002 and typhoon maemi in 2003. To reduce the realistic flood damage, it is urgently necessary to establish the non-structural measures that develops flood forecasting system that can forecast the possible flood in the future. Currently, a physical-based precipitation-discharge model that is used on working-level has large or small errors in the process of calculating the discharge amount from hydrological data. The purpose of this study is to develop an efficient flood forecasting system by real-time forecasting using Neuro-Fuzzy model and regression model which is a data-based model to complement problems that the physical-based precipitation-discharge model. Through this, by forecasting real-time flood risk that occurs by localized heavy rain, a system can be developed to make a safe society from flood disaster by protecting lives and properties of residents within the region. The Neuro-Fuzzy model started to be applied as a trial to make the best use of the advantages of a neuro network and fuzzy theory, and to complement the disadvantages that each technique has. Both the neuro network theory and fuzzy theory are interested in establishing a system that can work as a human for specific fields, but the fields in which the two theories has strong points are slightly different. The fuzzy theory is proper for processing and inferring ambiguous data under the logical base and is useful for highly difficult processing through natural linguistic expression. The neuro network theory has learning ability so there is a large flexibility in composition of a system, and it is excellent in data-based processing. Various combinations of input data were constructed by using rainfall and water level and discharge to select optimal combinations of input data for applying Neuro-Fuzzy model. The optimal combination of input data that composed of rainfall data at t time, rainfall data at t-1 time, water level data at t time, water level data at t-1 time, water level data at t-2 time was selected when current time is t time. Based on optimal combination of input data, training and checking events are selected by using various flood events for applying Neuro-Fuzzy model in Nakdong river basin. Neuro-Fuzzy system was constructed by estimating the number and parameters of membership function for each leadtime of selected training and checking events. This study composed and applied a combination of various conditions of input data on precipitation, water level and discharge to Nam River basin for selecting the optimum one for Neuro-fuzzy system and for real-time flood forecasting. Time scale for input data in a multiple linear regression model was determined in consideration of the input data selected this way. Besides, excluding the data of extensive amounts demanded by the existing rainfall-runoff model, on the subject of 8 major tributaries of Nakdong River basin, Neuro-fuzzy model and regression analysis model, which can predict the flood level and flood flow on major points of the basin with only data of precipitation, water level and discharge, were constructed and applied to the subject basin to forecast the flood level and flow by station. Applying the boundary condition of forecasted floods from tributary, river flood level in the mainstream was predicted by one-dimensional model, which was reviewed comparatively with the simulation result using the actual flood. The result of simulation results showed that forecasting of flow rate at the points after flowing the mainstream by tributaries well agreed with the survey values and the prediction of the highest water level for the whole mainstream of the Nakdong River was in accord with the result of simulation with a boundary condition of actual flood of observation. Through this, real-time river interpretation using Neuro-fuzzy flood prediction is considered to be used in the prediction of floods for the section of the Nakdong River basin.

하천 홍수터에서의 홍수예측을 위한 2차원 수리해석

최대진 경북대학교 대학원 2016 국내석사

GIS 基盤 洪水豫測地圖의 開發

김상호 韓京大學校 産業大學院 2006 국내석사

자연재해 중 홍수에 의한 재해는 우리나라의 경우 여름철 저기압, 장마전선, 태풍 등의 위험에 노출된 지역이 많이 있어 각종 산업시설을 보호하고 있는 하천제방 등의 홍수방지 시설물들은 그 설계와 관리에 만전을 기하고 있음에도 불구하고 홍수에 의한 범람 피해가 속출되고 있다. 또한 이러한 홍수범람으로 인하여 하천의 주변 지역에서는 많은 인명피해 및 재산피해 등 사회적 문제를 초래하고 있어 근본적인 대책이 필요한 실정이다. 한편, 최근 GIS의 발달과 수치지도 제작이 활발히 이루어지면서 GIS를 기반으로 하는 수리, 수문모형에 관한 연구가 활발히 이루어지고 있는데, 수치표고모형(Digital Elevation Model, DEM)에 기반을 둔 공간자료 구조를 가진 수문모형이 개발되었으며, 유역의 수문학적 특성을 자동으로 계산하고 분석하는 GIS도구와 수치지도모형(Digital Terrain Model, DTM)으로부터 하천단면을 추출하고 수리해석을 실시하며, 그 결과를 그래픽, 그래프, 또는 텍스트 등으로 나타낼 수 있는 GIS도구가 널리 보급되어 있다. 이러한 GIS와의 연계과정을 통한 홍수지도의 제작은 홍수정보를 공간적으로 표현하여 재난이 예상되는 지역에 대해 홍수로 인한 범람정보를 제공하고, 취약지점에 대한 집중적인 투자를 가능하게 하며, 실제 재난이 발생하였을 경우 대피운영, 복구 등에 있어 요구되는 각종 정보를 제공할 수 있다. 더 나아가 그러한 정보들을 효과적으로 처리, 표현함으로서 홍수지도를 누구나 쉽게 사용하도록 하는 것은 홍수로 인한 피해 경감 대책에 큰 도움이 될 수 있다. The objective of the study is to develop a GIS-based flood map. Hydraulic model (HEC-RAS) is linked with hydrologic model (HEC-HMS) for flood map. Geospatial data processors, HEC-Geo HMS and HEC-GeoRAS, are used for operating HEC-HMS and HEC-RAS. HEC-HMS was calibrated and validated at the Hwa- Ong watershed. HEC-HMS was used for calculating runoff from the Hwa-Ong watershed which consisted of Nam-Yang, Ja-An, U-Eun river sub-watersheds, and HEC-RAS was applied and validated for river flow routing at the Hwa-Ong watershed. The simulated results from HEC-HMS and HEC-RAS were reason ably good as compared with the observed data. HEC-RAS and HEC-HMS were applied to simulate flooding from probability rain fall at the Hwa-Ong watershed, and the simulated result was used to develop a flood map. Flood map developed in this research will be used for mitigating and predicting the flood damages.

하천 위험수위 예측을 위한 강우강도 산정에 있어서 GIS 적용

장수덕 경남대학교 산업대학원 2007 국내석사

The flood is the most general and frequently occurs among natural disasters. Generally flood by the rainfall which extends superexcellently for the occurrence but flash floods happen in a frequent Time, caused by local heavy rain, and the demage increases. So far, it's the mainstream to establish the system preventing disasters, focusing on big-sized rivers. Recent unusual changes of weather and geographical features make the flash fiood damage. Therefore, concerns about the countermeasures are addressed. Researches on predicting flash floods by small-sized rivers are not more than those on big-xized rivers yet. Even though the damage of big-sized river flood is complex and large in the economic respect, people raise the necessity of researches, flash flood model by mountainous rivers, connection between the models of big-sized rivers and mountainous rivers, and estimation of flooded amount. This paper aims to 1 hour real-time flash flood and predict possibibity at the area where is the possible flood will occur from the rainfall hour mountain after acquiring data in GIS(Geographic Information System) base by GcIUH(Geomorphoclmatic Instantaneous Unit Hydrograph). The flash flood occurrence is set up at 0.5m, 0.7m and 1.0m in standard depth. And this study suggests standard flood alarm which designed by probable flood according to duration time. The reseach result shows real-time flash flood evaluation has the suitable standard in the basin when comparing with the existing official warning announcement system considering topographical information. 본 연구는 각종 지형정보 자료를 획득하여 GIS기법으로 하천 위험수위 예측을 위한 강우강도 산정에 관한 연구로써 하천소유역에 대한 지형분석을 실시한 후 가능홍수량이 발생 할 수 있는 유출고를 강우-유출관계를 지형기후학적 순간단위유량도를 이용하여 산정하였으며 결과는 다음과 같다. 첫째, 홍수 산정기준을 인명피해가 발생할 수 있는 0.5m, 0.7m, 1.0m 이상의 수심을 기준으로 위험수위로 설정하였으며, 지속시간별-강우량별 홍수량을 산정하였다. 둘째, 도심지와 근접한 지역이라도 산악유역의 경우에는 측량성과를 가질 수없는 지역이 대부분이기 때문에 이에 대한 대안으로 대표단면을 이용하여 수심이 0.5m가 되는 경우의 홍수량을 Manning의 공식을 이용하여 한계유량을 산정하였다. 셋째, 홍수 지속시간에 따라 한계유량을 발생시키는 지속시간별 강우량을 지형기후학적 순간단위유량도와 GIS기법으로 산정하여 홍수예측 기준을 마련하였다. 본 연구에서 수행된 돌발홍수 예측 기준은 제방의 안정성과 범람에 의한 피해를 최소화시키는 것이 목적이 아니라, 계곡에서 발생 가능한 인명피해를 최소화하기 위한 것이다. 따라서 우량발생 시 첨두홍수량 발생시간을 지형기후학적 순간단위유량도를 적용한다면 최소한 20분 이상의 대피시간을 확보할 수 있는 것으로 나타났다. 또한 본 연구는 인명을 보호하는 목적을 최우선으로 두었기 때문에 하천의 수위뿐만 아니라 발생시간에 더 큰 비중을 두었으므로 그 의의가 크다고 할 수 있다.

실시간 수재해 예측과 대응을 위한 수치모형과 기계학습의 연계

김현일 경북대학교 대학원 2020 국내박사

Sudden flooding has been on the rise in the watersheds that have undergone significant urbanization, and this is attributable to climate change. This trend presents the potential of tremendous property damage and losses of lives, and diverse methods have been applied to address the issue with greater efficiency. Unlike the traditional numerical analysis models and statistical analysis methods, flood analysis using machine learning is a data-based analysis that can present results in a short period of time through prior learning. Flood analysis based on machine learning involves the use of a wide range of simulation-based and observational data, and thus may be viewed as a type of data-based analysis. Accordingly, it is important to apply rational machine learning techniques along with data that are appropriate in terms of quality and quantity, and this technique may be actively applied to analyze and respond to unforeseen hydrological events and disasters. In this study, a machine learning technique with the use of one- and two-dimensional numerical analysis models, diverse hydrological observation data, and deep learning was applied. As for the water disaster analysis and prediction, an analysis of the upper and lower streams of the Paldang dam and a flood analysis of a single urban drainage basin were carried out. With respect to an inflow of dams with a return period of 200 years or more, a flood map prediction for rising water levels of the Han-river was made with the application of the random-forest regression technique, and this is considered to be an efficient method to determine in real time the flooding pattern for the excessive dam inflow that can be expected. Using the range of flooding according to the amount of dam inflow under the probable maximum flood (PMF) condition, the utility of the flooding range prediction technique was proposed. The order of priority of the flood response to be made by each district of Seoul to counter the economic damage resulting from flooded buildings and casualties, and the level of importance of each factor affecting flood damage was estimated using the random forest technique. The proposed method can be used to select the districts in which flood response and restoration must be performed first in the event that a large-scale flood with a recurrence interval of 200 years or more occurs in the watershed of a metropolitan city. The flood analysis of urban drainage basin units was conducted on Samseong drainage basin, and a LSTM (Long Short Term Memory) neural network model capable of predicting the total amount of overflow in real time according to the occurrence of rainfall events was trained. A comparison with the verified one-dimensional urban runoff analysis model showed that the LSTM neural network model could even accurately model the peak overflow. Using the predicted total overflow, the flooding range and flood risk map were predicted. When predicting the flooding range, the range was predicted based on the flooding depth using the logistic regression technique, and while it failed to produce a high area fitness for a relatively high flooding depth, it was deemed to be a highly appropriate method for predicting the overall flooding range according to the rainfall intensity. For the analysis of the flood risk map, the formula for calculating the flood risk level presented in the Department for Environment, Food & Rural Affairs, Environment Agency (DEFRA)’s Flood Risks to People report was used, and the flood risk level was determined in grid units using the flooding depth and flow rate information obtained from the two-dimensional flood analysis. A flood risk map was generated by taking into consideration both scenarios in which the debris factor associated with water disasters was and was not considered, and predictions were made through random-forest classifications. The results of the prediction and analysis showed that a relatively high flood risk level was found in residential areas with many alleys when the debris factor was taken into account. The prediction results using the random forest were compared with the results of using the flood risk calculation formula, and a high degree of agreement was found. Thus, it was judged that suggested methodology could be used as providing the useful data for performing flood control of each urban drainage basin at a later time. Through flood prediction and analysis using diverse machine learning, it is possible to determine the danger zones in real-time in the event of a water disaster. The proposed methodology can be said to have high reliability and accuracy, as it uses a database built based on observational data and the results of applying verified numerical analysis models. Its practicality was also demonstrated by presenting the method of utilization with respect to the prediction and analysis results, and it is expected that the prediction model will help speed up the processes of the urban watershed flood control system if it is improved with more basic data. The aim of this study was to demonstrate the diverse uses of various observational data and numerical analysis models in connection with machine learning techniques and that they are accurate enough to be actively applied in practice to defend against flooding.