• Journal of The Korean Society of Agricultural Engineers
• /
• v.63 no.1
• /
• pp.11-25
• /
• 2021

The objective of this study was to analyze the impact of activation functions on flood forecasting model based on Artificial neural networks (ANNs). The traditional activation functions, the sigmoid and tanh functions, were compared with the functions which have been recently recommended for deep neural networks; the ReLU, leaky ReLU, and ELU functions. The flood forecasting model based on ANNs was designed to predict real-time runoff for 1 to 6-h lead time using the rainfall and runoff data of the past nine hours. The statistical measures such as R2, Nash-Sutcliffe Efficiency (NSE), Root Mean Squared Error (RMSE), the error of peak time (ETp), and the error of peak discharge (EQp) were used to evaluate the model accuracy. The tanh and ELU functions were most accurate with R2=0.97 and RMSE=30.1 (㎥/s) for 1-h lead time and R2=0.56 and RMSE=124.6~124.8 (㎥/s) for 6-h lead time. We also evaluated the learning speed by using the number of epochs that minimizes errors. The sigmoid function had the slowest learning speed due to the 'vanishing gradient problem' and the limited direction of weight update. The learning speed of the ELU function was 1.2 times faster than the tanh function. As a result, the ELU function most effectively improved the accuracy and speed of the ANNs model, so it was determined to be the best activation function for ANNs-based flood forecasting.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.200-200
• /
• 2020

본 연구에서는 SWAT 모형과 APEX-Paddy 모형의 연계 모델링을 통한 대표 BMP(Best management practice) 적용, 정식시기 및 벼 생육기간을 고려한 시나리오 적용을 통해 농업용수의 관리 및 수질환경 개선 등에 활용할 수 있는 저영향 영농활동을 분석하고자 하였다. 만경강 유역을 대상으로 SWAT 모형을 구축하고 유역 내에 위치한 논 시험포장을 대상으로 강우-유출 및 비점오염원 모니터링 자료를 활용하여 APEX-Paddy 모형을 구축하였다. SWAT 모형과 APEX 모형을 연계하여 유역의 수문, 수질에 대한 정밀한 모델링을 수행하였으며, 이는 저영향 영농활동을 분석하기 위한 필드단위의 정확한 결과를 유역차원에 반영하기 위함이다. 특히, 본 연구에 사용된 APEX-Paddy 모형은 농촌진흥청과 Texas A&M의 공동연구를 통해 개발된 새로운 모형으로서 한국의 논 영농활동 및 담수환경을 반영하여 논에서의 유출 및 비점오염원을 모의할 수 있다. 연계 모형의 적합성 평가를 위해 R² (Determine of Coefficient), RMSE (Root mean square error), NSE (Nash-sutcliffe efficiency)를 사용하였다. 적합성 평가 지표를 분석한 결과, 유출량은 R² 평균 0.91, RMSE 평균 2.87 mm/day, NSE 평균 0.78로 나타났다. T-N 부하량은 R² 평균 0.74, RMSE 평균 59.3 kg/ha/day, NSE 평균 0.50으로 나타났다. 저영향 영농활동 관리방안을 위한 시나리오로 1) 논의 물꼬높이(BMP) 관리 적용, 2) 벼 생육기간 조절을 고려하여 기온변화에 따른 정식시기, 벼 생육기간 등을 조정하여 적용하였다. 기후변화 시나리오는 10개 GCM 모델의 RCP 8.5 시나리오를 통해 분석하였으며, 유역차원의 미래 영향을 분석한 결과, 물꼬관리 BMP에 따라 담수심이 증가되며, 관개량이 감소하고 유출량 10.7%, T-N 11.2% 저감되는 것을 나타냈으며, 벼 생육기간 조절은 BMP보다 상대적으로 효과가 높진 않았지만, 유출량 1.4%, T-N 3.1%의 저감효과를 나타냈다. 따라서 두 가지의 저영향 영농활동 관리방안은 미래기간의 기후변화에 대응하여 농업용수 및 물관리에 도움이 될 것으로 사료된다. 하지만 본 연구결과는 모델링 결과에 의존한 것이며, 추후 지속적인 연구와 보완이 필요하다.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2020.06a
• /
• pp.262-262
• /
• 2020

새만금 유역 내에는 다수의 보 및 제수문이 위치하고 있으며, 관개, 배수, 오염원 등이 영향을 받고 있다. 선행연구 중에는 보 및 제수문을 고려하기 위해 모형의 소스코드를 일부 수정하여 연구되고 있으나 유역모형으로 구현하기에는 한계가 있으며, 이에 대한 연구는 미흡한 실정이다. 본 연구에서는 만경강 유역을 대상으로 유역 모형과 전산유체역학 모형을 이용하여 하류 제수문에 대한 유입, 유출 그리고 오염원 등의 영향을 분석하고자 한다. SWAT (Soil and water assessment tool)은 유역 모형으로 미국 농무부에서 농업유역의 수문순환 및 비점오염원을 모의하기 위해 개발한 모형이다. CFD (Computational fluid dynamics)는 전산유체역학 모형으로 구조물을 설계하고 유체, 기체 등을 모의할 수 있다. SWAT 모형을 이용하여 농업유역 하류 제수문 위치를 출구로 지정하여 수문을 모의하고 그 결과자료는 CFD에 입력할 수 있다. CFD는 하류 제수문 구조물을 설계하고 SWAT 모형의 수문자료를 입력하여 제수문의 영향을 평가할 수 있다. 우선, 만경강 유역을 대상유역으로 선정하고 부용, 황산, 상리, 고은교 등 제수문의 위치를 파악하였다. SWAT 모형 구축을 위해 2015-2018년까지 기상, 수위, 유량 관측자료를 수집하였으며, 보정기간과 검증기간은 각 2년이며, 모형 성능 검증에 사용한 적합성 평가 지수는 R² (Determine coefficient), RMSE (Root mean square error), 그리고 NSE (Nash-sutcliffe efficiency coefficient)를 사용하였다. 모형의 보정은 SWAT-CUP 자동보정프로그램을 사용하였으며, 모형의 보정지수는 NSE를 사용하였고, 1,000회 반복 수행을 통해 매개변수를 최적화하였다. 보정기간의 유출량 적합성 평가 지수는 R², RMSE 그리고 NSE가 각각 0.84, 2.96 mm/day, 0.70을 나타냈다. 검증기간의 유출량 적합성 평가 지수는 R², RMSE 그리고 NSE가 각각 0.72, 2.94 mm/day, 0.46을 나타냈다. 본 연구는 유역 차원과 구조물 차원의 모델링을 연계하는 것으로 향후 제수문 모니터링 자료를 활용하여 CFD 모형을 구축하고 유입량에 따른 제수문의 검보정 및 영향을 평가하고자 한다. 이러한 결과는 최근 기후변화에 따라 급격히 변화하는 유역환경에 대처할 수 있는 방안이 될 수 있을 것이며, 제수문 시설을 관리하는 기관에서도 합리적인 운영방안에 대한 기초자료로 기여할 수 있을 것으로 사료된다.

• Structural Engineering and Mechanics
• /
• v.84 no.2
• /
• pp.143-154
• /
• 2022

Corrosion of steel reinforcement is considered as the main cause of concrete structures deterioration, especially those under humid environmental conditions. Hence, fiber reinforced polymer (FRP) bars are being increasingly used as a replacement for conventional steel owing to their non-corrodible characteristics. However, predicting the shear strength of beams reinforced with FRP bars still challenging due to the lack of robust shear theory. Thus, this paper aims to develop an explicit data driven based model to predict the shear strength of FRP reinforced beams using multi-objective evolutionary polynomial regression analysis (MOGA-EPR) as data driven models learn the behavior from the input data without the need to employee a theory that aid the derivation, and thus they have an enhanced accuracy. This study also evaluates the accuracy of predictive models of shear strength of FRP reinforced concrete beams employed by different design codes by calculating and comparing the values of the mean absolute error (MAE), root mean square error (RMSE), mean (𝜇), standard deviation of the mean (𝜎), coefficient of determination (R²), and percentage of prediction within error range of ±20% (a20-index). Experimental database has been developed and employed in the model learning, validation, and accuracy examination. The statistical analysis illustrated the robustness of the developed model with MAE, RMSE, 𝜇, 𝜎, R², and a20-index of 14.6, 20.8, 1.05, 0.27, 0.85, and 0.61, respectively for training data and 10.4, 14.1, 0.98, 0.25, 0.94, and 0.60, respectively for validation data. Furthermore, the developed model achieved much better predictions than the standard predictive models as it scored lower MAE, RMSE, and 𝜎, and higher R² and a20-index. The new model can be used in future with confidence in optimized designs as its accuracy is higher than standard predictive models.

• Korean Journal of Remote Sensing
• /
• v.38 no.3
• /
• pp.265-282
• /
• 2022

Sea surface temperature (SST) is a factor that greatly influences ocean circulation and ecosystems in the Earth system. As global warming causes changes in the SST near the Korean Peninsula, abnormal water temperature phenomena (high water temperature, low water temperature) occurs, causing continuous damage to the marine ecosystem and the fishery industry. Therefore, this study proposes a methodology to predict the SST near the Korean Peninsula and prevent damage by predicting abnormal water temperature phenomena. The study area was set near the Korean Peninsula, and ERA5 data from the European Center for Medium-Range Weather Forecasts (ECMWF) was used to utilize SST data at the same time period. As a research method, Long Short-Term Memory (LSTM) algorithm specialized for time series data prediction among deep learning models was used in consideration of the time series characteristics of SST data. The prediction model predicts the SST near the Korean Peninsula after 1- to 7-days and predicts the high water temperature or low water temperature phenomenon. To evaluate the accuracy of SST prediction, Coefficient of determination (R²), Root Mean Squared Error (RMSE), and Mean Absolute Percentage Error (MAPE) indicators were used. The summer (JAS) 1-day prediction result of the prediction model, R²=0.996, RMSE=0.119℃, MAPE=0.352% and the winter (JFM) 1-day prediction result is R²=0.999, RMSE=0.063℃, MAPE=0.646%. Using the predicted SST, the accuracy of abnormal sea surface temperature prediction was evaluated with an F1 Score (F1 Score=0.98 for high water temperature prediction in summer (2021/08/05), F1 Score=1.0 for low water temperature prediction in winter (2021/02/19)). As the prediction period increased, the prediction model showed a tendency to underestimate the SST, which also reduced the accuracy of the abnormal water temperature prediction. Therefore, it is judged that it is necessary to analyze the cause of underestimation of the predictive model in the future and study to improve the prediction accuracy.

• Korean Journal of Remote Sensing
• /
• v.36 no.5_4
• /
• pp.1179-1194
• /
• 2020

This study investigates the feasibility of three algorithms, K-Nearest Neighbors (K-NN), Random Forest (RF) and Neural Network (NN), for estimating the air temperature of an unobserved area where the weather station is not installed. The satellite image were obtained from Landsat-8 and MODIS Aqua/Terra acquired in 2019, and the meteorological ground weather data were from AWS/ASOS data of Korea Meteorological Administration and Korea Forest Service. In addition, in order to improve the estimation accuracy, a digital surface model, solar radiation, aspect and slope were used. The accuracy assessment of machine learning methods was performed by calculating the statistics of R² (determination coefficient) and Root Mean Square Error (RMSE) through 10-fold cross-validation and the estimated values were compared for each target area. As a result, the neural network algorithm showed the most stable result among the three algorithms with R² = 0.805 and RMSE = 0.508. The neural network algorithm was applied to each data set on Landsat imagery scene. It was possible to generate an mean air temperature map from June to September 2019 and confirmed that detailed air temperature information could be estimated. The result is expected to be utilized for national disaster safety management such as heat wave response policies and heat island mitigation research.

• Journal of Korea Water Resources Association
• /
• v.38 no.12 s.161
• /
• pp.1061-1072
• /
• 2005

A hydrological grey model is developed to forecast short-term river runoff from the Naju watershed located at upstream of the Youngsan estuary dam in Korea. The runoff of the Naju watershed is measured in real time at the Naju streamflow gauge station, which is a key station for forecasting the upstream inflow and operating the gates of the estuary dam in flood period. The model's governing equation is formulated on the basis of the grey system theory. The model parameters are reparameterized in combination with the grey system parameters and estimated with the annealing-simplex method In conjunction with an objective function, HMLE. To forecast accurately runoff, the fifth order differential equation was adopted as the governing equation of the model in consideration of the statistic values between the observed and forecast runoff. In calibration, RMSE values between the observed and simulated runoff of two and six Hours ahead using the model range from 3.1 to 290.5 $m^{3}/s,\;R^2$ values range from 0.909 to 0.999. In verification, RMSE values range from 26.4 to 147.4 $m^{3}/s,\;R^2$ values range from 0.940 to 0.998, compared to the observed data. In forecasting runoff in real time, the relative error values with lead-time and river stage range from -23.4 to $14.3\%$ and increase as the lead time increases. The results in this study demonstrate that the proposed model can reasonably and efficiently forecast runoff for one to six Hours ahead.

• Journal of Korea Water Resources Association
• /
• v.47 no.10
• /
• pp.945-958
• /
• 2014

Soil temperature is one of the most important environmental factors that govern hydrological and biogeochemical processes related to diffuse pollution. In this study, considering the snowmelting and the soil freezing-thawing processes, a set of computer codes to estimate winter soil temperature has been developed for CAMEL (Chemicals, Agricultural Management and Erosion Losses), a distributed watershed model. The model was calibrated and validated against the field measurements for three months at 4 sites across the study catchment in a rural area of Yeoju, Korea. The degree of agreement between the simulated and the observed soil temperature is good for the soil surface ($R^2$ 0.71~0.95, RMSE $0.89{\sim}1.49^{\circ}C$). As for the subsurface soils, however, the simulation results are not as good as for the soil surface ($R^2$ 0.51~0.97, RMSE $0.51{\sim}5.08^{\circ}C$) which is considered resulting from vertically-homogeneous soil textures assumed in the model. The model well simulates the blanket effect of snowpack and the latent heat flux in the soil freezing-thawing processes. Although there is some discrepancy between the simulated and the observed soil temperature due to limitations of the model structure and the lack of data, the model reasonably well simulates the temporal and spatial distributions of the soil temperature and the snow water equivalent in accordance with the land uses and the topography of the study catchment.

• Journal of Korea Water Resources Association
• /
• v.51 no.9
• /
• pp.803-812
• /
• 2018

The purpose of this study is to develop Particle Swarm Optimization (PSO) automatic calibration algorithm with multi-objective functions by Python, and to evaluate the applicability by applying the algorithm to the Soil and Water Assessment Tool (SWAT) watershed modeling. The study area is the upstream watershed of Gongdo observation station of Anseongcheon watershed ($364.8km^2$) and the daily observed streamflow data from 2000 to 2015 were used. The PSO automatic algorithm calibrated SWAT streamflow by coefficient of determination ($R^2$), root mean square error (RMSE), Nash-Sutcliffe efficiency ($NSE_Q$), and especially including $NSE_{INQ}$ (Inverse Q) for lateral, base flow calibration. The results between automatic and manual calibration showed $R^2$ of 0.64 and 0.55, RMSE of 0.59 and 0.58, $NSE_Q$ of 0.78 and 0.75, and $NSE_{INQ}$ of 0.45 and 0.09, respectively. The PSO automatic calibration algorithm showed an improvement especially the streamflow recession phase and remedied the limitation of manual calibration by including new parameter (RCHRG_DP) and considering parameters range.

• Korean Journal of Remote Sensing
• /
• v.37 no.6_2
• /
• pp.1881-1890
• /
• 2021

Particulate matter (PM) affects the human, ecosystems, and weather. Motorized vehicles and combustion generate fine particulate matter (PM_2.5), which can contain toxic substances and, therefore, requires systematic management. Consequently, it is important to monitor and predict PM_2.5 concentrations, especially in large cities with dense populations and infrastructures. This study aimed to predict PM_2.5 concentrations in large cities using meteorological and chemical variables as well as satellite-based aerosol optical depth. For PM_2.5 concentrations prediction, a random forest (RF) model showing excellent performance in PM concentrations prediction among machine learning models was selected. Based on the performance indicators R², RMSE, MAE, and MAPE with training accuracies of 0.97, 3.09, 2.18, and 13.31 and testing accuracies of 0.82, 6.03, 4.36, and 25.79 for R², RMSE, MAE, and MAPE, respectively. The variables used in this study showed high correlation to PM_2.5 concentrations. Therefore, we conclude that these variables can be used in a random forest model to generate reliable PM_2.5 concentrations predictions, which can then be used to assess the vulnerability of schools to PM_2.5.