• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.294-294
• /
• 2019

The system for predicting flood of river at Flood Control Office is made up of a rainfall-runoff model and FLDWAV model. This system is mainly operating to predict the excess of the flood watch or warning level at flood forecast points. As the demand for information of the management and operation of riverside, which is being used as a waterfront area such as parks, camping sites, and bike paths, high-level forecasts of watch and warning at certain points are required as well as production of lowland flood forecast information that is used as a waterfront within the river. In this study, a technology to produce flood forecast information in lowland areas of the river used as a waterfront was developed. Based on the results of the 1D hydraulic analysis, a model for performing spatial operations based on high resolution grid was constructed. A model was constructed for Andong district, and the inundation conditions and level were analyzed through a virtual outflow scenarios of Andong and Imha Dam.

• Journal of Korea Water Resources Association
• /
• v.51 no.12
• /
• pp.1207-1216
• /
• 2018

Discharge or water level predictions at tidally affected river reaches are currently still a great challenge in hydrological practices. This research aims to predict water level of the tide dominated site, Jamsu bridge in the Han River downstream. Physics-based hydrodynamic approaches are sometimes not applicable for water level prediction in such a tidal river due to uncertainty sources like rainfall forecasting data. In this study, TensorFlow deep learning framework was used to build a deep neural network based LSTM model and its applications. The LSTM model was trained based on 3 data sets having 10-min temporal resolution: Paldang dam release, Jamsu bridge water level, predicted tidal level for 6 years (2011~2016) and then predict the water level time series given the six lead times: 1, 3, 6, 9, 12, 24 hours. The optimal hyper-parameters of LSTM model were set up as follows: 6 hidden layers number, 0.01 learning rate, 3000 iterations. In addition, we changed the key parameter of LSTM model, sequence length, ranging from 1 to 6 hours to test its affect to prediction results. The LSTM model with the 1 hr sequence length led to the best performing prediction results for the all cases. In particular, it resulted in very accurate prediction: RMSE (0.065 cm) and NSE (0.99) for the 1 hr lead time prediction case. However, as the lead time became longer, the RMSE increased from 0.08 m (1 hr lead time) to 0.28 m (24 hrs lead time) and the NSE decreased from 0.99 (1 hr lead time) to 0.74 (24 hrs lead time), respectively.

• Ecology and Resilient Infrastructure
• /
• v.10 no.4
• /
• pp.116-124
• /
• 2023

As the flood damage on small streams increase due to the increase in frequency of extreme climate events, the need to measure hydraulic data of them has increased for disaster risk management. National Disaster Management Institute, Ministry of Interior and Safety develops CADMT, a CCTV-based automatic discharge measurement technology, and operates pilot small streams to verify its performance and develop disaster risk management technology. The research selects two small streams such as the Neungmac and the Jungsunpil streams to develop the Nomograph by using the 4-Parameter Logistic method using only the observed rainfall data from the Automatic Weather System operated by the Korea Meteorological Agency closest to the small streams and discharge data collected by using the CADMT. To evaluate developed Nomograph, the research forecasts floods discharges in each small stream and compares the result with the observed discharges. As a result of the evaluations, the forecasted value is found to represent the observed value well, so if more accurate observed data are collected and the Nomograph based on it is developed in the future, the high-accuracy flood prediction and warning will be possible.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.4
• /
• pp.199-207
• /
• 2024

Local reservoirs are crucial sources for agricultural water supply, necessitating stable water level management to prepare for extreme climate conditions such as droughts. Water level prediction is significantly influenced by local climate characteristics, such as localized rainfall, as well as seasonal factors including cropping times, making it essential to understand the correlation between input and output data as much as selecting an appropriate prediction model. In this study, extensive multivariate data from over 400 reservoirs in Jeollabuk-do from 1991 to 2022 was utilized to train and validate a water level prediction model that comprehensively reflects the complex hydrological and climatological environmental factors of each reservoir, and to analyze the impact of each input feature on the prediction performance of water levels. Instead of focusing on improvements in water level performance through neural network structures, the study adopts a basic Feedforward Neural Network composed of fully connected layers, batch normalization, dropout, and activation functions, focusing on the correlation between multivariate input data and prediction performance. Additionally, most existing studies only present short-term prediction performance on a daily basis, which is not suitable for practical environments that require medium to long-term predictions, such as 10 days or a month. Therefore, this study measured the water level prediction performance up to one month ahead through a recursive method that uses daily prediction values as the next input. The experiment identified performance changes according to the prediction period and analyzed the impact of each input feature on the overall performance based on an Ablation study.

• Journal of Korea Water Resources Association
• /
• v.57 no.7
• /
• pp.471-479
• /
• 2024

Flood damage is becoming more serious due to the heavy rainfall caused by climate change. Physically based hydrological models have been utilized to predict stream water level variability and provide flood forecasting. Recently, hydrological simulations using machine learning and deep learning algorithms based on nonlinear relationships between hydrological data have been getting attention. In this study, the Long Short-Term Memory (LSTM) algorithm is used to predict the water level of the Seomjin River watershed. In addition, Climate Prediction Center morphing method (CMORPH)-based gridded precipitation data is applied as input data for the algorithm to overcome for the limitations of ground data. The water level prediction results of the LSTM algorithm coupling with the CMORPH data showed that the mean CC was 0.98, RMSE was 0.07 m, and NSE was 0.97. It is expected that deep learning and remote data can be used together to overcome for the shortcomings of ground observation data and to obtain reliable prediction results.

• Atmosphere
• /
• v.34 no.3
• /
• pp.305-317
• /
• 2024

The study examines the effects of parameters that define the characteristics of raindrops on the simulated precipitation during the summer season over Korea using the Weather Research and Forecasting (WRF) Double-Moment 6-class (WDM6) cloud microphysics scheme. Prescribed parameters, defining the characteristics of hydrometeors in the WDM6 scheme such as a_R, b_R, and f_R in the fall velocity (V_R) - diameter (D_R) relationship and shape parameter (𝜇_R) in the number concentration (N_R) - D_R relationship, presents different values compared to the observed data from Two-Dimensional Video Disdrometer (2DVD) at Boseong standard meteorological observatory during 2018~2019. Three experiments were designed for the heavy rainfall event on August 8, 2022 using WRF version 4.3. These include the control (CNTL) experiment with original parameters in the WDM6 scheme; the MUR experiment, adopting the 50th percentile observation value for 𝜇_R; and the MEDI experiment, which uses the same 𝜇_R as MUR, but also includes fitted values for a_R, b_R, and f_R from the 50th percentile of the observed V_R - D_R relationship. Both sensitivity experiments show improved precipitation simulation compared to the CNTL by reducing the bias and increasing the probability of detection and equitable threat scores. In these experiments, the raindrop mixing ratio increases and its number concentration decreases in the lower atmosphere. The microphysics budget analysis shows that the increase in the rain mixing ratio is due to enhanced source processes such as graupel melting, vapor condensation, and accretion between cloud water and rain. Our study also emphasizes that applying the solely observed 𝜇_R produces more positive impact in the precipitation simulation.

• Korean journal of applied entomology
• /
• v.14 no.3 s.24
• /
• pp.111-131
• /
• 1975

The study has been carried out to investigate the occurrence, damage, characteristics of the pathogen, environmental conditions affecting the disease outbreak, varietal resistance, forecasting, and chemical control of bacterial leaf blight of rice in Korea since 1964. Bacterial leaf blight of rice became a major disease in Korea since 1960. A correlation was found between the annual increase of epidemics and increase of cultivation area of susceptible varieties, Jinheung, Keumnampung etc. Areal damage within the country showed that the more was at southern province, Jeonnam, Gyeongnam and western coast, and at flooded rice paddy. Yield reduction directly related with the amount of infection on upper leaves at heading stage. Fifty per cent of reduction resulted when the lesion area was more than 60 per cent. Less than 20 per cent of lesion area, however, was not affected so much on yield loss One hundred and six isolates collected from all over the country were classified as 8 strains by using 4 different bacteriophages in 1973. It was, however, only two in 1965. There were some specificities on varietal distributions among the strains such as that the Jinheung attacked mainly by strain A, B, C and I, those attack Kimmaze were A, B, H and I. Most strains were found from Tongil except D and E, whereas Akibare was only variety that attacked by strain E. Low temperature, high humidity, heavy rainfall and insutficient daylight favored the disease epidemics. Especially, typhoon and flooding at heading stage were critical factors. The earlier transplanting the more disease was resulted, and more nitrogen fertilizer application accerelated the diseased development in general. The resistance to the disease varied by growing stage of the sane plants. All of recommended varieties in Korea were susceptible to the disease except Norm No. 6 and Sirogane which moderately resistant. The pathogen, Xanthomonas oryzae, was detectable from extract of healthy seedlings that were grown in the field with an heavy infection previous year. The more bacteriophage in irigation water resulted the more disease outbreak, and the existence of more than 50 bacteriophages in 1ml. of irrigation water were necessary to initiate the disease out break. The curves representing occurrence of bacteriophages and disease outbreak were similar with 15 days interval. The survey of bacteriophage occurrence can be utilized in forecasting of the disease two weeks ahead of disease outbreak. Three applications of chemicals, Phenazin and Sangkel, in weekly intervals at the early satage of out-break depressed the symptom development, and increased yield by 20per cent. Proper period for the chemical application was just before the number of bacteriophage reaches 50 in 1ml. of irrigation water.

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.18 no.4
• /
• pp.199-207
• /
• 2016

This study was conducted to develop a forest fire occurrence model using meteorological characteristics for practical forecasting of forest fire danger rate by reflecting the climate change for the time period of 2000yrs. Forest fire in South Korea is highly influenced by humidity, wind speed, temperature, and precipitation. To effectively forecast forest fire occurrence, we developed a forest fire danger rating model using weather factors associated with forest fire in 2000yrs. Forest fire occurrence patterns were investigated statistically to develop a forest fire danger rating index using times series weather data sets collected from 76 meteorological observation centers. The data sets were used for 11 years from 2000 to 2010. Development of the national forest fire occurrence probability model used a logistic regression analysis with forest fire occurrence data and meteorological variables. Nine probability models for individual nine provinces including Jeju Island have been developed. The results of the statistical analysis show that the logistic models (p<0.05) strongly depends on the effective and relative humidity, temperature, wind speed, and rainfall. The results of verification showed that the probability of randomly selected fires ranges from 0.687 to 0.981, which represent a relatively high accuracy of the developed model. These findings may be beneficial to the policy makers in South Korea for the prevention of forest fires.