Taesoon Kim

Monthly Precipitation Forecasting with a Neuro-Fuzzy Model

Quantitative and qualitative monthly precipitation forecasts are produced with ANFIS. To select the proper input variable set from 30 variables, including climatological and hydrological monthly recording data, the forward selection method, which is a wrapper method for feature selection, is applied. The error analysis of the results from training and checking the data sets suggests that 3 variables can be used as a suitable number of inputs for ANFIS, and the best five 3-input-variable sets were selected. The quantitative monthly precipitation forecasts were computed using each 3-input-variable set, and the ensemble averaging method over the five forecasts was used for calculations to reduce the uncertainties in the forecasts and to remove the negative rainfall forecasts. A qualitative forecast that is computed with the quantitative forecast also produced three types of categories that describe the next month’s precipitation condition and was compared with data from the weather agency of Korea.

Application of multi-objective genetic algorithms to multireservoir system optimization in the Han River basin

The objective of multireservoir system optimization is to achieve an optimal reservoir operating plan by the effective use of water resources. Many optimization techniques have been applied for the last decades, and researchers have recently interested in the heuristic approaches like evolutionary computation. This study proposes a methodology for applying multi-objective genetic algorithms (MOGAs) to a multireservoir system optimization in the Han River basin. The second generation evolutionary multiobjective technique, NSGA-II, is used. The simulation model is applied to the Han River basin and the performance of the model is compared with the historical reservoir operation records. Two different cases are performed to evaluate the applicability of NSGA-II. Case 1 shows the basic performance of NSGA-II as applied to multireservoir system optimization, and Case 2 presents the methodology to discriminate the critical decision variables. In addition, the alternative releases and storages by NSGA-II are compared with the historical releases and storages. Cases 1 and 2 show that NSGA-II can be applied to multireservoir system optimization, and the alternative releases and storages computed using the results from NSGA-II can be used as the possible resrvoir operating plans that supply more water resources to downstream than the historical releases.

Improving accuracy of IDF curves using long- and short-duration separation and multi-objective genetic algorithm

Multi-objective genetic algorithm (MOGA) and cumulative distribution function (CDF) are used to improve the accuracy of IDF curve. Rainfall durations are divided into short- and long-duration using root mean squared error (RMSE) and relative root mean squared error (RRMSE) between rainfall quantiles by IDF curve and at-site frequency analysis. RMSE could be used for estimating parameters of relatively long-duration, and RRMSE for short-duration. The compromised solutions could be ahieved through MOGA with two multi-objective functions. The duration separating technique called COMBI_1 is suggested and the comparison with the five different parameter estimation methods provides COMBI_1 is superior to the other methods.

Optimization of Tank Model Parameters Using Multi-Objective Genetic Algorithm (I): Methodology and Model Formulation

The objective of this study is to evaluate the applicability of multi-objective genetic algorithm(MOGA) in order to calibrate the parameters of conceptual rainfall-runoff model, Tank model. NSGA-II, one of the most imitating MOGA implementations, is combined with Tank model and four multi-objective functions such as to minimize volume error, root mean square error (RMSE), high flow RMSE, and low flow RMSE are used. When NSGA-II is employed with more than three multi-objective functions, a number of Pareto-optimal solutions usually becomes too large. Therefore, selecting several preferred Pareto-optimal solutions is essential for stakeholder, and preference-ordering approach is used in this study for the sake of getting the best preferred Pareto-optimal solutions. Sensitivity analysis is performed to examine the effect of initial genetic parameters, which are generation number and Population size, to the performance of NSGA-II for searching the proper paramters for Tank model, and the result suggests that the generation number is 900 and the population size is 1000 for this study.

Regional Rainfall Frequency Analysis by Multivariate Techniques

Regional rainfall quantile depends on the identification of hydrologically homogeneous regions. Various variables relevant to precipitation can be used to form regions. Since the type and number of variables may lead to improve the efficiency of partitioning, it is important to select those precipitation related variables, which represent most of the information from all candidate variables. Multivariate analysis techniques can be used for this purpose. Procrustes analysis which can decrease the dimension of variables based on their correlations, are applied in this study. 42 rainfall related variables are decreased into 21 ones by Procrustes analysis. Factor analysis is applied to those selected variables and then 5 factors are extracted. Fuzzy-c means technique classifies 68 stations into 6 regions. As a result, the GEV distributions are fitted to 6 regions while the lognormal and generalized logistic distributions are fitted to 5 regions. For the comparison purpose with previous results, rainfall quantiles based on generalized logistic distribution are estimated by at-site frequency analysis, index flood method, and regional shape estimation method.

Adjustment of the Mean Field Rainfall Bias by Clustering Technique

Fuzzy c-means clustering technique is applied to improve the accuracy of G/R ratio used for rainfall estimation by radar reflectivity. G/R ratio is computed by the ground rainfall records at AWS(Automatic Weather System) sites to the radar estimated rainfall from the reflectivity of Kwangduck Mt. radar station with 100km effective range. G/R ratio is calculated by two methods: the first one uses a single G/R ratio for the entire effective range and the other two different G/R ratio for two regions that is formed by clustering analysis, and absolute relative error and root mean squared error are employed for evaluating the accuracy of radar rainfall estimation from two G/R ratios. As a result, the radar rainfall estimated by two different G/R ratio from clustering analysis is more accurate than that by a single G/R ratio for the entire range. keywords : radar rainfall, mean field bias, fuzzy c-means clustering, G/R bias adjustment ..............................................................................................................................................................................................

Rainfall Quantile Estimation Using Scaling Property in Korea

In this study, rainfall quantile was estimated using scale invariance property of rainfall data with different durations and the applicability of such property was evaluated for the rainfall data of South Korea. For this purpose, maximum annual rainfall at 22 recording sites of Korea Meteorological Administration (KMA) having relatively long records were used to compare rainfall quantiles between at-site frequency analysis and scale invariance property. As the results, the absolute relative errors of rainfall quantiles between two methods show at most 10 % for hourly rainfall data. The estimated quantiles by scale invariance property can be generally applied in the 8 of 14 return periods used in this study. As an example of down-scaling method, rainfall quantiles of minutes duration were estimated by scale invariance property based on index duration of 1 hour. These results show less than 10 % of absolute relative errors except 10 minutes duration. It is found that scale invariance property can be applied to estimate rainfall quantile for unmeasured rainfall durations.

Prognostic clinical factors associated with remission and relapse in bullous pemphigoid

Prognostic factors for remission and relapse of bullous pemphigoid (BP) remain uncertain.

Optimization of Tank Model Parameters Using Multi-Objective Genetic Algorithm (II): Application of Preference Ordering

Preference ordering approach is applied to optimize the parameters of Tank model using multi-objective genetic algorithm (MOGA). As more than three multi-objective functions are used in MOGA, too many non-dominated optimal solutions would be obtained thus the stakeholder hardly find the best optimal solution. In order to overcome this shortcomings of MOGA, preference ordering method is employed. The number of multi-objective functions in this study is 4 and a single Pareto-optimal solution, which is 2nd order efficiency and 3 degrees preference ordering, is chosen as the most preferred optimal solution. The comparison results among those from Powell method and SGA (simple genetic algorithm), which are single-objective function optimization, and NSGA-II, multi-objective optimization, show that the result from NSGA-II could be reasonalby accepted since the performance of NSGA-II is not deteriorated even though it is applied to the verification period which is totally different from the calibration period for parameter estimation.

Input variables selection of artificial neural network using mutual information.

Input variable selection is one of the various techniques for improving the performance of artificial neural network. In this study, mutual information is applied for input variable selection technique instead of correlation coefficient that is widely used. Among 152 variables of RDAPS (Regional Data Assimilation and Prediction System) output results, input variables for artificial neural network are chosen by computing mutual information between rainfall records and RDAPS` variables. At first the rainfall forecast variable of RDAPS result, namely APCP, is included as input variable and the other input variables are selected according to the rank of mutual information and correlation coefficient. The input variables using mutual information are usually those variables about wind velocity such as D300, U925, etc. Several statistical error estimates show that the result from mutual information is generally more accurate than those from the previous research and correlation coefficient. In addition, the artificial neural network using input variables computed by mutual information can effectively reduce the relative errors corresponding to the high rainfall events.