• Journal of Korea Water Resources Association
• /
• v.46 no.1
• /
• pp.47-58
• /
• 2013

Recently, an increase in the occurrence of sudden local flooding of great volume and short duration due to global climate changes has occasioned the significant danger and loss of life and property in Korea as well as most parts of the world. Such a local flood that usually occurs as the result of intense rainfall over small regions rises quite quickly with little or no advance warning time to prevent flood damage. To prevent the local flood damage, it is important to quickly predict the flood severity for flood events exceeding a threshold discharge that may cause the flood damage for inland areas. The aim of this study is to develop the NFI (New Flood Index) measuring the severity of floods in small ungauged catchments for use in local flood predictions by the regression analysis between the NFI and rainfall patterns. Flood runoff hydrographs are generated from a rainfall-runoff model using the annual maximum rainfall series of long-term observations for the two study catchments. The flood events above a threshold assumed as the 2-year return period discharge are targeted to estimate the NFI obtained by the geometric mean of the three relative severity factors, such as the flood magnitude ratio, the rising curve gradient, and the flooding duration time. The regression results show that the 3-hour maximum rainfall depths have the highest relationships with the NFI. It is expected that the best-fit regression equation between the NFI and rainfall characteristics can provide the basic database of the preliminary information for predicting the local flood severity in small ungauged catchments.

• Journal of Korea Water Resources Association
• /
• v.47 no.11
• /
• pp.1095-1105
• /
• 2014

A spatial downscaling method using the Support Vector Machine (SVM) Regression for 25 km Tropical Rainfall Measuring Mission (TRMM) Monthly precipitation is proposed. The nonlinear relationship among hydrometeorological variables and precipitation was effectively depicted by the SVM for predicting downscaled grid precipitation. The accuracy of spatially downscaled precipitation was estimated by comparing with rain gauge data from sixty-four stations and found to be improved than the original TRMM data in overall. Especially the positive bias of the original TRMM data was effectively removed after the downscaling procedure. The spatial distributions of 25 km and 1 km grid precipitation were generally similar, while the local spatial trend was better detected by 1 km grid precipitation. The downscaled grid data derived from the proposed method can be applied in hydrological modelling for higher accuracy and further be studied for developing optimized downscaling method incorporation other regression methods.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.18 no.8
• /
• pp.378-385
• /
• 2017

This study investigated the design of a snap fit, which is widely used for fastening plastic parts. We analyzed the assembly mechanism of a lock snapfit, measured the assembly force and separation force based on the design of experiments, and derived a regression equation through an analysis of variance. The response surface methodology was also used. Polybutylene terephthalate was used to fabricate specimens, and the assembly force and separation force were measured using a micro-tensile tester. The length, width, thickness, and interference were considered as factors. A second-order regression model was used to derive the regression equation. The assembly force decreased with increasing length and width, but it increased with increasing thickness and interference. The finite element method was used to analyze the assembly mechanics. The width decreased the assembly force by increasing the ductility. The influences of the factors for low assembly force and high release force were shown to be opposite to each other. It was necessary to design a structure that minimized the assembly force while maintaining an appropriate level of separation force.

• Korean Journal of Food Science and Technology
• /
• v.28 no.3
• /
• pp.411-416
• /
• 1996

Supercritical fluid extraction of ${\beta}$-carotene from carrot was optimized to maximize ${\beta}$-carotene (Y) extraction yield. A central composite design involving extraction pressure ($X_1$ 200-,100 bar), temperature ($X_2,\;35-51^{\circ}C$) and time ($X_1$$ 60-200min) was used. Three independent factors ($X_1,\;X_2,\;X_3$) were chosen to determine their effects on the various responses and the function was expressed in terms of a quadratic polynomial equation,$Y={\beta}_0+{\beta}_1X_1+{\beta}_2X_2+{\beta}_3X_3+{\beta}_11X_12+{\beta}_22X_3^2+{\beta}_-12X_1X_2+{\beta}_12X_1X_2+{\beta}_13X_1X_3+{\beta}_23X_2X_3,$ which measures the linear, quadratic and interaction effects. Extraction yields of ${\beta}$-carotene were affected by pressure, time and temperature in the decreasing order, and linear effect of tenter point (${\beta}_11$) and pressure (${\beta}_1$) were significant at a level of 0.001(${\alpha}$). Based on the analysis of variance, the model fitted for ${\beta}_11$-carotene (Y) was significant at 5% confidence level and the coefficient of determination was 0.938. According to the response surface of ${\beta}$-carotene by cannoical analysis, the stationary point for quantitatively dependent variable (Y) was found to be the maximum point for extraction yield. Response area for ${\beta}$-carotene (Y) in terms of interesting region was estimated over $10,611{\mu}g$ Per 100 g raw carrot under extraction.

• Korean Journal of Food Science and Technology
• /
• v.35 no.5
• /
• pp.914-917
• /
• 2003

This study was undertaken to find the optimum condition for the Maillard browning reaction of tagatose and glycine model solution by applying the response surface methodology. Independent variables were pH (3, 5, 7), temperature (70, 85, $100^{\circ}C$), and time (60, 180, 300 min), while the dependent variables were absorbance, yellowness, color difference, and organoleptic score. The quadratic models with the cross-product proved to be suitable, due to the high coefficients of determination and the lack of fit results. Since all the dependent variables had saddle points, the optimal points were determined through ridge analysis. For absorbance, yellowness, and color difference, the optimal points were the lowest values; in contrast, the optimal point of organoleptic score was the highest value.

• Korean Journal of Food Science and Technology
• /
• v.35 no.5
• /
• pp.884-892
• /
• 2003

This study was conducted to optimize the water homogenization process of mandarin orange peel for colored rice. Four variables were used to determine the optimum conditions for homogenization speed, time, temperature, and water volume with a five level central composite design and response surface methodology. The process was optimized using the combination of EI and b values of rice coated with water extract of the mandarin orange peel. The effect of water volume was the most significant compared to the other variables on the quality of water homogenate. The regression polynomial model was a suitable (p>0.05) model by lack-of-fit analysis showing high significance. To optimize the process, based on surface response and contour plots, individual contour plots for the response variables were superimposed. The optimum conditions for manufacturing water extract from mandarin orange was with 8,500 rpm homogenization speed, 2.8 min time, $53^{\circ}C$ temperature, and 42 mL water volume with the maximum of restricted variables of EI above 400 and h value above 24.

• Food Engineering Progress
• /
• v.21 no.1
• /
• pp.79-87
• /
• 2017

This study was carried out to extract ginsenosides in by-products from honeyed red ginseng. Response surface methodology (RSM) was used to optimize the extraction conditions. Based on D-optimal design, independent variables were ethanol (extraction solvent) concentration (30-90%, v/v), extraction temperature ($25-70^{\circ}C$), and extraction time (5-11 h). Extraction yield (Y1) and total ginsenosides (Y2) in the extract were analyzed as dependent variables. Results found that extraction yield increased with increasing extraction temperature and time, whereas it was decreased with increasing ethanol concentration. Similar trends were found for the content of ginsenosides in the extracts, except for ethanol concentration, which was increased with increasing ethanol concentration. Regression equations derived from RSM were suggested to coincide well with the results from the experiments. The optimal extraction conditions for extraction yield and total ginsenosides were an extraction temperature of $56.94^{\circ}C$, ethanol concentration of 57.90%, and extraction time of 11 h. Under these conditions, extraction yield and total ginsenoside contents were predicted to be 84.52% and 9.54 mg/g, respectively.

• The Korean Journal of Applied Statistics
• /
• v.34 no.3
• /
• pp.309-327
• /
• 2021

Dynamic treatment regimes (DTRs) are decision-making rules designed to provide personalized treatment to individuals in multi-stage randomized trials. Unlike classical methods, in which all individuals are prescribed the same type of treatment, DTRs prescribe patient-tailored treatments which take into account individual characteristics that may change over time. The Q-learning method, one of regression-based algorithms to figure out optimal treatment rules, becomes more popular as it can be easily implemented. However, the performance of the Q-learning algorithm heavily relies on the correct specification of the Q-function for response, especially in observational studies. In this article, we examine a number of double-robust weighted least-squares estimating methods for Q-learning in high-dimensional settings, where treatment models for propensity score and penalization for sparse estimation are also investigated. We further consider flexible ensemble machine learning methods for the treatment model to achieve double-robustness, so that optimal decision rule can be correctly estimated as long as at least one of the outcome model or treatment model is correct. Extensive simulation studies show that the proposed methods work well with practical sample sizes. The practical utility of the proposed methods is proven with real data example.

• Journal of Korea Water Resources Association
• /
• v.53 no.12
• /
• pp.1143-1157
• /
• 2020

In this study, a criteria for the SWAT model calibration method in SWAT-CUP which considers multi-site and multi-variable observations was presented. For its application, the SWAT model was simulated using long-term observed flow, soil moisture, and evapotranspiration data in Yongdam study watershed, investigating the hydrological runoff characteristics and water balance in the water cycle analysis. The model was calibrated with different parameter values for each sub-watershed in order to reflect the characteristics of multiple observations through one-by-one calibration, appropriate settings of model simulation run/iteration number (1,000 simulation runs in the first iteration and then 500 simulation runs for the following iterations), and executions of partial and all run in SWAT-CUP. The flow simulation results of watershed outlet point, ENS 0.85, R2 0.87, and PBIAS -7.6%, were compared with the analysis results (ENS 0.52, R2 0.54, and PBIAS -22.4%) applied in the other batch (i.e., non one-by-one) calibration approach and showed better performances of proposed method. From the simulation results of a total of 15 years, it was found that the total runoff (streamflow) and evapotranspiration rates from precipitation are 53 and 39%, and the ratio of surface runoff and baseflow (i.e., sum of lateral and return flow, and recharge deep aquifer) are 35 and 65%, respectively, in Yongdam watershed. In addition, the analytical amount of available water (i.e., water yield), including the total annual streamflow (daily average 21.8 m3/sec) is 6.96 billion m3 per year (about 540 to 900 mm for sub-watersheds).

• Korean Journal of Agricultural and Forest Meteorology
• /
• v.4 no.4
• /
• pp.203-212
• /
• 2002

This study was conducted to remove the urban heat island effects embedded in the interpolated surfaces of daily minimum temperature in the Korean Peninsula. Fifty six standard weather stations are usually used to generate the gridded temperature surface in South Korea. Since most of the weather stations are located in heavily populated and urbanized areas, the observed minimum temperature data are contaminated with the so-called urban heat island effect. Without an appropriate correction, temperature estimates over rural area or forests might deviate significantly from the actual values. We simulated the spatial pattern of population distribution within any single population reporting district (city or country) by allocating the reported population to the "urban" pixels of a land cover map with a 30 by 30 m spacing. By using this "digital population model" (DPM), we can simulate the horizontal diffusion of urban effect, which is not possible with the spatially discontinuous nature of the population statistics fer each city or county. The temperature estimation error from the existing interpolation scheme, which considers both the distance and the altitude effects, was regressed to the DPMs smoothed at 5 different scales, i.e., the radial extent of 0.5, 1.5, 2.5, 3.5 and 5.0 km. Optimum regression models were used in conjunction with the distance-altitude interpolation to predict monthly normals of daily minimum temperature in South Korea far 1971-2000 period. Cross validation showed around 50% reduction in terms of RMSE and MAE over all months compared with those by the conventional method.conventional method.