• Journal of Korea Water Resources Association
• /
• v.54 no.12
• /
• pp.1295-1304
• /
• 2021

For the efficient use and management of water resources, a reliable rainfall-runoff analysis is necessary. Still, continuous hydrological data and rainfall-runoff data are insufficient to secure through measurements and models. In particular, as part of the reasonable improvement of a rainfall-runoff model in the case of an ungauged watershed, regionalization is being used to transfer the parameters necessary for the model application to the ungauged watershed. In this study, the GR4J model was selected, and the SCEM-UA method was used to optimize parameters. The rainfall-runoff model for the analysis of the correlation between watershed characteristics and parameters obtained through the model was regionalized by the Copula function, and rainfall-runoff analysis with the regionalized parameters was performed on the ungauged watershed. In the process, the intermediate state variables of the rainfall-runoff model were extracted, and the correlation analysis between water level and the ground water level was investigated. Furthermore, in the process of rainfall-runoff analysis, the Standardized State variable Drought Index (SSDI) was calculated by calculating and indexing the state variables of the GR4J model. and the calculated SSDI was compared with the standardized Precipitation index (SPI), and the hydrological suitability evaluation of the drought index was performed to confirm the possibility of drought monitoring and application in the ungauged watershed.

• Journal of Korea Water Resources Association
• /
• v.56 no.9
• /
• pp.563-574
• /
• 2023

Recently, climate change has affected functional responses of river basins to meteorological variables, emphasizing the importance of rainfall-runoff simulation research. Simultaneously, the growing interest in machine learning has led to its increased application in hydrological studies. However, it is not yet clear whether machine learning models are more advantageous than the conventional conceptual models. In this study, we compared the performance of the conventional GR6J model with the machine learning-based Random Forest model across 38 basins in Korea using both gauged and ungauged basin prediction methods. For gauged basin predictions, each model was calibrated or trained using observed daily runoff data, and their performance was evaluted over a separate validation period. Subsequently, ungauged basin simulations were evaluated using proximity-based parameter regionalization with Leave-One-Out Cross-Validation (LOOCV). In gauged basins, the Random Forest consistently outperformed the GR6J, exhibiting superiority across basins regardless of whether they had strong or weak rainfall-runoff correlations. This suggest that the inherent data-driven training structures of machine learning models, in contrast to the conceptual models, offer distinct advantages in data-rich scenarios. However, the advantages of the machine-learning algorithm were not replicated in ungauged basin predictions, resulting in a lower performance than that of the GR6J. In conclusion, this study suggests that while the Random Forest model showed enhanced performance in trained locations, the existing GR6J model may be a better choice for prediction in ungagued basins.

• Journal of Korea Water Resources Association
• /
• v.37 no.12
• /
• pp.1019-1032
• /
• 2004

Flood frequency estimate is an essential index for determining the scale of small and middle hydraulic structure. However, this flood quantity could not be estimated directly for practical design purpose due to the lack of available flood data, and indirect method like design rainfall-runoff method have been used for the estimation of design flood. To give the good explain for design flood estimates, regional flood frequency analysis was performed by flood index method in this study. First, annual maximum series were constructed by using the collected data which covers from Japanese imperialism period to 1999. Wakeby distribution recommended by WMO(1989) was used for regional flood frequency analysis and L-moment method by Hosking (1990) was used for parameter estimation. For the homogeneity of region, the discordance and heterogeneity test by Hosking and Wallis(1993) was carried for 4 major watersheds in Korea. Physical independent variable correlated with index flood was watershed area. The relationship between specific discharge and watershed area showed a type of power function, i.e. the specific discharge decreases as watershed area increases. So flood quantity according to watershed area and return period was presented for each watershed(Han rivet, Nakdong river, Geum river and Youngsan/Seomjin river) by using this relation type. This result was also compared with the result of point frequency analysis and its regionalization. It was shown that the dam construction couldn't largely affect the variation of peak flood. The property of this study was also examined by comparison with previous studies.

• Journal of Korea Water Resources Association
• /
• v.40 no.3
• /
• pp.205-214
• /
• 2007

The objective of this study is to generate the regional scale runoff scenarios by using IPCC SRES A2 climate change scenario for analyzing the spatial variation of water resources in Korea. The PRMS model was adopted to simulate long-term stream discharge. To estimate the PRMS model parameters on each sub-basin, the streamflow data at 6 dam sites and Rosenbrock's scheme are used for model parameter calibration and those parameters are translated to ungauged catchments by regionalization method. The other 3 dam sites are selected for the verification of the adequateness of regionalized model parameters in ungagued catchments. The statistical results show that the simulated flows by using regionalized parameters well agree with observed ones. The generated runoff scenarios by climate change are compared with observed data on 4 dam sites for the reference period. The consequences show that the selection of climate station for generating climate scenario affects the reliability of climate scenario at sub-basin. The comparison results of the stream flows between the 30-year baseline period (1971-2000) and future 90-year (2001-2030, 2031-2060, 2061-2090) show that the long-term mean annual runoff in the Han River has increasing trend, while the Nakdong, the Gum, the Youngsan and the Sumjin Rivers have decreasing trend.

• Journal of Korea Water Resources Association
• /
• v.54 no.6
• /
• pp.395-405
• /
• 2021

When issuing forecasts and alerts for agricultural drought, the relevant ministries only rely on the observation data from the reservoirs managed by the Korea Rural Community Corporation, which creates gaps between the drought analysis results at the local (si/gun) governments and the droughts actually experienced by local residents. Closing these gaps requires detailed local geoinformation on reservoirs, which in turn requires the information on reservoirs managed by local governments across Korea. However, installing water level and flow measurement equipment at all of the reservoirs would not be reasonable in terms of operation and cost effectiveness, and an alternate approach is required to efficiently generate information. In light of the above, this study validates and calibrates the parameters of the TANK model for reservoir basins, divided them into groups based on the characteristics of different basins, and applies the grouped parameters to unmeasured local government reservoirs to estimate and assess inflow. The findings show that the average determinant coefficient and the NSE of the group using rice paddies and inclinations are 0.63 and 0.62, respectively, indicating better results compared with the basin area and effective storage factors (determinant coefficient: 0.49, NSE: 0.47). The findings indicate the possibility of utilizing the information regarding unmeasured reservoirs managed by local governments.