연구보고서
- 저자
- 김종필 박사
- 작성일
- 2016.01.23
- 조회
- 442
- 요약
- 목차
Estimation of long-term streamflow from watersheds is very important to manage water resources effectively and to cope with water-related disasters, including flooding and drought. Long-term streamflow prediction is generally achieved through hydrological modeling, which includes preparation of input datasets, an implementation of a hydrological model, and model calibration. For the watershed of interest, the hydrological model simulates streamflow and an optimal parameter set is estimated by comparing the output data with measurements. This parameter set can be applied to predict another runoff event within the same watershed.
However, daily hydrometeorogical observations, including streamflow, air temperature, relative humidity, sunshine, and wind speed are insufficient for hydrological modeling in many watersheds, even though precipitation records are relatively abundant. This lack of appropriate data interferes with the accurate and reliable prediction of streamflow. Moreover, complex nonlinear equations are added to recently developed hydrological models to simulate more accurate watershed responses. Thus, many parameters in the models may require estimation through model calibration because of the deficit of measurements.
Model parameters can be transferred by using different regionalization methods, which transfer a parameter set from the nearest watershed to the watershed of interest, consider the streamflow from a different watershed with the most similar properties, and use the mean values of the parameters from the neighboring watersheds.
This study employed linear regression equations between the model parameters and physiographic attributes of watersheds, such as drainage area, elevation, slope, land cover, and soil type, which is the most widely used regionalization method. A distributed hydrological model, the Coupled Routing and Excess Storage (CREST) model, was employed for streamflow simulation in the study areas, namely the Chunju Dam basin (CJDB), the Soyanggang Dam basin (SYDB), the Andong Dam basin (ADDB), the Imha Dam basin (IHDB), the Namgang Dam basin (NGDB), the Miryanggang Dam basin (MRDB), the Yongdam Dam basin (YDDB), and the Juam Dam basin (JADB). Two of these basins, MRDB and JADB, were considered ungauged basins for validation.
The Tropical Rainfall Measuring Mission (TRMM) 3B42v6 from the National Aeronautics and Space Administration (NASA) and the Potential Evapotranspiration (PET) from the Famine Early Warning System Network (FEWSNET) were applied to the CREST model as two meteorological input forcings. Topographic parameters were estimated from the Digital Elevation Model (DEM) of the Shuttle Radar Topography Mission (SRTM).
Streamflow simulations were implemented by the CREST model from 01 Jan 2004 to 31 Dec 2009 at a daily time-step and the optimal parameter sets were found using the Adaptive Random Search (ARS) method. The relationships between the model parameters and the basin properties, including drainage area, elevation, slope, longest path, river length, elongation ratio, impervious area, forest area, paddy field, and crop land, were computed using the multiple linear regression method for the six gauged basins. Then, new parameter sets were generated by the regression equations and the physiographic properties of the six gauged basins and the two ungauged basins.
Finally, the new parameters were validated for all the study basins and assessed by four indices, namely the Nash-Sutcliffe Coefficient of Efficiency (NSCE), Percent Bias (PBIAS), Root Mean Squared Error – Observation Standard Deviation Ratio (RSR), and the Pearson’s Correlation Coefficient (PCC).
The results showed that the CREST hydrological model and the proposed regression equations can acceptably simulate streamflow, considering NSCE, RSR, and PCC in both the six gauged and the two ungagued basins. However, they provided somewhat biased streamflow simulations for all the study basins. In further studies, these biases should be reduced by adding other basins and by finding basin properties that are highly related with the model parameters.