Climate Change and Its Impacts on Streamflow: WRF and SCE-Optimized SWAT Models - Data Assimilation for Atmospheric, Oceanic and Hydrologic Applications - page 410

17.2.2.1 Model Setup

In this study, the main objective was to quantify the impacts of climate change
to hydrological stream flow over a long period of time. SWAT was chosen as the
hydrological model to which the precipitation output of a regional climate model
was applied as input to simulate stream flow. The input for ArcSWAT included a
spatial reference map such as the DEM (Digital Elevation Model) with a resolution
of 250 m 250 m, a land use map and a soil map (converted to raster format with
the same resolution) and meteorological data (precipitation and temperature time-
series of all stations in daily format). The DEM was obtained from Department of
Survey and Mapping (DSM) of Vietnam. The land use map was obtained from the
Forest Investigation and Planning Institute (FIPI), of Vietnam for the year 2000. The
soil map was obtained through the Ministry of Agriculture and Rural Development
(MARD) of Vietnam categorized by the FAO (Food and Agriculture Organization).

17.2.2.2 Model Sensitivity Analysis

The sensitivity analysis is a method that analyzes the sensitivity of model parameters
to the model performance. This method entails to filter the model parameters that
either have or have not any significant influence on the model results. On the other
hand, it also aims to reduce the number of parameters required in fitting to a model
input-output. Traditional methods of sensitivity analysis have been classified by
Saltelli et al. 2000 . They are (1) Local method ( Melching and Yoon 1996 )(2)
Integration of local to global method using Random One-Factor-At-a-Time (OAT)
proposed by Morris (1991) and (3) Global methods like Monte Carlo and Latin-
Hypercube (LH) simulation ( McKay et al. 1979 ; McKay 1988 ). By studying the
advantages and disadvantages of each of the above methods, van Griensven and
Meixner ( 2006 ) developed the LH-OAT method which performs LH sampling
followed by OAT sampling. This method samples the full range of all parameters
using LH design along with the precision of OAT sampling to ensure that the
changes in each model output could be attributed to the changed parameter. The
LH-OAT design has been coupled to the SWAT model for sensitivity analysis
module. Model parameters are analysed based on the performance of its output
compared against observed data and the model itself. In the SWAT model, there are
26 parameters sensitive to water flow, 6 parameters sensitive to sediment transport
and other 9 parameters sensitive to water quality. In this study, since the stream flow
is the main focus, 10 most sensitive parameters out of the available 26 options are
analysed (Table 17.1 ).

17.2.2.3 Auto-Calibration by ParaSol Method (Parameter Solution) Using

SCE-UA Algorithm

SWAT model has the options to choose either manual or auto-calibration. Calibra-
tion is applied to those most sensitive parameters to yield the optimal set of values