A hydro-environmental model chain in the Doam dam basin, Korea, was developed for an impact assessment under the Intergovernmental Panel on Climate Change's A1B scenario. The feasible downscaling scheme composed of an artificial neural network (ANN) and non-stationary quantile mapping was applied to the GCM (Global Climate Model) output. The impacts under climate and land use change scenarios were examined and projected using the Soil and Water Assessment Tool (SWAT) model. The daily SWAT model was calibrated and validated for 2003–2004 and 2006–2008, respectively. Meanwhile the monthly SS (suspended solids) was calibrated and validated for 1999–2001 and 2007–2009, respectively. The simulation results illustrated that under the assumption of 1–5% urbanization of the forest area, the hydrologic impact is relatively negligible and the climate change impacts are dominant over the urbanization impacts. Additionally the partial impacts of land use changes were analyzed under five different scenarios: partial change of forest to urban (PCFUr), to bare field, to grassland, to upland crop (PCFUp), and to agriculture (PCFA). The analysis of the runoff change shows the highest rate of increase, 73.57% in April, for the PCFUp scenario. The second and third highest rate increases, 37.83% and 31.45% in May, occurred under the PCFA and PCFUr scenarios, respectively.

INTRODUCTION

From the recent Intergovernmental Panel on Climate Change (IPCC) 4th Assessment Report, the climate change not only results in increasing atmospheric water vapor content; changing precipitation patterns, intensity and extremes; and changing soil moisture and runoff but also affects water quality and exacerbates many forms of water pollution such as nutrients, sediments, and pathogens with possible negative impacts on ecosystems and human health (IPCC 2007; Bates et al. 2008). The effects of non-point source (NPS) pollutants, which are caused by extreme rainfall events and associated land surface runoff, are well documented in other studies (Novotny & Olem 1994; Carpenter et al. 1998).

The Doam dam basin has been designated as a reference area for non-point pollution sources because large amounts of materials causing turbid water have collected here due to its unique geological structure and the land use pattern of upland cultivation. In other words, the basin is an area in which natural soil loss occurs more actively along the upland slope compared to other areas, with such basin characteristics as the soil, landforms and land cover conditions creating favorable conditions for sediment detachment from upland surfaces. During the rainy season, the turbidity of the reservoir water increases rapidly during rainfall events. The water being discharged into the Namdaecheon Stream in Gangneung for hydroelectric power generation leads to water quality deterioration in the downstream river courses.

In this study, the Doam dam watershed located upstream of the Song stream basin was selected as a study area. The hydro-environmental impacts on the Doam dam watershed caused by climate and land use changes were examined and projected using the watershed runoff model Soil and Water Assessment Tool (SWAT) and properly downscaled Regional Climate Model (RCM) outputs under a moderate climate change scenario (A1B). Accurate modeling of future runoff regimes is generally challenging with limited current and historical runoff data. Setegn et al. (2009, 2011) tried to apply the SWAT model to simulate the hydrology and impact of climate change on the hydroclimatology of the Lake Tana Basin in Ethiopia under the temporal and spatial coarse hydrologic data availability. They used low resolution topographic information such as a 90 m resolution Shuttle Radar Topography Mission digital elevation model (DEM) and a 1:50,000 scale land use map. Similarly, Jayakrishnan et al. (2005) used the SWAT model to project runoff in the Sondu River basin in Kenya. Joyce et al. (2000), as a part of the national assessment of climate change, stated that the land use percentage would likely shift between forest and agricultural areas as climate change adaptation proceeded. For the sake of regional bias correction for Global Climate Model (GCM) projection, the IPCC's A1B GCM output was downscaled using artificial neural networks (ANNs) and non-stationary quantile mapping. The SWAT model was calibrated and validated with runoff measured at the outlet of the Doam dam watershed.

METHODOLOGY

Study area

The research basin of this study is the Doam dam watershed with an area of 149.42 km2 and a channel length of 22.72 km. The main stream is Song stream that has tributaries that include Changhang stream, Daegwallyeong stream, and Yongpyeong stream. The watershed consists of 62.0% forest area and 31.7% agriculture area. The upland agricultural and built-up area that can be a major source of sediment yield is occupied at a high rate of approximately 34%. The Doam dam is located at the Pyeongchang-gun, Gangwon-do, in the northeastern part of South Korean territory (Figure 1) and was constructed to generate hydroelectric power in 1989. However, operation has been suspended since 2001 because it has been recognized as the main source of turbid water downstream. Figure 1 shows sub-watersheds for hydrologic modeling, the main river channel, weather station and quality monitoring station. The land use percentage in the Doam dam watershed shows that forest and agricultural fields occupy areas of 61.96% and 16.53%, respectively, as of 2007 (Table 1). The surface slope between 7 and 30% occupies 65.5% of the area, which illustrates the steep mountainous topographical characteristics of the watershed (Table 2).
Table 1

Land use in the Doam dam watershed (MoE 2007a)

  Agricultural area
 
     
Variety Built-up area Rice paddy Upland crop Other Forest area Sward Waters Others Total 
Area (ha) 277.0 175.7 2,470.0 2,092.5 9,258.1 329.6 164.7 174.4 14,942.0 
Percentage (%) 1.85 1.18 16.53 14.00 61.96 2.21 1.10 1.17 100.00 
  Agricultural area
 
     
Variety Built-up area Rice paddy Upland crop Other Forest area Sward Waters Others Total 
Area (ha) 277.0 175.7 2,470.0 2,092.5 9,258.1 329.6 164.7 174.4 14,942.0 
Percentage (%) 1.85 1.18 16.53 14.00 61.96 2.21 1.10 1.17 100.00 
Table 2

Area percentage by surface slope of upland crop in Doam dam watershed (MoE 2007b)

Total < 7% 7–15% 15–30% > 30% 
2,470.1 ha (100%) 837.4 ha (33.9%) 1,069.3 ha (43.3%) 542.2 ha (22.2%) 15.2 ha (0.6%) 
Total < 7% 7–15% 15–30% > 30% 
2,470.1 ha (100%) 837.4 ha (33.9%) 1,069.3 ha (43.3%) 542.2 ha (22.2%) 15.2 ha (0.6%) 
Figure 1

Meteorological and hydro-environmental measurement sites in the Doam dam watershed.

Figure 1

Meteorological and hydro-environmental measurement sites in the Doam dam watershed.

Climate change scenarios

The runoff and water pollution loads at the reservoir are computed using a SWAT model. The climate data used in the SWAT model include the rainfall (mm), maximum and minimum temperatures (°C), solar radiation quantity (MJ/m2), wind velocity (m/sec) and humidity (%). The input data of the reservoir inflows estimated using the reservoir stage-volume curve are used for calibration and validation. The calibration and validation of SWAT modeling was implemented from January 1, 1991 to December 31, 2004 and January 1, 2006 to December 31, 2008 using the meteorological data provided by Daegwallyeong station.

This study predicted the basin-wide climate change using RCM projection under IPCC's A1B scenario, which is regarded as a moderate projection among the IPCC's emission scenarios. Before being used for the impact models, the original GCM ECHO-G output was downscaled by the Korea Meteorological Administration's MM5 RCM. The A1B scenario is a moderate scenario among the IPCC's AR4 scenarios and has been chosen as representative for the regional impact studies in Korea (Ahn et al. 2009). To correct regional biases of the original RCM scenario, the ANN model was established. The suggested ANN has a 3-layer perceptron and back-propagation learning algorithm. The rainfall (mm), maximum and minimum temperature (°C), and humidity (%) simulated by the regional climate model were used as predictors and the observed rainfall at Daegwallyeong station was used as a predictand for the ANN model. The ANN learning process was performed from January 1991 to December 2005, and the ANN verification process was performed from January 2006 to December 2010. The ANN predictors selected through the sensitivity analysis were used to establish a final ANN model for the Doam dam watershed. Through the ANN model, the daily temperature, precipitation and humidity from 2011 to 2100 were predicted. The watershed-wide prediction data of the climate change scenario were input data for the SWAT modeling and were used to estimate available water resources during the projection periods. Figure 2 shows a schematic diagram of this study.
Figure 2

Schematic diagram of this study.

Figure 2

Schematic diagram of this study.

Geomorphologic characteristics of the watershed

Characterizing the watershed geomorphology is a critical step in the simulation of runoff and pollutant transport within a watershed. The geomorphology consists of watershed characteristics such as topography (slope, aspect, and elevation), land use, and soil types. Hydrologic geomorphologic modeling was carried out using watershed boundary, stream network, and DEM. The watershed boundary and stream network are provided by the National Geographic Information Institute. The DEM layer was prepared at 30 × 30 m resolution. Accurate land use data were essential for accurately estimating the impacts of NPS pollution within the Doam dam watershed due to its rapid population growth and land use change. The land use map (1/25,000 scale in the shape polygon format) from the Environmental Geographic Information System and a detailed soil map (1/25,000 vector map) from the National Institute of Agricultural Science and Technology with land use and soil tables produced in 2005 were used in this study (Figure 3).
Figure 3

Thematic map of geographic information system.

Figure 3

Thematic map of geographic information system.

The sub-watersheds were further divided into hydrological response units (HRUs) consisting of land use and soil types with similar attributes (Manguerra & Engel 1998; USEPA 2001). The division of sub-watersheds enables the modeler to separate the effects of the major components in soil and land use type for each HRU from those of topography and rainfall distribution. Runoff is predicted for each HRU and routed to obtain the total runoff for each of the 21 spatial sub-watersheds (USEPA 2001). This method enhances the accuracy over the single HRU method and gives a better physical description of the water balance (Manguerra & Engel 1998). Sub-basin delineations and HRU definition are required before parameterizing and executing the model. The Doam dam watershed was divided into 21 sub-basins based on DEM and the location of the stream gauges.

Water quality data

The change in water quality concentration in Song stream during the 2000–2009 period was analyzed using the data from the water quality monitoring stations managed by the Ministry of Environment (Figure 4). For runoff data at the Doam outlet, the daily data provided by Water Resources Management Information System were used. For water quality data at Song-cheon 1 site, SS (suspended solids), total phosphorus (T-P), and T-N (total nitrogen), measured monthly and transformed into loads considering the collection dates, were used for model calibration and validation. Song-cheon 1 is located upstream of Doam dam, and Song-cheon 2 is located downstream of Doam dam.
Figure 4

Time series plot of water contamination loads in Doam dam watershed (KWF 2011).

Figure 4

Time series plot of water contamination loads in Doam dam watershed (KWF 2011).

Effective planning of water resource use and protection under changing environments requires a watershed runoff model that can simulate hydro-environmental flow regimes under IPCC's different emission scenarios. The in-stream water quality monitoring data are necessary for water quality model calibration and validation. Table 3 shows the available hydrologic and water quality data in the Doam dam watershed.

Table 3

Hydrologic and water quality data

Item Gauging station Collection period Source 
Hydrologic Dam-inflow (m3/sec) Doam dam watershed outlet 1991–2010 K-Water 
Water quality SS (mg/L) Song stream #2 1994–2009 Water Information System 
T-N (mg/L) 
T-P (mg/L) 
Item Gauging station Collection period Source 
Hydrologic Dam-inflow (m3/sec) Doam dam watershed outlet 1991–2010 K-Water 
Water quality SS (mg/L) Song stream #2 1994–2009 Water Information System 
T-N (mg/L) 
T-P (mg/L) 

Land-use change scenario

The land-use changes in the Doam dam watershed during 1985 and 2000 are depicted and summarized in Figure 5 and Table 4. The land-use changes in the Doam dam watershed indicates that urban land, bare land and grassland have increased, but forest, paddy field and upland crop have decreased. The increase was the most significant for the grassland and urban land areas. Large-scale farm and resort facilities for ski resorts and golf courses have increased, too.
Table 4

Land-use changes in the Doam dam watershed

  Area (km2) (%)
 
Year Water Urban Bare field Grassland Forest Paddy field Upland crop 
1985 0.81 (0.55) 0.12 (0.08) 5.84 (3.95) 7.00 (4.73) 109.37 (73.92) 24.12 (16.30) 0.70 (0.47) 
2000 0.81 (0.55) 0.32 (0.22) 7.12 (4.81) 7.75 (5.24) 108.50 (73.32) 22.84 (15.43) 0.64 (0.43) 
  Area (km2) (%)
 
Year Water Urban Bare field Grassland Forest Paddy field Upland crop 
1985 0.81 (0.55) 0.12 (0.08) 5.84 (3.95) 7.00 (4.73) 109.37 (73.92) 24.12 (16.30) 0.70 (0.47) 
2000 0.81 (0.55) 0.32 (0.22) 7.12 (4.81) 7.75 (5.24) 108.50 (73.32) 22.84 (15.43) 0.64 (0.43) 
Figure 5

Land use changes in the Doam dam watershed between 1985 and 2000.

Figure 5

Land use changes in the Doam dam watershed between 1985 and 2000.

The land-use changes in the Doam dam watershed in the past were analyzed from 1990 to 2000 (Figure 6). During this period, forest and upland crop areas decreased 3–5%, whereas urban areas and bare fields increased 15–25% and specifically increased by 158% in 2000.
Figure 6

Changing rate by land use attributes in the Doam dam watershed.

Figure 6

Changing rate by land use attributes in the Doam dam watershed.

The Doam dam watershed has not been an area of dramatic land-use change up to recently. However, in the future, activities accompanying land-use change are expected in the area due to the construction of urban infrastructures and complexes in preparation for the 2018 Pyeong-Chang Winter Olympics and other future events (MoE 2007a). The expanded urban area constructed in the vegetated areas, e.g., forest, pasture and upland agricultural area, can accelerate the hydro-environmental change in the region.

The forestry that occupies the majority of the Doam dam watershed area is currently under development to build the infrastructures and facilities for the 2018 Pyeong-Chang Winter Olympics. The impact on and sensitivity to the soil erosion from land use changes were implemented under the following five different land use scenarios: partial change of forest to urban (PCFUr), partial change of forest to bare field (PCFB), partial change of forest to grassland (PCFG), partial change of forest to upland crop (PCFUp), and partial change of forest to agriculture (PCFA).

Model description

ANN

ANN is a family of statistical learning algorithms that are used to estimate or approximate functions that can depend on a large number of inputs. Arrangements of computational nodes form the ANN, which has very simple neuron-like processing elements, weighted connections between the processing elements, highly parallel processing and distributed control and automatic learning of internal representations. Typically, the sums of each node are weighted, and the sum is passed through a non-linear function known as an activation function or transfer function (Haykin 1994).

The neural networks for the flood periods (June–October) and the non-flood periods (November–May) were constructed separately, which enhances the ANN training performance by separating rainfalls caused by typhoons from other types of rainfalls to use them for learning in the neural networks. However, it is more or less impractical to train the ANN with daily data, which shows considerably low performance because of the high variability of daily precipitation. Therefore, daily accumulated values were converted to monthly accumulated values to increase the ANN training and validation performance. The final daily projection can be obtained by transforming monthly data into daily data. In the cases of the highest and lowest temperatures, the learning period was adjusted to be ten years (1996–2005) to reduce the calculation loads. The validation results for the ANN downscaling model are shown in Figure 7 with a 0.65 correlation coefficient between observation and the downscaled RCM outputs that were improved from that between observation and the original RCM outputs.
Figure 7

Validation of ANN modeling.

Figure 7

Validation of ANN modeling.

Non-stationary quantile mapping

To remove systematic biases, quantile mapping is used to match the simulated variable to the observed variable so that the empirical probability distributions for the two data sets would have equivalent distributions (Hashino et al. 2006). 
formula
1
where is the cumulative distribution function of the observed monthly volumes and is the cumulative distribution function of the corresponding simulation.

In essence, the quantile mapping approach replaces the simulated ensemble volume with the observed flow that has the same non-exceedance probability. Instead of fitting a mathematical model to the cumulative distribution functions, we use a simple one-to-one mapping of order statistics of the observed and simulated monthly volumes from the historical record. A corrected ensemble volume is obtained from interpolation between the order statistic pairs. Similar approaches have been used by Leung et al. (1999), Wood et al. (2002), and others to correct biases in temperature and precipitation forecasts from atmospheric models for hydrologic forecasting. In this research, a non-stationary trend in the hydrologic time series is expected due to climate change, so non-stationary quantile mapping with parameters varying with time was applied to produce a more realistic projection. The monthly data were disaggregated into daily data using the change factor (CF) method.

The CF method can be applied only when equivalent observational and GCM data exist. For example, CFs could not be used to directly estimate changes in significant wave heights or air quality without recourse to an intermediate model. Furthermore, because CFs are calculated for specific time-slices, the method cannot be used to explore transient changes in the local climate scenario. However, key advantages of the monthly CF approach are the ease and speed of application and the direct scaling of the scenario in line with changes suggested by the GCM or RCM (Arnell & Reynard 1996; Pilling & Jones 1999; Diaz-Nieto & Wilby 2005). This paper explores the relative merits of the CF (Table 5).

Table 5

Relative strengths and weaknesses of the CF method for climate scenario generation

Scenario technique Strengths Weaknesses 
CFs 
  • Station-scale scenarios

  • Computationally straightforward and quick to apply

  • Local climate change scenario is directly related to changes in the regional climate model output

 
  • Depends on realism of the climate model providing the CFs

  • Temporal structure is unchanged for future climate scenarios

  • Step changes in scaling at the monthly interface

  • Restricted to time-slice scenarios

 
Scenario technique Strengths Weaknesses 
CFs 
  • Station-scale scenarios

  • Computationally straightforward and quick to apply

  • Local climate change scenario is directly related to changes in the regional climate model output

 
  • Depends on realism of the climate model providing the CFs

  • Temporal structure is unchanged for future climate scenarios

  • Step changes in scaling at the monthly interface

  • Restricted to time-slice scenarios

 

SWAT model

The SWAT model incorporates features of several Agricultural Research Service models and is a direct outgrowth of the Simulator for Water Resources in Rural Basins model (Williams et al. 1985; Arnold et al. 1990). SWAT is an operational or conceptual model that operates on a daily time step. The objective in model development was to predict the impact of management on water, sediment and agricultural chemical yields in large ungauged watersheds. To satisfy the objectives, the model: (a) does not require calibration (calibration is not possible for ungauged watersheds); (b) uses readily available inputs for large areas; (c) is computationally efficient to operate on large basins in a reasonable time; and (d) is continuous in time and capable of simulating long periods for computing the effects of management changes (Arnold & Fohrer 2005). The SWAT model uses the Modified Universal Soil Loss Equation for estimating sediment loads from land surfaces. 
formula
2
where Qsed is the sediment yield on a given day (tons/acre), Q is the surface runoff volume (m3), qp is the peak runoff (m3/sec), KUSLE is the soil erodibility factor, LSUSLE is the topographic factor, CUSLE is the cover management factor, and PUSLE is the support practice factor.
In large sub-basins with a time of concentration greater than one day, only a portion of the surface runoff will reach the main channel on the day it is generated. The SWAT model incorporates a surface runoff storage feature to lag a portion of the surface runoff release to the main channel. Sediment in the surface runoff is lagged as well. Once the sediment load in the surface runoff is calculated, the amount of sediment released to the main channel is calculated (Arnold & Fohrer 2005): 
formula
3
where sed is the amount of sediment discharged to the main channel on a given day (tons), sed′ is the amount of sediment load generated in the HRU on a given day, sedstor,i−1 is the sediment stored or lagged from the previous day, surlag is the surface runoff lag coefficient, and tconc is the time of concentration for the HRU (hrs).
The SWAT model allows the lateral and groundwater flow to contribute sediment to the main channel. The amount of sediment contributed by lateral and groundwater flow is then calculated (Arnold & Fohrer 2005): 
formula
4
where sedlat is the sediment loading in the lateral and groundwater flow (metric tons), Qlat is the lateral flow for a given day (mm H2O), Qgw is the groundwater flow for a given day (mm H2O), areahru is the area of the HRU (km2), and concsed is the concentration of sediment in the lateral and groundwater flow (mg/L).

RESULTS

Trend of climate change scenarios

The variables were projected by applying an ANN bias correction model with a monthly base from 2011 until 2100, and then the trends for the non-flood periods (January–May, October–December) and the flood periods (June–October) were analyzed (Table 6). According to the trend analysis results, we can see that the projected hydrologic values show a general tendency to increase more or less during 2011–2100. In the case of precipitation in particular, the values are shown to increase by 76% during the flood periods of 2100 compared to 2011. The monthly variation in Figure 8 shows that the precipitation peaks shifted from July in baseline to August in projection. The humidity decreased during the fall and winter seasons in the more distant future, but the differences are insignificant. Other factors, such as the highest and lowest temperatures, were found to increase more in the distant future. However, they did not increase significantly compared to the precipitation. The trend analysis results regarding the prediction results for the Doam dam watershed are shown in Table 6.
Figure 8

Trend analysis of monthly prediction of hydrologic components for ANN predictors.

Figure 8

Trend analysis of monthly prediction of hydrologic components for ANN predictors.

Table 6

Projection of hydrologic components for non-flood and flood seasons

  Season classification Baseline 2011–2040 2041–2070 2071–2100 
Precipitation (mm/month) Non-flood season 57.4 58.8 62.1 66.1 
Flood season 215.9 242.1 275.4 311.2 
Humidity (%) Non-flood season 48.0 48.2 48.3 48.3 
Flood season 58.2 58.4 59.0 59.8 
Maximum temperature (°C) Non-flood season 13.7 13.9 14.7 15.7 
Flood season 26.8 27.1 27.8 28.5 
Minimum temperature (°C) Non-flood season 0.6 0.8 1.5 2.3 
Flood season 15.8 16.5 17.6 18.6 
  Season classification Baseline 2011–2040 2041–2070 2071–2100 
Precipitation (mm/month) Non-flood season 57.4 58.8 62.1 66.1 
Flood season 215.9 242.1 275.4 311.2 
Humidity (%) Non-flood season 48.0 48.2 48.3 48.3 
Flood season 58.2 58.4 59.0 59.8 
Maximum temperature (°C) Non-flood season 13.7 13.9 14.7 15.7 
Flood season 26.8 27.1 27.8 28.5 
Minimum temperature (°C) Non-flood season 0.6 0.8 1.5 2.3 
Flood season 15.8 16.5 17.6 18.6 

Calibration and validation of runoff and sediment in Doam dam watershed

The Doam dam watershed has 149.42 km2 of area, and the national monitoring network data are not sufficient. Thus, a spatially lumped parameter approach was tried, and basin-wide calibration was carried out using the outlet point of the watershed. The simulation period for model stabilization was set from 1990 to 2010. The periods for calibration and verification of the model were set to two years from 2003 to 2004 and to three years from 2006 to 2008 during which the quantity and quality of measurement data were sufficient. The land use map used in this research is provided from year 2000 data. Major land use changes due to preparing for the Winter Olympics are expected to start after 2011, the year the hosting was confirmed. The land use changes before 2011 are insignificant. Therefore, the temporally changing factors during the calibration and validation processes were neglected.

The calibration of the model was performed using the parameters selected through Latin hypercube one factor at a time sensitivity analysis. In general, CN2 (coefficient of roughness) is the most sensitive, followed by ESCO (soil evaporation compensation factor) and SOL_AWC (available water capacity of the soil layer), as parameters related to the surface water flow in the SWAT models. GWQMN (threshold depth of water in the shallow aquifer required for return flow to occur), GW_REVAP (groundwater ‘revap’ coefficient), REVAPMN (threshold depth of water in the shallow aquifer for percolating to the deep aquifer to occur) and other factors are sensitive as parameters related to the groundwater flow. ALPHA_BF (base alpha factor) and others are known to be sensitive as recession curve parameters. The previous study for the upland area with similar geographic and hydro-climatic characteristics was referred to for the model sensitivity analysis (Heo et al. 2005). The calibrated set of the parameters are summarized in Table 7.

Table 7

Hydrologic and sediment parameters of SWAT

  Input file Parameter Definition Reference range Calibrated value 
Flow .gw ALPHA_BF Baseflow recession constant 0–1 0.5 
 GW_DELAY Delay time for aquifer recharge 0–500 50 
.hru ESCO Soil evaporation compensation coefficient 0–1 0.1 
.mgt CN2 SCS runoff curve number 35–98 −0.5 
.sol SOL_AWC Available water capacity of the soil layer 0–1 +0.05 
Sediment .rte CH_EROD Channel erodibility −0.05–0.6 −0.02 
.hru LET_SED Sediment concentration in alteral flow and groundwater flow 0–5000 
.mgt USLE_P USLE equation support practice (P) factor 0.1-–1 0.1 
  Input file Parameter Definition Reference range Calibrated value 
Flow .gw ALPHA_BF Baseflow recession constant 0–1 0.5 
 GW_DELAY Delay time for aquifer recharge 0–500 50 
.hru ESCO Soil evaporation compensation coefficient 0–1 0.1 
.mgt CN2 SCS runoff curve number 35–98 −0.5 
.sol SOL_AWC Available water capacity of the soil layer 0–1 +0.05 
Sediment .rte CH_EROD Channel erodibility −0.05–0.6 −0.02 
.hru LET_SED Sediment concentration in alteral flow and groundwater flow 0–5000 
.mgt USLE_P USLE equation support practice (P) factor 0.1-–1 0.1 

The determination coefficient (R2) and the root mean square error (RMSE) were used as objective functions to judge the feasibility of the simulation and correlation of the models, and the Nash Sutcliffe Efficiency (NSE) coefficient suggested by Nash & Sutcliffe (1970) was used to validate the model efficiency. The selected performance indices have been used in a number of previous studies for domestic dam watersheds that applied SWAT for runoff estimation (Choi et al. 2009; Park et al. 2009; Kang et al. 2013).

The results of model performance during the calibration and verification periods are shown in Table 8. There are various suggestions for the criteria for judgment indices. For a reasonable range for R2 and NSE, 0.5 or higher and 0.4 or higher, respectively, were suggested by Green et al. (2006), and 0.6 or higher and 0.5 or higher, respectively, were suggested by Ramanarayanan et al. (1997) and Santhi et al. (2001a, 2001b). Considering these criteria, the model performance for runoff at the Doam dam site can be regarded as reasonable based on the performance indices of R2 and NSE being 0.82 and 0.66 during the calibration period and 0.81 and 0.73 during the validation period, respectively (Table 8). The RMSEs during the calibration and verification periods were 6.04 and 4.34 mm/day, respectively. Figure 9 shows the time series plot of the simulated and observed inflow of the Doam dam reservoir, and Figure 9 shows the calibration and validation of the SS (suspended sediments). During the calibration (1999–2000) and validation (2008–2009) periods, R2 was 0.39 and 0.53, respectively, and NSE was 0.85 and 0.93. The coefficient of runoff was influenced for heavy rainfall at the flood season for the rainfall characteristics in Korea. We predicted a whole trend of annual runoff change in this study. The observed coefficient of runoff is somewhat different than the simulated coefficient of runoff at calibration and validation. However, when the validity of the calibration and validation is determined by R2, RMSE, and NSE, the coefficient of runoff is deemed suitable to apply the SWAT model to the Doam dam watershed.
Figure 9

Calibration and validation results at Doam dam water quality. (a) Calibration (Obs. vs Sim.), (b) Validation (Obs. vs Sim.).

Figure 9

Calibration and validation results at Doam dam water quality. (a) Calibration (Obs. vs Sim.), (b) Validation (Obs. vs Sim.).

Table 8

Model performance at the Doam dam site

Runoff
 
  Discharge (mm)
 
Runoff ratio (%)
 
   
Year Precipitation (mm) Obs. Sim. Obs. Sim. R2 RMSE (mm/day) NSE 
Calibration 
2003 2,685.6 1,700.8 2,404.8 63.3 89.5 0.79 7.40 0.53 
2004 1,815.5 1,302.0 1,470.2 71.7 81.0 0.84 4.68 0.79 
Average 2,250.6 1,501.4 1,937.5 67.5 85.3 0.82 6.04 0.66 
Validation 
2006 2,112.9 1,410.2 1,714.7 66.74 81.16 0.80 6.63 0.79 
2007 1,401.1 961.4 1,065.1 68.62 76.02 0.80 2.72 0.80 
2008 1,128.6 517.9 753.5 45.89 66.77 0.83 3.57 0.58 
Average 1,547.5 963.2 1,177.8 60.4 74.7 0.81 4.30 0.73 
Sediment
 
Year  Corr.  R2  NSE   
Calibration 
1999  0.99  0.98  0.82   
2000  0.99  0.97  0.88   
Average  0.99  0.97  0.85   
Validation 
2008  0.95  0.91  0.83   
2009  0.99  0.98  0.83   
Average  0.98  0.95  0.93   
Runoff
 
  Discharge (mm)
 
Runoff ratio (%)
 
   
Year Precipitation (mm) Obs. Sim. Obs. Sim. R2 RMSE (mm/day) NSE 
Calibration 
2003 2,685.6 1,700.8 2,404.8 63.3 89.5 0.79 7.40 0.53 
2004 1,815.5 1,302.0 1,470.2 71.7 81.0 0.84 4.68 0.79 
Average 2,250.6 1,501.4 1,937.5 67.5 85.3 0.82 6.04 0.66 
Validation 
2006 2,112.9 1,410.2 1,714.7 66.74 81.16 0.80 6.63 0.79 
2007 1,401.1 961.4 1,065.1 68.62 76.02 0.80 2.72 0.80 
2008 1,128.6 517.9 753.5 45.89 66.77 0.83 3.57 0.58 
Average 1,547.5 963.2 1,177.8 60.4 74.7 0.81 4.30 0.73 
Sediment
 
Year  Corr.  R2  NSE   
Calibration 
1999  0.99  0.98  0.82   
2000  0.99  0.97  0.88   
Average  0.99  0.97  0.85   
Validation 
2008  0.95  0.91  0.83   
2009  0.99  0.98  0.83   
Average  0.98  0.95  0.93   
Table 8 shows definitely the annual average of simulated discharge is higher than observation. This is because some highly simulated flood flows tend to impact on the overall performance for the whole flow regime including the low flows which dominate in terms of occurring periods. Considering the discharges of 50% and 90% exceeding probability, the observed ones were 3.3 and 1.3 m3/s and the simulated ones were 4.1 and 0.7 m3/s for the calibration period (2003–2004), which shows reasonable agreements considering the uncertainties from various sources in the long-term continuous streamflow simulation (Figure 10).
Figure 10

Calibration and validation results at Doam dam runoff. (a) Runoff calibration, (b) Runoff validation, (c) Sediment yield calibration, (d) Sediment yield validation.

Figure 10

Calibration and validation results at Doam dam runoff. (a) Runoff calibration, (b) Runoff validation, (c) Sediment yield calibration, (d) Sediment yield validation.

Compared with observed data, the simulated stream flow has a high correlation with the observed data. The results are shown in Figure 10. The correlation of the calibration and validation were 0.804 and 0.781, respectively.

Predicting change and variability of runoff and water quality under climate change scenarios

In this study, the climate projection data downscaled by the ANN model were applied to the SWAT model to evaluate the potential impacts on surface runoff and non-point pollution loads under various climate change scenarios. The projection of monthly and seasonal values generated from the daily projection data were suggested from 2011 to 2100, grouped by 30-year periods. The prospective environmental changes in the watershed, such as future land use and vegetation changes, were addressed separately. The future changes in the hydro-environmental components due to climate change are expected to be significant with respect to not only hydrologic surface runoff itself but also non-point pollutant runoff (Lee et al. 2012).

Figure 11 and Table 9 show the monthly inflow changes under the IPCC A1B emission scenario. The runoff projection shows the highest increasing ratio in the future during August and September. The increasing ratios themselves are high during the winter season, too, but relatively small runoff does not grant much confidence in those results. Figure 12 and Table 8 illustrate the future variation of hydrologic components with reference to the baseline (1990–2010) under the A1B scenario. The annual precipitation shows a 9.5% increase in the near future (2011–2040) and a 36.9% increase in the long-term future period (2070–2100). However, it is likely that the projection for the long-term future has a higher degree of uncertainty. The variation trend for hydrologic runoff including direct runoff and baseflow shows a pattern similar to precipitation. The hydrologic runoff is expected to increase approximately 10% in the near future (2011–2040). However, the projection for evapotranspiration estimated by the SWAT model shows a relatively low increase (Table 10). The physical interpretation for that phenomenon would require understanding the micro and macroscopic evapotranspiration process and accurate modeling. The complementary relationship and advection-aridity model were one of the trials to explain the two-way interaction between evapotranspiration and atmospheric humidity and their impacts on the relationship between actual and potential evapotranspiration (Hobbins et al. 2001a, 2001b). Based on the projection results, water quality problems could be expected to occur due to increasingly turbid water flowing into the reservoir water body as a result of increased rainfall intensity during the flood season, leading to increased sediment detachment from land surfaces and the upland agricultural area in particular. The turbid water in the reservoir tends to be deposited and discharged downstream along spillways or outlets, causing a high degree of suspended solids (SS) along the stream. Table 11 shows a seasonal projection of monthly sediment under the A1B scenario with respect to the baseline. The increasing ratio in July, when the amount of sediment product is greatest, is 25.9% during the near future period and 113.2% during the long-term future period.
Figure 11

Projection of monthly dam inflow under climate change scenarios.

Figure 11

Projection of monthly dam inflow under climate change scenarios.

Figure 12

Projection by hydrologic components under climate change scenarios.

Figure 12

Projection by hydrologic components under climate change scenarios.

Table 9

Projection of monthly dam inflow by periods and seasons under climate change scenarios

  1990–2010 (Baseline)
 
2011–2040
 
2041–2070
 
2071–2100
 
Month  Inflow (m3/s) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) 
Winter 12 39.0 69.9 +79.0 97.4 +149.6 150.6 +285.7 
14.0 23.5 +68.2 38.2 +173.7 86.8 +522.0 
19.4 17.7 −9.1 44.8 +130.4 123.2 +534.2 
Mean 24.1 37.0 +46.0 60.1 +151.2 120.2 +447.3 
Spring 161.7 162.5 +0.4 192.9 +19.3 210.2 +30.0 
187.6 200.1 +6.7 176.3 −6.0 195.1 +4.0 
221.0 180.8 −18.2 186.8 −15.5 231.5 +4.8 
Mean 190.1 181.1 −3.7 185.3 −0.7 212.3 +12.9 
Summer 356.7 276.6 −22.5 295.3 −17.2 492.2 +38.0 
478.9 600.0 +25.3 744.8 +55.5 930.8 +94.4 
339.5 812.5 +139.3 982.2 +189.3 1,374.7 +304.9 
Mean 391.7 563.0 +47.4 674.1 +75.9 932.6 +145.8 
Autumn 172.6 397.0 +130.0 497.3 +188.1 903.4 +423.3 
10 103.5 142.0 +37.2 195.4 +88.9 322.6 +211.8 
11 39.0 174.4 +346.8 185.4 +374.9 249.2 +538.3 
Mean 105.0 237.8 +171.3 292.7 +217.3 491.7 +391.1 
Annual 2,133.1 3,057.0 +43.3 3,636.9 +70.5 5,270.4 +147.1  
  1990–2010 (Baseline)
 
2011–2040
 
2041–2070
 
2071–2100
 
Month  Inflow (m3/s) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) 
Winter 12 39.0 69.9 +79.0 97.4 +149.6 150.6 +285.7 
14.0 23.5 +68.2 38.2 +173.7 86.8 +522.0 
19.4 17.7 −9.1 44.8 +130.4 123.2 +534.2 
Mean 24.1 37.0 +46.0 60.1 +151.2 120.2 +447.3 
Spring 161.7 162.5 +0.4 192.9 +19.3 210.2 +30.0 
187.6 200.1 +6.7 176.3 −6.0 195.1 +4.0 
221.0 180.8 −18.2 186.8 −15.5 231.5 +4.8 
Mean 190.1 181.1 −3.7 185.3 −0.7 212.3 +12.9 
Summer 356.7 276.6 −22.5 295.3 −17.2 492.2 +38.0 
478.9 600.0 +25.3 744.8 +55.5 930.8 +94.4 
339.5 812.5 +139.3 982.2 +189.3 1,374.7 +304.9 
Mean 391.7 563.0 +47.4 674.1 +75.9 932.6 +145.8 
Autumn 172.6 397.0 +130.0 497.3 +188.1 903.4 +423.3 
10 103.5 142.0 +37.2 195.4 +88.9 322.6 +211.8 
11 39.0 174.4 +346.8 185.4 +374.9 249.2 +538.3 
Mean 105.0 237.8 +171.3 292.7 +217.3 491.7 +391.1 
Annual 2,133.1 3,057.0 +43.3 3,636.9 +70.5 5,270.4 +147.1  
Table 10

Hydrologic water balance at the Doam dam

  Annual. prec. (mm) Total runoff (mm) Direct runoff (mm) Base F. (mm) E.T. (mm) 
1991–2010 1,486.1 1,037.7 792.8 242.7 385.2 
2011–2040 1,627.3 (9.5%) 1,162.7 (12.0%) 881.3 (11.2%) 278.0 (14.5%) 392.7 (1.9%) 
2041–2070 1,817.3 (22.3%) 1,317.2 (26.9%) 1,002.9 (26.5%) 308.9 (27.3%) 420.9 (9.3%) 
2071–2100 2,033.8 (36.9%) 1,486.4 (43.2%) 1,143.9 (44.3%) 334.3 (37.7%) 461.5 (19.8%) 
  Annual. prec. (mm) Total runoff (mm) Direct runoff (mm) Base F. (mm) E.T. (mm) 
1991–2010 1,486.1 1,037.7 792.8 242.7 385.2 
2011–2040 1,627.3 (9.5%) 1,162.7 (12.0%) 881.3 (11.2%) 278.0 (14.5%) 392.7 (1.9%) 
2041–2070 1,817.3 (22.3%) 1,317.2 (26.9%) 1,002.9 (26.5%) 308.9 (27.3%) 420.9 (9.3%) 
2071–2100 2,033.8 (36.9%) 1,486.4 (43.2%) 1,143.9 (44.3%) 334.3 (37.7%) 461.5 (19.8%) 
Table 11

Seasonal projection of monthly sediment yield under IPCC A1B climate change scenario

  1990–2010 (Baseline)
 
2011–2040
 
2041–2070
 
2071–2100
 
Month  Sediment (ton) Sediment (ton) Change (%) Sediment (ton) Change (%) Sediment (ton) Change (%) 
Winter 12 383 516 +34.9 511 +33.5 1,484 +287.6 
320 429 +34.1 474 +48.1 554 +73.2 
675 385 −43.0 438 −35.2 413 −38.8 
Spring 498 435 −12.6 439 −11.8 389 −21.8 
874 882 +0.9 824 −5.7 741 −15.3 
735 828 +12.7 902 +22.7 1,067 +45.2 
Summer 2,705 2,017 −25.4 1,999 −26.1 2,654 −1.9 
5,332 6,713 +25.9 8,230 +54.3 11,370 +113.2 
8,090 7,760 −4.1 11,062 +36.7 14,160 +75.0 
Autumn 2,086 2,263 +8.5 2,999 +43.8 4,586 +119.8 
10 894 999 +11.8 1,269 +42.0 1,507 +68.6 
11 383 661 +72.8 732 +91.1 899 +134.9 
Annual 22,976 23,888 +4.0 29,881 +30.0 39,825 +73.3 
  1990–2010 (Baseline)
 
2011–2040
 
2041–2070
 
2071–2100
 
Month  Sediment (ton) Sediment (ton) Change (%) Sediment (ton) Change (%) Sediment (ton) Change (%) 
Winter 12 383 516 +34.9 511 +33.5 1,484 +287.6 
320 429 +34.1 474 +48.1 554 +73.2 
675 385 −43.0 438 −35.2 413 −38.8 
Spring 498 435 −12.6 439 −11.8 389 −21.8 
874 882 +0.9 824 −5.7 741 −15.3 
735 828 +12.7 902 +22.7 1,067 +45.2 
Summer 2,705 2,017 −25.4 1,999 −26.1 2,654 −1.9 
5,332 6,713 +25.9 8,230 +54.3 11,370 +113.2 
8,090 7,760 −4.1 11,062 +36.7 14,160 +75.0 
Autumn 2,086 2,263 +8.5 2,999 +43.8 4,586 +119.8 
10 894 999 +11.8 1,269 +42.0 1,507 +68.6 
11 383 661 +72.8 732 +91.1 899 +134.9 
Annual 22,976 23,888 +4.0 29,881 +30.0 39,825 +73.3 
Table 12

Change in runoff for land-use change scenarios

  2000 yr PCFUr
 
PCFB
 
PCFG
 
PCFUp
 
PCFA
 
  Inflow (m3/s) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) 
Dec 95.0 93.8 −1.24 93.8 −1.21 93.5 −1.50 71.2 −25.03 88.6 −6.69 
Jan 51.4 51.3 −0.15 51.1 −0.48 51.1 −0.51 41.1 −20.07 49.5 −3.71 
Feb 113.8 113.7 −0.09 113.7 −0.12 113.6 −0.21 115.3 +1.30 113.4 −0.34 
Mar 160.0 159.7 −0.22 159.7 −0.25 159.7 −0.24 152.1 −4.95 156.2 −2.38 
Apr 141.4 141.2 −0.17 141.0 −0.31 141.0 −0.27 140.1 −0.92 136.9 −3.18 
May 193.2 194.4 +0.64 194.1 +0.47 194.1 +0.50 200.0 +3.55 188.9 −2.20 
Jun 394.4 397.4 +0.79 396.8 +0.62 396.9 +0.65 412.1 +4.51 391.5 −0.73 
Jul 788.7 790.2 +0.18 789.2 +0.06 789.6 +0.11 820.6 +4.04 787.5 −0.16 
Aug 1,108.8 1,109.3 +0.05 1,108.4 −0.03 1,108.7 −0.01 1,132.1 +2.11 1,106.5 −0.20 
Sep 672.0 672.0 +0.00 671.8 −0.03 671.3 −0.11 662.2 −1.47 664.9 −1.06 
Oct 210.7 209.9 −0.42 210.6 −0.07 209.3 −0.69 173.9 −17.47 199.9 −5.16 
Nov 162.4 161.0 −0.87 161.4 −0.62 160.5 −1.16 143.3 −11.74 153.7 −5.32 
  2000 yr PCFUr
 
PCFB
 
PCFG
 
PCFUp
 
PCFA
 
  Inflow (m3/s) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) Inflow (m3/s) Change (%) 
Dec 95.0 93.8 −1.24 93.8 −1.21 93.5 −1.50 71.2 −25.03 88.6 −6.69 
Jan 51.4 51.3 −0.15 51.1 −0.48 51.1 −0.51 41.1 −20.07 49.5 −3.71 
Feb 113.8 113.7 −0.09 113.7 −0.12 113.6 −0.21 115.3 +1.30 113.4 −0.34 
Mar 160.0 159.7 −0.22 159.7 −0.25 159.7 −0.24 152.1 −4.95 156.2 −2.38 
Apr 141.4 141.2 −0.17 141.0 −0.31 141.0 −0.27 140.1 −0.92 136.9 −3.18 
May 193.2 194.4 +0.64 194.1 +0.47 194.1 +0.50 200.0 +3.55 188.9 −2.20 
Jun 394.4 397.4 +0.79 396.8 +0.62 396.9 +0.65 412.1 +4.51 391.5 −0.73 
Jul 788.7 790.2 +0.18 789.2 +0.06 789.6 +0.11 820.6 +4.04 787.5 −0.16 
Aug 1,108.8 1,109.3 +0.05 1,108.4 −0.03 1,108.7 −0.01 1,132.1 +2.11 1,106.5 −0.20 
Sep 672.0 672.0 +0.00 671.8 −0.03 671.3 −0.11 662.2 −1.47 664.9 −1.06 
Oct 210.7 209.9 −0.42 210.6 −0.07 209.3 −0.69 173.9 −17.47 199.9 −5.16 
Nov 162.4 161.0 −0.87 161.4 −0.62 160.5 −1.16 143.3 −11.74 153.7 −5.32 

Prediction of change and variability of runoff and water quality under land-use change scenarios

The impacts of land-use change were evaluated with respect to the land use in 2000. The urbanization in the future was assumed to change 1%, 3% and 5% of the forest area into urban area considering future development. Figure 13 shows that the annual mean runoffs due to climate change under the A1B scenario and 5% land use change are 3,500, 4,050, and 4,600 m3/s during the near future (2011–2040), mid-term future (2041–2070), and long-term future (2071–2100), respectively. The simulations illustrated that under the assumption of 1–5% urbanization of the forest area, the hydrologic impact is relatively negligible and the runoff variation due to climate change is more dominant than the urbanization impacts.
Figure 13

Combined impacts of climate and land-use change. (a) Land-use change, (b) Climate change.

Figure 13

Combined impacts of climate and land-use change. (a) Land-use change, (b) Climate change.

To analyze the effects of the parameter sensitivity on runoff related to six types of land use, the amounts of runoff were compared under five scenarios. Five land-use change scenarios were applied in this study. These scenarios were constructed assuming 20% of the forest area, which occupies the majority of the area at present, was replaced by urban land (PCFUr), PCFB, PCFG, PCFUp, and PCFA. Under each scenario, the runoff was projected and compared to that under the climate change scenario.

The runoff projections due to land-use changes are shown in Figure 14 and Table 12. In the PCFA scenario, the runoff decreases significantly during the non-flood periods in particular but increases during the flood periods. This result shows increased infiltration loss from the agricultural area. However, in the PCFUr scenario, the runoff increased during the flood season because of the increased surface runoff from the impermeable urban area. The PCFUp scenario shows the greatest variability from a 4.51% decrease up to a 25.03% increase, of which the seasonal pattern is similar to that of the PCFUr scenario, but the high runoff during flood season can be explained by the orographic rainfall increase in the mountainous area.
Figure 14

Runoff projections under land-use change scenarios.

Figure 14

Runoff projections under land-use change scenarios.

In the case of the PCFUp scenario, the highest runoff increase during the flood period increases sediment during the flood season. The 4.51% of the highest increasing runoff occurred during flood season along with a 73.57% sediment yield increase (Figure 15 and Table 13). The source of sediment increase cannot be explained solely by runoff increase due to the changes of upland crop area, but there would be close relationship between active sediment detachment and highly intense rainfall. The results illustrate that the sensitivity of the sediment yield would be significantly higher than that of the hydrologic surface runoff.
Figure 15

Sediment yield projection under land use change scenarios.

Figure 15

Sediment yield projection under land use change scenarios.

Table 13

Change in sediment yield for land-use change scenarios

  2000 yr PCFUr
 
PCFB
 
PCFG
 
PCFUp
 
PCFA
 
 Sediment yield (tons/d) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) 
Dec 433.7 230.3 −46.89 265.9 −38.69 323.2 −25.48 556.0 +28.22 344.6 −20.55 
Jan 324.4 332.1 +2.35 297.8 −8.20 329.8 +1.67 403.9 +24.49 346.8 6.89 
Feb 1,407.0 1,514.8 +7.66 1,306.5 −7.15 1,506.7 +7.08 1,685.4 +19.79 1,554.2 +10.46 
Mar 1,391.7 1,648.6 +18.45 1,405.0 +0.95 1,647.9 +18.40 1,870.9 +34.43 1,690.5 +21.47 
Apr 968.7 1,262.2 +30.30 1,154.3 +19.16 1,278.8 +32.01 1,681.4 +73.57 1,323.6 +36.63 
May 1,554.4 2,043.2 +31.45 1,763.8 +13.47 2,073.1 +33.37 2,691.9 +73.18 2,142.4 +37.83 
Jun 4,192.8 5,259.5 +25.44 4,645.2 +10.79 5,338.1 +27.32 6,176.6 +47.31 5,453.7 +30.07 
Jul 9,527.4 11,120.8 +16.72 10,653.3 +11.82 11,780.4 +23.65 13,302.5 +39.62 12,380.2 +29.94 
Aug 12,358.3 8,063.2 −34.75 8,774.1 −29.00 11,212.4 −9.27 17,251.2 +39.59 13,792.6 +11.61 
Sep 6,112.6 2,369.7 −61.23 2,759.5 −54.86 4,033.9 −34.01 8,861.3 +44.97 4,270.1 −30.14 
Oct 1,233.7 516.7 −58.12 588.7 −52.28 783.4 −36.50 1,991.8 +61.45 830.8 −32.66 
  2000 yr PCFUr
 
PCFB
 
PCFG
 
PCFUp
 
PCFA
 
 Sediment yield (tons/d) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) Sediment yield (tons/d) Change (%) 
Dec 433.7 230.3 −46.89 265.9 −38.69 323.2 −25.48 556.0 +28.22 344.6 −20.55 
Jan 324.4 332.1 +2.35 297.8 −8.20 329.8 +1.67 403.9 +24.49 346.8 6.89 
Feb 1,407.0 1,514.8 +7.66 1,306.5 −7.15 1,506.7 +7.08 1,685.4 +19.79 1,554.2 +10.46 
Mar 1,391.7 1,648.6 +18.45 1,405.0 +0.95 1,647.9 +18.40 1,870.9 +34.43 1,690.5 +21.47 
Apr 968.7 1,262.2 +30.30 1,154.3 +19.16 1,278.8 +32.01 1,681.4 +73.57 1,323.6 +36.63 
May 1,554.4 2,043.2 +31.45 1,763.8 +13.47 2,073.1 +33.37 2,691.9 +73.18 2,142.4 +37.83 
Jun 4,192.8 5,259.5 +25.44 4,645.2 +10.79 5,338.1 +27.32 6,176.6 +47.31 5,453.7 +30.07 
Jul 9,527.4 11,120.8 +16.72 10,653.3 +11.82 11,780.4 +23.65 13,302.5 +39.62 12,380.2 +29.94 
Aug 12,358.3 8,063.2 −34.75 8,774.1 −29.00 11,212.4 −9.27 17,251.2 +39.59 13,792.6 +11.61 
Sep 6,112.6 2,369.7 −61.23 2,759.5 −54.86 4,033.9 −34.01 8,861.3 +44.97 4,270.1 −30.14 
Oct 1,233.7 516.7 −58.12 588.7 −52.28 783.4 −36.50 1,991.8 +61.45 830.8 −32.66 

CONCLUSION

The future climate and land-use change impacts on the runoff and sediment yield at the Doam dam watershed were evaluated using GCM outputs downscaled by ANN and non-stationary quantile mapping model under IPCC's A1B scenario. The runoff and sediment yield were simulated using the SWAT model. The following major conclusions were derived from the present study.

Through ANN modelling and quantile mapping, the monthly variation of the hydrologic components of runoff and sediment yield from the present to 2100 was projected in the Doam dam watershed. The projection was carried out separately during flood and non-flood periods. During the flood season, most predictors including rainfall, relative humidity, and maximum and minimum temperatures showed increasing patterns of precipitation in the watershed.

Calibration and validation were conducted for the dam inflow at the outlet point of the watershed. The runoff at the Doam dam watershed was sensitive to CN2 (coefficient of roughness) and soil evaporation compensation factor (ESCO) as surface water parameters.

The evaluation of the correlation coefficient showed that NSE during the calibration and validation periods were 0.66 and 0.73, respectively, satisfying the 0.5 threshold. The R2 were at 0.82 and 0.81, respectively, satisfying the over 0.6 condition. Because the urbanization impacts before 2011 are insignificant, the temporally changing factors could be neglected in the calibration and validation processes.

Using the SWAT model, the impacts of climate change and urbanization were compared. The simulations illustrated that under the assumption of 1–5% urbanization of the forest area, the hydrologic impact is relatively negligible, and the climate change impacts govern over the urbanization impacts. However, the impacts of urbanization influence the sediment yield more than runoff.

Using five land-use change scenarios, the runoff variation by land uses and sensitivity to soil erosion were analyzed. The PCFUp scenario shows the greatest variability, from a 4.51% decrease up to a 25.03% increase in runoff, illustrating the high runoff during flood season explained by the orographic rainfall increase in the mountainous area. The 4.51% of the highest increasing runoff occurred during flood season along with a 73.57% sediment yield increase.

The disaggregation of the monthly RCM output into daily data for use as an input to the SWAT model enhanced model performance of the sediment yield computational process due to high-intensity storm rainfall. However, there is still a limitation, and the model requires a sub-hourly disaggregation scheme.

In the future, the discharge of pollutants in the Doam dam watershed can be predicted temporally and spatially under specific revised climate change scenarios through downscaling and SWAT model chain. Additionally, such a projection of coordinated model output may be used as data for sustainable management of the hydro-environment in the reservoirs.

ACKNOWLEDGEMENTS

This research was supported by a grant (14AWMP-B082564-01) from Advanced Water Management Research Program funded by Ministry of Land, Infrastructure and Transport of Korean government.

REFERENCES

REFERENCES
Ahn
S. R.
Park
M. J.
Park
G. A.
Kim
S. J.
2009
Assessing future climate change impact on hydrologic components of Gyeongancheon watershed
.
JKWRA
42
,
33
50
.
Arnold
J. G.
Williams
J. R.
Nicks
A. D.
Sammons
N. B.
1990
SWRRB: A Basin Scale Simulation Model for Soil and Water Resources Management
.
Texas A&M University Press
,
College Station, TX
.
Bates
B. C.
Kundzewicz
Z. W.
Wu
S.
Palutikof
J. P.
(eds)
2008
Climate Change and Water, Technical Paper of the Intergovernmental Panel on Climate Change
.
IPCC Secretariat
,
Geneva
.
210 pp
.
Carpenter
S. R.
Caraco
N. F.
Correll
D. L.
Howarth
R. W.
Sharpley
A. N.
Smith
V. H.
1998
Nonpoint pollution of surface waters with phosphorus and nitrogen
.
Ecological Applications
8
,
559
568
.
Choi
D.
Keem
M. S.
Kim
N. W.
Kim
S.
2009
An analysis of the effect of climate change on Byeongseong stream's hydrologic and water quality responses using CGCM's future climate information
.
JKWRA
42
,
921
931
.
Green
C. H.
Tomer
M. D.
Di Luzio
M.
Arnold
J. G.
2006
Hydrologic evaluation of the Soil and Water Assessment Tool for a large tile-drained watershed in Iowa
.
Transactions of the ASAE
49
,
413
4220
.
Hashino
T.
Bradley
A. A.
Schwartz
S. S.
2006
Evaluation of bias-correction methods for ensemble stream flow
.
Hydrology and Earth System Sciences Discussion
3
,
561
594
.
Haykin
S.
1994
Neural Networks: A Comprehensive Foundation
.
Macmillan
,
New York
.
Heo
S. G.
Kim
K. S.
Sagong
M.
Ahn
J. H.
Lim
K. J.
2005
Evaluation of SWAT applicability to simulate soil erosion at highland agricultural lands
.
Rural Planning
11
,
67
74
.
IPCC
2007
Climate Change 2007
.
The Physical Science Basis, IPCC Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change
,
Cambridge University Press
,
Cambridge
,
United Kingdom
and New York, NY, USA.
Jayakrishnan
R.
Srinivasan
R.
Santhi
C.
Arnold
J.
2005
Advances in the application of the SWAT model for water resources management
.
Hydrol. Processes
19
,
749
762
.
Joyce
L. A.
Aber
J.
McNulty
S.
Dale
V. H.
Hansen
A.
Irland
L. C.
Neilson
R. P.
Skog
K.
2000
Potential consequences of climate variability and change for the forests of the United States
. In:
National Assessment Synthesis Team, Comps
.
Climate change impacts on the United States: The potential consequences of climate variability and change: Foundation
.
Cambridge University Press
,
Cambridge
,
UK
, pp.
489
522
.
Korea Water Forum (KWF)
2011
Development of combined turbid water management system and plan suggestion of optimum water quality improvement for problem solution of turbid water in Doam Dam, 6–7.
Korea Water Forum, Pyeongchang-gun
,
Republic of Korea
.
Lee
J. M.
Kim
Y. D.
Kang
B. S.
Lee
H. S.
2012
Impact of climate change on runoff in Namgang dam watershed
.
Journal of Korea Water Resources Association
45
,
517
529
.
Leung
L. R.
Hamlet
A. F.
Lettenmaier
D. P.
Kumar
A.
1999
Simulations of the ENSO hydroclimate signals in the Pacific Northwest Columbia River Basin
.
Bull. Am. Meteorol. Soc.
80
,
2313
2328
.
Manguerra
H. B.
Engel
B. A.
1998
Hydrologic parameterizatin of watersheds for runoff prediction using SWAT
.
J. Am. Water Resour. Assoc.
34
,
1149
1162
.
Ministry of Environment (MoE)
2007a
Non-point Pollution Source Management Countermeasures in Doam Reservoir Watershed
,
4
.
Ministry of Environment
,
Gwacheon-si
,
Republic of Korea
.
Ministry of Environment (MoE)
2007b
The Study for Setting of Non-point Pollution Source Management Area Designation Criteria, 113
.
Ministry of Environment
,
Gwacheon-si
,
Republic of Korea
.
Nash
J. E.
Sutcliffe
J. V.
1970
River flow forcasting through conceptual models: Part I. A discussion of principles
.
Journal of Hydrology
10
,
283
290
.
Novotny
V.
Olem
H.
1994
Water Quality: Prevention, Identification, and Management of Diffuse Pollution
.
Van Nostrand Reinhold
,
New York
.
Park
J. Y.
Park
M. J.
Ahn
S. R.
Kim
S. J.
2009
Watershed modeling for assessing climate change impact on stream water quality of Chungju dam watershed
.
JKWRA
42
,
877
889
.
Pilling
C.
Jones
J. A. A.
1999
High resolution climate change scenarios: implications for British runoff
.
Hydrol. Processes
13
,
2877
2895
.
Ramanarayanan
T. S.
Williams
J. R.
Dugas
W. A.
Heuck
L. M.
McFarland
A. M. S.
1997
Using APEC to Identify Alternative Practices for Animal Waste Management
.
Minneapolis, MN
,
ASAE97-2209
.
Santhi
C.
Arnold
J. G.
Williams
J. R.
Dugas
W. A.
Srinivasan
R.
Hauck
L.
2001a
Validation of the SWAT model on a large river basin with point and nonpoint sources
.
Journal of the American Water Resources Association
37
,
1169
1188
.
Santhi
C.
Arnold
J. G.
Williams
J. R.
Hauck
L.
Dugas
W. A.
2001b
Application of a watershed model to evaluate management effects on point and nonpoint source pollution
.
Transactions of the ASAE
44
,
1559
1570
.
Setegn
S. G.
Srinivasan
R.
Melesse
A. M.
Dargahi
B.
2009
SWAT model application and prediction uncertainty analysis in the Lake Tana Basin, Ethiopia
.
Hydrol. Processes
23
,
3738
3750
.
Setegn
S. G.
Rayner
D.
Melesse
A. M.
Dargahi
B.
Srinivasan
R.
2011
Impact of climate change on the hydroclimatology of Lake Tana Basin, Ethiopia
.
Water Resour. Res.
47
,
W04511
.
doi:10.1029/2010WR009248
USEPA
2001
Better Assessment Science Integrating Point and Nonpoint Sources. Office of Science and Technology
.
EPA 823-B-01-001, USEPA
,
Washington, DC
.
Williams
J. R.
Nicks
A. D.
Arnold
J. G.
1985
Simulator for water resources in rural basins
.
ASCE Journal of the Hydraulics Division
111
,
970
986
.
Wood
A. W.
Maurer
E. P.
Kumar
A.
Lettenmaier
D. P.
2002
Long-range experimental hydrologic forecasting for the eastern United States
.
J. Geophys. Res. Atmos.
107
,
4429
.