3. RESULTS
3.1 SWAT Calibration
SWAT was calibrated to verify its applicability of simulating the
watershed hydrology of the Geum River basin. The SWAT calibration was
performed from 2006 to 2015 while 4 years (2002~2005)
served as the warm-up period. Calibration and validation were conducted
from 2006 to 2010 and from 2011 to 2015, respectively. The calibration
was carried out in daily time steps based on the observed inflow data at
five monitoring points, two multipurpose dams (YDD, DCD) and three
multifunction weirs (SJW, GJW, BJW). Besides, GA and FG condition of the
2010s were input to the SWAT during the calibration.
SWAT parameters adjusted to calibrate the model are the SCS curve
number, Manning‘s “n” value, soil evaporation compensation factor,
groundwater delay time, threshold depth for the return flow in the
shallow aquifer, baseflow recession constant, saturated hydraulic
conductivity, snowfall temperature, snow melt base temperature, and
hydraulic structure sources (Table 1). Default value was used for other
parameters that are not mentioned in Table 1.
[Insert Table 1]
The model performance was evaluated by coefficient of determination
(R2), Nash-Sutcliffe efficiency (NSE), root mean
square error (RMSE), and percent bias (PBIAS). The annual evaluation
results were averaged for the calibration, validation, and the total
period. Table 2 presented the statistical summary of dam calibrations.
The calibration result of the total period achieved R2values of 0.73 and 0.82, NSE values of 0.81 and 0.77, RMSE values of
2.35 mm/day and 1.57 mm/day, and PBIAS values of -2.51% and -8.44% for
YDD and DCD, respectively. Table 3 shows the statistical summary of weir
calibrations. The calibration result of the total period showed
R2 values of 0.79, 0.81, and 0.82; NSE values of 0.75,
0,75 and 0.77; RMSE values of 0.53 mm/day, 0.54 mm/day, and 0.58 mm/day;
and PBIAS values of 3.21%, 11.74%, and 10.73% for SJW, GJW, and BJW,
respectively. Graphical comparisons between observed data and calibrated
result of YDD, DCD, SJW, GJW, and BJW are shown in Figure 5.
[Insert Table 2]
[Insert Table 3]
[Insert Figure 5]
3.2 Hydrologic Responses to Groundwater Abstraction and
Forest Growth
Calibrated model ran 10-year hydrology of the target watershed. The
condition of watershed hydrology was explained by hydrological
components including evapotranspiration (ET), surface runoff (SR),
percolation (PE), soil moisture (SM), groundwater flow (GF),
infiltration (IN), groundwater recharge (GWR), and lateral flow (LF).
The hydrologic components in the 1980s were used as a standard data to
estimate hydrologic responses. Three scenarios were applied to
understand how GA and FG separately and concurrently affected watershed
hydrology. Scenario 1, 2, 3 considered decadal GA change (GA1990s,
GA2000s, GA2010s), decadal FG change (FG1990s, FG2000s, FG2010s), and
decadal change for both GA and FG (1990s, 2000s, 2010s), respectively.
When estimating three scenarios, the decadal condition of GA and FG
corresponding to each decade was applied to SWAT while weather condition
was fixed on the 2010s so that the impact of two factors can be
observed.
Based on the watershed hydrology estimated from the 1980s and three
scenarios, the hydrologic responses were examined at the watershed
outlet subbasin to understand the impact of GA and FG on the whole
catchment. Table 4 summarized the hydrologic responses of the seven
hydrological components and total runoff (TR) calculated from SR, FG,
and LF. In GA scenario, GF has temporally decreased and GWR has
temporally increased. Resultingly, TR has temporally decreased and
showed the decrease percentage of 0.9% (7.0 mm/year), 3.1% (24.1
mm/year), and 5.8% (44.9 mm/year) in the GA1990s, the GA2000s, and the
GA2010s, respectively. In FG scenario, the growth of forest has
temporally increased ET with the value of 0.5% (2.9 mm/year), 1.5%
(8.6 mm/year), 2.4% (13.4 mm/year) in the FG1990s, the FG2000s, and the
FG2010s, respectively. Influenced by the increase of ET, other
hydrological components showed decreasing trend which led to the
temporal decrease of TR. TR in the FG scenario showed the decrease
percentage of 0.4% (3.4 mm/year), 1.3% (10.4 mm/year), and 2.0% (15.8
mm/year) in the FG1990s, the FG2000s, and the FG2010s, respectively.
When both GA and FG were considered, the result showed decreasing trend
of SR, PE, SM, GF, and LF while ET and GWR showed increasing trend.
Consequently, TR was decreased by 1.3% (10.1 mm/year), 4.4% (34.2
mm/year), and 7.8% (60.3 mm/year) in the 1990s, the 2000s, and the
2010s, respectively.
[Insert Table 4]
From the hydrologic response at the watershed outlet, monthly response
of hydrologic components to GA, and FG in the 2010s compared with the
1980s was estimated and presented in Figure 6. As a result, the
hydrologic response of all components showed high value in June and July
while GF showed the biggest hydrologic response with the average value
of 25.61%. The fluctuation of all hydrological beside GF showed a
tendency to increase from January to June and decrease from June to
December. In case of GF, although it showed same tendency that the
decrease rate showed high value in June and July, there observed
additional up and down tendency from October to March.
[Insert Figure 6]
To comprehend the impact of GA and FG on the annual streamflow
condition, flow-duration curves of the 1980s and three scenarios were
additionally analyzed. Using the descending order of streamflow data,
the flow rate corresponding to the time duration of 90, 180, 275, and
355 day was estimated. Table 5 showed the temporal reduction percentage
of TR in different time durations compared with the 1980s. Q90, Q180,
Q275, and Q355 stand for the flow rate according to the time duration of
90, 180, 275, and 355, respectively. From Table 5, two noticeable trends
were observed. First, the reduction percentage has temporally increased
in all scenarios. The reduction percentage in time duration of 355 in
Scenario 3 showed 4.3%, 10.6%, and 16.8% in the 1990s, 2000s, and
2010s, respectively. Secondly, the streamflow in bigger duration showed
higher value of decrease percentage. In case of the 2010s in Scenario 3,
the time duration of 90, 180, 275, and 355 showed 7.3%, 8.3%, 9.5%,
and 16.8% of reduction percentage, respectively. These two increasing
trends were equally applied to all scenarios and time durations.
[Insert Table 5]
3.3 Spatial Vulnerability of Total Runoff Decrease
Based on the 78 subbasins delineated by SWAT, the spatial vulnerability
of TR caused by GA and FG was observed using the decrease percentage of
TR in each subbasin. The percentage was estimated from the TR difference
of each subbasin between the 1980s and three scenarios. The results are
presented in Figure 7. Commonly, the spatial vulnerability of TR loss
has temporally intensified in all scenarios. In GA scenario, the maximum
percentage of TR decrease in a subbasin reached 3.0%, 8.2%, 12.2% in
the GA1990s, the GA2000s, and the GA2010s, respectively. In FG scenario,
the percentage showed 1.6%, 4.4%, and 6.3% in the FG1990s, the
FG2000s, and the FG2010s, respectively. In Scenario 3, the percentage
became 3.6%, 10.3%, and 14.9% in the 1990s, the 2000s, and the 2010s,
respectively.
Spatially, the vulnerability distribution between GA and FG scenario was
different. The decadal GA mainly influenced the downstream and
West-Northern subbasins while that of FG imposed the decrease of TR
mostly on the West-Northern subbasins. In Scenario 3, these
vulnerabilities were spatially integrated and resultingly showed that
the subbasins vulnerable to TR loss were prevalent along the downstream
and West-Northern part of the target watershed.
[Insert Figure 7]