Many hydraulic projects such as reservoirs, ponds and paddy fields as well as soil and water conservation engineering projects have been constructed to improve utilization of water resources upstream of the Wudaogou station basin in Northeast China in recent years. As a result, the local hydrological characteristics of the basin and the flood runoff and process have been changed. These changes in the basin characteristics make basin hydrological forecasting more difficult. In order to model and assess this situation, the TOPMODEL, which includes the dynamic soil moisture storage capacity (DSMSC-TOPMODEL), is used in this study to simulate the flood impact of hydraulic projects. Furthermore, the Bayesian method is used to evaluate model parameter uncertainty and assess the TOPMODEL's performance over the basin. Flood simulation results show that accuracy is significantly improved when the stock version of TOPMODEL is replaced with DSMSC-TOPMODEL, with the qualified ratio of forecasting runoff yield increasing from 65% in the former to 88% in the latter. Moreover, these flood simulations are more suitable for helping observers visualize the process.

INTRODUCTION

Exploiting and utilizing water resources have increased greatly with the rapid economic development of urbanization in China. However, factors related to infiltration, pan evaporation and runoff are always changing with the continuous construction of hydraulic projects in upstream catchments (Callow & Smettem 2009; Assani et al. 2011). Hydraulic project construction presents mutability and nonhomogeneity (Wang et al. 2009), both of which exert a dominant influence on floods. Therefore, it is imperative to understand the influence and relative importance of hydraulic projects with respect to runoff, which is critical to the management of regional water resources (Batallaa et al. 2004; Brath et al. 2006; Kezer & Matsuyama 2006; Ma et al. 2010).

Deitch et al. (2013) indicate that the cumulative impact of existing small reservoirs may exert a strong influence on early-season streamflow, but the impact becomes less over time. Moreover, some researchers have attempted to describe the effects of small reservoirs and low-head dams by estimating reservoir storage volumes. Rodrigues et al. (2012) develop a simple method that allows the estimation of reservoir storage volumes as a function of their surface areas. However, studies of the effect of hydraulic projects on flood forecasting are rarely seen.

In China, upstream of the Wudaogou station basin, a major tributary of the Songhua River basin (SRB) located in the northeast of China exists. A large area of wetlands in the SRB has degraded, while farmland has been increasing rapidly (Meng & Mo 2012; Mu et al. 2012; Li et al. 2014). This region is unique because it is studded with hundreds of medium- and small-sized reservoirs, ponds, and paddy fields (Zhang et al. 2012). Cao et al. (2011) find that hydraulic project construction can affect the runoff of floods. The above-mentioned methods can only quantify the effects of hydraulic projects as a whole, while the influence process cannot be reflected or described quantitatively.

This paper aims to simulate floods impacted by hydraulic project activity in the Wudaogou basin from 1956 to 2010, based on the selected technique of data analysis as well as the TOPMODEL with the dynamic soil moisture storage capacity (DSMSC). The findings of these analyses provide a flood forecasting method that simulates flood runoff and the processes impacted by hydraulic project construction.

CASE STUDY AREA AND DATA

Basin description

This study is performed in the 42,500 km2 basin named the Fengman reservoir, which is located in the Jinlin province in Northwest China. The primary river of this basin is the Songhua River. During the past 50 years, due to pressure from population increase, agricultural growth has boomed and a large number of water projects have been constructed for irrigation. Flourishing water projects have caused the region to have the highest density of small- and medium-sized reservoirs in Northeast China. In 2002, the Fengman Hydropower Plant charted lakes using satellite data and found 9,335 lakes spread over 4,000 square meters. The basin is divided into three subareas, with the Fengman reservoir holding areas (I), the Upstream of Wudaogou station (II) and the Upstream of Baishan reservoir (III). The Upstream of Wudaogou station (UWB) and the main distribution area of water projects covering an area of 11,900 km2 was selected as the study area (Figure 1).
Figure 1

Location, hydro-meteorological stations and types of hydraulic projects in UWB.

Figure 1

Location, hydro-meteorological stations and types of hydraulic projects in UWB.

The 2006 investigation showed that there are one large reservoir, 12 mid-sized reservoirs, 476 small-sized reservoirs and countless ponds in the study area, totaling a capacity of 9.58 × 108 m3. Table 1 shows the statistics of the cumulative storage and control area of each type of hydraulic projects. Table 1 also shows that the control area of the middle- and small-sized reservoirs accounts for about 83% of the total area. Consequently, it is necessary to consider the flood detention and discharge of middle- and small-sized reservoirs. Information on the operation of mid-sized reservoirs can be collected easily. However, information regarding small reservoirs is much more difficult to obtain because of their large number and wide distribution. Because of concern about the combined effect of these reservoirs, in this paper we quantify the combined effect as the maximum influence quantity Vmax using the quantitative method as follows.

Table 1

The statistical cumulative storages and control areas of each type of hydraulic project in UWB

Reservoir typePondsSmall (II)Small (I)Small totalMiddleLargeSummation
Classification capacity (106 m3≤0.1 0.1–1 1–10 0.1–10 10–100 ≥100 
Numbers  390 86 476 12 489 
Capacity (106 m354.62 152.63 255.07 407.70 179.82 316.00 958.14 
Of the total capacity (%) 5.70 15.93 26.62 42.55 18.77 32.98 100.00 
Control areas (km22252.06 1989.37 1902.19 3891.56 729.65 548.00 7421.27 
Of the total areas (%) 30.35 26.81 25.63 52.44 9.83 7.38 100.00 
Reservoir typePondsSmall (II)Small (I)Small totalMiddleLargeSummation
Classification capacity (106 m3≤0.1 0.1–1 1–10 0.1–10 10–100 ≥100 
Numbers  390 86 476 12 489 
Capacity (106 m354.62 152.63 255.07 407.70 179.82 316.00 958.14 
Of the total capacity (%) 5.70 15.93 26.62 42.55 18.77 32.98 100.00 
Control areas (km22252.06 1989.37 1902.19 3891.56 729.65 548.00 7421.27 
Of the total areas (%) 30.35 26.81 25.63 52.44 9.83 7.38 100.00 

The maximum impact ability of hydraulic projects on flood

  1. Large reservoirs: According to historical information, the combined effects of hydrologic characteristics and their design parameters read out as ‘detention’ when the reservoir water level is low. Consequently, the reservoir water level may be higher than the normal storage level but will reach flood control level by the end of the flood. So the maximum influence quantity of large reservoirs can be calculated as the utilizable capacity.

  2. Mid-sized reservoirs: The analysis of existing data on mid-sized reservoirs reveals that utilizable capacity accounts for about 70% of total capacity due to a lack of operation rules. Therefore, the maximum influence quantity of mid-sized reservoirs can be calculated as 70% of total capacity.

  3. Small-sized reservoirs: To account for small-sized reservoirs while under the constraint of poor manual control ability, the maximum influence quantity can be calculated as total capacity.

Using the maximum influence quantity calculation method and the statistics of cumulative storage for each type of water project in each year, each year's maximum influence quantity can be calculated (see Table 2).

Table 2

The maximum influence quantity of each year's hydraulic projects to floods

 Total capacity (106 m3)
Maximum influence quantity (106 m3)
  
YearsLargeMiddle and smallPondsLargeMiddle and smallPondsSum (106 m3)Runoff depth (mm)
1950s  61.0 54.6 0.0 42.7 54.6 97.3 7.9 
1960s  198.0 54.6 0.0 138.6 54.6 193.2 15.7 
1970s 316 405.1 54.6 124.0 283.6 54.6 462.2 37.6 
1980s 316 463.5 54.6 124.0 324.4 54.6 503.0 40.9 
1990s 316 525.7 54.6 124.0 368.0 54.6 546.6 44.4 
2000s 316 587.5 54.6 124.0 411.3 54.6 589.9 48.0 
 Total capacity (106 m3)
Maximum influence quantity (106 m3)
  
YearsLargeMiddle and smallPondsLargeMiddle and smallPondsSum (106 m3)Runoff depth (mm)
1950s  61.0 54.6 0.0 42.7 54.6 97.3 7.9 
1960s  198.0 54.6 0.0 138.6 54.6 193.2 15.7 
1970s 316 405.1 54.6 124.0 283.6 54.6 462.2 37.6 
1980s 316 463.5 54.6 124.0 324.4 54.6 503.0 40.9 
1990s 316 525.7 54.6 124.0 368.0 54.6 546.6 44.4 
2000s 316 587.5 54.6 124.0 411.3 54.6 589.9 48.0 

Dataset

The annual climate and streamflow datasets compiled for the DSMSC-TOPMODEL include the annual precipitation and potential evapotranspiration series in UWB and the annual floods series of Wudaogou hydrologic stations. The annual floods series of the Wudaogou hydrologic station are obtained from the Songliao River Conservancy Commission (SLRCC). The Hydrographic and Water Resources Bureau of Jilin province provides the daily precipitation data from 1953 to 2010 for 15 stations. The other daily meteorological data (temperature, solar radiation, humidity and wind speed) from seven stations are obtained from the China Meteorological Data Sharing Service System. The long-term annual precipitation is obtained from the arithmetic mean of daily gauged records for 15 stations from 1953 to 2010, while potential evapotranspiration is estimated using the method proposed by Xu et al. (2005), which provides a better definition and has been proved to provide a satisfactory estimation of potential evapotranspiration.

From the Data Center for Resources and Environmental Sciences Chinese Academy of Sciences (RESDC), digital elevation data are collected to calculate the topographic index in the TOPMODEL study area. The storage and control areas of each type of hydraulic projects are collected from SLRCC to investigate the impact of hydraulic project construction and activities on floods.

DSMSC-TOPMODEL

TOPMODEL

TOPMODEL is a semi-distributed variable-contributing area hydrological model. It is based on simple physics and assumes that there is a steady transfer of water in the saturated zone along a hill slope, with a water table nearly parallel to the ground surface. TOPMODEL was proposed firstly in 1979 by Beven & Kirkby (1979).

The structure of the original version of TOPMODEL is shown in Figure 2, which shows that the total runoff is generally the sum of two major flow components. The main runoff generation mechanisms must be equivalent to the saturation overland flow (i.e., precipitation on soils saturated ‘from below’) and base flow feed by shallow groundwater.
Figure 2

The structure of TOPMODEL.

Figure 2

The structure of TOPMODEL.

 
formula
1
a0 is defined as the unsaturated soil water storage saturation degree before flood happens, the ratio of which has a dependency relation which influences the hydraulic projects of runoff, and Suz means the depth of the unsaturated store and SD means the local saturated deficit.

Parabolic curve of DSMSC

Surface runoff generated by precipitation after a long drought is usually held by numerous middle- and small-sized reservoirs, which makes the actual flood entering the reservoir less than the predicted value. Consequently, reservoir operation decisions could be made that result in early or enlarged discharge by referring to the predicted value. The water supply and power generation in the following year will be damaged by such operation decisions if at any point in the future there is no longer any large runoff. If prophase precipitation is sufficient and further precipitation occurs, the small reservoirs will be discharged and the dam will break, resulting in a predicted flood much smaller than the actual value.

To investigate and analyze 26 floods over 55 years (1956–2010), Figure 3 shows that the runoff coefficient and the unsaturated soil water storage saturation degree have a positive relationship. The correlation coefficient is 0.7. Therefore, is used to identify the saturation of hydraulic before floods happen. Then the impact of hydraulic project activity on runoff can be recognized (see Figure 4). In the model, hydraulic project activity can be described as the change of SD.
Figure 3

The trend chart of runoff coefficient and the unsaturated soil water storage saturation degree from 1956 to 2010.

Figure 3

The trend chart of runoff coefficient and the unsaturated soil water storage saturation degree from 1956 to 2010.

Figure 4

The parabolic curve of DSMSC.

Figure 4

The parabolic curve of DSMSC.

In Figure 4, the horizontal axis x shows the SD process, where the vertical axis y represents the unsaturated soil water storage saturation degree and Vmax is the biggest impact quantity of hydraulic projects on runoff. As reservoir storage changes from empty to saturated, and the comprehensive effects of hydraulic projects on flood change from holding to discharging, the capacity of hydraulic projects tends to reach an intermediate region of stability. At this point, A1 and A2 are key control nodes representing or , respectively, where the simulated effect of hydraulic projects holds and SD increases. When , the simulated effect of the aggregated reservoir discharges and SD increases. The value range of A1 and A2 is (0, 1). B is a parameter of the parabolic index.

Uncertainties in model parameters of DSMSC-TOPMODEL

To evaluate the effects of hydraulic projects on hydrologic processes using TOPMODEL, it is necessary to analyze the uncertainty effects due to the model parameters on the model performance (Homberger & Spear 1981; Engeland et al. 2005). In this study, the Sobol method (Zhang et al. 2013) is used to perform sensitivity analysis on model parameters. The uncertainty interval at each time step is the major result by the generalized likelihood uncertainty estimation (GLUE) technique (Beven & Freer 2001) in terms of evaluating hydrological modeling uncertainty.

For any model-predicted variable Z, the predicted quantities for any value z can be calculated from the cumulative distribution 
formula
2
where , the value of variable Z at any time t, is simulated by model with the behavioral parameter set and the likelihoods scaled such that the sum over all behavioral models is always unified. 
formula
3
where is a specified prior likelihood for the model, is a likelihood measure calculated for the model over period T with input YT and observed variable ZT, and C is a scaling constant that ensures a cumulative posterior likelihood of unity. The posterior likelihoods are always conditional upon the input and observation data used in the evaluation, and the choice of likelihood measure and choice of any threshold is used for rejection.

The 90% confidence intervals for streamflow due to parameter uncertainty are computed from the streamflow samples with 10,000 parameter sets generated by the GLUE algorithm.

RESULTS AND DISCUSSION

A break point in the runoff series

Line 1 in Figure 5 shows the cumulative annual hydraulic projects capacity, Line 2 shows the cumulative annual precipitation, and Line 3 shows cumulative annual runoff from 1953 to 2010 over the UWB. From Line 2, it can be observed that the cumulative annual precipitation curves are nearly straight lines. From this point, it may be inferred that there is no abrupt change in annual precipitation. K means the slope of the trend line. Using the trend analysis method, one break point at 1956 is detected in the runoff record. The slope of the trend line before 1956 is 0.40, which is greatly reduced after 1956. From comparison with Line 1, it is determined that decreases in runoff can be attributed to hydraulic project activity. Changes in runoff between years can be inferred from lines 2 and 3 and correlate with hydraulic project construction. Accordingly, this study period is divided into two periods, 1953–1955 and 1956–2010. Eight natural floods without the impact of hydraulic project activity are used for model parameters calibration and verification from 1953 to 1955.
Figure 5

Long-term variations in annual precipitation, streamflow and cumulative capacity of hydraulic projects of UWB.

Figure 5

Long-term variations in annual precipitation, streamflow and cumulative capacity of hydraulic projects of UWB.

Uncertainties in model parameters

In this paper, such uncertainty analyses over the benchmark period 1953–1955 (i.e., eight floods) are conducted using GLUE inference to analyze the uncertainties of hydrological models. The eight natural floods are also used for model natural parameters calibration and verification.

The 90% confidence intervals of the simulated flood runoff in the period before 1956 due to the parameter uncertainty is shown in Table 3 and Figure 6 (i.e., the gray interval).The observed runoff is shown in Figure 6 (i.e., the black line). These figures indicate that the influence of model parameter uncertainty on streamflow simulation is minimal, which further confirms that the possibility of ‘equifinality’ for TOPMODEL would be small. Therefore, it is determined that changes in floods can be attributed to hydraulic project activity.
Table 3

The results of model calibration and verification

TypesNumberObserved flood runoff (mm)Simulated flood runoff (mm)
Simulated error (%)
The 90% confidence intervals5%95%5%95%
Calibration 80.6 65.8 73.7 −18 −9 
173.3 172.9 177.7 
40.3 36.9 44.6 −8 11 
55.2 48.9 49.8 −11 −10 
Verification 86.8 80.7 94.4 −7 
93.6 77.9 89.6 −17 −4 
81.6 76.1 81.3 −7 
87.1 98.1 101.9 13 17 
TypesNumberObserved flood runoff (mm)Simulated flood runoff (mm)
Simulated error (%)
The 90% confidence intervals5%95%5%95%
Calibration 80.6 65.8 73.7 −18 −9 
173.3 172.9 177.7 
40.3 36.9 44.6 −8 11 
55.2 48.9 49.8 −11 −10 
Verification 86.8 80.7 94.4 −7 
93.6 77.9 89.6 −17 −4 
81.6 76.1 81.3 −7 
87.1 98.1 101.9 13 17 
Figure 6

The 95% confidence intervals for simulated streamflow due to the parameter uncertainty of flood in 1953–1956.

Figure 6

The 95% confidence intervals for simulated streamflow due to the parameter uncertainty of flood in 1953–1956.

Flood simulations results based on natural TOPMODEL

Using the data from the impact of 26 floods on hydraulic projects after 1956, the simulated runoff without the consideration of hydraulic project activity is shown in Table 4. It can be observed that the forecasted runoff of nine floods is inaccurate; the accuracy of the simulation is 65%. By analyzing the inaccurately forecasted floods, Figure 7 shows that the of 20010704, 19960720, 19640811, 19600824, 19710805, 19600806 and 19850817 floods is 37%, 28%, 52%, 8%, 14%, 21% and 47%, respectively, which means that the surface runoff is generated by precipitation after a long drought and, therefore, that hydraulic projects are empty. The simulated effect of these hydraulic projects holds. In contrast, the of 19910722 and 19640816 floods are both 99%, which means reservoirs under saturated storage and small reservoirs would be discharged.
Table 4

Flood simulation results based on DSMSC-TOPMODEL

Flood numberObserved flood runoff (mm)α0ΔSDSimulated runoff ‘before’a (mm)Simulated runoff ‘after’b (mm)Simulated error ‘before’a (%)Simulated error ‘after’b (%)
19560723 81.6 71 0.0 89.0 89.0 
19570823 117.5 63 0.0 110.6 110.6 − 6 − 6 
19600806 87.1 21 15.7 108.4 93.6 24 
19600824 59.8 15.7 82.8 67.0 38 12 
19630718 106.5 24 15.7 120.4 97.8 13 − 8 
19640811 50.6 52 15.7 65.7 53.8 30 
19640816 137.5 99 − 15.6 106.2 107.9 − 23 − 21 
19710805 76.8 14 37.6 91.8 90.2 20 17 
19750802 102.4 53 0.0 114.9 114.9 12 12 
19850808 90.7 62 0.0 77.7 77.7 − 14 − 14 
19850817 99.1 47 40.9 119.4 105.1 21 
19860804 112.7 60 0.0 104.2 104.2 − 8 − 8 
19910722 73.6 99 − 44.1 55.9 65.3 − 24 − 11 
19910731 83.5 66 0.0 73.3 73.3 − 12 − 12 
19940710 59.2 66 0.0 67.8 67.8 15 15 
19940818 91.0 55 0.0 89.1 89.1 − 2 − 2 
19950731 267.6 23 44.4 253.0 232.0 − 5 − 13 
19960720 29.5 28 44.4 56.6 47.1 92 60 
19960726 39.6 74 0.0 43.5 43.5 10 10 
19960808 33.0 65 0.0 35.2 35.2 
19960813 61.8 62 0.0 60.2 60.2 − 3 − 3 
19980807 28.5 79 0.0 27.2 27.2 − 5 − 5 
20010704 29.2 37 48.0 45.9 39.5 57 35 
20100729 109.6 16 48.0 123.7 107.0 13 − 2 
20100810 81.1 63 0.0 86.2 86.2 
20100824 78.1 49 0.0 80.0 76.6 − 2 
Flood numberObserved flood runoff (mm)α0ΔSDSimulated runoff ‘before’a (mm)Simulated runoff ‘after’b (mm)Simulated error ‘before’a (%)Simulated error ‘after’b (%)
19560723 81.6 71 0.0 89.0 89.0 
19570823 117.5 63 0.0 110.6 110.6 − 6 − 6 
19600806 87.1 21 15.7 108.4 93.6 24 
19600824 59.8 15.7 82.8 67.0 38 12 
19630718 106.5 24 15.7 120.4 97.8 13 − 8 
19640811 50.6 52 15.7 65.7 53.8 30 
19640816 137.5 99 − 15.6 106.2 107.9 − 23 − 21 
19710805 76.8 14 37.6 91.8 90.2 20 17 
19750802 102.4 53 0.0 114.9 114.9 12 12 
19850808 90.7 62 0.0 77.7 77.7 − 14 − 14 
19850817 99.1 47 40.9 119.4 105.1 21 
19860804 112.7 60 0.0 104.2 104.2 − 8 − 8 
19910722 73.6 99 − 44.1 55.9 65.3 − 24 − 11 
19910731 83.5 66 0.0 73.3 73.3 − 12 − 12 
19940710 59.2 66 0.0 67.8 67.8 15 15 
19940818 91.0 55 0.0 89.1 89.1 − 2 − 2 
19950731 267.6 23 44.4 253.0 232.0 − 5 − 13 
19960720 29.5 28 44.4 56.6 47.1 92 60 
19960726 39.6 74 0.0 43.5 43.5 10 10 
19960808 33.0 65 0.0 35.2 35.2 
19960813 61.8 62 0.0 60.2 60.2 − 3 − 3 
19980807 28.5 79 0.0 27.2 27.2 − 5 − 5 
20010704 29.2 37 48.0 45.9 39.5 57 35 
20100729 109.6 16 48.0 123.7 107.0 13 − 2 
20100810 81.1 63 0.0 86.2 86.2 
20100824 78.1 49 0.0 80.0 76.6 − 2 

a‘before’ means without considering hydraulic project activity.

b‘after’ means considering hydraulic project activity.

Figure 7

The Suz and D-value of SD and Suz before the flood impacted by hydraulic project activity.

Figure 7

The Suz and D-value of SD and Suz before the flood impacted by hydraulic project activity.

Flood simulations results based on DSMSC-TOPMODEL

The parameters are optimized by the PSO method (Kennedy & Eberhart 1995). A1 = 0.53, A2 = 0.80, B = 0.80. The DSMSC-TOPMODEL simulated results consider hydraulic project activity in Table 4 and Figure 8. It can be observed that the accuracy of the simulation has greatly improved and the number of inaccurate flood runoff forecasts has decreased to three. The overall accuracy of the simulation has increased from 65 to 88%.
Figure 8

Flood simulation results based on DSMSC-TOPMODEL.

Figure 8

Flood simulation results based on DSMSC-TOPMODEL.

Changes in floods can be attributed to hydraulic project activity. The flood runoff after the rainy season begins and before the rainy season ends usually holds up in hydraulic projects. In a high flow year, the small reservoirs would be discharged and the dam would break, resulting in a predicted flood runoff much smaller than the actual value. These changes of runoff coefficients would reflect hydraulic project activity. After considering the impact of hydraulic project activity on floods, it can be seen from Figure 9 that the flood process simulated by the DSMSC-TOPMODEL is more close to the observed process than TOPMODEL.
Figure 9

Flood processes simulated by DSMSC-TOPMODEL.

Figure 9

Flood processes simulated by DSMSC-TOPMODEL.

CONCLUSIONS

A semi-distributed variable contributing area hydrological model, i.e., TOPMODEL, is used to evaluate the impacts of hydraulic projects on the changes of runoff in Wudaogou basin in northern China. The uncertainty analysis indicates that TOPMODEL can consistently simulate the flood runoff over the study area. Floods that occur after the rainy season begins and before the rainy season ends are usually held back by hydraulic projects, which are discharged in a high flow year. Considering the impact of hydraulic project activity on flood, the runoff coefficient and the unsaturated soil water storage saturation degree have a positive relationship. A parabolic curve describing DSMSC is therefore calculated to simulate the impacts of hydraulic project activity on floods.

By using the DSMSC-TOPMODEL to simulate 26 floods impacted by hydraulic project activity, the accuracy of simulation has greatly improved from 65 to 88%. The simulated flood process is thus more close to the observed process.

ACKNOWLEDGEMENTS

This work is supported by the Major International (Regional) Cooperation Project (Grant No. 51320105010), the National Natural Science Foundation of China (Grant No. 51379027, 51409043, 91547111) and the Natural Science Foundation of Liaoning Province (Grant No. 2015020608).

REFERENCES

REFERENCES
Batallaa
R. J.
Gomez
C. M.
Kondolf
G. M.
2004
Reservoir-induced hydrological changes in the Ebro River basin (NE Spain)
.
Journal of Hydrology
290
,
117
136
.
Beven
K.
Kirkby
M. J.
1979
A physically-based variable contributing area model of basin hydrology
.
Hydrology Science Bulletin
24
(
1
),
43
69
.
Cao
M. L.
Zhou
H. C.
Zhang
C.
Zhang
A. J.
Li
H. Y.
Yang
Y.
2011
Research and application of flood detention modeling for ponds and small reservoirs based on remote sensing data
.
Science China Technological Sciences
54
,
2138
2144
.
Deitch
M. J.
Merenlender
A. M.
Feirer
S.
2013
Cumulative effects of small reservoirs on streamflow in Northern Coastal California catchments
.
Water Resources Management
27
,
5101
5118
.
Engeland
K.
Xu
C. Y.
Gottschalk
L.
2005
Assessing uncertainties in a conceptual water balance model using Bayesian methodology
.
Hydrological Sciences Journal-Journal Des Sciences Hydrologiques
50
(
1
),
45
63
,
doi:10.1623/hysj.50.1.45.56334
.
Homberger
G. M.
Spear
R. C.
1981
An approach to the preliminary analysis of environmental systems
.
Journal of Environmental Management
12
,
7
18
.
Kennedy
J.
Eberhart
R.
1995
Particle swarm optimization[C]//IEEE Int. 1 Conf. on Neural Networks
.
Perth
,
Australia
, pp.
1942
1948
.
Kezer
K.
Matsuyama
H.
2006
Decrease of river runoff in the Lake Balkhash basin in Central Asia
.
Hydrological Processes
20
(
6
),
1407
1423
.
Rodrigues
L. N.
Sano
E. E.
Steenhuis
T. S.
Passo
D. P.
2012
Estimation of small reservoir storage capacities with remote sensing in the Brazilian Savannah region
.
Water Resources Management
26
,
873
882
.
Xu
J.
Haginoya
S.
Saito
K.
Motoya
K.
2005
Surface heat balance and pan evaporation trends in eastern Asia in the period 1971 to 2000
.
Hydrological Processes
19
, (
11
),
2161
2186
.