2.1. Study Area
The Qinhuai River basin, located on the south bank of the Yangtze river between longitude 118°39′~119°19′E and latitude 31°34′~32°10′N, was selected to study in this paper (
Figure 1). The area of the basin is 2631 km
2. It is one of the core areas of the Yangtze River Delta. There are two outlets of the study basin, and there are gauge stations for the two outlets, which are the Qinhuaixinhe station (QHXHS) and the Wudingmen station (WDMS). The total discharge of the watershed is the sum of the discharge of the two outlets.
The research area is located in the semi-humid climate region. The average annual temperature is approximately 15.4 °C and precipitation is about 1047 mm. The main land use types include paddy field, urban land, and dry land. The dry land in the study area mainly includes dry farming land and bare land. According to the Harmonized World Soil Database, there are mainly six kinds of soil types in the Qinhuai River basin, Dystric Regosols (RGd), Eutric Fluvisols (FLe), Cumulic Anthrosols (ATc), Eutric Gleysols (GLe), Haplic Luvisols (LVh), and Eutric Planosols (PLe). Because of the weather characteristics and underlying surface characteristics, flood disasters occur frequently in the study area, mainly in the heavy rainy season (April to September).
The Qinhuai River basin is relatively flat with hilly areas surrounding it, and there are low-lying plains in the center. The basin is a typical place of plain water system area, with well-developed, crisscrossed and wide rivers as well as scattered lakes and reservoirs. Based on the above characteristics, the basin has a certain ability of peak regulation, peak flow is smaller and the flood processes are comparatively gentle and last longer.
The data used in this study is presented in
Table 1.
2.2. Description of Urban Agglomeration Polders
Urbanization is the main cause of the existence of UAPFCP. Given the existence of city circle polders, the policy-making department tends to give priority to urbanized area planning inside polders, instead of the areas outside the polders. This leads to the increase of the urbanized areas ratio inside the polders. For instance, in 2007, the document “overall urban planning in Nanjing” proposed that the development of urbanized areas inside the Qianhancun and Dongshan polders needed to be expanded and accelerated. The agricultural land, such as paddy fields and garden plots, inside the polders needed to be transferred outside, and the areas of residential, commercial and first class industrial lands inside the polders needed to be increased. As for the regions surrounding the polders, the areas of the second and third class industrial lands were planned, and the population settlements were expanded and increased. Besides, road construction needed to be strengthened to intensify the connection between the regions inside and outside the polders.
A city circle polder is a closed unit. The urban area which needs protection in the polder is divided within the basin river system by dikes. There is no direct interaction between the runoff in the polders and the river system outside. The connection is through a pump station and sluice gate [
11]. The main land use types in the city circle polder are urban land and dry land, with low storage [
8]. However, the main land use types in the general polder in China are farmland, woodland, and water body. Water surface ratio in the general polder is higher than the ones outside, so that there are certain storage capacities in the general polder. The main function of the city circle polder is city flood control with certain safety standards. Drainage modulus represents the drainage capacity. Maximum allowable water depth represents the flood control safety standard. In order to ensure the safety of urban flood control, the polders need to drain away flood in a timely manner. Due to these reasons, the drainage modulus of the city circle polder should be larger than the general polder and the maximum allowable water depth of a city circle polder should be less than a general polder. The water in the polder will not be pumped out until it reaches the maximum allowable water depth. Urban agglomeration polders are composed of several city circle polders.
There are four groups of city circle polders (i.e., Jurong, Lishui, Qianhancun, and Dongshan) in the study area. The four groups of polders formed the urban agglomeration polders of the Qinhuai River basin. The areas of Jurong, Lishui, Qianhancun, and Dongshan polders are 318.1 km
2, 256.7 km
2, 216.8 km
2, and 238.1 km
2 respectively. The distribution of urban agglomeration polders can be seen in
Figure 1.
2.3. Model Setup
This paper used HEC-HMS to simulate the process of rainfall-runoff. Each model run consists of the meteorological model, the basin model, the control specifications, and the data systems [
16].
The spatio-temporal precipitation was determined by the Specified Hyetograph method. The hyetograph of each sub basin was specified on the basis of the Thiessen Polygons, which was constructed according to the seven rain gage stations in the studied basin [
10]. According to the location of the center of the gravity of the sub basin, the corresponding rain gage station of each sub basin can be determined.
Figure 2a shows the distribution of sub basins and Thiessen Polygons of the Qinhuai River basin.
Surface runoff was predicted from daily rainfall by the Natural Resource Conservation Service (NRCS) Curve Number method based on land use data, soil type data, cumulative precipitation data, and antecedent soil water content [
17], with two parameters of impervious rate and curve number (CN) value [
18]. The CN value is an important parameter of the NRCS method, which is determined by land use, the initial soil moisture condition, and the hydrological unit of the soil type. The NRCS Curve Number method determined three initial soil moisture conditions, which are the dry condition (wilting point), semi moist condition, and moist condition (field capacity). According to the research results of Xu et al. [
19], the initial soil moisture condition in the Qinhuai River basin is generally the semi moist condition. The hydrological unit of a kind of soil type is determined by the final constant infiltration rate (
Y), which reflects the hydrological properties of the soil type. The integrated CN value of each sub basin can be determined according to the land use and soil condition of the sub basin [
20]. The empirical formula for calculating the final constant infiltration rate is as follows [
21]:
where
M is the average particle diameter of soil.
The SCS Unit Hydrograph method was used in the present study to estimate direct runoff with the parameter of lag time (
tlag), which is defined as the time difference between the center of mass of rainfall excess and the peak of the unit hydrograph [
22].
tlag can be calculated with the following equation:
where
C is the conversion coefficient,
Ct is the basin coefficient,
L is the distance from the riverhead to the outlet section of the main channel,
Lc is the distance from the outlet section of the main channel to the center of the basin.
The Recession model was adopted to calculate the base flow and explain the drainage from natural storage in a watershed, with three parameters of base flow threshold ratio to peak, recession constant, and initial value [
23]. It defines the relationship of the base flow
Qt at any time t to an initial value
Q0 as:
where
E is an exponential decay constant. A threshold flow, after the peak of the direct runoff, should be specified either as a flow rate or as a ratio to the computed peak flow when applying the recession model.
The Muskingum method was adopted for channel flow routing, with two parameters of travel time through the reach (
K) and Muskingum weighting factor (
X, 0 ≤
X ≤ 0.5). The method uses the following equation:
where
I1,
I2 are the inflows to the routing reach at the beginning and end of the computation step respectively,
Q1 and
Q2 are the outflows from the routing reach at the beginning and end of the computation step respectively, and
is the calculation step.
The urban agglomeration polders were gradually constructed, and would not change the drainage area and the general rainfall-runoff characteristics of the study area. Thus, the HEC-HMS model of the studied basin with urban agglomeration polders was constructed based on the model without polder. Under the same urbanization scenario, in order to reflect the characteristics of urbanization of the city circle polder, according to the planning and construction condition of city circle polders in the study area, this paper increased the urbanized area ratio of the sub basins inside the polder by 20%, which resulted in an increase of impervious ratio and a change of the CN value. However, the other parameters, such as lag time, K, and X etc., were not changed. As for the sub basins outside the polder, all the parameters were kept unchanged compared with the model without polder.
According to the characteristics of the city circle polders in the Qinhuai River basin, this research assumed the polders to be flat bottom reservoirs in the model. In the model setting, the runoff inside the polder directly flows into the reservoir. When the water volume reaches the maximum allowable volume (corresponding to the maximum allowable water depth), the reservoir starts to drain out the water with its drainage capacity defined by the drainage modulus, and the drainage is then stopped immediately after the water is emptied. The polder will start draining again until the next time the water volume reaches its threshold. Based on the actual data, the drainage modulus in the model is set to 4 m3/(s·km2) and the maximum allowable water depth is set to 0.1 m.
Figure 2b,c shows the HEC-HMS models of the Qinhuai River basin under both without and with polder conditions. The main differences between the models are circled in red.
2.4. Calibration of HEC-HMS
This paper used four evaluation criteria (i.e., Correlation coefficient (
R), Nash–Sutcliffe efficiency (
NSE), Relative flood volume error (
Dv), and relative peak flow error (
Dp)), to evaluate model performance.
R was used to test the correlation of the change trend between the simulated results and observed data.
NSE [
24] was used to measure how well the plot of observed against the simulated flows fits the 1:1 line [
9]. The equations for
R,
NSE,
Dv, and
Dp are as follows:
where
Qsi and
Qoi are the
ith simulated and observed stream flows at time
i,
and
are the average simulated and observed stream flows,
Qsp and
Qop are the peak flows of simulated and observed hydrographs of the simulated flood events.
The acceptable values for
R and
NSE are greater than 0.8, and for
Dv and
Dp less than 20% [
17]. Five parameters (i.e.,
tlag, base flow threshold ratio to peak, recession constant,
K, and
X) need to be calibrated in HEC-HMS. A series of model parameter sets was estimated using the automated optimization tool provided by HEC-HMS by selecting several objective functions, and for the whole calibration period,
NSE was computed for each set of parameters to examine the calibration results. The peak-weighted root mean square error was used as the objective function to evaluate the calibrated model parameters. Validation was then performed; the parameters used during the calibration were not changed during the validation period.