**1. Introduction**

It is undisputed that frequent rainstorms driven by climate change and land-use change caused by high urbanization have resulted in urban flooding in many countries and regions [1–18]. As one of the most serious natural hazards, urban flooding, especially in highly urbanized areas, threatens lives and hinders society's sustainable development nowadays [19–24]. See, for example, the major flooding that occurred throughout most of the Brisbane River's catchment in January 2011, with more than 20 deaths and an economic loss of \$2.55 billion [25]. During the 2011 rainy season, Thailand encountered a large flood of a 50-year return period and a total of 65 provinces were flooded; more than 700 people died and the economic loss reached up to \$41.2 billion [26]. In China, an extreme storm attacked Beijing on 21 July 2012 and a flash flood was triggered in the urban area, causing 79 deaths and a direct economic loss of \$1.86 billion, damaging infrastructure like roads with trapped cars and buses, bridges, and collapsed buildings [27–29]. Therefore, urban flooding is a hot topic in the field of disaster research at present and has attracted much attention in both developed and developing countries [30–37].

The middle and lower reaches of a river are usually more prosperous than the upper reach, and consequently, flooding problems are more serious. Therefore, more effective measures of flooding should be carried out to deal with increasing flooding risk with high urbanization [38]. In general, there are two ways of flood control: engineering and non-engineering measures. Engineering measures include building reservoirs, dikes, detention basins, pumping stations, and spillways. Non-engineering measures usually include flood forecasting or simulation, land-use management, land acquisition and relocation plans, flood emergency planning and response, and post-flood recovery [39]. Among these, flood forecasting or simulation analysis is essential for flood control. The state-of-the-art method for flood simulation includes a one-dimensional river flow model coupled with a two-dimensional surface flow model [40]. For example, Yazdi et al. coupled MIKE 11 with the NSGA-II model for a small watershed in the central part of Iran, showing that optimal designs of multi-reservoir systems can efficiently reduce construction costs, flood peaks and their corresponding damage costs at the downstream reaches of the basin [41]. For flood control measures, there are usually scientific urban plans, sufficient funds, effective laws and regulations in developed countries to put the flood control system on the right track, while the blind and disordered development in developing or undeveloped countries usually cause river flooding problems [42–44]. Rapid urban growth in developing countries usually results in the proliferation of informal settlements. The housing within informal settlements is virtually always built without the consent of the official planning authorities and rarely conforms to official planning guidelines, building regulations, and construction standards [45]. In most cases, there are many buildings along rivers in highly urbanized areas and there is not enough space for the construction of flood control. Meanwhile, the height of river levees differ from one place to another. To solve the problem of river flooding, many developing countries have paid much attention to flooding control measures. Ali Reza Shokoohi has studied the effect of constructing feasible detention dams in urban areas and found that they had good operational effects [46]. Marcelo et al. (2009) focused upon the use of a wide range of different flood control measures in the Joana River watershed, located at the northern region of Rio de Janeiro City, Brazil, and pointed out that distributed detention reservoirs in upstream reaches, parks, public squares, or at urban sites, are very important flood control alternatives [47]. In Turkey, to cope with floods and decrease any further damage, local authorities have designed a set of measures aiming to improve stream conveyance capacity by straightening reaches, lining channels, and building hydraulic structures such as new dams [48]. While in Vietnam, the Red River Delta, the central part of the country, and the Mekong Delta all have completed master plans for flood defenses including upgrading dikes [49]. However, these conventional flood control measures such as widening rivers and increasing embankment height are usually not the best flood control measure for small and medium-scale rivers in highly urbanized areas, nor do they control floods in such a way that other rivers within the same region are considered and fully used to reduce the flood hazards of the target river [50]. When widening the river, due to a large number of residential buildings around rivers, the surrounding residents need to be compensated with land compensation, resettlement compensation and compensation for attachments and young crops, which is a huge financial burden for the government. In the process of land acquisition and compensation allocation, there are often some distribution disputes. If the levee is raised, it will also not be coordinate with the surrounding buildings. Accordingly, the traditional measures are time-consuming and expensive due to the large population and dense buildings, making it very difficult to build a flood control system for small and medium-scale rivers in highly urbanized areas. Unlike large rivers, it is impossible to carry out unified planning for small rivers. Decision making and optimization should be carried out according to the local and actual situation. From a whole region perspective, river networks in the same region are usually connected to each other, and some might suffer floods while the others may not during a storm event; it is possible to utilize the rivers having enough flood bearing capacity to reduce flood hazards of the flooded rivers. However, such idea is seldom proposed in either the research community or the engineering field; how to achieve this idea remains unsolved at present and requires further study.

Therefore, this paper aims to explore a new flood control measure by fully considering the flood bearing capacities of all rivers within the same region. We specifically looked to use the surplus flood bearing capacities of the rivers free from flooding to reduce flood hazards of the flooded rivers. To demonstrate such an idea, we chose a tributary of TieShan River, i.e., the SheGong River located in the comprehensive development zone of Guangzhou Baiyun international airport in China as the case study, and integrated MIKE 11 with MIKE 21 and MIKE FLOOD [47] to simulate flooding situations of all the rivers in the case region. We expect to provide a new idea of flood control for highly urbanized areas around the world.

#### **2. Study Area and Data**

### *2.1. Study Area*

In this study, we chose the river SheGong River located in GuangZhou BaiYun airport development zone in China as the study case. It is a tributary of the TieShan River originated from TieShan New Village in Huashan Town, flowing into the Tieshan River from north to south. The total length of the SheGong River is 7.98 km with a drainage area of 9.58 km2, and the average slope of the river is 3.4%. The Tieshan River Basin covers an area of 58.3 km2, and the topographic map of the basin is shown in Figure 1. The SheGong River, located in the central part of Huashan Town, constitutes the main channel for irrigation and flood channel. Many small reservoirs were built along the river channel and these reservoirs could store massive amounts of water. With the rapid development and utilization of land in the river basin, the irrigation function has been gradually weakened, and the request for ecological landscape water use is prevailing. At present, although the TieShan River has been regulated based on the standard of 20-year return period, the SheGong River is basically in a natural state. The twisted, narrow, and seriously silted channel and the reservoirs frequently lead to flooding after rainstorms; the accumulation of water in the plains of the two sides has caused serious impacts on local production and life. The map of land-use type of the study area in the year 2000 and 2019 are showed in Figure 2.

**Figure 1.** Topographic and water system in the study area.

**Figure 2.** Land-use type in 2000 and 2019.

According to the topographical conditions and possible submergence range, we finally determined the boundary of the study area: the north boundary reaches the Sanjia sluice of the TieShan River and the source of SheGong River and the south boundary reaches the SheGong River estuary, forming a relatively complete river basin (Figure 3). In the study area, the section of the SheGong River is determined as 7.94 km long and that of the TieShan River is 13.4 km.

**Figure 3.** River generalization map of the study area.

#### *2.2. Data*

The land-use type of the study area was obtained from satellite remote sensing image maps in the years 2000 and 2019. The data was acquired by combining and correcting the remote sensing data of the study area, and by comprehensive classification methods such as supervised classification and decision tree classification.

The water system planning of Huadu District in Guangzhou indicates that the flood control standard of the SheGong River is a 20-year return period in the near future and a 50-year return period in the long term. Therefore, according to the flood control standard of China (GB50201-2014), the flood control (tide) standard of Guangdong Province (Trial), and the Huadu District water system planning of Guangzhou, the flood standard of this project is a 20-year return period, with the protection level of 4 in both levee and the main building, and level 5 in both the secondary building and the temporary building.

According to the requirements of the code for design of levee engineering of China (GB50286-98), the safety heightening value of level 4 in the levee engineering is as follows: 0.6 m for the levee engineering that is not allowed to cross the waves, and 0.3 m for the levee engineering that is allowed to cross the waves. The elevation of dike top of this project is designed according to the principle that no overtopping is allowed; in this case, the safe heightening is set to 0.6 m. The calculation height induced by run-up wave is 0.2 m and the super elevation of the dike top is 0.8 m. Therefore, the design flood level plus 0.8 m super elevation of the dike top is the elevation value of the dike top.

The design river bottom elevation at 0 + 000 section at the beginning of the regulation in the SheGong River is 22.2 m, and that of the 5 + 600 section at the key point of regulation is 9.1 m. According to the measured data in 2015, the current river bottom elevation of the 0 + 000 section is 22.54 m, and that of the 5 + 600 section is 9.38 m.

A hydrological atlas of Guangdong Province published by Guangdong Hydrological Bureau was used in the study. The lower boundary condition in the simulation model, i.e., river stages, came from the Comprehensive Planning Report of River Basin in Huadu District of Guangzhou City provided by the Guangzhou Water Conservancy Bureau. The topographic map of the Huadu District was provided by the Guangzhou Surveying and Mapping Institute, while that of SheGong River with a measuring scale of 1:2000 and the cross- and vertical section of the river were given by a commissioned surveying and mapping company.

#### **3. Methodology**

#### *3.1. Design Flood Calculation*

According to the manual regarding the use of rainfall-runoff curves in Guangdong Province and the standard of design flood calculation for water conservancy and hydropower engineering (SL44-2006), the design peak flow of small watershed is as follows [48,49]:

$$Q\_m = \frac{0.278 \psi S\_p}{\pi^n} F \tag{1}$$

$$
\pi = \frac{0.278L}{m f^{\frac{1}{5}} Q^{\frac{1}{4}}} \tag{2}
$$

where *Qm* is design peak flow (m3/s); ψ is runoff coefficient of flood peak ; *Sp* is design rainfall intensity; *F* is catchment area of the basin (km2); τ is routing duration; *m* is routing parameter, *L* is river length (km); *J* is average slope of the river basin. These parameters are further given as

$$
\psi = 1 - \frac{\mu}{S\_p} \tau^n (t\_\varepsilon \ge \tau) \tag{3}
$$

$$
\psi = n \left( \frac{t\_c}{\tau} \right)^{1-n} \tag{4}
$$

$$
\mu = RS(rl)p\tag{5}
$$

$$S\_p = H\_{24p} t^{n-1} \tag{6}$$

$$t\_c = \left[\frac{(1-n)S\_p}{\mu}\right]^{\frac{1}{n}}\tag{7}$$

where μ is the soil infiltration rate; *R* is the loss coefficient; *rl* is the loss index (for the calculation of rainstorm peak in a small watershed); *H*24*<sup>p</sup>* is the maximum 24-h rainfall; *n* is the rainstorm decay index (by referring to the Hydrographic Atlas of Guangdong Province).

The cross-section inputs are set up according to the observed data in the SheGong River and the calculated data in the TieShan River channel. The cross sections are set at an average spatial distance of 50 m, and there are 427 sections in total. Since there are no smaller sub-basins in the river channel and the river channel is long enough, the inflow between two sections is regarded as uniform. The results of design peak flow for each section are shown in Table 1.


**Table 1.** Discharge of each calculated section of river.

#### *3.2. Flood Simulation Model*

MIKE FLOOD is used to couple the one-dimensional MIKE 11 model and the two-dimensional MIKE 21 model, for simulating the flood prone area with different return periods.

#### 3.2.1. MIKE 11

MIKE 11 is an implicit finite difference model for one dimensional unsteady flow computation, the basic equations of which are Saint–Venant equations that are solved by the Abbott–Ionescu six-point implicit difference method and are used to simulate flood processes in a river channel [50]. The general steps of model setup include river network generalization, river section setting, boundary conditions setting, and determination of hydraulic parameters (Figure 4). In this study, we generalized the river network based on geographic data of the study area (Figure 3). Based on observed data, a total of 427 river sections were set up, and for boundary conditions, the upstream boundary was selected as the flow boundary, whilst the downstream boundary was selected as the water level boundary; through analysis and demonstration, we selected the design flood level with a 5-year return period as the boundary water level. The setting of hydraulic parameters mainly included the determination of a river roughness coefficient based on the current status of the river and the related data.

#### 3.2.2. MIKE 21

MIKE 21 belongs to the free two-dimensional surface flow model, and uses the finite volume method to solve the planar two-dimensional shallow water equation. The general steps of model setup include grid division, elevation interpolation, boundary conditions setting, and determination of hydraulic parameters (Figure 4). In this study, the study area was divided into 28,800 grids with grid size of 50 × 50 m. We also performed elevation interpolation based on the measured terrain data. For the boundaries connected to rivers in MIKE 21, we set them as open boundaries, and other boundaries were set as closed boundaries. The setting of hydraulic parameters was to determine the surface roughness coefficient based on the type of land-use [51].

**Figure 4.** Flow chart of flooding simulation model.
