Application of Remote-Sensing-Based Hydraulic Model and Hydrological Model in Flood Simulation

Abstract

Floods are one of the main natural disaster threats to the safety of people’s lives and property. Flood hazards intensify as the global risk of flooding increases. The control of flood disasters on the basin scale has always been an urgent problem to be solved that is firmly associated with the sustainable development of water resources. As important nonengineering measures for flood simulation and flood control, the hydrological and hydraulic models have been widely applied in recent decades. In our study, on the basis of sufficient remote-sensing and hydrological data, a hydrological (Xin’anjiang (XAJ)) and a two-dimensional hydraulic (2D) model were constructed to simulate flood events and provide support for basin flood management. In the Chengcun basin, the two models were applied, and the model parameters were calibrated by the parameter estimation (PEST) automatic calibration algorithm in combination with the measured data of 10 typical flood events from 1990 to 1996. Results show that the two models performed well in the Chengcun basin. The average Nash–Sutcliffe efficiency (NSE), percentage error of peak discharge (PE), and percentage error of flood volume (RE) were 0.79, 16.55%, and 18.27%, respectively, for the XAJ model, and those values were 0.76, 12.83%, and 11.03% for 2D model. These results indicate that the models had high accuracy, and hydrological and hydraulic models both had good application performance in the Chengcun basin. The study can a provide decision-making basis and theoretical support for flood simulation, and the formulation of flood control and disaster mitigation measures in the basin.

1. Introduction

Flooding is an extreme hydrological phenomenon [1,2,3,4]. Floods are one of humanity’s biggest challenges because of their highly destructive attributes [5]. Recently, under the influence of climate change, extreme flood events have occurred frequently, posing a serious threat to the safety of human life and property [6,7]. Due to its special geographical location and climatic conditions, China has suffered from the threat and impact of floods for a long time, which has an adverse effect on the sustainable development of society and the environment [8,9,10]. Clearly, effective flood management methods are urgently needed to minimize flood-related losses, which not only has an important theoretical contribution, but also has strong practical significance.

The most commonly used nonengineering measures in flood simulation and management mainly include hydrological and hydraulic models [11,12]. The hydrological model is a technology for flood process simulation that is generally acknowledged [12,13,14]. Since the 1950s, the hydrological models have been continuously enriched and improved [15,16,17]. Numerous hydrological models have been developed and applied extensively around the world with good results, such as the XAJ [18], TOPMODEL [19], VIC [20], TANK [21], GR3 [22], GR4 J [23], EasyDHM [24], PDM [25], ARNO [26], SAC [27], SWAT [28,29], HBV [30], SWM [31], and G2G [32] models; more details about these models are given in

Table 1. Model characteristics.

Hydrological Model	Developed/Maintained by	Type	Runoff Characteristic	Applications
XAJ	Zhao, 1984	Lumped	Saturation—excess runoff	Humid/semihumid Hourly/daily
TOPMODEL	Beven and Kirkby, 1979	Semidistributed	Saturation—excess runoff	Humid/arid Daily
VIC	Liang et al., 1994	Distributed	Infiltration—excess runoff	Daily/monthly River basin
TANK	Sugawara, 1961	Lumped	Infiltration—excess runoff	Single rainfall simulation Humid/arid
GR3	Edijaton, 1999	Lumped	Saturation—excess runoff	Hourly/daily Small/medium watershed
GR4 J	Perrin et al., 1993	Lumped	Saturation—excess runoff	Hourly/daily Small/middle watershed
EasyDHM	Lei, X., et al., 2010	Distributed	Saturation—excess/infiltration—excess	Wide applicability
PDM	University of Reading, UK, 1998	Lumped	Saturation—excess runoff	Daily Continuous simulation A macroscale globe model
ARNO	Ciarapica, Todini, 1996, 2002	Lumped/ distributed	Saturation-excess runoff	Large range of spatial scales Continuous simulation
SAC	Burnash, Ferral, Ferral, 1970s	Lumped	Saturation—excess runoff	Long-time continuous simulation Daily/6 h Large/medium watershed Humid/arid
SWAT	Arnold et al., 1995	Distributed	Infiltration—excess runoff	Up to large river basin Daily/monthly
HBV	Gardelin, 1997	Lumped/ Distributed	Saturation—excess runoff	Large/medium watershed Daily/donthly
SWM	Linsley, Crawford,1959	Lumped	Saturation—excess runoff	Continuous simulation Hourly

Reference: [13,24,37,38,39,40,41,42].

. These models have been tested and applied through long-term practice, and have high reliability and practicability [33]. Among them, the XAJ model has been widely applied in China and achieved good results [34,35,36].

With the development of numerical simulation technology and the introduction of systems theory, hydraulic methods with higher simulation precision have been gradually developed [43,44,45]. With the rapid development of computer technology, the numerical modelling in Saint-Venant equations has been fully developed and extensively applied in scientific research, which has been a central topic in hydraulic research [46,47,48]. On the basis of these studies, 2D models have been extensively employed in flood simulation and water resource management [44,49,50]. These models can not only provide more hydrological spatial information but also better simulate the hydraulic properties. In recent years, benefiting from the continuous enrichment of remote sensing and 3S (GIS, RS, and GPS) technology, hydrologists and hydraulic researchers are paying increasing attention to 2D models, and hydraulic models were rapidly enhanced [51,52,53,54,55].

In this paper, the XAJ and 2D models were established to compare and explore the performance of the hydrological and hydraulic models. The Chengcun basin was taken as the study area. Results were evaluated by comparing observed and calculated hydrographs for 10 typical flood events. The purpose of this study was to provide a theoretical basis for flood control and disaster reduction. This paper is structured as follows. Section 2 describes the data, study area, and the methods. Section 3 discusses the results of the XAJ and 2D models. In Section 4, we summarize the conclusions.

2. Materials and Methods

2.1. Hydrological Model (XAJ Model)

As a lumped rainfall–runoff model, the XAJ model based on a saturation–excess runoff generation mechanism was developed by Zhao et al. [18]. Figure 1 shows the flowchart of the XAJ model. The inputs of the XAJ model are precipitation (P) and potential evaporation (EM), and the output is the discharge (Figure 1). The model consists of four parts: evapotranspiration, runoff generation, runoff separation, and runoff routing.

Evapotranspiration: A three-layer evapotranspiration approach was adopted to calculate the evaporation, and the soil profile was divided into upper, lower, and deep soil layers. Their evaporation values were EU, EL, and ED, respectively. Total evaporation E could be obtained as follows:

E = EU + EL + ED,

(1)

Runoff generation: Runoff generation is one of the most important modules in the XAJ model, and the saturation–excess mechanism was employed, which can be represented by the following equations:

$$\frac{\mathrm{fa}}{FA}=1-{\left(1-\frac{W_{\mathrm{m}}^{\prime }}{W_{\mathrm{mm}}^{\prime }}\right)}^{\mathrm{B}},$$

(2)

$$\{\begin{array}{cc}if P-E+A<W_{\mathrm{mm}}^{\prime }& Then: R=P-E-WM+W_0+WM{\left[1-\left(P-E+A\right)/W_{\mathrm{mm}}^{\prime }\right]}^{1+B}\\ if P-E+A\ge W_{\mathrm{mm}}^{\prime }& Then:R=P-E-WM+W_0\end{array},$$

(3)

where $\mathrm{fa}$ is the partial area where soil moisture storage capacity is less than or equal to $W_{\mathrm{m}}^{\prime }$ (km²), $FA$ is the basin area (km²), $W_{\mathrm{m}}^{\prime }$ is the storage capacity at a point (mm), $W_{\mathrm{mm}}^{\prime }$ is the maximal value of $W_{\mathrm{m}}^{\prime }$, $\mathrm{B}$ is the exponent of the spatial distribution curve of tension water storage capacity, $P$ is rainfall (mm),$E$ is evapotranspiration (mm), $A$ is the initial state of the basin, $R$ is runoff yield (mm), $WM$ is the average tension water capacity (mm), $W_0$ is the soil water storage at the beginning (mm).

Runoff separation: After runoff generation, the total runoff is divided into surface runoff, interflow, and groundwater runoff with a free water capacity distribution curve:

$$\frac{\mathrm{FS}}{FR}=1-{\left(1-\frac{SM{F}^{\prime }}{\mathrm{SMMF}}\right)}^{\mathrm{EX}},$$

(4)

where $\mathrm{FS}$ is the area of the basin free water storage capacity that is less than or equal to $SM{F}^{\prime }$ (km²), $FR$ is the area where runoff yield is more than 0 (km²), $SM{F}^{\prime }$ is the free water storage capacity (mm), $\mathrm{SMMF}$ is the maximal value of $SM{F}^{\prime }$ (mm), and $\mathrm{EX}$ is the exponent of the free water capacity curve influencing the development of the saturated area.

Runoff routing: The surface runoff is routed by the instantaneous unit hydrograph or by the lag-and-route method [56,57], while the interflow and groundwater runoff are routed by linear reservoirs [58]. The outflow of each sub-basin is routed to the basin outlet with the Muskingum method [59].

2.2. Two-Dimensional Hydraulic Model (2D Model)

The 2D model solves the 2D shallow water equations (SWEs) on the basis of the conservation of mass principles and momentum conservation under steady-state flow conditions. The two-dimensional unsteady flow equation is composed of mass balance and the momentum balance, and can be represented as follows [54,60,61]:

$$\frac{\partial q}{\partial \mathrm{t}}+\frac{\partial \mathrm{f}}{\partial \mathrm{x}}+\frac{\partial \mathrm{g}}{\partial \mathrm{y}}=\mathrm{st},$$

(5)

where $q$ represents flow, f is the flux in the x direction, g is the flux in the y direction $, \mathrm{t}$ denotes the time, and $\mathrm{st} \mathrm{stands} \mathrm{for}$ the source terms in which

$$q=\left[\begin{array}{c}d\\ du\\ dv\end{array}\right],\mathrm{f}=\left[\begin{array}{c}du\\ u^{2}d+\frac{1}{2}\\ uvd\end{array}g{d}^2\right], \mathrm{g}=\left[\begin{array}{c}dv\\ dvu\\ d{v}^2+\frac{1}{2}g{d}^2\end{array}\right]$$

(6)

$$\mathrm{st}=\left[\begin{array}{c}\begin{array}{c}r\\ -\frac{{\phi }_x}{\rho }-gdB{F}_x\end{array}\\ -\frac{{\phi }_y}{\rho }-gdB{F}_y\end{array}\right],$$

(7)

$$B{F}_x=\frac{\partial sp}{{\partial }_x},B{F}_y=\frac{\partial \mathrm{sp}}{{\partial }_y},$$

(8)

$${\phi }_x=\rho C_{r}u\sqrt{u^2+v^2},{\phi }_y=\rho C_{r}v\sqrt{u^2+v^2},$$

(9)

where $d$ is the depth of the flow, $u$ and $v$ are flow velocities,$g$ is gravitational acceleration, r is the surface runoff, ${\tau }_{mx}$ and ${\tau }_{my}$ are the bed friction stresses, $\rho$ is the density of water, $B{F}_x$ and $B{F}_y$ is the bottom friction in the $x$ and $y$ directions, ${\phi }_x$ and ${\phi }_y$ are the bed friction stresses in the $x$ and $y$ directions, $\mathrm{sp}$ is the slope, $C_r$ is the bed roughness coefficient ($C_r=g{n}^2/d^{\frac{1}{3}}$), and $n$ is Manning’s coefficient.

The above equations are solved by the Godunov-type scheme and finite volume methods:

$${{\displaystyle \int }}_{\mathsf{\Omega }} \frac{\partial q}{\partial \mathrm{t}}\mathrm{dV}+{{\displaystyle \int }}_{\mathsf{\Gamma }} \mathrm{f}\left(q\right)\cdot n\mathsf{d}\mathsf{\Gamma }+{{\displaystyle \int }}_{\mathsf{\Gamma }} \mathrm{g}\left(q\right)\cdot n\mathsf{d}\mathsf{\Gamma }={{\displaystyle \int }}_{\mathsf{\Omega }} \mathrm{s}\left(q\right)\mathrm{dV},$$

(10)

where $\mathsf{\Omega }$ is the computational domain, $\mathrm{n}$ is the normal vector of the boundary, $\mathrm{and}$ $\mathrm{dV}$ and $\mathrm{d}\mathsf{\Gamma }$ are the area integration element and line integration element.

Equation (10) can be discretized as follows:

$$q^{n+1}=q^n-\frac{∆t}{A}{{\displaystyle \sum }}_{i=1}^m\left[f_n^i\left(q\right)+g_n^i\left(q\right)\right]L^i+∆t\cdot \mathrm{s}\left(q\right),$$

(11)

where $\mathrm{n}$ is the outer normal vector of the boundary of cell, ∆t is the present time step,$A$ is the cell area (m²), i is the cell edge’s index, $m$ is the present time level, and $L$ is the corresponding cell edge length.

The HLLC approximate Riemann solver was adopted to solve the numerical fluxes because it is suited to cases involving a wet/dry interface.

2.3. Study Area and Data Description

Chengcun basin is situated in the south of Huangshan city, Anhui province, China (29°32′–29°44′ N, 117°38′–117°53′ E) (Figure 2). The basin area is 290 square kilometers, and the longest river is 36 km. The Chengcun basin has a subtropical monsoon humid climate with four distinctive seasons. Average annual precipitation in the basin is about 2100 mm with uneven rainfall distribution. Annual precipitation is mainly concentrated in the flood season, which is from April to September, accounting for 70% of the total precipitation. The terrain is high-altitude in the south, and low-altitude in the north with a large relative height difference, and the average elevation is about 583 m. The mean slope gradient of the river channel is 0.95%. The vegetation in the basin is good (Figure 3), and the vegetation types mainly include evergreen coniferous, deciduous broadleaved, and mixed forests [62].

The dataset included rainfall, evaporation, discharge, DEM, and land use. The hydrological data (rainfall, evaporation, and discharge) were taken from regional hydrological administrations or the hydrological yearbooks. The DEM dataset was provided by the site of the Geospatial Data Cloud, Computer Network Information Center, Chinese Academy of Sciences (http://www.gscloud.cn) (accessed on 22 April 2022). Land use data come from a NASA website (https://earthexplorer.usgs.gov) (accessed on 22 April 2022).

2.4. Data Preprocessing and Model Settings

XAJ model data preprocessing and model settings mainly included sub-basin division, hydrological data processing, and parameter calibration. According to the modeling idea of the XAJ model, the entire study area was divided into sub-basins according to DEM and GIS. The average areal rainfall was calculated with the Thiessen polygonal method [63]. The parameter estimation (PEST) automatic calibration algorithm was adopted to calibrate the model with a MATLAB environment [64]. This algorithm had the advantages of the inverse Hessian method and the steepest descent method, and can obtain the optimal parameter results through fewer model runs [65]. During parameter optimization, we considered three elements: the parameter boundaries, the objective function, and the termination conditions. The parameter boundaries were identified by the recommended parameter range and their physical meaning (

Table 2. Optimized model parameters based on PEST.

	Parameter	Physical Meaning	Range and Units [66,67,68]	Value
				(Hourly Model) [69]
Evapotranspiration	K	Evaporation coefficient	0.5–1.1 (-)	0.8
	C	Evaporation coefficient of the deep layer	0.1–0.3 (-)	0.1
	WUM	Tension water capacity of upper layer	5–100 (mm)	17
	WLM	Tension water capacity of lower layer	50–300 (mm)	87
	WDM	Tension water capacity of deep layer	5–100 (mm)	33
Runoff generation	B	Representation of the nonuniformity of the spatial distribution of the tension water capacity	0.1–2 (-)	0.3
	IMP	Proportion of impervious surface	0.01–0.1 (%)	0.01
Runoff separation	SM	Mean free water storage capacity	5–100 (mm)	97
	EX	The distribution of free water storage capacity	1–1.5 (-)	1.5
Runoff routing	KSS	Outflow coefficient of the free water storage reservoir to interflow	0.01–0.7 (-)	0.4
	KG	Outflow coefficient of the free water storage reservoir to groundwater	0.01–0.7 (-)	0.3
	KKI	Recession coefficient of the interflow	0.05–0.95 (-)	0.9
	KKG	Recession coefficient of the groundwater	0.9–0.999 (-)	0.98

Sum KSS + KG was kept as 0.7.

). The objective function was presented as follows:

$$Obj={{\displaystyle \sum }}_{i=1}^n{\left(Q_{sim,i}-Q_{obq,i}\right)}^2,$$

(12)

where $Obj$ is the objective function, $Q_{sim,i}$ is the simulated discharge at time $i$, and $Q_{obq,i}$ is the observed discharge at time $i$.

In our study, 10 flood events were used for model calibration (six events) and validation (four events) at a 1 h time step. The details can be found in Appendix A. The calibration parameters are summarized in Table 2.

The data preprocessing of the 2D hydraulic model mainly included land use and Manning’s coefficient processing, rainfall data processing, and model setting. (1) According to 30 m Landsat image data, considering the influence of land use on roughness [58], the supervised classification method was used to classify the land use types into five categories: water, cultivated land, urban land, forest, and grass land. The Manning’s coefficient of different land uses is shown in

Table 3. Manning’s coefficients corresponding to different land use types [54,70].

Land Use	Manning’s Coefficient (s/[m1/3])
Urban land	0.013
Water	0.021
Grassland	0.031
Cultivated land	0.041
Forest land	0.139

. (2) Rainfall after deducting vertical loss was used as the surface flow of the 2D model. (3) The 2D model adopted the Godunov-type scheme to calculate Equations (5)–(11). On the basis of the Courant–Friedrichs–Lewy (CFL) criterion, the adaptive time step method was implemented. The output time step was 1 h.

2.5. Modeling Evaluation Criteria

According to the accuracy standard in China, the Nash–Sutcliffe efficiency (NSE), percentage error of flood volume (RE), percentage error of peak discharge (PE), and the time difference of arrival flood peak ($∆\mathrm{T}$) were employed to estimate the model performance.

$$\mathrm{NSE}=1-\frac{{\displaystyle \sum }{{(Q}_o{-Q}_s)}^2}{{\displaystyle \sum }{{(Q}_o-\overline{Q_o})}^2} ,$$

(13)

$$\mathrm{RE}=\left|\frac{{\displaystyle \sum }{Q}_s-{\displaystyle \sum }{Q}_o}{{\displaystyle \sum }{Q}_o}\times 100\%\right| ,$$

(14)

$$\mathrm{PE}=\left|\frac{MAX\left(Q_s\right)-MAX\left(Q_o\right)}{MAX\left(Q_o\right)}\times 100\%\right|,$$

(15)

$$∆\mathrm{T}=\left|T_{MaxOb}-T_{MaxSi}\right| ,$$

(16)

where $Q_o$ is the observed discharge, $Q_s$ is the simulated discharge, $\overline{Q_o}$ is the mean of observed discharge, $T_{MaxOb}$ is the arrival time of observed peak discharge, and $T_{MaxSi}$ is the arrival time of simulated peak discharge.

According to the hydrological modelling accuracy standard and the time scale of this paper [71], if the PE/RE was less than 20%, the simulation results are acceptable. The results were classified into three levels as follows: the first level (PE/RE ≤ 10%; NSE ≥ 0.8), the second level (10% < PE/RE ≤ 15%; 0.7 ≤ NSE < 0.8) and the third level (15% < PE/RE ≤ 20%; 0.6 ≤ NSE < 0.7). Otherwise, the results were unsatisfying.

3. Results and Discussion

3.1. Land Use Change in Chengcun Basin

The stability and consistency of rainfall runoff in the basin greatly impacted the model’s results, which were also strongly related to land use change. Hence, it was critical to analyze the land use change in the basin before analyzing the flood simulation results. The change in land use during the study period is shown in

Table 4. Comparison of land use areas over the study period.

Land Use	1990		1991		1992		1993		1994		1995		1996
	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)	Area $\left(\mathbf{k}{\mathbf{m}}^2\right)$	Ratio(%)
Cultivated land	1.44	0.51	1.62	0.58	1.70	0.61	1.80	0.64	1.80	0.64	1.75	0.62	1.68	0.60
Forest	279.33	99.43	279.14	99.36	279.06	99.33	278.95	99.30	278.96	99.30	279.02	99.32	279.09	99.35
Grass land	0.11	0.04	0.11	0.04	0.09	0.03	0.09	0.03	0.09	0.03	0.08	0.03	0.08	0.03
Water	0.01	0.00	0.02	0.01	0.03	0.01	0.02	0.01	0.01	0.00	0.02	0.01	0.01	0.00
Urban land	0.04	0.01	0.05	0.02	0.06	0.02	0.07	0.02	0.07	0.02	0.07	0.02	0.07	0.02

. There were five land use types, and forest land accounted for the largest proportion (about 99%). The land use type is very suitable for the saturated runoff mechanism. Furthermore, the degree of land-use change was not obvious over the years, so it had little impact on the consistency of the hydrological data. These characteristics were very beneficial for our research, and this is another reason why we chose this study area.

3.2. Comparing the Results of the Hydrological and Hydraulic Models

Considering the availability and consistency of hydrological data, 10 flood events were selected to validate the models. Results were evaluated with the four statistical indices mentioned in Section 2.5. The indices were calculated and are shown in

Table 5. Performance of the XAJ and 2D hydraulic models.

Flood Code	Rainfall (mm)	RE (%)		PE (%)		NSE		$\mathbf{∆}\mathbf{T} \left(\mathbf{h}\right)$
		XAJ	2D	XAJ	2D	XAJ	2D	XAJ	2D
19900614	112.88	31.61	36.13	10.58	13.54	0.85	0.77	0	0
19900626	246.09	18.23	4.68	18.86	6.02	0.90	0.85	7	3
19910518	158.97	8.24	26.17	13.80	7.21	0.86	0.64	1	3
19920516	81.58	19.45	19.75	25.45	17.27	0.83	0.56	0	3
19920701	225.62	31.45	1.16	6.25	15.03	0.87	0.57	0	6
19930629	566.69	17.16	2.51	14.62	11.61	0.90	0.83	1	4
19940608	574.15	14.96	10.38	5.14	19.53	0.83	0.80	2	7
19950519	500.78	17.10	9.02	24.17	12.56	0.69	0.64	1	3
19950701	160.98	26.76	6.41	18.70	11.13	0.79	0.80	1	3
19960619	169.21	2.99	19.12	20.84	1.81	0.62	0.91	0	4

.

As shown in Table 5: (1) The REs of the XAJ and 2D models were mostly within 20% (XAJ: 7/10, 2D: 9/10), indicating that the two models had achieved satisfactory accuracy in water balance. However, both models had large errors during the first flood event. This phenomenon may have been due to the antecedent precipitation data, and may have needed further improvement. (2) In terms of PE, the 2D model obtained better results with all qualified simulations, while the XAJ model achieved a 60% qualification rate. This demonstrated the strengths of the 2D hydraulic model in confluence. (3) In terms of ΔT, the XAJ model achieved good simulation performance of the peak occurrence time (within 3 h except 19900626, which may have been because of the uneven distribution of rainfall or errors in the meteorological data). (4) From the perspective of NSE, the results of the XAJ model were better. NSE values of XAJ ranged from 0.62 to 0.90 with a mean of 0.81, while they ranged from 0.56 to 0.91 with a mean of 0.75 in the 2D model. These results indicate that the XAJ model was able to reproduce the flood process once the optimal parameters had been obtained. The hydrological model had better overall discharge results. (5) According to the hydrological modeling accuracy standard, in terms of NSE, the percentages of the first, second, and third levels were 60%, 10% and 30%, respectively, for the XAJ model, and those values were 40%, 30%, and 10% for the 2D model. Taken as a whole, both models showed good performance that could reflect the magnitude and trend of the flood process.

The scatter plot of the simulated and observed discharges is shown in Figure 4. The plot indicates that the results of the both models were within acceptable ranges ((R²: 0.85 and 0.77, respectively). Among them, the fitting line of XAJ model was close to the 1:1 line, indicating that the simulated discharge was closest to the observed one. The fitting line of the 2D model was also within a reasonable range. However, the simulated value was too large at low flow, and was low at high flow. These errors were due to both anthropogenic activities (agricultural irrigation, Figure 3 and Table 4) and rainfall intensity (Table 5) [69,72].

The discharge hydrographs are shown in Figure 5. The observed and simulated discharges agreed well, indicating that the determined parameters could basically reflect the characteristics of runoff generation and confluence in the basin. The XAJ model did not fit the regression process well for some flood events due to its generalization in confluence. How to integrate the 2D model into the confluence of the XAJ model is a point worth investigating.

In addition to the discharge at the basin outlet, the 2D model could also provide more information about the water depth and inundation range within the basin. The maximal water depth of the basin was analyzed and presented in Figure 6. The maximal submerged water depth of the flood events (19,930,629) was about 2 m. The flow was distributed in the river channel, and there were also varying degrees of inundation in the surrounding areas of the river channel. The flood spread significantly to both sides across the river, and the residential areas were considered to be at risk of flooding. These results provide additional information for understanding flood movement and flood mitigation.

4. Conclusions

Flood disasters pose a tremendous threat to human life and property. Flood simulation and prediction play a huge role in flood control and disaster mitigation. Hydrological and hydraulic models are major nonengineering methods for flood analysis and management, and their importance is self-evident. In this paper, a hydrological (XAJ) model and a two-dimensional hydraulic (2D) model were established on the basis of measured hydrological, surface-elevation, land-use, and other data in the study area. The simulation performance of the two models for flood volume, flood peak, and arrival time (flood peak) was compared. The results (Table 5) show that our approach is a useful strategy for flood simulation in the Chengcun basin. Compared with the XAJ model, the 2D model achieved much better performance in terms of flood peak (mean PE:11.00% < 16.45%). The XAJ model, on the other hand, performed better in terms of goodness of fit (Figure 4, mean NSE:0.79 > 0.76). These results demonstrate that the XAJ model had more advantages in the overall effect of the flood simulation, while the 2D model was more prominent in flood local feature simulation. The spatial inundation map could also be primarily captured by the 2D model. The residential areas are at risk of being in inundation (Figure 6), which indicates that these areas urgently need to strengthen the river channel.

This study provides theoretical reference for the combined application of hydrological and hydraulic models. However, due to limited conditions, there are still things that need further research or improvement: The Manning’s coefficients of different land use types were directly adopted from the related research. These coefficients may be different from the actual roughness of the basin. In addition, the effect of antecedent precipitation on the results should be further considered. Furthermore, due to the limitations of the data, the flood events that could be used in this paper were limited, so the obtained conclusions may have certain errors. With the future supplementation of data and the improvement of the models, these conclusions should be further tested and revised.