Impacts of River Network Connectivity on Flood Signatures and Severity Regulated by Flood Control Projects

Abstract

The operation of hydraulic projects within plain river networks to mitigate floods can alter river network connectivity patterns, subsequently affecting flood processes. This study employed the MIKE 11 model to simulate flood processes under three different river network connectivity scenarios. Based on the simulations, we propose a method to evaluate flood intensity severity by integrating three flood characteristic indices: Slope of the Flow Duration Curve (SFDC), Rising Climb Index (RCI), and Flashiness Index (FI). These indices assess the overall magnitude of change, the rate of rise, and process fluctuations, respectively. Results indicate that changes in river network connectivity significantly impact RCI and SFDC, more than FI. Compared to the natural river network connectivity mode, changes in urban or watershed river network connectivity resulted in a significant decrease in RCI values by 3–37% or 18–38% across various return periods, with the rate of change in RCI values increasing as the return period lengthened. The impact of urban river network connectivity changes on SFDC within the Changzhou urban area was more pronounced under high-magnitude storm conditions, causing a 61% reduction. Furthermore, changes in watershed river network connectivity had a larger effect on SFDC under low-magnitude storm conditions than under high-intensity storms. Over 80% of the rivers under natural connectivity conditions exhibited flood intensity severity of Level III or higher, particularly in the Chenshu–Qingyang area. The alterations in connectivity significantly decreased flood intensity severity, with 85% to 91% of rivers showing the lowest flood intensity severity of Level I. Under a 100-year rainstorm scenario, flood risk shifted from within the flood protection envelope to outside it in the Changzhou urban area. The results will provide an important scientific basis for regional flood management in plains with dense rivers.

1. Introduction

With the continuous expansion of towns and cities, high-intensity human activities have dramatically altered the morphology and structure of rivers within delta regions. This has led to significant changes in the connectivity of river networks, a critical factor that defines how water flows through and between different components of a watershed [1,2]. River network connectivity refers to the physical and functional linkages between rivers, streams, and other water bodies, which determine the movement and distribution of water, sediments, and nutrients. When connectivity is disrupted or altered it can lead to fragmentation of the water system, which in turn affects hydrologic processes within the watershed [3,4].

Numerous hydraulic projects have been constructed in the river network to mitigate flooding threats [5,6,7]. To protect urban areas from flooding, many regions have developed a large number of water conservancy projects in key urban areas, creating relatively closed and adjustable flood control units [8,9]. The construction of such projects is mainly concentrated along rivers, lakes, and delta areas affected by tides, particularly in the Yangtze River Delta [10,11]. With the acceleration of urbanization and the gradual improvement of flood control construction standards, various flood control projects have become essential components of regional flood prevention and control efforts [12,13,14].

In recent years, the impact of artificial regulation, particularly through hydraulic engineering construction, on river network connectivity has been increasing [15,16,17,18]. River fragmentation caused by the construction of hydraulic projects (such as dams and reservoirs) is a major contributor to the loss of river connectivity [19]. River network connectivity also plays an important role in flood formation. While flood control projects have successfully reduced flood peaks, they have also complicated water circulation paths and altered the structure and connectivity of river networks throughout the region [20,21]. For example, Zhang et al. (2015) established a relationship between the river fractal dimension and the frequency of flood events, revealing that in the Hangzhou region, the frequency of floods tends to increase as the fractal dimension of the river network decreases [22].

Our aim is to explore the variations of flood characteristics and flood intensity severity under different river network connectivity patterns regulated by the operation of hydraulic projects. The research results will provide an important scientific basis for regional flood management and disaster prevention and mitigation.

2. Materials and Methods

2.1. Study Area

The Taihu Lake Basin, situated in Eastern China, comprises the water conservancy sub-districts: Wu-Cheng-Xi-Yu (WCXY), Yang-Cheng-Dian-Mao (YCDM), Hang-Jia-Hu (HJH), Hu-Xi (HX), Zhe-Xi (ZX), Pu-Dong and Pu-Xi (PD and PX) (Figure 1a). Among these, the WCXY District in the northern part of the basin is characterized by a typical plain river network (Figure 1b). To mitigate severe flood disasters and protect vital urban areas such as Changzhou and Wuxi, extensive urban-based flood control projects (UFCPs) have been implemented. The Changzhou UFCP covers approximately 156 km², while the Wuxi UFCP covers about 144 km² within the WCXY District. Additionally, several watershed-based flood control projects (WFCPs) are positioned along the Yangtze River to manage floodwater discharge. Details of UFCPs and WFCPs are provided in Table S1, and their spatial distributions are illustrated in Figure 1c. Various hydraulic projects throughout the region regulate river flows, altering the connectivity pattern of the river network.

2.2. Methods

2.2.1. Modeling Scenarios of River Network Connectivity Regulated by Engineering

Flood protection projects in river network areas can be categorized into two main types: watershed-based projects that enhance drainage capacity and urban-based projects that ensure flood safety in key urban areas [23]. Based on this classification, the connectivity modes of river networks are generalized into three scenarios, considering the locations of hydraulic protection projects:

Scenario 1: The flooding process is simulated by the model under natural river network connectivity conditions, without any flood control projects.

Scenario 2: The flooding process is simulated by the model with changes in river network connectivity inside and outside urban areas when urban-based flood control projects are activated. The changes in watershed connectivity occur in this scenario.

Scenario 3: The flooding process is simulated by the model with changes in river network connectivity inside and outside the watershed when watershed-based flood control projects are activated. The changes in urban connectivity occur in this scenario.

In this study, the MIKE 11 model is employed to simulate the flooding process under different river network connectivity scenarios. The MIKE 11 model is highly effective in simulating the flow processes of both tributaries and main streams, making it a valuable tool for analyzing the hydrological behavior of river channels [10,24]. Moreover, the model is coupled with the NAM (Nedbør–Afstrømnings Model), which considers the impact of land use on runoff generation [25]. The NAM enhances the model’s ability to simulate how different land cover types influence the hydrological processes, particularly in generating surface runoff. This study quantitatively evaluates the model’s simulation performance using the relative peak error of flood level (PE), the relative root mean square error (RRMSE), and the Nash–Sutcliffe efficiency coefficient (NSE). PE represents the difference between the simulated and measured peak water levels, with smaller values indicating better simulation accuracy. An NSE value greater than 0.75 and an RRMSE value less than 0.2 indicate that the model effectively simulates the flood process.

2.2.2. Flood Signatures

The actual shape of the hydrologic curve visually characterizes changes in water levels over time [26]. The Slope of the Flow Duration Curve (SFDC), Rising Climb Index (RCI), and Flashiness Index (FI) are three commonly used indices to characterize floods in terms of the overall magnitude of change, the rate of water rise, and the volatility of change, respectively [27].

The RCI indicates the speed of flood peak formation, with larger values suggesting a faster rise in flood water levels over a short period [28]. The SFDC is calculated by determining the slope of the water level ephemeral curve, where higher SFDC values correspond to higher overall flood process values [29,30,31]. The FI reflects the magnitude of fluctuations in flood levels, with larger values indicating more intense fluctuations [32]. The equations for these three indices are as follows.

$$\mathrm{R}\mathrm{C}\mathrm{I}={\displaystyle \frac{w_p-w_0}{\mathsf{\Delta }t}},$$

(1)

$$\mathrm{S}\mathrm{F}\mathrm{D}\mathrm{C}={\displaystyle \frac{\ln \left(w33\%\right)-\ln \left(w66\%\right)}{0.66-0.33}},$$

(2)

$$\mathrm{F}\mathrm{I}={\displaystyle \frac{\sum _{i=1}^n\left|w_i-w_{i-1}\right|}{\sum _{i=1}^n{w}_i}}$$

(3)

where w_p is the peak water level of the hydrological process; w₀ is the starting level; Δt is the interval from the starting level to the peak level. w33% and w66% are the 33rd and 66th quartiles of the water level process, respectively; w_i is the water level at the ith moment.

2.2.3. Severity Assessment of Flood Intensity

In this study, three flood signature indices were utilized to explore variations in flood processes under different river network connectivity modes with artificial regulation. However, a single flood index cannot fully capture the complexity of flood processes. Therefore, this study combines the three indices to propose a quantitative method for assessing flood intensity severity. Figure 2 shows a schematic diagram of flood intensity severity.

Generally, a flood process characterized by a high overall water level, a fast rate of rise, and high fluctuations indicates a high probability of triggering a flood disaster. In other words, the severity of the flood intensity is high. Thus, this study classifies flood intensity severity into eight levels based on the combination of high or low values of RCI, SFDC, and FI, considering the rate of water rise, the overall magnitude of change, and fluctuations, as shown in Figure 2. The combination of high values for RCI, SFDC, and FI indicates greater flood intensity severity. To address inconsistencies in magnitude between the flood characteristic indices, RCI, SFDC, and FI are standardized to a range of 0 to 1 in this study.

Level VIII: When the values of the RCI, SFDC, and FI indicators for a flood process are greater than 0.5, it indicates that the flood process has a high rate of rise, high variability, and a high average water level. This makes it highly susceptible to flooding, indicating a flood process with high severity of flood intensity.

······

Level I: When the values of the RCI, SFDC, and FI indicators for a flood process are less than 0.5, it indicates that the flood process has a low rate of rise, low fluctuation, and low average water level. This corresponds to a flood process with low severity of flood intensity.

The severity of flood intensity increases in order from Level I to Level VIII.

2.3. Data Description

Daily rainfall and water level data from 1980 to 2020 were obtained from the hydrological yearbook of the Taihu Lake Basin and were used to estimate design rainfall and water level for return periods of 10, 20, 50, and 100 years. Hourly rainfall and water level data from 2015 to 2020 were obtained from telemetry station data and used in the construction of the MIKE model. River network data were derived from a 1:50,000 digital map from the 2010s and refined through high-resolution remote sensing imagery calibration.

3. Results

3.1. Model Design

3.1.1. Calibration and Validation

The study selected observed data from two flood events (21–25 June 2016 and 1–19 July 2016) to calibrate the parameters of the MIKE 11 model. Additionally, observed data from three flood events (14–19 September 2016, 28 September–4 October 2016, and 25–30 October 2016) were used to validate the model parameters. During the calibration period, the average NSE, PE, and RRMSE were 0.85, 2%, and 0.02, respectively, across the seven sites. Similarly, during the validation period, these metrics were 0.84, 4%, and 0.03, demonstrating effective simulation by the constructed MIKE 11 model. Figure 3 presents flood process curves for selected stations during both calibration and validation periods, showing generally consistent trends between simulated and observed water levels.

3.1.2. Scenario Setting

Based on the annual maximum rainfall records from each rainfall station in the WCXY District spanning 1980 to 2020, the Generalized Extreme Value (GEV) function was employed to estimate design rainfall for return periods of 10, 20, 50, and 100 years. This design rainfall served as input for the MIKE 11 model to assess how changes in river network connectivity affect flooding dynamics under varying intensities of heavy rainfall. As depicted in Figure 4, the spatial distribution of rainfall across different return periods exhibits a gradual decrease from west to east. The “20150616” rainstorm event, notable for its high intensity, short duration, and widespread impact, represents a rare heavy rainfall event in the study area. Therefore, this event (15–17 June 2016) can be used as a time period allocation for modeling rainfall events of different magnitudes. The MIKE model simulates changes in flood processes, including alterations in flood signatures and the severity of flood intensities, under different storm magnitudes across three connectivity modes, illustrated in Figure 5.

3.2. Impact of Changes in Connectivity Modes on Flood Processes

3.2.1. Water Level Processes at Representative Stations

Figure 6 illustrates the variation in flood characteristics across different return periods of rainfall. Adjacent to the violin plot, a rotated kernel density plot shows the probability density distribution at various values, with wider areas indicating the distribution’s range. Thick black bars represent the interquartile range, and white dots denote the medians. As rainstorm intensity increases (from 10-year to 100-year events), significant strengthening is observed in flood peaks (PW) and RCI values under the three river network connectivity modes, accompanied by a notable shortening of peak occurrence times (PT). Conversely, changes in SFDC and FI indices are less pronounced (Figure 6c,e).

The flood process curves for different river network connectivity patterns are shown in Figure 7. The flooding process intensifies as the rainstorm’s intensity increases (from 10-year to 100-year storms). In Scenario 2, involving changes in connectivity within and outside the watershed, the activation of WFCPs has minimal impact on hydrological processes at regional representative stations. This limited impact is likely due to the site’s distance from the Yangtze River, reducing their exposure to the hydraulic projects along the river. However, in Scenario 3, after the activation of UFCPs, connectivity within and outside the urban river network is disrupted, leading to more variable flood processes at the stations. For instance, the flood processes at SBJ and NM stations, located within the UFCP coverage area, changed significantly, with peak water levels dropping substantially (Figure 7e,f). Specifically, the peak water level at SBJ station decreased by 35, 47, 49, and 59 cm for 10-, 20-, 50-, and 100-year rainstorms, respectively, and at NM station, the peak water level decreased by 18, 28, 39, and 46 cm, respectively.

3.2.2. Spatial Variations in Flood Signatures

The spatial distribution of flood signature indices for different return periods is shown in Figure 8 and Figure 9. After changes in river network connectivity, the alteration of flood characteristics is primarily concentrated in the Qingyang–Chenshu area, Changzhou urban area, and the regions along the river. The rates of change in flood characteristics for these areas are detailed in

Table 1. Variations of RCI in the different conditions.

	Scenario 1				Variations between Scenario 2 vs. Scenario 1 (%)				Variations between Scenario 3 vs. Scenario 1 (%)
Region	10a	20a	50a	100a	10a	20a	50a	100a	10a	20a	50a	100a
Chenshu–Qingyang	0.74	0.86	1.19	1.22	14.6	2.5	−3.4	8.0	−21	−18	−38	−25
Changzhou core region	0.64	0.68	0.80	0.89	−3	−24	−28	−37	−2.5	−0.2	−3.4	−2.5
Along Yangtze River	0.68	0.89	1.12	1.14	5.0	0.0	1.1	−0.7	−6.9	−9.5	−17	−7.2

and

Table 2. Variations of SFDC in the different conditions.

	Scenario 1				Variations between Scenario 2 vs. Scenario 1 (%)				Variations between Scenario 3 vs. Scenario 1 (%)
Region	10a	20a	50a	100a	10a	20a	50a	100a	10a	20a	50a	100a
Chenshu–Qingyang	0.29	0.29	0.29	0.29	−3.6	−4.4	−4.4	−1.5	−32	−21	0.1	0.9
Changzhou core region	0.36	0.36	0.36	0.35	−4.2	−8.0	−23	−61	−19	−12	−0.4	1.5
Along Yangtze River	0.30	0.30	0.31	0.30	0	0	0	0	−12	−10	−5.0	−2.6

. Compared to the simulation results of natural river network connectivity in Scenario 1, the change in urban river network connectivity in Scenario 2 led to a significant decrease in the RCI values within the influence of the Changzhou UFCP. The RCI values decreased by 3–37% across different rainstorm return periods, with the rate of change in the RCI values increasing as the return period lengthened. In Scenario 3, with changes in watershed river network connectivity regulated by the WFCP, the RCI values decreased by 18–38% compared to the natural river network connectivity results. For instance, the average RCI value in the Qingyang–Chenshu area during natural river network connectivity was 1.2 (under 50- and 100-year scenarios). After the changes in river network connectivity of Scenario 3, the average RCI values decreased to 0.8, marking a 33% reduction on average, and the flood level rise was significantly reduced. Additionally, in this scenario, the connectivity changes inside and outside the watershed also reduced RCI values along the river, although the effect was less pronounced than in the Qingyang–Chenshu area.

In Scenario 2, changes in urban river network connectivity significantly affected the SFDC in the Changzhou core area of influence under high-magnitude storm conditions (100-year rainfall), resulting in a 61% decrease in the SFDC in this area. In Scenario 3, changes in river network connectivity within and outside the watershed under low-magnitude storm conditions (10- and 20-year rainfall) reduced SFDC values in the Changzhou urban area and the Ganlu–Chenshu area. Conversely, under high-magnitude storm conditions, changes in river network connectivity within and outside the watershed had little effect on SFDC values.

3.3. Variations in Flood Intensity Severity Influenced by Different Connectivity Modes

Figure 10 shows the spatial variation in the severity of flood intensity due to changes in river network connectivity enabled by flood control projects (WFCP and UFCP). Under the natural river network connectivity condition, the region, especially the Chenshu–Qingyang area, experiences high overall flood intensity (Figure 10a). More than 80% of the rivers in the region exhibit high flood severity levels (Level III or higher, i.e., Level III to Level VIII). Among these, under high-intensity rainstorm scenarios (50-year and 100-year events), about 10% of the rivers face flood hazards of Level VII and Level VIII, indicating a high likelihood of severe flooding and significant associated hazards.

Following the activation of flood protection projects, changes in river network connectivity have led to a significant reduction in flood intensity severity in the region, particularly under the 10-year, 20-year, and 50-year storm scenarios (Figure 10b). Post-intervention, 85% to 91% of the rivers exhibit the lowest flood severity of Level I. It is important to note that Level I does not imply an absence of flooding but rather a lower severity of flood intensity compared to other areas. While rivers within the Changzhou urban area are protected from flooding under the 100-year storm scenario, rivers outside the Changzhou urban area continue to exhibit higher flood severity.

4. Discussion

4.1. Method Design

Numerous studies have examined the response mechanisms of flood processes under the regulation of hydraulic engineering [33,34,35,36]. These studies primarily describe flood process changes in terms of flood line shape, flood peak, and peak present time, with fewer analyses focusing on flood process line characteristics [37,38,39]. Additionally, previous research has often centered on changes in a single flood characterization index, which can be insufficient for comprehensively characterizing flood process changes. For instance, the SFDC index shows minimal change under different storm scenarios, despite the expectation that flood intensity should increase with storm intensity. This limitation arises because SFDC describes the flooding process generally but does not capture peak water level intensity. Therefore, it is necessary to combine SFDC and RCI indices, as higher values of both SFDC and RCI indicate greater flood intensity. Generally, a higher overall water level, faster water rise rate, and greater fluctuation in the water level process line increase the likelihood of triggering flood disasters and elevate flood risk.

In this study, three flood signature indices are combined to explore changes in flood intensity severity under different scenarios, considering the rate of rise, overall magnitude of change, and variability of fluctuations in the flood process. This method overcomes the limitations of using a single index and better reflects flood intensity. The spatial map of flood hazard classification developed in this study can also support regional flood warnings. For areas with high flood risk, it is crucial to provide early warnings of heavy rainfall and floods and implement related water conservancy facilities to protect people’s lives and properties.

4.2. Implication and Limitation

The UFCP functions primarily to protect against higher-magnitude storm flooding. For low-magnitude storm conditions, the UFCP is generally not activated, or only some of the sluices are activated. Consequently, the internal and external connectivity of the urban river network changes less, resulting in minimal impact on the flooding process. In recent years, the construction of numerous UFCPs has forcibly cut off original river connections, reducing regional storage capacity. During high-intensity rainstorms, UFCPs can effectively reduce the river’s flood level within the city, delay peak appearance times, and alleviate urban flood danger to some extent. However, maintaining low safe water levels within urban areas can create significant water level differences between inside and outside the UFCP, transferring flood risk from within the city to outside areas, thus increasing flood pressure outside the city.

This study’s methodology has some limitations. In addition to indicators that characterize the flooding process, such as flood depth and flood duration, flood severity is usually related to the extent of damage to infrastructure (e.g., roads, bridges, buildings, and utility systems). Additionally, the number of deaths and injuries, as well as population displacement, are important indicators of flood severity. The economic impact of floods, including property damage, loss of livelihoods, and recovery costs, is another critical aspect of flood severity. Floods can also have serious ecological consequences, affecting vegetation, wildlife, and water quality. Assessing these impacts contributes to a comprehensive understanding of flood severity. In future assessments, we will consider multiple aspects to provide a comprehensive evaluation. Understanding flood severity helps authorities and communities better prepare for and respond to future flood events and implement effective mitigation measures.

5. Conclusions

In this study, the impacts of different river network connectivity modes on flood processes, including flood signatures and severity, under flood control project scheduling were simulated using a hydrological and hydrodynamic model. The main conclusions are as follows.

(1)

As rainstorm intensity increases (from 10-year to 100-year events), significant strengthening is observed in flood peaks and RCI values under the three connectivity modes, accompanied by a notable shortening of peak occurrence times. Compared to RCI and SFDC, FI is less variable across the three connectivity modes. After changes in river network connectivity, alterations in flood characteristics are primarily concentrated in the Qingyang–Chenshu area, Changzhou core area, and regions along the river.

(2)

Compared to the natural river network connectivity mode, changes in urban connectivity resulted in a 3–37% decrease in RCI values for storms with varying return periods, with the rate of change increasing alongside the return period. Changes in watershed connectivity led to an 18–38% reduction in RCI values. Changes in urban connectivity significantly affected the SFDC in the Changzhou core area of influence under high-magnitude storm conditions (100-year rainfall), resulting in a 61% decrease in the SFDC in this area. Conversely, under high-magnitude storm conditions, changes in river network connectivity within and outside the watershed had little effect on SFDC values.

(3)

The region experiences high flood severity levels under the natural river network connectivity condition. More than 80% of the rivers in the region exhibit high flood severity levels (Level III or higher). Following the activation of flood protection projects, changes in river network connectivity have led to a significant reduction in flood intensity severity in the region, with 85% to 91% of the rivers exhibiting the lowest flood severity of Level I. While rivers within the Changzhou urban area are protected from flooding under the 100-year storm scenario, rivers outside the Changzhou urban area continue to exhibit higher flood severity.

Our study found that the alteration of river network connectivity at the commissioning of flood control projects has a significant impact on flood characteristics. In future studies, efforts should be made to effectively meet the requirements of urban flood control by optimizing the artificial regulation strategies of flood control projects while ensuring the sustainable use and management of regional water resources.