Potential of Fish Habitat Resilience Under Hydrodynamic Regulation of a Plain Urban River Network in Shanghai City, China

Abstract

Cities in plain areas have small slopes at the bottoms of rivers, with weak hydrodynamics, heavy pollution and poor self-purification capacities for the restoration of biological habitats. Hydrodynamic and water quality improvements are effective means for the ecological restoration of plain urban rivers. The potential for fish habitat resilience in a typical urban river network plain (more than 130 river sections) in the Dianbei part of China was studied. The tolerant fish, Carassius auratus (C. auratus), and the sensitive fishes Trachidermus fasciatus (T. fasciatus) and Anguilla japonica (A. japonica), were selected as the protection targets, and hydrodynamic factors, river morphology and water quality factors were chosen as environmental indicators. With the fish habitat suitability index, a fish habitat resilience potential evaluation model was established. The response of the habitat resilience potential index (HRPI) to hydrodynamic regulation was subsequently analyzed, and the HRPI indicated an increased habitat resilience potential with its value increasing from 0 to 1. Overall, the resilience potential of tolerant fish species was greater than that of sensitive species in the Dianbei. For the HRPI of C. auratus adults (tolerant species), approximately 62.8% of the river sections were above 0.6 (high resilience level) and were concentrated in the northwest area of the river network. While for the resilience potential of A. japonica adults and T. fasciatus adults (sensitive species), only 60% of the river sections exhibited moderate resilience level (HRPI > 0.5). The average dimensionless habitat resilience potential index (AHRPI) was enhanced by water diversion with its values increased by 10.3%, 9.3% and 12.7% for C. auratus adults, T. fasciatus adults and A. japonica adults, respectively. The habitat resilience potential of C. auratus changed little during the spawning period, which indicated that the effect of hydrodynamic regulation was limited. This study provides a scientific basis for managers to restore urban river network habitats in plain areas.

1. Introduction

River ecosystems function in regulating climate, maintaining hydrological cycles and nutrient cycles and providing primary production [1]. Urban rivers refer to rivers or river sections that originate within a city or flow through a city, and are usually adapted from natural rivers. Urban rivers have the strongest function in the coordination of human activities and natural processes [2,3]. Human disturbances to rivers, including river channelization or constraints by sluice pumps, are increasing with urbanization [4,5]. This type of disturbance changes hydrodynamics, deteriorates water quality, and decreases the stability of urban river ecosystems [6,7]. The rivers located in plain areas are particularly impacted by urbanization due to the dense population and developed industry.

In China, most plain river network areas are highly urbanized, and their water environment and ecological problems are serious [8]. Flat plain terrain and small slopes at the river bottom result in small water level differences, weak hydrodynamic conditions, and backflow and overflow in plain river networks [9,10,11,12]. The spatial structure diversity of the river network is weakened, simplifying the exchange between urban rivers and surrounding ecosystems [13]. In densely populated areas, only a few very long rivers remain free-flowing [14], and many hydraulic structures, such as gates and pumping stations, exist in urbanized areas. The flow of water bodies is easily affected by dispatching, and the river system becomes disordered [15], decreasing the self-purification ability of the river [16]. The pollutant loads from rain-washing runoff on impervious surfaces and the overall flow of combined sewers in urbanized areas are usually high [17,18,19]. The loss of river species diversity is caused by species-regional effects associated with urbanization [20], hindering the restoration of biological habitats such as those of river fish.

The main strategies for ecological restoration in plain urban rivers include reducing the pollutant load, replenishing water, and remediating the environment [21]. Water replenishment by means of water diversion projects is one of the most frequently used technologies to improve the water quality of urban rivers [22]. Water diversion has been implemented in many urban rivers, such as the Sumida River in Tokyo, the Seonakdong River in Korea, and the Severn and Thames rivers in Britain [23,24]. In China, water diversion has been applied in Taihu Lake, the Pearl River and other areas [25,26,27]. Hydrodynamic regulation through the joint dispatch of sluice pumps has also been applied in plain urban river networks, significantly improving hydrodynamic conditions and water quality [24,28,29]. On the one hand, the hydrodynamic characteristics of plain river networks, such as water velocity and water level, improved quickly because of hydrodynamic regulation. On the other hand, the river re-oxygenation ability was enhanced, thereby increasing the pollutant degradation rate. The diffusion and dilution of pollutants were also promoted through water diversion [30].

At present, most evaluations of the effects of hydrodynamic regulation engineering have focused on improving water quality [31,32,33], while the resilience potential of fish habitats with improved hydrodynamic environments has been less addressed [34,35]. To more easily consider nonlinear dynamics in observable ecosystems, the concept of “resilience” was introduced into ecology by Holling in 1973 for the first time; that is, the capacity of a system to recover and return to its original state under pressure [36]. Over time, studies on resilience have been carried out; however, the concept is unclear, and resilience is divided into engineering resilience and ecological resilience [37]. The habitat resilience potential was quantified through continuous ecosystem modelling studies, and new criteria were added based on their own research characteristics and tendencies [38,39]. The habitat resilience potential will be induced and applied in this study.

Considering the weak hydrodynamic power, poor water quality, destruction of biological habitats and many hydrodynamic regulation projects in plain urban rivers, characterizing the impact of hydrodynamic regulation on the ecological restoration of fish habitats is necessary. In this study, the following three steps were taken: (1) establish a model for fish habitat restoration potential to evaluate river health; (2) explore the effects of hydrodynamics on fish habitat restoration potential; and (3) identify key factors limiting fish habitat restoration.

2. Materials and Methods

2.1. Study Area

The study area is the urban river network plain in the Dianbei part of Shanghai (31°7′~31°16′ N, 121°15′~121°28′ E), which is one of the 14 water conservation parts in Shanghai. With the Dianpu River and Suzhou River as the north-south boundaries, and the Xiaolaigang River as the west boundary, all the rivers flow into the Huanpu River (Figure 1a). The study area is approximately 163.35 km² with a population of approximately four million people, and it is located in the central area of Shanghai city. Gates and flinty riverbanks have been built in all the major rivers; some riverbanks have been overrun, and the water-crossing sections are narrow in the river network (Figure 1b). Small fishes, such as Gambusia affinis and Belontiidae, which can live in surface waters for a long time, have appeared in the studied rivers.

2.2. Water Diversion, Monitoring Programs and Data Collection

The water diversion was mainly driven by the tidal power of the Huangpu River, which is a tidally influenced river that injects into the East China Sea. The tidal rhythm both provides the required water and the power for water movement. The enhanced water diversion was conducted in Dianbei from 25 December 2015 to 4 January 2016. Water was drawn from the Suzhou River of the northern boundary and the Dianpu River of the southern boundary at the two high tides until the water level reached 3.5 m above sea level inside the gates in the Suzhou River, with the water gates along the Huanpu River of the eastern boundary locked. Water was drained once at the second low tide, with the water gates along the Huanpu River opening (Figure 1c). Routine water diversion was conducted from 4 January to 8 January 2016. Water was drawn from the Suzhou River and the Dianpu River at the two high tides until a water level of 3.1 m was reached at the inside of the gates in the Suzhou River, with the water gates along the Huanpu River of the eastern boundary locked. Water was drained twice at the two low tides with the water gates along the Huanpu River opened.

Hydrological and hydrodynamic data, including data on the operation of the fifteen sluice-pump stations, water level, and flow velocity, were continuously collected by the local water conservancy bureau during the period of water diversion to validate the hydrodynamic model. Water samples at eleven sample sites were collected across the river network (Figure 1). Water samples were collected on 24 December (before water diversion) 2015, 30 December 2015 (enhanced water diversion), and 7 January 2016 (routine water diversion) to evaluate the improvement in water quality caused by enhanced water diversion. At each sampling site, a surface water sample was taken at 0.5–1.0 m depth underwater using a 5 L Plexiglas water sampler, and the water samples were stored in an icebox, brought to the laboratory, and analyzed immediately. The analyzed water parameters included dissolved oxygen (DO), ammonia nitrogen (NH₃-N), the permanganate index (COD_Mn), and total phosphorus (TP). The DO concentration was monitored on site via portable YSI electrodes (Xylem Co. New York, NY, USA), and the NH₃-N concentration was detected via ion chromatography (ICS-2000, Dionex Company, Sunnyvale, CA, USA). The COD_Mn was determined via back titration with organic matter in raw water samples oxidized by potassium permanganate. The TP concentration was determined via the acid-molybdenum-blue colorimetric method with persulfate digestion of raw water samples.

Geographic data was also collected from 1:2000 electronic maps and digital elevation models (DEMs), high-precision remote sensing images or aerial photographs and land use maps of the research area in 2015. The fish data was collected from the Fishes Compendium in Shanghai, fishery resource monitoring reports from the Institute of Hydrobiology of the Chinese Academy of Sciences, and Shanghai Ocean University. The fish catalogs include national or local protected species, rare and endangered species, endemic species, long-distance migratory fishes, and other protection targets. The life stage of fish species has also been described in the Fishes Compendium fishery resource monitoring reports.

2.3. Evaluation Model for Fish Habitat Restoration Potential

2.3.1. Indicator Species

Fish can indicate the stages of river ecological restoration. In general, common fishes are considered to be characteristic fishes during river ecological restoration, and local sensitive fishes that have appeared before but disappeared are considered characteristic fishes in the future stage of restoration. One of local tolerant fish species, C. auratus, was selected as the targeted species for fish habitat restoration. With respect to the survival of local sensitive fish species in Shanghai, T. fasciatus and Anguilla japonica (A. japonica) were selected. T. fasciatus and A. japonica have been found to enter the sea to reproduce from freshwater rivers and then migrate upstream to freshwater rivers for growth. Therefore, three fish indicator species were selected: (1) the tolerant fish C. auratus (spawning and feeding period), and (2) the sensitive fish T. fasciatus (adult) and A. japonica (adult).

2.3.2. Survey of Key Habitat Factors

The hydrological indicators (depth and velocity), river morphology (substrate) and water quality factors (NH₃-N and DO) were selected as the key habitat factors for fish growth and reproduction [40,41], as shown in

Table A1. Key habitat factors for fish habitats.

Type	Index	Range of Study Area	Ecological Significance
Hydrological	velocity (m/s)	0–2 m/s	It can affect the concentration of nutrients in the water body, stimulate the swimming and route of migratory fish, and provide suitable conditions for floating fish eggs to hatch downstream.
	depth (m)	0–25 m	Water depth provides a suitable moving space for bottom-dwelling fish and a hatching environment for sinking eggs.
River morphology	substrate	rock or hard channels	One of the sources of nutrients needed by aquatic animals; the distribution of bottom-dwelling locations and spawning grounds that fish species choose are impacted of substrate styles.
Water quality	DO (mg/L)	0.6–8 mg/L	Breathing and feeding activities are reduced and, at the same time, immunity, growth, and metabolism activities are affected by hypoxic stress of fish when the DO content in the river is too low. However, bubble disease in fish will be caused by a DO concentration that is too high.
	NH3-N (mg/L)	0.8–4.5 mg/L	Fish species poisoning will take place, their respiratory function will be inhibited, and their nerve centers will be damaged, which is manifested by a significant reduction in food intake. Moreover, their immune function will be impaired, various abnormal physiological and biochemical indicators will be shown, and the fish will die.

in Appendix A. The river network was segmented according to the key habitat factors and then measured in accordance with the flow state of the river on a free scale. The critical environmental factors that affect the habitat restoration of target fishes were collected, including the abovementioned factors and other indicators, such as the characteristics of the riverbank, discharge, near land use, and botanical data. The habitat demand range for each influencing factor of the target fish species was determined based on expert opinions and historical data, and the factors were distinguished and graded. A field survey plan (

Table A2. Large-scale field survey plan for fish habitats in the plain urban river network.

Habitat Type	Used Type			Urbanization Level		Discharge
Stream segment:		(Shipping)	□	Very low	□	Very low	□
Beginning coordinate:		Electricity	□	Low	□	Low	□
Ending coordinate:	Water supply		□	Low to moderate	□	Low to moderate	□
Fishway:				Moderate	□	Moderate	□
Yes				Moderate to high	□	Moderate to high	□
No				Very high	□	Very high	□
DO (mg/L)	CODMn (mg/L)			NH3-N (mg/L)		TP (mg/L)
≥7.5	□	≤2	□	≤0.15	□	≤0.02	□
6–7.5	□	2–4	□	0.15–0.5	□	0.02–0.1	□
5–6	□	4–6	□	0.5–1	□	0.1–0.2	□
3–5	□	6–10	□	1–1.5	□	0.2–0.3	□
2–3	□	10–15	□	1.5–2	□	0.3–0.4	□
<2	□	>15	□	>2	□	>0.4	□
Depth	Velocity			Substrate
0–30 cm	□	0–10 cm/s	□	Mucky clay, <0.063 mm			□
20–50 cm	□	10–30 cm/s	□	Silt, 0.063–2.0 mm			□
40–70 cm	□	20–50 cm/s	□	Small gravel, 2.0–6.0 mm			□
60–120 cm	□	40–70 cm/s	□	Moderate gravel, 6.0–20.0 mm			□
100–200 cm	□	>70 cm/s	□	Big graval, 2.0–6.0 cm			□
>200 cm	□			Big cobblestone, 12.0–20 cm Boulder, >20.0 cm Rock or hard channel			□
Porosity filling		Suspended solids
1		1
2		2
3		3
4		4
5		5
Riparian vegetation:						Riverbank type:
Coverage rates are higher than 80% and more than 5 plant types						Cement channel
The planting coverage rate is between 60~80%, and there are 4 or 5 plant types						Stone assembled
The planting coverage rate is between 40~60%, and there are 3 or 4 plant types						Vegetation
The planting coverage rate is between 20~40%, and there are 2 or 3 plant types						Nature
The penetration rate is less than 20% and there was only a uniform plant type						Mucky clay

in Appendix A) for large-scale fish habitats in the plain urban river network was established, which provided data support for the application of fuzzy rules and subsequent model establishment.

2.3.3. Hydrodynamic Model

The river simulation software Infoworks Integrated Catchment Management (ICM, version 7.0) was used for hydrodynamic calculations. The ICM discretizes the Saint-Venant equations based on the Preissman four-point implicit scheme, in which the upstream boundary of the model uses the discharge as the control condition and the downstream boundary uses the water level. The hydrodynamic model for the river network covers a total area of 178 km², with a total of approximately 120 river stations constructed. All dispatch data during the field observation period were used. After calibration and verification, the simulated results from a one-dimensional hydrodynamic mathematical model in Dianbei was used for subsequent habitat simulation.

2.3.4. Evaluation Model for Fish Habitat Resilience Potential Based on Fish Habitat Suitability

The evaluation model for fish habitat resilience potential in the urban river network plain was based on fish habitat suitability and was divided into two parts: (1) a fish habitat suitability model, and (2) an evaluation model for fish habitat resilience potential. In the model, data on water depth and flow velocity from hydrodynamic modelling results were used, monitoring data on water quality parameters in the major river sections during the water diversion period, and data on riverbank characteristics and riverbed geomorphology from the field survey. The river network was divided into 738 sections to construct a data platform for simulating the potential for fish habitat restoration.

For the fish habitat suitability model (HSM), flow velocity (v), water depth (d) and substrate (sub) were selected as indicators of the habitat in natural rivers less impacted by pollutants. For the fish habitat resilience potential model (HRPM), factors of water quality were extended in HSIM due to their importance in urban rivers in this research. Based on expert knowledge and literature data, fuzzy sets were formulated for three indicator species of different life stages, namely, C. auratus (feeding and spawning period), T. fasciatus adults, and A. japonica adults. Fuzzy sets were defined and described by membership functions. Unlike classical sets with definite boundaries, the boundaries of the membership functions of fuzzy sets usually overlap. The key indicators were classified into three levels: low (L), moderate (M), and high (H). The habitat suitability index (HSI) was classified into five levels: very low (VL), low (L), moderate (M), high (H), and very high (VH). Fuzzy rules have two parts: conditions (IF), and results (THEN). The fuzzy reasoning process adopted the Mamdani method [42] to realize nonlinear mapping from the input factors to the habitat suitability index. The defuzzification process adopted the center of gravity method, which has good tolerance. The HSI that was simulated was characterized by a dimensionless value of 0–1, where 0 represented “very low” suitability and 1 represented “very high” suitability. These values represented the lower and upper thresholds of habitat suitability, respectively. The general steps of the fish habitat suitability model are shown in Figure 2a, and the fuzzy sets and fuzzy rules for C. auratus during the spawning period are shown in Figure 2b.

For the evaluation model of the fish habitat resilience potential, a three-stage calculation method (Figure 2c) was used to construct a habitat resilience potential index (HRPI). The HSI together with two water quality parameters (NH₃-N and DO) were selected as the input parameters for the next stage to establish a new round of fuzzy set formulation. Finally, the simulated HRPI was achieved as the model output.

For comparison, the weighted usable area (WUA) was divided by the total plaque area to obtain the dimensionless average habitat resilience potential index (AHRPI) via the following equation, with the retuned values ranging from 0 to 1.

$$AHRPI={\displaystyle \frac{WUA}{{\sum }_{j=1}^N{A}_j}}$$

where A_j = the area of unit j, (m²); HSI_j = the habitat suitability index of unit j; and N = the number of units.

3. Results

3.1. Habitat Characteristics and Water Quality Changes with Water Diversion

According to the large-scale habitat survey (Figure 3), 72.6% of the river sections had a velocity of level 1, that is, 0~0.1 m/s. Only 1.3% of the river sections were at level 5, with the Huangpu River at over 0.7 m/s. The Caohejing and Puhuitang Rivers flow quickly in the network. For water depth, 2.2% of the river sections ranged from 0 to 0.5 m, and 56.5% ranged from 1.5 to 3 m. The depths of the Suzhou River, Dianpu River, and Huangpu River in the northern, southern and eastern boundaries were greater than 3 m.

Riverbanks were divided into five classes: cement channels, assembled stone, silts, vegetation, and natural vegetation for slope protection. The riparian vegetation was divided into five grades according to the vegetation type and coverage rate (Table A2). Substrates of most of the riverbed were cement channels with small differences in the river network.

The results for DO, COD_Mn, NH₃-N and TP at the eleven sampling sites are shown in Figure 4. The DO concentration at the nine sampling sites was less than 5.0 mg/L, and NH₃-N pollution was severe, with all sampling sites exceeding 2.0 mg/L before water diversion. The COD_Mn concentration at all the sampling sites was greater than 5.0 mg/L. The TP concentration at ten sampling sites before water diversion was greater than 0.2 mg/L. The DO concentration at most of the monitoring sites increased during the period of enhanced water diversion and decreased during the period of routine water diversion. The NH₃-N concentration at most of the monitoring sites decreased during the period of enhanced water diversion but did not recover during the period of routine water diversion. Overall, COD_Mn did not significantly increase or decrease during water diversion. However, TP presented an increasing trend during water diversion.

3.2. Fish Habitat Restoration Potential Resilience (HRPI) with Water Diversion

In terms of tolerant fish species, the HRPI was generally high in the Dianbei. For the HRPI of C. auratus adults (Figure 5a–c), approximately 62.8% of the indices were above 0.6 and were concentrated in the northwest area of the river network. Most of the area exhibited moderate resilience potential during the spawning period of C. auratus (Figure 5d–f), with an HRPI between 0.4 and 0.6. The larger rivers in the boundary, e.g., the Suzhou River, Dianpu River, and Huangpu River, presented high levels of HRPI. The HRPI of the two sensitive adult fishes was lower than that of the tolerable species. The resilience potential of A. japonica adults (Figure 5g–i) and T. fasciatus adults (Figure 5j–l) was lower, with only approximately 60% of the river sections exhibiting moderate resilience.

The fish habitat resilience potential improved to different degrees in response to water diversion for C. auratus adults, T. fasciatus adults, and A. japonica adults. The level of HRPI during the period of enhanced water diversion increased greatly, with the number of river sections at high levels increased, and the number of river sections at low levels decreased. An obvious improvement occurred in the southwestern area of the river network. For C. auratus adults, the proportion of river sections at high and very high levels increased from 62.8% to 97.6% (Figure 5a–c). When the hydrodynamic conditions were changed to the case of routine water diversion, the HRPI only slightly decreased compared with the period of enhanced water diversion. However, for C. auratus in the spawning period, the HRPI did not obviously increase with water diversion (Figure 5d–f).

Table 1. AHRPI of three target fish species in three phases.

Date	CAA	CAS	AJA	TFA
24 December 2015	0.6759	0.5389	0.5172	0.5269
30 December 2015	0.7709	0.5221	0.5505	0.5645
7 January 2016	0.7455	0.5440	0.5650	0.5938

Notes: CAA indicates C. auratus adults, CAS indicates C. auratus in the spawning period, AJA indicates A. japonica adults (AJA), and TFA indicates T. fasciatus adults.

summarizes the AHRPI of the target fish species before water diversion, enhanced water diversion, and routine water diversion. With water diversion, the AHRPI increased by 10.3%, 9.3%, and 12.7% for C. auratus adults, T. fasciatus adults and A. japonica adults, respectively, except for that of C. auratus during the spawning period.

4. Discussion

4.1. Feasibility of Improving Fish Habitat Resilience Potential Under Hydrodynamic Regulation

The fish habitat resilience potential could be improved by hydrodynamic regulation. For example, the construction of overflowing dams positively affects fish habitats by improving water quality conditions. However, the natural conditions for fish spawning require a certain velocity and complex flow patterns. River segments with turbulent flow, variable velocity and direction are usually favorable. C. auratus requires a stimulated velocity of approximately 1.1 m/s for spawn, but the average flow rate in the river network is less than 0.6 m/s, which is too low to meet the ecological demand for spawn. Only several river sections, e.g., the Suzhou River and Huangpu River, had higher velocities exceeding 1.1 m/s, with higher habitat resilience potential according to the model calculations (Figure 5).

Tolerant fishes have low hydrological and water quality requirements, and they can survive in environments with high levels of pollution and low levels of dissolved oxygen. Except for the spawning period, their demand for velocity is not rigid. For sensitive fishes, A. japonica has relatively high water and sediment quality requirements, and it is difficult for them to survive in environments with low dissolved oxygen. T. fasciatus has stricter requirements for pollutant concentrations, with high ammonia nitrogen greatly impacting these species, while they are relatively more tolerant to the substrate.

Water diversion can increase the water level and velocity in some river sections, increase the dissolved oxygen concentration, and dilute pollutants such as ammonia, all of which would benefit fish habitat restoration. However, some pollutants, such as COD_Mn and TP, did not decrease. This is probably caused by the disturbance of sediments during hydrodynamic regulation. If the return of sensitive fishes is expected, further water quality improvement and substrate repair measures could be necessary. In the studied area, most river channels, except several large rivers, cannot meet the ecological needs of C. auratus during the spawning period.

During the period of enhanced water diversion, the sluices on the southern bank of the Dianpu River and the northern bank of the Suzhou River draw water at full capacity twice at the high tides of one day, and the sluices along the western bank of the Huangpu River fully discharge once at the low tide of the night. Hydrodynamic regulation enhanced the water flow of the river network. The flow velocity was greater than 0.05 m/s in 65% of the river sections during the water-drawing stage and in 75% of the rivers during the water-draining stage. The habitat resilience potential of C. auratus adults, A. japonica and T. fasciatus improved during hydrodynamic regulation (Figure 5), particularly in the southeastern area of the river networks, where the resilience potential increased by one level. During the period of routine water diversion, the resilience potential of the three-indicator species subsequently decreased. These data can support the subsequent determination of the threshold of ecological restoration in the Dianbei.

As water flows in the river network change with the tide cycle in the Huangpu River, the fish habitat restoration potential fluctuates with the tidal rhythm. When the water level decreases, the fish habitat resilience potential decreases. When the water level increases, the fish habitat resilience potential increases. Therefore, the suitability of the target fish habitat and the response curve of the hydrodynamic parameters are positively correlated. However, improvements in the resilience potential of fish habitats via hydrodynamic regulation are still limited. Managers could change the water diversion strategy and intensity according to requirements for ecological restoration in different periods, combined with further improvements in the water quality and suitability of the substrate.

4.2. River Habitat Restoration Potential Evaluation Model Feedback

The protection and recovery of endangered species are important for restoring biological diversity. The responses of aquatic species to habitats must be determined [43,44], with hydrological conditions and water quality as two principal aspects of habitat metrics. It is necessary to improve or at least maintain habitats for the restoration of aquatic ecological systems [43]; hence, the environmental factors affecting river systems need to be discerned [45], and suitable indicators must be chosen to evaluate habitat suitability [46]. Previous models have emphasized the prediction of species-level indices more than the selection of the most important habitat factors for restoration [47]. In this research, further simulated scenarios were conducted, in which water quality and hydrodynamics were used as individual control variables (

Table 2. AHRPI of three target fish species with hydrodynamic factors and water quality as individual control variables.

	Hydrodynamic Factors as the Control Variable				Water Quality as the Control Variable
	CAA	CAS	AJA	TFA	CAA	CAS	AJA	TFA
24 December 2015	0.6759	0.5389	0.5172	0.5269	0.6759	0.5389	0.5172	0.5269
30 December 2015	0.7610	0.4727	0.5494	0.5678	0.7154	0.4489	0.5190	0.5113
7 January 2016	0.7416	0.4346	0.5673	0.6030	0.7142	0.4833	0.5144	0.5171

Notes: CAA indicates C. auratus adults, CAS indicates C. auratus in the spawning period, AJA indicates A. japonica adults (AJA), and TFA indicates T. fasciatus adults.

). Then, simulation results of these two scenarios were compared with the simulation results from the integrated variables of water quality and hydrodynamics. Water quality was found to be the most important contributing factor for the resilience potential of fish habitats. With no change in hydrodynamic conditions, improved water quality obviously increased the habitat resilience potential of the three indicator fishes in the adult stage. For the sensitive adult fish, T. fasciatus, the potential increased by 14.4% on average in the river network. However, adjusting only the hydrodynamic conditions had a poor effect on the habitat resilience of sensitive fish. For C. auratus in the spawning period, which has high hydrodynamic requirements, the integrated regulation of hydrodynamics and water quality can achieve positive results, although it only results in an increase of 0.95%. However, independent regulation of water quality or hydrodynamic conditions results in a negative increase of more than 10%. The area of the river channel is believed to account for these results. The advantage of the increased hydrodynamic force of the slender channel of C. auratus during the spawning period cannot offset the slight decline in hydrodynamic conditions in the main channel. Therefore, water quality is the preferred habitat factor for habitat resilience potential.

The relative importance of hydrodynamic conditions and water quality can also be observed in different areas of the river network. The water quality in the southeastern areas was worse, while the water flow conditions were better. The HRPI of the three indicator fishes in the adult period was not high, but that of C. auratus in the spawning period was greater because hydrological conditions such as velocity and water level were the priority habitat factors for this fish in the spawning stage. For C. auratus adults and the two sensitive adult fishes, hydrological conditions had little effect, whereas water quality was the limiting factor. During water diversion, the DO concentration increased, and the NH₃-N concentration decreased in the southeastern area. However, owing to hydrological limitations in this area, the habitat resilience potential of C. auratus during the spawning period has decreased. In terms of the average fish habitat resilience potential changing with tides (Figure 6), the sensitive A. japonica responded poorly to changes in water level. Because A. japonica needs to migrate and desires to inhabit soil holes and stone crevices, better sediment conditions and flow rate stimulation are considered important for their recovery. For T. fasciatus, the connectivity of the river networks was reduced by building dams or water gates, and their migration routes were cut off as a cause of their endangerment.

Innovatively, this research has established the evaluation model for fish habitat resilience potential to explore changes in the eco-health of urban river network. Particularly, the urban river network is featured in crisscrossed rivers with weak flow, hydrodynamic regulation with a great number of sluice gates, and heavy pollution from urbanized areas.

According to the simulation results, hydrodynamic regulation not only needs to consider improvements in hydrodynamic conditions and water quality, it also needs to pay attention to the activities of fish at different stages of the life cycle, which could be crucial for enhancing the potential for habitat restoration. For river management, the connectivity in areas with poor resilience potential can be determined through analysis of the spatial distribution of the resilience potential of fish habitats across the river network. If fishes are unable to complete life activities in some areas, a simulation with local parameter adjustment can be conducted to obtain parameter thresholds, which are fed back to the restoration plan. It is worth noting that this research is limited in the simulation of a short-term improvement in the fish habitat by water diversion, while fish recovery is a long-term process. On this point, the results need to be further verified in future field observation. In addition to this, the ecological flow rate in the river networks could be further calculated to guide the accurate operation of water diversion for the restoration of the fish habitat.

Although the restoration potential of the fish habitat has been improved by water diversion in this research, several important facts for fish habitat restoration have to be emphasized: (1) Water quality is crucial, and great efforts should be made to cleaner water by continuously reducing pollution. (2) Fish migration barriers should be removed for free fish migration. (3) Hydromorphological restoration of rivers that have been altered by man are necessary, aiming for natural rivers where possible with natural banks, meandering, etc.

5. Conclusions

Based on the concept of resilience and theories related to fish habitat suitability, an evaluation model for fish habitat resilience potential was established to explore changes in the eco-health of urban river networks under hydrodynamic regulation. It was demonstrated that the habitat resilience potential of urban river networks can be improved through hydrodynamic improvement. River managers can obtain data to support the restoration of urban river networks by means of water replenishment or other hydrodynamic regulation. The spatial distribution of the fish habitat restoration potential can be used for analysis, the poor parameters can be adjusted, and then the ecological water demand scheduling threshold can be calculated and fed back to the restoration plan. In addition, this model could be applicable to other indicator organisms in other rivers, but different index factors should be selected for different types. Despite the improvement of fish habitats by water diversion in this research, the important facts in the restoration of fish habitats have to be emphasized; for instance, better water quality, removing barriers to free fish migration, and hydromorphological restoration of rivers toward natural rivers where possible.