Drifting with Flow versus Self-Migrating-How Do Young Anadromous Fish Move to the Sea?

The downriver migration process of young anadromous fish has a far-reaching impact on their survival rate and the efficacy of hatchery-reared fish release, but it is poorly understood. Moreover, the impact of dams on the fish remains unclear. The Chinese sturgeon is an anadromous and dam-affected fish in the Yangtze River. Here, we propose a novel theoretical framework to reveal the migration process of young Chinese sturgeon. We clarify the effects of active swimming of fish and water flow on the downriver migration and the parametric traits of the migrational stages. Then, we show that the young fish migrate downriver along the inshore waters in a gradually transforming manner from passive drift to active swimming. Lastly, we evaluate the impact of the Gezhouba Dam (GD) on the migration of the young fish, as well as demonstrate the life cycles of Chinese sturgeon in the Yangtze River pre- and post-GD.

Introduction

Understanding the mechanisms driving an aquatic organism's movement is an essential component in the conservation and management of species and ecosystems. The migration pattern of aquatic organisms plays a fundamental role in the survival of their populations, especially for migratory fish, which usually depend on their swimming ability and the towing capacity of the water flow or both. In the past, water flow was presumed to dominate organisms' movements. However, at present there is a consensus that active swimming, even at seemingly trivial speeds, could have profound consequences for the movements, fitness, and distribution of marine organisms (Fossette et al., 2015, Putman and Mansfield, 2015, Putman et al., 2016). The situation for river organisms, especially anadromous fish, remains poorly understood.

The anadromous fishes, comprising 110 species that live in seas and migrate into fresh water to spawn, play a significant role in linking the river-sea ecosystem (Kynard et al., 2002, Braaten et al., 2008, Braaten et al., 2012, Stoll and Beeck, 2012, Huang and Wang, 2018). The downriver migration process of young fish has far-reaching impacts on the survival of fish and the efficacy of hatchery-reared fish release. There is minimal and fragmentary information concerning local river reach (Braaten et al., 2008, Braaten et al., 2012) mainly due to technical obstacles in sampling, identifying individual ages, tracking the fish (Braaten and Fuller, 2007), and the fact that there is no robust theoretical model. Larvae or juveniles are assumed to act as passive bodies, traveling with the river's flow; however, this method underestimates the weak active swimming ability of young fish (YARSG, 1988, Auer and Baker, 2002, Stoll and Beeck, 2012). Moreover, dams are regarded as a serious threat to anadromous fish, and the mechanism by which dams affect the young fish remains unclear.

Fisheries restocking programs have primarily been applied to bolster stocks by rearing fish in hatcheries and releasing them into the wild. This is at a time when the world's fish species are under threat from habitat degradation and over-exploitation. However, the behavioral deficits displayed by hatchery-reared fish and the resulting poor survival rates in the wild have been noted for over a century (Brown and Day, 2002). In China, artificial restocking of fish and release has been used as the sole remedial measure of dam construction for rescuing rare and endangered fish, and its efficacy has been controversial. Brown and Day (2002) emphasized that the focus of fisheries research must shift from husbandry to improving post-release behavioral performance. Thus, how to assess and improve post-release performance of cultured fish is closely related to a fundamental issue: detailing the migration process of juveniles in the river.

Chinese sturgeon (Acipenser sinensis) is a typical example of the 16 species of anadromous sturgeon globally and is a flagship species of the Yangtze River. The sturgeon was listed as critically endangered by the International Union for Conservation of Nature (IUCN) in 2010 and included in Appendix II of Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES) in 2015. Before the river closure by the Gezhouba Dam (GD), the crucial first and lowermost water project on the Yangtze River on 4 January 1981 (pre-GD), spawning Chinese sturgeons spread in the 800-km reach between Maoshui (Yibin City, Sichuan Province) and Wanzhou (Chongqing Municipality) and spawned at 19 sites. The Chinese sturgeon enters the Yangtze River between June and August every year when the gonadal development of the adult fish reaches stage III (males, 8 years old; females, 13 years or older) while fasting all along the way and migrating 2,850 km to the upstream spawning ground. After breeding in autumn of the following year, 15 months after initially entering the Yangtze River, the adult fish returns to the sea within 1 month (Huang and Wang, 2018). Hatched larvae begin to migrate downriver in October or November each year and reach the estuary in the next summer, all the while foraging along the way and adapting to the saltwater environment of estuarine areas such as the eastern beach of Chongming Island. After 1981 (post-GD), the spawners were forced to lay eggs in a less suitable spawning ground about 30 km downstream of the GD (Figures 1A and 1B). However, with the operation of the cascade dams on the upper reaches of the Yangtze River, especially the Three Gorges and Xiluodu dams, the population of Chinese sturgeon has continually decreased to the verge of extinction (Wu et al., 2015).

Figure 1

Zoning of Migration Path and Spawning Grounds of Chinese Sturgeon Pre- and Post-GD

(A) The Yangtze River basin showing spawning grounds pre- and post-GD and the migration zoning of the juveniles. Here, the bold purple line indicates the migration path of the Chinese sturgeon. River kilometers (km) are used as the unit of distance, days (d) as the unit of time, and number of individuals (ind) as fish quantity. The navigation channel mileage was used to calculate the coordinates of the main cities along the river. Numbers in parentheses below city ports are river distances from the Yangtze mouth (km).

(B) The GD is the first and the lowest dam built on the main stem of the Yangtze River. After 1981 (post-GD), the spawners had to lay eggs in a less suitable spawning ground about 30 km downstream of the GD.

(C) The tidal-affected region below Nanjing consists of freshwater and brackish subregions, demarcated at Wusong or Pu Town indicated by yellow stars. The red stars at Xupu and the eastern beach of Chongming Island indicate the sampling sites of juveniles in the estuarine area. According to estuarine salinity, the tidal-affected region can be further divided into freshwater and brackish subregions. The saltwater intrusion caused by tidal current affects not only the flow velocity but also the salinity. The range of the brackish subregion depends on the interaction between the Yangtze River runoff and the salt tide, near the mouth of the estuary. North and south branches bifurcate the estuary of the Yangtze River. The south is the mainstream area, but the salt tide invades the north with high salinity. In the dry season, the saltwater of the north flows backward and invades the south from the bifurcation point. Therefore, we consider the south branch as the migration path of juveniles from May to August of each year instead of the north branch, which was confirmed by an investigation between 1982 and 1993 showing no juvenile Chinese sturgeon in the north branch (Yi, 1994). Bao and Zhu (2017) calculated the horizontal and vertical salinity distribution of the estuary during the spring tide and the low tide in the 1950s and the 1970s, and in 2012. They concluded that, according to the drinking water salinity standard of 0.45 psu, the upper boundary of saltwater intrusion in the south branch is near Wusong and Pu Town, about 30 km from the East Beach of Chongming Island. However, salinity had less effect on the swimming ability of juveniles over 7 months of age (He et al., 2013).

See also Figure S1.

Before the 1990s, there was insufficient information concerning the population migration or dynamics of Chinese sturgeon except for fishing or bycatch records (YARSG, 1988). Since then, biotelemetry, hydroacoustic detection, and mark-recapture techniques have been widely used in the study of the Chinese sturgeon (Kynard et al., 1995, Yang et al., 2005, Lin, 2008, Wang et al., 2012, Wang et al., 2014). Only the mark-recapture technique can be used to obtain the interval speed between the releasing place to the recapture site of larvae or juveniles, because their body sizes are too small to be monitored (tracked or detected) by biotelemetry or hydroacoustic detection (Yang et al., 2005, Wang et al., 2014). Therefore, we cannot clarify the mechanism of the downriver migration for larvae or juveniles with existing technology. Taking the Chinese sturgeon as a model organism, Huang and Wang (2018) proposed the Migration Dynamics Model (MDM) for anadromous or dam-affected fish and successfully applied it to the spawning upriver and post-spawning downriver migration processes of adult Chinese sturgeon in the Yangtze River. Theoretically, this model can be used for young fish, but juveniles have more complicated behavior than adults, such as avoiding predation and feeding; this is related to ontogenetic behavior and the hydrodynamic impacts. Here we attempt to establish a novel framework for theoretical modeling to analyze the downriver migration process of young Chinese sturgeon and to evaluate the impact of the dams on the migration process.

We used the number of days post-hatching (dph) to characterize the age of the fish rather than the number of days after fertilization. The early life stages of the fish vary greatly in morphology and physiology and are usually divided into three developmental stages, namely: early larva (0–11 dph), late larva (12–39 dph), and juvenile (40 + dph) (Zhuang, 1999, Zhuang et al., 2002). The juveniles more than 40 dph old have external characteristics like the adult fish. Here we also used an overall term, young fish, to refer to these fish at all developmental stages from the hatching site to the estuary, and we set up a coordinate system by taking the Yangtze mouth as the origin point and tracing upriver (Figure 1A).

Results

Considering the mortality of young fish along the migratory path, we derived a modified Migration Dynamics Model (MDM) for the young fish. The key to solving the MDM is to estimate the following three parameters: the migration velocity (U), the diffusion coefficient (D), and the mortality rate (K). Of these, U and D are directed toward the fish as a living body, rather than a passive body (Stoll and Beeck, 2012). Previous studies considered only the speed of the water flow and neglected the swimming ability of the larvae (Erwin and Jacobson, 2015). The swimming ability of the larvae increases with age, whereas the effect of the current on U and D diminishes. Therefore, we considered that the migration speed and the diffusion coefficient of juveniles depend on both the age of the larvae and the water flow. According to the ontogenetic development and ecological behavior of the fish, we divided the downriver migration process into three phases: drift stage, cover stage, and self-migration stage (Figure 2). Of the three, the last stage is predominant in length and time. Based on the river zoning and including the effects of riverbanks and tides on the migration of young fish, we built formulas of divided functions on the three parameters at each stage and then calculated the migration processes of young fish by numerical methods (see Transparent Methods).

Figure 2

Schematic Diagram of the Downriver Migration Process of Young Chinese Sturgeon

(A) The vertical (top) and side (bottom) views of the young fish migration path in the Yangtze River. Here we assume the Yangtze River as an open channel. The dotted lines represent the lateral (top) and vertical (bottom) velocity distributions of the current. The elevation view shows that, in the drift stage, the free-living embryos or early larvae migrate from the middle of the river to both shores while they are moving downward after hatching, and then enter the cover stage during which the early larvae hide on the inshore bottom of the river bed. After 18 dph, the late larvae enter the self-migration stage and then migrate downstream along the inshore waters. The side view shows that the juveniles move vertically to the water surface at 1–2 dph of age while drifting with the current and then move to the riverbed at 3–7 dph until at 8–18 dph when they enter the cracks of riverbed substrate for hiding and avoiding predation. After 18 dph of age, the juveniles migrate downstream, depending on their swimming ability; they are also affected by the current.

(B) The location of the water layer (relative water depth) of the early larvae at different ages (dph) during the drift stage. According to the water depth preference of early larvae (Zhuang, 1999, Zhuang et al., 2002), we assume that the vertical swimming height (dotted line) of larvae is linear with time (or age) at the population level. The larvae reach the middle water layer (relative depth y/H = 0.5) at 0 dph and the water surface (relative depth y/H = 0) at 2 dph. Afterward, the vertical position of the larvae decreases linearly with time (or age), and the larvae reach the bottom of the river at 8 dph, and then enter the cover stage.

(C) During the drift stage, the early larvae migrate laterally from the thalweg to the shores and vertically from the bottom to the surface and then down to the bottom. Their longitudinal migration speed at the population level is the same as the current speed calculated by Equation 4, and their lateral and vertical speeds are age dependent for reaching different locations near the shores (red triangle).

(D) Migration stage division corresponding to ontogenetic development of the Chinese sturgeon after hatching (top) (Zhuang, 1999, Zhuang et al., 2002) and linear length growth of juveniles with age (bottom).

See also Figures S2 and S3, Tables 1 and S1, Video S1.

Table 1

Division of Migration Stages and Model's Parameters of Juvenile Chinese Sturgeon

Daily Age (dph)	Migration Stage/Young Fish	Swimming Traits	Relative Water Depth (y/H)	Migration Speed of Fish U (km/day)		Diffusion Coefficient of Fish D (km2/day)
				Pre-GD	Post-GD
0	Hatchling	Drift with current, accompanied by vertical swimming and transverse swimming up to the water surface and toward the shores due to weak phototaxis	0.5	–	–	–
1	Drift stage/Free-living embryo or early larva		0.25	102.14	73.48	36.3
2			0	78.19	56.25	42.4
3		Drift with current, accompanied by vertical swimming down from the upper water layer to the bottom water layer followed by swimming to the shores due to phototaxis	0.17	52.9	38.06	50.9
4			0.33	32.8	23.59	63.6
5			0.5	17.7	12.74	84.8
6			0.67	7.43	5.34	127.2
7			0.83	1.68	1.21	254.4
8		Enter the bottom water layer in the shores	1	0	0	254.4
–		Sum of drifting distance (km)	–	292.24	210.67	–
9–11	Cover stage/Early larva	Hide in the cracks of riverbed gravel-cobble substrate	1	0	0	254.4
12	Cover stage/Early larva start feeding
13–18	Cover stage/Late larva
19–39	Self-migration stage/Late larva	Restart downriver migration along the inshore waters and bottom water layer relying on the swimming ability, although affected by current	Bottom of riverbed	Equation 12 in Transparent Methods		Equation 20 in Transparent Methods
40–150	Self-migration stage/Juvenile
150–270	Self-migration stage/Juvenile	Enter tidal-affected region

Migration Stage Division and Its Distinguishing Features

(1) Drift stage (0–8 dph): This stage involves mainly the downstream area of the spawning ground. The spawners lay eggs that adhere to the rocks at the bottom of the spawning ground, attracting predators. Five days later, the fertilized eggs hatch and start to move downstream with the water current; the early larvae survive through endogenous nutrition and do not forage. Figure S1A shows the velocity vector of the early larvae, longitudinally drifting with the current, while laterally swimming to shores, as well as vertically going up to the water surface or down to the riverbed (Figure 2A). The longitudinal U of larvae mainly depends on the uneven velocity distribution of the current at different water layers where the larvae stay (Figure 2B). The larvae gradually go up to the surface and then down to the bottom (Figure 2C). Therefore, the drift velocity of the larvae depends on the spatial position of the river cross-section. When they drift longitudinally with the current, the larvae move horizontally to shallow waters near the shore through a weak swing due to their phototaxis. The larvae prefer shallow water for survival and future feeding. Hence, the riverbank has a significant effect on the speed of larval movement.

Figures 2B and 2C show the larvae's locations in cross-sections of the water layer after hatching. According to the water depth preference of early larvae, we assumed that the vertical swimming distance of the larvae is linear with time (or age) at the population level. The larvae reach the middle water layer (relative depth y/H = 0.5) at 0 dph and water surface (relative depth y/H = 0) at 2 dph. Afterward, the vertical position of the larvae decreases linearly with time (or age), and the larvae reach the bottom of the river at 8 dph and then enter the cover stage. Table 1 shows the migration speed (U1) and the diffusion coefficient (D1) of larvae at different ages. We have applied the perched water layer of relative depth value y/H obtained in the laboratory (Zhuang, 1999, Zhuang et al., 2002) to the Yangtze River. This method has been supported by Kynard et al. (2007) and Braaten et al. (2008), whereby the vertical distribution of larval pallid sturgeons in laboratory and field experiments show similar characters. The diffusion coefficient increases rapidly in the drift stage, indicating that the larvae disperse quickly to avert the threat of predators.

(2) Cover stage (9–18 dph): Zhuang (1999) and Zhuang et al. (2002) reported that the early larvae begin to hide in the cracks of gravel-cobbles at 7 dph and reach peak individual numbers at 8–10 dph. The probability of hiding at 11 dph starts to decrease until 18 dph, when most of the larvae leave the cracks. In this stage, the larvae initiate feeding when the yolk sac is exhausted at 11–12 dph. An increase in the duration of the capability to resist the current is observed at the onset of exogenous feeding by the larvae. We considered that the larvae hide in cracks at the riverbed from 9 to 18 dph in the lower reaches of the spawning ground but show diffusion behavior owing to the local eddies. Therefore, the time-averaged velocity of the current (U2) = 0. We took the diffusion coefficient at 7 dph when larvae are near the bottom of the river bed as D2 = 254.4 km2/day (Table 1, also see Transparent Methods), indicating that the larvae are scattered as far as possible in the cover stage, especially after the start of feeding, to reduce the risk of predation.

(3) Self-migration stage (after 18 dph): Based on studies in the laboratory by Zhuang (1999) and Zhuang et al. (2002), we inferred that, after 18 dph, the late larvae begin their inshore migration downstream while searching for rich food in rearing areas on their way. In the self-migration stage, the juveniles' path can be divided into the river region, where the current is mainly determined by the upstream inflow, and the tidal-affected region, where the current is affected by both upstream inflow and the tidal current. Because of the spatiotemporal variability of runoff and the tidal current, the lengths and the origin-destination of the two regions vary with the complicated interaction of river runoff and tidal current. The stronger the runoff is, the more powerfully the freshwater suppresses the tidal current. Then the upper boundary of the tidal-affected region moves downward, or vice versa. Xu et al. (2012) found that the upper boundary of the tidal-affected region should be between Wuhu and Zhenjiang. According to calculations considering the combination of flood season and high tide, the average cross-section velocity below Nanjing is reduced by the tidal current. The time for juveniles to reach the estuary is between May and August. Therefore, the tidal effect on the migration speed of juvenile should occur below Nanjing (347 km). Wang et al. (2014) showed that the downriver speed of juvenile Chinese sturgeon decreased when they entered the tidal-affected region. Therefore, to simplify the calculations, we regarded Nanjing as the fixed demarcation point between the river region and the tidal-affected region, where the latter can be further divided into freshwater and brackish subregions (Figures 1C and S1).

Classification methods of the fish swimming speed correspond to different definitions and indices of swimming speed, such as critical swimming speed, maximum sustained swimming speed, and optimum cruising swimming speed (Wang et al., 2010). The migration speed (U) in the MDM refers to the swimming speed of the juveniles over the ground at the population level, which is distinguished from the critical swimming speed that is widely used. Based on the growth data of juvenile Chinese sturgeon (Zhuang et al., 2002, He et al., 2013), the relationship between full length and age of a juvenile is expressed as L = 0.1256 t +1.4338. Other species of sturgeon show a similar linear relationship (Braaten and Fuller, 2007). When the juveniles are less than 12.5 months old, the critical swimming speed and the age can be approximately expressed as a linear relationship, despite the difference in test conditions leading to different formulae (Figure S2C). Therefore, we assumed that the migration speed of juveniles is a linear function of age (Figure S2D).

Owing to the limitations of observation techniques, it is difficult to obtain from the field environment the spatiotemporal distribution of migration speed (U3) of juveniles after 18 dph. Theoretically, the swimming ability of the fish is related to their body condition, including health, body length, tail length, and swing frequency, as well as environmental conditions such as water temperature, velocity, and velocity gradient of the current. We used the fishing and mark-recapture data to obtain similar expressions for the speed function and took the average value as the migration speed of juveniles. Meanwhile, we assumed that the migration speed of juveniles from Nanjing to the estuary is reduced by 30%, and we introduced a tidal influence coefficient (γ) in the tidal-affected region. In the self-migration stage, the diffusion coefficient (D3) is synthetically determined by the swimming ability and the current speed; the current effect on the juvenile migration speed decreases with age. Therefore, we divided D3 into two parts: the fish-related diffusivity, denoted by Df, is determined by the juvenile swimming ability and increases with age, and the current-related diffusivity (Dw) is determined by the current speed and is attenuated with an increase in age. We have estimated D3 in the self-migration stage (t > 18 dph) (see Transparent Methods).

Vital Functions of Weak Swing or Swimming of Larval Fish

The spawners of Chinese sturgeon usually lay adhesive eggs in the rapids. The eggs are deposited while being fertilized and then adhere to the rocks at the bottom of spawning ground, normally attracting large numbers of predators. Five days later, the fertilized eggs hatch, and the early larvae start to drift downriver with the current. The early larvae, due to their phototaxis, skillfully utilize the bend flow and the mechanical interaction between their weak swing and the current, amplifying the weak swing by dint of the water current, to reach the littoral zone at the end of the drift stage (Figure 2). The larvae have evolved a unique swimming pattern as a survival tactic. Conversely, if the larvae had drifted with the maximum water velocity of the thalweg without approaching the shore by weak swing, they would have reached the brackish subregion at the estuary within about half a month and then would have certainly died, as they would be unable to feed or would be too small to adapt to the saltwater environment. Here, we highlight that the weak swing of early larvae or swimming ability of larvae, ignored in the past, plays a crucial role in leaving the rapids and in antipredator behavior. The larvae prefer shallow waters with rich food for survival and feeding.

Figure 3 shows that the total mortality rate of Chinese sturgeon from egg to 9 months of age is about 99.98% before they enter the sea, which is consistent with the egg-to-1-year mortality rate range of 99.96%–100% for the Gulf sturgeon (A. oxyrinchus desotoi) (Pine et al., 2001). The mortality of young Chinese sturgeon in the Yangtze River plays a crucial role in population recovery. Their mortality risk sources change from mostly predators in the drift and cover stages to complex factors such as starvation (Caroffino et al., 2008), water pollution (Hu et al., 2009), bycatch (Chang, 1999), and a variety of hydrological conditions, including the effects of saltwater in the estuary (He et al., 2009, Zhao et al., 2011, Zhao et al., 2015) in the self-migration stage. Before the self-migration stage, the survival rate of larvae is the lowest at 0.39%, mainly owing to predators devouring eggs and larvae. Therefore, for the wild larvae of Chinese sturgeon, their anti-predator behavior in the drift and cover stages is the most important factor affecting the population recovery; this also underlines the ecological significance of weak swing or free swimming of the wild larvae.

Figure 3

The Mortality Rate of Wild Young Chinese Sturgeon After Spawning in the Yangtze River

To study the migration process of young Chinese sturgeon in the Yangtze River, we need to assume the initial number of sturgeon eggs and then estimate the number of young fishes through the age-specific mortality rate as the initial conditions for the calculations. The egg production of the fish is usually estimated by sampling predators and dissection statistics of the eggs devoured, but this results in a broad range of annual variation (Wei, 2003, Chang, 1999, Yu et al., 2002), even when the pre-GD had a stable adult population size. According to our estimation of the population size of adult Chinese sturgeon (Huang et al., 2017, Huang and Wang, 2018), the GD construction had resulted in a reduction of the yearly effective breeding population from 1,009 individuals of pre-GD to 244 post-GD, corresponding to 24.2% of the original. Considering the sex ratio (1:1) of the fish and the average fecundity (400,000 eggs per female), the total egg production for pre-GD and post-GD are 202 million eggs and 48.8 million eggs, respectively.

See also Transparent Methods and Video S1.

Parametric Traits of Migration Stages

Figures 4A and 4B show the migration speed and diffusion coefficient of juveniles over time in the downriver migration process. First, after hatching the early larvae leave the bottom and rapidly depart the spawning ground with the water current. The larvae's drift speed mainly depends on the velocity distribution of water flow and the spatial position of the larvae accompanying their weak vertical and horizontal swimming ability. The diffusion coefficient increases rapidly in the drift stage, indicating that the larvae disperse quickly to avoid predators. Second, during the cover stage, the larvae are mainly distributed in the inshore riverbed, hiding in cracks of gravel-cobbles and starting to feed at 11–12 dph. At this stage, the larvae stay in the bottom substrate without time-averaged migration velocity. However, the diffusion coefficient reaches the maximum owing to the inhomogeneity of the boundary layer with turbulent eddies, indicating that the larvae are scattered as far as possible, especially after the start of feeding, to reduce the risks of predation and food competition. Third, during the self-migration stage, the larvae start the course of migration along the inshore waters and search for food in the rearing area on the way. They move randomly in the Yangtze River at the individual level but always move downstream at the population level. With the increase of age, the larvae's swimming ability grows and the effect of current on their migration speed gradually weakens. The fish-related diffusivity is proportional to age squared, and the constant Peclet number of the fish migration shows that the convective term is about nine times the fish-related diffusivity term in the self-migration stage of juveniles. Overall, the migration speed and diffusion coefficient of juveniles gradually increase. After entering the tidal-affected region (below Nanjing), the juveniles slow down. However, the diffusion coefficient increases with the age of the fish.

Figure 4

Model Parameters and Calculated Results of Downriver Migration Processes of Young Chinese Sturgeon in the Yangtze River

(A and B) Migration velocity U and diffusion coefficient D of young Chinese sturgeon varied with time in the ① drift stage; ② cover stage; and ③ self-migration stage. Pre-GD (A) was similar to post-GD (B), but with the difference in spawning ground, the flow velocity was higher in the upper Yangtze River than that below the GD in the middle reach of the Yangtze River.

(C and D) Normal migration processes of juvenile Chinese sturgeon for pre-GD (C) and post-GD (D), indicating that the peak density of juveniles passed through the main cities.

(E) Sankuanshi, one of the three famous spawning sites pre-GD, is located about 45 km upstream of Yibin City.

(F) The standing spawning site of post-GD below the GD.

(G and H) Comparison of juvenile densities at the starting site of the self-migration stage and in the river mouth for pre-GD (G) and post-GD (H), showing that the aggregation effect of the juveniles in the 10-km-long East Beach of Chongming Island occurred and that the GD has caused a considerable drop in the population size of juveniles.

See also Figures S1–S4, Tables 1 and S1–S3, Video S1.

The downriver migration reflects some characteristics successively as drifting with the current→ hiding→ self-migrating (acceleration-deceleration) along the inshore waters and indicates certain gradually transforming manners, from passive movement (drifting with the flow) to active swimming (self-migrating), and from the rapids of the thalweg to the quiet flow area of the littoral zone. Furthermore, we revealed that the spatiotemporal density of the juveniles evolved along the migration path into a normal distribution.

Panoramas of Migration Processes

Regardless of the impact of the GD on juvenile mortality, we can estimate the number of surviving larvae or juveniles at all stages (Figure 3). At the end of the cover stage, the number of early larvae used as the initial conditions of MDM calculation was 780,000 individuals pre-GD and 185,000 post-GD. Figure 3 shows that pre-GD there were 39,000 individuals entering the sea and post-GD the number was 9,250. Pine et al. (2001) reported that the annual mortality rates of Gulf sturgeon (A. oxyrinchus desotoi) were 25% for those 1–3 years old and 16% for those 4–25 years old. Assuming that the annual mortality rates of Chinese sturgeon in the sea are the same as those of Gulf sturgeon, we can estimate the maximum number of the potential recruit population in 18 years (average age of female and male adults, corresponding total survival rate of 3%) to be 1,170 individuals pre-GD and 278 post-GD, equivalent to the numbers of annual recruitment. These numbers are consistent with the theoretical estimates (Huang and Wang, 2018) and the estimated results from the field tests in the early years of post-GD (Chang, 1999, Wei, 2003). In a word, the total natural survival rate from eggs to mature adults that can live to return to the Yangtze River averaged about 6 × 10−6 for the wild Chinese sturgeon, implying that each female spawner with a fecundity of 1 million eggs can contribute only six surviving recruits.

Figure 4C shows that in pre-GD, the peak density of juveniles at Nanjing occurred in July, meaning that they were entering the tidal-affected region at 8 months old when their pioneers arrived at the estuary. The peak time of the fish reaching the estuary (Shanghai) was in August when they were 9 months old. The final time for the juveniles to leave the estuary was in early August. Figure 4D shows that post-GD there was a peak density at Nanjing at the beginning of May, when the fish were 5 months old, with pioneers reaching the estuary. However, the peak time at the estuary was mid-June, when fish were at the age of 6.5 months, and the time to enter the sea was in late August. Figures 4C and 4D show that the density distribution curve of juveniles gradually flattens along the path and bulges in the estuarine area, showing an aggregation effect of the fish; the GD has shortened the migration distance of Chinese sturgeon by 1,175 km, causing the juveniles to reach the estuary 1.5 months earlier while posing an extra mortality risk related to the saltwater adaption.

We can estimate that pre-GD had the migration distance of the drift stage for 292 km, from the spawning site at Sankuaishi (Figure 4E) to the hiding site between Hejiang in Sichuan Province and Lanjiatuo in Chongqing Municipality. However, post-GD had the fish drifting for 211 km, from the standing spawning site below the GD (Figure 4F) to a hiding site between Jinzhou and Shishou in Hubei Province (Figure 1A), which was verified empirically by frequent bycatch of local fishermen. The number of larvae hatched post-GD was only a quarter of that pre-GD, because the GD reduced the size of the spawning ground. Therefore, the peak density of juveniles post-GD (Figure 4G) was only one-half of that pre-GD (Figure 4H) in the estuary.

Finally, we describe the overall life cycle of Chinese sturgeon in the Yangtze River pre- and post-GD, involving the migration of wild adult and young fish upriver or downriver (Figure 5A). Meanwhile, we demonstrate that the gonadal development stage of Chinese sturgeon is a vital sign of the fish entering and departing the estuary (Figure 5B) as a result of evolutionary adaption.

Figure 5

Chinese Sturgeon Life Cycle and Gonadal Development

(A) Overall life cycles of the migration processes for the adult and the young Chinese sturgeon pre- and post-GD. The red dotted rectangle indicates the location and size of the spawning ground. The adult Chinese sturgeon entered the Yangtze estuary from June to August (normally distributed with July 15 as the median) each year. In the following autumn, reproduction was completed under suitable hydrological conditions. After spawning, the adults quickly migrated to the ocean. The GD shortened the downriver migration distance by 1,175 km; thus, the adults reached the estuary 10 days earlier than normal (Huang and Wang, 2018). The downriver migration of juveniles takes about 9 months, and the occurrence time of juveniles in the estuary shows a normal distribution. The GD shortened the migration time of juveniles to reach the estuary by 1.5 months. On the abscissa (in order): J, July; A, August; S, September; O, October; N, November; D, December; J, January; F, February; M, March; A, April; M, May; J, June.

(B) Gonadal development in the Yangtze and the sea. After birth in October-November, the larvae move downriver stepwise along the path at gonadal stage 0. Gonads reach stage I at 9 months old when the larvae leave the river mouth in August-September. In the ocean, at 1.5–2.2 years (males) or 2.5–3.0 years (females) old, gonads develop into stage II, and at least at age 8 years for males or 13 years old for females into stage III, at which the fishes become spawners and start to enter the Yangtze River in June-August. They go upstream to reach the spawning ground while there is complete gonadal development from stage III to IV. Spawners remain in the spawning ground for 3 months until the gonads grow from stage IV to IV2 and become mature. Gonads develop fast from stages IV2 to VI with suitable hydrological stimuli, and then mating occurs. After breeding, gonads drop from stages VI to II (YARSG, 1988, Chen et al., 2004, Huang and Wang, 2018).

Discussion

Vertical Distribution of Juveniles

Data concerning the vertical distribution of juvenile Chinese sturgeon is lacking for inshore waters. Caroffino et al. (2009) studied the vertical distribution of the larval lake sturgeon (A. fulvescens) at a total length of 16–22 mm and found an uneven vertical distribution; the density of the upper layer was higher than that of the lower layer. In a 150-cm deep artificial stream tube, shortnose sturgeon (A. brevirostrum) larvae moved downstream, but the majority swam above the bottom at an average height of 100 cm (Kynard and Horgan, 2002). In a similar stream tube experiment, pallid sturgeon (Scaphirhynchus albus) and shovelnose sturgeon (S. platorhynchus) larvae drifted mostly downstream at the surface (Kynard et al., 2002), whereas white sturgeon (A. transmontanus) larvae moved downstream at an average depth of 4–58 cm above the bottom (Kynard and Parker, 2005). Therefore, we inferred that the perched waters of juveniles are within 2–5 m in depth during the self-migration stage, and that the early juveniles mainly migrate along with the bottom layer (Zhuang et al., 2002). With the age-dependent increase of swimming ability, the juveniles switch to an even distribution vertically in the inshore waters. Here we reflect the average swimming behavior of the young fish at a large timescale, without considering the details of diel rhythm.

Backcasting Estimation of the New Spawning Place

The Yangtze cascade dams have had a significant impact on the Chinese sturgeon, which has had its spawning activities decrease from continuous to occasional since the operation of the Three Gorges Dam in 2008 and have disappeared since the Xiluodu Dam in 2013 (Wu et al., 2017b). Despite spawners being mainly distributed within the 30 km below the GD during the spawning season, a small number of the fish may also be scattered in the Wuhan-Jiujiang section (Huang and Wang, 2018). Therefore, a large amount of spawning activity in the traditional spawning ground may cover up the fragmentary, small numbers of spawning fish in other sites. We can infer that other spawning sites may exist if there are appropriate water temperatures, substrates, and hydrological factors. Only four wild juveniles were caught in the estuary on April 16–25, 2015, earlier than normal and far less in number than in previous years. This demonstrated that a small amount of spawning activity in 2014 occurred in an unknown place downstream far from the GD, instead of at the traditional site (Zhuang et al., 2016). Here we estimate that the spawning area in 2014 was probably located between Wuhan and Jiujiang, most likely in the Huangshi section. If we have more collected data in the estuary, such as the peak time of juvenile density, we can calculate backward the spawning time and site more accurately.

Improvements of Artificial Restocking

On 4 January, 1981, the GD dammed the Yangtze River, causing a hot dispute over if—and by how much—the GD may have influenced the river's aquatic life. From then on, China listed the Chinese sturgeon as the sole target of GD's fish rescue and started an artificial restocking program as a remedial measure. From the mid-1980s to the present, more than 6 million individuals of different sizes have been continually released into the Yangtze River, but this has so far achieved little in the recruitment of the population due to an inappropriate strategy—“emphasis on reproduction technique and neglect of post-release behavior,” resulting in a lack of effective evaluation of the artificial restocking program (Brown and Day, 2002). For example, the traditional view states that the larger or older the individual released is, the higher the survival rate; is this true? How can one balance the economics of hatchery-reared fish number or size and their post-release survival rate for a cost-effective outcome? Wu et al., (2017a) reported that 61 cultured juvenile Chinese sturgeon 3 years of age were tagged and released below the GD on April 12, 2015. The fish migrated downriver 1,500 km, and finally only 21 individuals (34%) reached the Yangtze estuary half a month later. As a result of exceeding expectations, they were unable to explain why 66% of the tagged fish were “lost” during their seaward migration. Similar situations occurred in the subsequent years.

Based on the laboratory results of juvenile hybrid sturgeon (Huso duricus ♂ × A. schrencki ♀) (Li et al., 2011) and Chinese sturgeon (Zhuang et al., 2017), there are four main movement patterns in the flowing environment of juvenile fish: upstream, still, countercurrent backward, and drift-downward movements. The upstream movement indicates that juvenile fish swim against the current and move forward owing to their rheotaxis; the still movement means that the fish remains in a motionless state over the ground, and the countercurrent backward movement pattern indicates receding relative to the ground. The only pattern of downstream swimming is the drift-downward movement, indicating that the juvenile fish move downstream with the current without adverse-current behavior. We can assume that the individual juvenile Chinese sturgeon in the wild will display the four movement patterns at different velocities of water flow during the self-migration stage, namely, upstream, still, countercurrent backward, and drift-downward movements (Figures S3A–S3E). Among the four-movement patterns, the predominance of the countercurrent backward in a running water environment implies that a juvenile must consume a great deal of energy compared with a drift-downward movement that would save energy. Experiments in the laboratory reported that an individual juvenile usually shows complex diurnal swimming behavior (Kynard et al., 2002, Zhuang et al., 2002). However, the juveniles generally move downstream at an average daily swimming speed at the population level (Figure S3F). We can infer that the released fish migration downriver is characterized by the four movement patterns mentioned earlier and that the juveniles must consume a great deal of energy. If the Yangtze River cannot provide suitable food along the way for the released fish, this will lead to high mortality. Therefore, improving post-release behavioral performance requires understanding the migration process. The findings of this paper can contribute to the improvement of artificial restocking for the endangered Chinese sturgeon and other anadromous fish species in the world.

Limitations of the Study

Here, we reflect on the migration characteristics of juvenile fish at a daily scale and the population level. The swimming behavior at the individual level, or hourly scale or segment scale, remains unclear. For the self-migration stage, if we can characterize the distribution of juvenile-specific bait-organisms along the Yangtze River, we can combine the MDM with the habitat model to obtain a more detailed spatiotemporal distribution of juvenile fish. In any event, the findings of this paper can provide useful information to determine the key areas of protection for the management of juvenile fish along the Yangtze River, and the model can be used to assess the influence of dams on the migration of juvenile fish.

Additionally, we do not consider the influence of changing river hydrological conditions on the migration process or the navigational mechanism of the long-distance migration. Studies have shown that long-distance migrants can use geomagnetic information to navigate. Species studied include Pacific salmon (Oncorhynchus spp) (Putman et al., 2014b), Chinook salmon (Oncorhynchus tshawytscha) (Putman et al., 2018), steelhead trout (Oncorhynchus mykiss) (Putman et al., 2014a), loggerhead sea turtles (Caretta caretta) (Putman and Mansfield, 2015), and European eels (Anguilla anguilla) (Naisbett-Jones et al., 2017). An inherited magnetic map (i.e., an ability to extract positional information from Earth's magnetic field) exists in these organisms to guide their migration processes. These above-mentioned two components merit further study to improve our model for the Chinese sturgeon and other migratory species.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.