The distribution of fish spawning habitat is usually determined by the adaptation of reproductive physiology and species response to the surrounding environmental conditions (Gallego et al., 2007; Zorica et al., 2013; Gonzalez-Irusta and Wright, 2016; Reglero et al., 2017). The spawning distribution could be influenced by various environmental factors, e.g., temperature, salinity, topography (Opdal, 2010; Dumont et al., 2011; Shi et al., 2014). There is evidence that compared with adults, the early life history stages of fish species seem to be more sensible and vulnerable to the variations in the environmental conditions (Portner and Farrell, 2008). To ensure more survival of eggs and larvae, spawning activity usually happens in the recurrent hydrographic features (Munk et al., 2009) such as estuaries, upwellings, or tidal mixing front (Tian et al., 2009; Zorica et al., 2013; Gibson et al., 2018). The estuaries, as important fish spawning and nursery habitats, are usually characterized by complex hydrological systems and spatially varying environmental conditions (Andrew et al., 2009). A study on fish spawning habitat is the fundamental step for the fish recruitment process (Lelievre et al., 2014), especially for the estuarine species. Examining fish spawning activity and its response to the spatial non-stationary environment is important for gaining a comprehensive understanding of spawning habitat selection mechanism, which helps to understand fish stock fluctuations and, hence, establish appropriate protection strategies for estuarine fish.

To evaluate statistical relationships between the spawning habitat and environmental variables, “global” regression models (Brunsdon et al., 1996; Windle et al., 2010) such as generalized additive model (GAM) and generalized linear model (GLM) are fully developed and used (Bacheler et al., 2010; Asch and Checkley, 2013; Gonzalez-Irusta and Wright, 2016). These models treat the entire data set as “whole map” and function under the assumption of spatial stationarity (Brunsdon et al., 2002; Windle et al., 2010). In the recent years, more and more studies have focused on spatial heterogeneity and the changes in species distribution caused by variations in the environmental conditions (Li et al., 2018). The geographically weighted regression (GWR) model provides a promising approach for investigating spatial non-stationarity effect and could make good predictions of fish species distribution (Lauri et al., 2014). The GWR model has been applied successfully to both the marine and freshwater habitats, e.g., North Atlantic Ocean (Windle et al., 2010), Gulf of Maine (Li et al., 2018), and Lake Malawi (Likongwe et al., 2015). However, rarely study has been reported that applies the GWR model method to demonstrate the non-stationary effect of the environment on the fish spawning habitat.

The Yangtze Estuary in China is the largest estuary in the western Pacific Ocean with three brunches and four outlets to the East China Sea (Figure 1). It is characterized by high biological productivity associated with relatively extreme and varying environmental conditions (Shan et al., 2010). It serves as an important spawning habitat, nursing ground, and migration channel for varieties of fish species because of the advantaged conditions (Zhuang et al., 2006; Yu and Zhang, 2017). More than 300 fish species have been found inhabiting the Yangtze Estuary (Zhuang et al., 2006). Among them, Coilia mystus is considered one of the most important commercial fish species in the Yangtze Estuary (Shi and Wang, 2002) with an annual yield of more than 1,500 tons in the 1990s (Ni, 1999). However, in 2019, China had to announce that it would stop issuing special fishing licenses for the wild populations of C. mystus because of the severe depletion of the resources (Zhai and Pauly, 2019, 2020) caused by overfishing and the habitat degradation (Ni et al., 1999; Liu et al., 2013).

Coilia mystus belongs to Clupeiformes, with a short lifespan, seasonal migrations, and long spawning periods (Zhuang et al., 2006; Zhao et al., 2020). Also, it is a semi-anadromous fish (Elliott et al., 2007) that generally lives in the shallow marine area and spawns in the estuarine area (Zhuang et al., 2006; Zhao et al., 2020). The South Branch of the Yangtze Estuary is considered the main spawning location for C. mystus (Ni et al., 1999). The spawning period of C. mystus starts from late April to September and peaking in late May and June (Zhuang et al., 2006). In the spawning season, adults of C. mystus migrate to the brackish water in the Yangtze Estuary to spawn, but will never enter the freshwater areas above the Jiangyin section of the river (Yang et al., 2006b). There are few studies on the selection mechanism of spawning habitat of C. mystus and its response to environmental conditions, except for some previous work examining the migration pattern and biology trait (Yang et al., 2006a; Zhao et al., 2020). It is generally believed that the sea surface temperature (SST) and the sea surface salinity (SSS) would influence the spawning habitat of C. mystus (Jiang et al., 2006; Jiang and Shen, 2006). However, some geographical variables such as the distance to the coast (DTC) and the depth (DEP) of water, which were involved in other estuarine ichthyoplankton study (Locke and Courtenay, 1995; Halvor et al., 2007), were barely explored. An updated study on spawning activity and specific locations of spawning habitat in the area is necessary. Especially, understanding how the spawning habitat of C. mystus shifts over subtle spatial changes in the environment is the key problem.

In this study, the spawning habitat selection mechanism under the locally varying environmental conditions was explored based on a case study of C. mystus in the Yangtze Estuary. The spatial GWR model was applied to investigate the non-stationary effect of the environmental variables, i.e., the SST, the SSS, and geographical variables, i.e., the DTC and the DEP on the distribution of spawning habitat of C. mystus. The conclusions are helpful in gaining a comprehensive understanding of fish spawning habitat selection mechanism in the estuarine areas and conductive to the protection of the population.

The spatial regression GWR model was used to explore the spawning habitat of C. mystus and its response to the rapidly changing abiotic variables in the Yangtze Estuary. The estimated coefficients of each variable changed locally across the study area, indicating non-stationary effects caused by environment heterogeneity in the Yangtze Estuary. Seasonal variations can be found in environmental factors that affect the distribution of the spawning habitat of C. mystus. In the spring, SSS, DTC, and DEP were the main impact factors and saltwater intrusion in the South Branch might lead to the trend that the spawning habitat moved further upstream to the Changxing Island. SST was most important in the summer and the relatively lower SST was more crucial in the spawning habitat selection than DTC and DEP. The predicted spawning habitat of C. mystus was mainly distributed in the middle of the South Branch, near the Chongming, Changxing, and Hengsha Islands, which basically corresponded with the known area of spawning habitat defined in previous studies (Ni et al., 1999; Zhuang et al., 2006). However, there was possibility that the spawning habitat of C. mystus moved further upstream (Figure 3: May 2020) and adults of C. mystus might have the habit of spawning in the freshwater (Yang et al., 2006a,b).

The non-stationary effects of abiotic variables were reflected by the spatially varying relationship between the independent variables and dependent variables (Li et al., 2018). Both SST and SSS represented the primary habitat environment in defining C. mystus spawning habitat (Jiang et al., 2006; Jiang and Shen, 2006; Shi et al., 2014). In the reproductive season, adults of C. mystus would spawn to the brackish water with salinity from 0.14 to 7.51 and temperature of 18–28°C in the Yangtze Estuary (Ni et al., 1999; Zhuang et al., 2006; Ding, 2011).

The temperature across the study area in May 2019 and May 2020 (19–24°C) was within the range of suitable temperature for spawning, resulting in SSS the decisive environmental factor in defining the spawning habitat in the spring (Jiang and Shen, 2006). Although SSS was also involved in August 2019, the non-stationary effect of SSS was only significant in spring. Several studies based on the stepwise model and GAM have shown a universal negative correlation between SSS and the number of C. mystus eggs (Jiang et al., 2006; Ding, 2011; Hu et al., 2021). However, results from the GWR models indicated that the strength of the negative relationship between SSS and egg density varied spatially upstream of the Hengsha Island in May 2020 (Figure 5). There was saltwater intrusion of the South Branch in this month, as high salinity water was found along the Jiuduan sandbank (Figure 4B; Chen and Zhu, 2014). To avoid that eggs being transported to the unfavorable salinity waters (Fortier and Gagné, 1990), more adult fish might choose to spawn further upstream of the Changxing Island (Figure 3) and resulted in the strength difference. There was also a strong difference in the negative effect of SSS between May 2020 and May 2019. Overfishing and environmental degradation were the main reasons for the resource decline of C. mystus (Ni et al., 1999; Liu et al., 2013). The policy that stopping issuing special fishing licenses of C. mystus in 2019 might contribute to more egg production in 2020 and lead to the strength difference.

Sea surface temperature was the dominant environmental factor in summer, as the strength of the effect of SST was much stronger than that of SSS in August 2019. The conclusion was contradicted with that of Jiang and Shen (2006), who thought that SSS influenced greater than SST in summer. It might because the average SST in the summer of 2002 was lower than 28°C (Jiang and Shen, 2006). Previous studies have proved that the increase of temperature would accelerate fish gonadal maturation and, thus, promoting spawning activity (Jiang et al., 2006; Jiao et al., 2019). The widely distributed positive coefficients estimated of SST in August 2019 agreed with the conclusion. However, under higher SST in the North Channel, the area around the Changxing Island, and the South Passage in August 2020 (Figure 4A), the relationship between SST and egg density was negative (Figure 5), which indicated that the “extreme” temperature might restrain the spawning activity or harm the survival of eggs. Ni et al. (2020) reported that under the gradual warming of the Yangtze Estuary, the yield of adults of C. mystus have declined dramatically since a surge by 2°C occurred in 1997 and did not recover in the recent years. The regional temperature rise in the South Branch might be one of the reasons that influence the recruitment success of C. mystus. More attention should be paid to such areas in the future surveys.

There has been no obvious alteration in the distribution of C. mystus spawning habitat over the years. Homing of mature adults to their natal area is observed in a huge variety of migratory fishes such as diadromous fish and Gadiformes (Stabell, 1984; Robichaud and Rose, 2001; Knutsen et al., 2007). However, cues for the homing behavior were still uncertain and complex to explain. In this study, DTC and DEP were applied to explore the geographic effect on the distribution of spawning. It was shown that the relationship between DTC and egg density is positive in the upstream of the Hengsha Island and negative in the rest of subareas A and B (Figure 5). The spatially varying relationship suggests that adult C. mystus might prefer to spawn in the middle of the river upstream of the Hengsha Island, while close to the coast near waters around the Hengsha Island (Yang et al., 1990). The less egg production around the Hengsha Island caused by weaker runoff in May 2020 might also account for the smaller area of the negative effect of DTC in the North Channel (Figure 5). In August 2020, the non-stationary effect of DTC around the Changxing and Hengsha Islands was different from that in the other months. The choice of relatively lower temperature areas in August 2020 rather than the suitable DTC might be the main reason for the “abnormal” phenomenon. Since it is possible to ensure the survival of eggs under “extreme” temperature, the environmental condition of SST would determine the spawning habitat of C. mystus. The reproductive migration process of C. mystus was from offshore to inshore waters and finally stopped upstream of the Hengsha Island (Zhuang et al., 2006); based on the relationship analyzed above, the gradually reducing DTC might be one cue for the homing process of adults of C. mystus.

It was shown that the estimated coefficients of DEP were positively correlated to egg density in most of the study areas (Figure 5) in all the 3 months, which validated the result that C. mystus would spawn in deep waters (Zhou et al., 2011). The non-stationary effect caused by DEP was obvious in May 2019 and August 2020. As the deepest DEP yielded approximately the same level of egg density as other areas where water depth was relatively shallow (Figures 1A, 3), the strength of the positive effect of DEP was weaker in the upstream of the Hengsha Island in May 2019 (Figure 5). The negative relationship between DEP and egg density was due to the eggs found in the shallow waters at the bottom of subarea C in May 2019. However, these eggs might be a result of the transport of local current (Ye et al., 2004; Wan and Wang, 2018) rather than a choice of spawning, as the environmental conditions are unsuitable for spawning. In August 2020, there were small areas of negative relationship in waters near the Changxing and Hengsha Islands. Such a phenomenon might also be the result of lower temperature preference in summer, as eggs were found in the shallow water of relatively lower SST rather than the deep water with “extreme” waters.

In conclusion, the specific location of the spawning habitat of C. mystus was predicted and defined by the GWR model in the Yangtze Estuary. The GWR model showed capability in explaining the response of species to the environmental variation in an estuarine habitat and the non-stationary effect of local variables had been analyzed for better understanding the spawning habitat selection mechanism of C. mystus. However, it was difficult to predict the actual spawning habitat specifically based on our limited data. In the future, more surveys covering the upstream of the South and North Branch throughout the spawning season should be carried out. Meanwhile, there was inference about the potential spawning habitat of C. mystus in the North Branch (Zhao et al., 2020). The potential spawning habitat can be defined as “a habitat where the environmental conditions are suitable for spawning” (Planque et al., 2010). We have collected a few eggs in the North Branch and subarea C; however, such phenomenon did not appear repeatedly and the environmental condition in the North Branch seemed unsuitable for spawning. The member/vagrant hypothesis proposed that the successful stock recruitment depended heavily on the local current system for preventing the eggs from being dispersed to unfavorable conditions (Fortier and Gagné, 1990). The eggs found might be under the transport of local currents to the nursing ground (Hu et al., 2021), as residual currents in the North Channel to the offshore area have the direction first to East then to the North (Ye et al., 2004; Wan and Wang, 2018). Thus, a study on the transport mechanism of the early life stage of C. mystus is necessary to verify the inference and hypothesis.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

DW and ZL conceived and designed the study. RW, JZ, LZ, and SZ were involved in revising the manuscript for important intellectual content. XL and PS collected the data. DW compiled and analyzed the data and drafted the manuscript. All authors approved the final version of the manuscript and agreed to be accountable for all the aspects of the work.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.