Integrating Remote Sensing and Ecological Modeling to Assess Marine Habitat Suitability for Endangered Chinese Sturgeon

Abstract

The Chinese sturgeon (Acipenser sinensis), a critically endangered anadromous fish species, spends over 90% of its life cycle in marine habitats, yet research on its marine ecology and habitat requirements is limited due to sparse data. To address this, we integrated satellite remote sensing with ecological modeling to assess spatiotemporal dynamics in marine habitat suitability across China’s continental shelf (2003–2020). Nine key habitat factors were derived from multi-source remote sensing data and inverted transparency algorithms. Species occurrence data were coupled with the Maximum Entropy (MaxEnt) model to evaluate habitat preferences and seasonal shifts. Results revealed distinct environmental preferences: shallow depths (≤20 m), sea surface and bottom temperature (10–30 °C and 10–25 °C), salinity (10–35‰), transparency (0.40–3.00 m), eastward and northward seawater velocity (−0.20–0.15 m/s and −0.20–0.20 m/s), moderate productivity (1000–3000 mg/m²), and zooplankton carbon (0.20–6.00 g/m²). Habitat factor importance varied seasonally—salinity, depth, and net primary productivity dominated in spring; bottom temperature and productivity in summer/autumn; salinity and transparency in winter. Spatially, high-suitability areas peaked in autumn (70% total suitable habitat), concentrating near the Yangtze Estuary, northern Jiangsu coast, and Zhoushan Archipelago. This study emphasizes the need to prioritize these areas for protection and inform proliferation and release schemes for Chinese sturgeon. It also demonstrates the efficacy of remote sensing for mapping essential habitats of migratory megafauna in complex coastal ecosystems and provides actionable insights for targeted conservation strategies.

1. Introduction

The Chinese sturgeon (Acipenser sinensis), a critically endangered anadromous fish species endemic to China, holds significant value in the study of fish evolution and biodiversity [1]. As a unique large-scale migratory fish species in China, Chinese sturgeon spends over 90% of its life cycle in marine habitats—growing to sexual maturity in coastal waters before returning to freshwater to spawn [2]. Alarmingly, their wild populations have collapsed critically since the 1980s due to overfishing and river fragmentation by dams, prompting intensive conservation efforts [3]. However, existing research and conservation measures predominantly focus on its freshwater habitats, leaving significant knowledge gaps regarding its marine ecology and habitat requirements due to the limitations in marine environmental data and species distribution information [4].

To date, research on the biological characteristics and habitat selection mechanisms of Chinese sturgeon during its marine life stages remains limited. Preliminary reports have documented its migratory distribution and feeding behaviors along the eastern coast of China, but these studies are scarce [5,6,7]. Currently, domestic research on the migration mechanisms and distribution characteristics of Chinese sturgeon during its marine life stages remains in its infancy [8]. The available data on the marine distribution of Chinese sturgeon remain sparse, both in quantity and spatial–temporal coverage. Current knowledge of the distribution and migratory patterns of Chinese sturgeon is primarily derived from bycatch records, release recapture or tag pop-up (conventional/satellite tag), and observations of its feeding behaviors [9,10,11,12]. While these sources have provided initial insights into the marine migration and distribution of Chinese sturgeon, a clear and comprehensive understanding of its marine habitat and preferences remains elusive due to low tag recovery rates, poor temporal continuity, and restricted spatial coverage [8,13]. In recent decades, the deterioration of water quality and environmental conditions, frequent natural disasters, and sustainable utilization capacity in certain coastal waters of China may have significantly impacted the marine habitat of Chinese sturgeon [14,15,16]. Against the backdrop of global climate change and intensified human activities, it is uncertain whether the current environmental conditions of marine habitats and bait resources can continue to meet the needs of Chinese sturgeon for fattening and growth.

To address these challenges, this study aims to employ geographic information systems (GIS), remote sensing technology, and species habitat suitability modeling to systematically reveal the temporal and spatial variation characteristics of marine habitat environment and the habitat suitability of Chinese sturgeon. Remote sensing technology has gradually emerged as a crucial tool for marine observation, offering advantages such as large-scale coverage, multitemporal, high-frequency data acquisition, and cost-effectiveness [17]. It is widely used in the fields of marine ecology, resource monitoring, disaster management, environmental prediction, and security [18,19]. In particular, ocean color remote sensing utilizes spaceborne or airborne sensors to capture optical information from water-leaving radiation, which is processed through a water bio-optical model to retrieve the concentrations of water color components [20]. The continuous development of various water color sensors and the emergence of abundant hyperspectral data have promoted the advancement of ocean color remote sensing technology [21]. In addition, active remote sensing technologies such as radar can monitor the dynamic environment of ocean tides, currents, and waves to some extent [22].

To investigate the habitat suitability of Chinese sturgeon, it is necessary to employ a species habitat suitability model to assess its habitat preference and overall suitability, based on important marine habitat elements derived from remote sensing data. Over the past several decades, habitat suitability models have been widely applied and continuously refined in ecological research, with the Habitat Suitability Index (HSI) and Maximum Entropy (MaxEnt) models being among the most commonly used. The Habitat Suitability Index (HSI) model, grounded in habitat selection theory, niche differentiation, and limiting-factor concepts, constructs functional relationships between species and environmental factors [23]. However, it relies heavily on expert judgment and assumes linear species–environment relationships, which limits its applicability in large-scale or data-poor contexts [24,25,26,27]. In contrast, MaxEnt requires only presence records and combines occurrence points with environmental variables and background data to estimate habitat suitability [28]. Its capacity to handle limited data, coupled with high predictive accuracy and ease of use, makes MaxEnt a widely adopted tool in fish habitat assessment and species distribution modeling [29]. Its suitability is particularly evident in the case of Chinese sturgeon. As a long-distance migratory species with broad but poorly documented marine distribution, satellite tagging has provided only limited presence data, with no reliable absence information. Furthermore, the scarcity of marine ecological studies on Chinese sturgeon makes it difficult to construct robust HSI models due to a lack of expert consensus and input variables. Under such conditions, MaxEnt’s strengths in handling presence-only data and learning complex species–environment relationships make it especially appropriate. Similar applications have demonstrated its effectiveness in modeling the distributions of mosquito fish in Jiangsu Province and amphioxus in the coastal waters of Xiamen, both involving species with limited ecological data or lacking reliable absence records [30,31].

To address these limitations, this study innovatively integrates satellite remote sensing with the MaxEnt model to analyze the spatiotemporal dynamics of marine habitat suitability for Chinese sturgeon. We collected marine distribution data from the published literature covering the period from 2003 to 2020, which was selected as the study period [7,11]. This study aims to assess the environmental adaptability of Chinese sturgeon and reveal its suitability distribution in marine habitats. The results contribute to the accumulation of data on the marine life history of Chinese sturgeon and offer a reference for in-depth analysis of its habitat preference mechanisms.

2. Materials and Methods

2.1. Study Area

The Chinese sturgeon is widely distributed across China’s marine continental shelf waters, from the Yellow Sea and the west of the Korean Peninsula in the north, to the Sea of Japan off Kyushu in the east, and to the Wanning Sea off Hainan in the south across the Taiwan Strait [8]. The monitor tagging data of the Chinese sturgeon shows that the pop-up and recapture locations are primarily concentrated along the coastal areas of the East China Sea, with a notable cluster near the mouth of the Yangtze River [6,32]. This region lies in the interaction zone between the East Asian continent and the ocean, including the Bohai Sea, the Yellow Sea, and the East China Sea (Figure 1). As an important continental shelf system in the western Pacific, it extends from nearshore shallow waters to the Okinawa Trough and receives inputs from major rivers like the Yangtze and Yellow Rivers [33]. The area features depth variations across the region, with deeper waters offshore and shallower waters near the river mouths and coastal areas. The area spans tropical, subtropical, warm temperate zones. It experiences southeast winds in the hot, rainy summer and north winds along the coastline in winters, with an annual average temperature of about 13 °C. Precipitation occurs mainly in the summer, causing higher coastal runoff into the sea, while it is much lower in winter [33]. Influenced by oceanic currents such as the Taiwan Warm Current, Zhejiang Coastal Current, and Yellow Sea Coastal Current, along with the Yangtze River diluted water and the Yellow Sea Cold Water Mass, China’s eastern coastal waters exhibit high biological productivity and support diverse ecosystems, including estuarine wetlands, mangrove forests, seagrass beds, and coral reefs [34]. These ecosystems play an essential role in sustaining marine biodiversity by providing key functional zones such as feeding grounds, spawning sites, nursery areas, and migratory corridors for a wide range of fish species [35,36]. As an integral part of marine ecosystems, the eastern coastal seas not only sustain key ecological functions—such as energy flow, nutrient cycling, and climate regulation—but also influence the spatial distribution of fish populations in relation to their environment [37,38]. However, human activities, river discharges, seasonal monsoons, and tidal effects have significantly impacted the marine environment, leading to frequent red and green tides, sub-healthy seawater quality, and increasing areas of poor-quality seawater, making it one of the world’s most turbid marine regions with a complex hydrological and ecological environment [15,39,40].

2.2. Data Sources

2.2.1. Habitat Factor Selection

As a typical anadromous fish, Chinese sturgeon spends most of its life in the ocean. Therefore, changes in the marine environment significantly impact its survival, growth, reproduction, and migratory behavior [8]. Temperature and salinity directly influence the survival, fitness, and physiological behavior of Chinese sturgeon, which are key parameters in assessing habitat suitability [41,42]. While the Chinese sturgeon has a diverse diet, it primarily feeds on mesopelagic and bottom-dwelling fish [12]. However, direct assessment of benthic biomass in the ocean is limited by methodological constraints, so this study uses phytoplankton primary productivity and zooplankton carbon content as alternative metrics [43,44]. These indirectly reflect the bait resource dynamics through the nutrient cascade effect [45]. Seawater transparency, which affects underwater light distribution, influences fish visual perception and behavior [46]. The inshore current system transports nutrients and plankton, affecting bait resource distribution [44]. The current rate in the appropriate direction can reduce the energy expenditure of Chinese sturgeon during migration, thereby facilitating the completion of the migration process [26]. At the same time, areas with concentrated coastal currents, currents, and eddies often become critical habitats for juvenile fish [11]. Water depth also impacts the vertical and horizontal distribution of Chinese sturgeon in marine environments [6].

Based on these factors, this study selected nine habitat factors to study the marine environment for Chinese sturgeon and categorized them into three types: physiological (Sea Surface Temperature (SST), Bottom Temperature (BT), and Salinity), physical oceanographic (Transparency, Eastward Seawater Velocity (EV), Northward Seawater Velocity (NV), and Depth), and biological (Net Primary Productivity (NPP) and Mass Content of Zooplankton expressed as Carbon (ZOOC)) (

Table 1. Selection of habitat factors for Chinese sturgeon marine habitat.

Habitat Factor Type	Factor	Spatiotemporal Resolution	Source	Time Series Length
Physiology-based	Sea Surface Temperature (SST)	day/0.083°	NASA ocean color	2003–2020
	Bottom Temperature (BT)	day/0.083°	CMEMS
	Salinity	day/0.083°
Biological	Net Primary Productivity (NPP)	day/0.083°
	Mass Content of Zooplankton expressed as Carbon (ZOOC)	day/0.083°
Physical oceanographic	Eastward Seawater Velocity (EV)	day/0.083°
	Northward Seawater Velocity (NV)	day/0.083°
	Depth	60 arc-second	NOAA
	Transparency	day/0.083°	Inversion

).

2.2.2. Habitat Factor Data Source

The selected habitat factor data (BT, salinity, EV, NV, NPP, ZOOC) were sourced from the Global Ocean Physics Reanalysis dataset provided by the Copernicus Marine Environment Monitoring Service (CMEMS) (https://marine.copernicus.eu/, accessed on 20 December 2024). These habitat factor data have a temporal resolution of 1 day and a spatial resolution of 0.083° × 0.083° (approximately 9 km × 9 km), with a time series spanning from 2003 to 2020 (Table 1). Ocean depth data were obtained from the Global Topography Model ETOPO-2 provided by the NOAA National Centers for Environmental Information (https://www.ncei.noaa.gov/, accessed on 27 December 2024), with a spatial resolution of 60 arc-second. Water transparency, commonly represented by Secchi disk depth (${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$), is defined as the extent to which light penetrates the water column and is currently retrieved primarily through remote sensing inversion methods based on remote sensing reflectance data (${\mathrm{R}}_{\mathrm{rs}}$) [46]. Given the limitations of existing nearshore transparency datasets, this study used an improved semi-analytical algorithm and remote sensing reflectance data to generate the transparency dataset for China’s eastern seas from 2003 to 2020 [46]. The remote sensing reflectance data used to derive water transparency, along with sea surface temperature (SST) data, were obtained from the ocean color remote sensing Level 2 products provided by NASA’s Ocean Biology Processing Group (https://oceancolor.gsfc.nasa.gov/, accessed on 27 December 2024). Both datasets were collected between January 2003 and December 2020, with a time resolution of 1 day and a spatial resolution of 1 km. The specific steps of remote sensing data processing and transparency inversion method are described in Section 2.3.2 of this study, while the detailed inversion algorithm workflow and verification results are provided in the work of Cao et al. [47].

2.2.3. Biological Data Source

This study obtained the spatiotemporal distribution dataset of the Chinese sturgeon in coastal waters through a literature review [7,11]. The dataset includes records of tagged releases and subsequent recapture times and locations from 2003 to 2018, with 78 original datasets collected (Figure 1). Data filtering was conducted based on the defined study area: (1) Non-study area records, such as inland waters and the Sea of Japan, were excluded, retaining only data from the East China Sea region (116°29′–127°20′E, 22°56′–41°05′N). (2) Data points near release sites were excluded to enhance model accuracy, as tags might have detached prematurely due to unstable marking devices, operational stress, or strong water flow [48]. The dataset includes occurrence records of tagged releases and subsequent recapture times and locations from 2003 to 2018, with 78 original datasets collected (Figure 1). In addition, an independent dataset consisting of 36 occurrence records from 2022 to 2024 was obtained from the publicly available literature [49] and used to externally validate the predicted suitable habitats for the Chinese sturgeon.

2.3. Methods

2.3.1. Remote Sensing Inversion Method for Transparency

Existing large-scale nearshore water transparency remote sensing datasets are limited in spatial coverage, temporal and spatial resolution, and time series length, falling short of this study’s requirements. Consequently, we employed remote sensing inversion technology to generate a daily transparency remote sensing dataset for China’s eastern coast from 2003 to 2020, with a spatial resolution of 0.083° × 0.083° [47]. Among the available inversion methods, the semi-analytical Quasi-Analytical Algorithm (QAA) developed by Lee et al. [46] is one of the most widely used bio-optical models for estimating water transparency. Among the available inversion methods, the semi-analytical Quasi-Analytical Algorithm (QAA) developed by Lee et al. [46] is one of the most widely used bio-optical models for estimating water transparency. It enables the derivation of the total absorption coefficient ($\mathsf{\alpha }$) and total backscattering coefficient (${\mathrm{b}}_{\mathrm{b}}$) from remote sensing reflectance. Based on these outputs, the minimum diffuse attenuation coefficient (${\mathrm{K}}_{\mathrm{d}}$) within the 400–700 nm spectral range can be calculated to estimate ${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$. Adopting the nomenclature of Xiang et al. (2023), we hereafter refer to this algorithm as “Lee15” [50], which can be expressed as follows:

$${\mathrm{Z}}_{\mathrm{sd}}={\displaystyle \frac{1}{2.5\mathrm{Min}({\mathrm{K}}_{\mathrm{d}}(412,443,488,531,547,555,667))}}\ln \left({\displaystyle \frac{\left|0.14-{\mathrm{R}}_{\mathrm{rs}}^{\mathrm{PC}}\right|}{0.013}}\right)$$

(1)

where $\mathrm{Min}({\mathrm{K}}_{\mathrm{d}})$ represents the minimum diffuse attenuation coefficient of the 412 nm, 443 nm, 488 nm, 531 nm, 547 nm, 555 nm, and 667 nm bands in coastal waters. ${\mathrm{R}}_{\mathrm{rs}}^{\mathrm{PC}}$ denotes the remote sensing reflectance in the same band as the minimum ${\mathrm{K}}_{\mathrm{d}}\left(\mathsf{\lambda }\right)$.

The Lee15 algorithm classifies water into two categories—clear water (${\mathrm{QAA}}_{\mathrm{clear}}$) and turbid water (${\mathrm{QAA}}_{\mathrm{turbid}}$)—based on the value of remote sensing reflectance at 670 nm. Numerous studies have demonstrated that this dichotomous scheme performs robustly in optically clear waters, whereas its applicability is constrained under turbid conditions [46,51]. To address this limitation, Chen et al. (2022) introduced a logistic function into the Lee15 framework, enabling a continuous transition between ${\mathrm{QAA}}_{\mathrm{clear}}$ and ${\mathrm{QAA}}_{\mathrm{turbid}}$, thereby effectively mitigating data discontinuities caused by fixed threshold segmentation [52]. The enhancement was achieved by fitting a binary logistic regression model [52], in which the ratio of ${\mathrm{QAA}}_{\mathrm{clear}}$ approaches 0 under extremely low ${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$ and converges to 1 under extremely high ${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$. A coefficient corresponding to ${\mathrm{QAA}}_{\mathrm{clear}}$ is used to construct the logistic function, which can be expressed as follows:

$${\mathrm{C}}_{\mathrm{clear}}=1/(1+{\mathrm{e}}^{-\mathrm{k}({\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{clear}}}-{\mathrm{x}}_0)})$$

(2)

where ${\mathrm{C}}_{\mathrm{clear}}$ and ${\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{clear}}}$ represent the proportion of ${\mathrm{QAA}}_{\mathrm{clear}}$ and the estimated ${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$, respectively; k and ${\mathrm{x}}_0$ denote the steepness and the midpoint of the logistic curve; the sum of the proportions of ${\mathrm{QAA}}_{\mathrm{clear}}$ and ${\mathrm{QAA}}_{\mathrm{turbid}}$ equals 1.

Finally, the ${\mathrm{Z}}_{\mathrm{s}\mathrm{d}}$ can be expressed as follows:

$${\mathrm{Z}}_{\mathrm{sd}}={\mathrm{C}}_{\mathrm{clear}}\times {\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{clear}}}+{\mathrm{C}}_{\mathrm{turbid}}\times {\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{turbid}}}$$

(3)

where ${\mathrm{C}}_{\mathrm{clear}}$ and ${\mathrm{C}}_{\mathrm{turbid}}$ are the weighting coefficients of ${\mathrm{QAA}}_{\mathrm{clear}}$ and ${\mathrm{QAA}}_{\mathrm{turbid}}$, respectively; ${\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{clear}}}$ and ${\mathrm{Z}}_{{\mathrm{sd}}_{\mathrm{turbid}}}$ are the estimated values derived from Equation (2). The specific parameters and threshold values follow the study by Chen et al. [52].

2.3.2. Processing Methods for Remote Sensing Data

The global ocean physical dataset provided by CMEMS includes variables such as BT, salinity, EV, NV, NPP, and ZOOC, all with a spatial resolution of 0.083° × 0.083° (approximately 9 km × 9 km). Both the remote sensing reflectance data used for retrieving transparency and the SST products have a native spatial resolution of 1 km. Prior to analysis, SST data and remote sensing reflectance data were first mosaicked and extracted using the study area mask to ensure spatial consistency. Quality control bands were then applied to remove the effects of clouds, water vapor, and other noise factors. To unify the spatial resolution across all remote sensing datasets, the remote sensing reflectance data and SST data were resampled to match that of the other datasets using a stepwise resampling strategy. Specifically, the original 1 km resolution data (approximately 0.0083°) were first resampled to an intermediate scale of 0.01° and subsequently downsampled to a standardized grid of 0.083° to meet the spatial requirements for regional-scale habitat modeling. To mitigate potential information loss due to invalid values (NaN) during this process, a non-mean aggregation algorithm was applied. This algorithm computes the arithmetic mean only among valid fine-resolution pixels (1 km) within each coarse-resolution grid cell, assigning the result as the representative value. The resulting 0.083° dataset follows an 8 × 8 (64:1) spatial aggregation scheme and was generated using the GDAL library (Python 3.9). The resampled remote sensing reflectance data was processed through the aforementioned improved model to generate a standardized transparency dataset. This workflow ensures spatial consistency and provides standardized environmental covariates for subsequent habitat modeling. Finally, the monthly, seasonal, and annual average data of each habitat factor were derived by calculating the arithmetic mean.

2.3.3. Preference Analysis of Habitat Factors for Chinese Sturgeon

To determine the adaptability and preference range of Chinese sturgeon to various habitat factors, we analyzed the relationships between habitat factors at release points, tag shedding points, and recapture points in the marine environment. To align Chinese sturgeon occurrence records with satellite-derived environmental data, we extracted the environmental values corresponding to the exact release, recapture, or tag pop-up points and their associated dates. Recognizing the spatiotemporal continuity of marine conditions, we enlarged the sampling window to the focal cell plus its 24 neighboring grid cells and extended the temporal window to one week before and after each event, thereby maximizing the number of valid matches. The Sankey diagram, created using the ggsankey package in RStudio (version 4.4.2), is utilized to visually display the relationship between habitat factors at release points and tag releasing/recapture points. A statistical method based on cumulative probability distribution is employed to define the preference range of Chinese sturgeon for habitat factors. Continuous intervals with a cumulative probability exceeding 70% and an individual interval probability not less than 10% are designated as the preference range for the respective habitat variable.

2.3.4. Habitat Suitability Prediction for Chinese Sturgeon via MaxEnt Model

The MaxEnt model, an ecological niche model based on the maximum entropy principle, has demonstrated significant advantages in predicting suitable habitats for species. This model integrates species distribution records with multiple environmental variables to model the potential distribution of species. Compared to other prediction models, MaxEnt requires minimal sample sizes and delivers robust predictive performance, making it widely used in fish habitat prediction [53,54]. Given the limited data available on Chinese sturgeon marine tagged datasets, monthly analyses might introduce uncertainty. Thus, we categorized the data into four seasons—spring, summer, autumn, and winter—and paired them with corresponding habitat factor data for analysis.

This study used the MaxEnt species distribution model software (version 3.4.4) (https://biodiversityinformatics.amnh.org/open_source/maxent/, accessed on 5 March 2025). The model output was based on the Logistic model to obtain probability distribution values of Chinese sturgeon potential habitats within the 0–1 range, with other parameters set to default. To comprehensively evaluate the model’s predictive performance, this study employed the Receiver Operating Characteristic (ROC) curve and its Area Under Curve (AUC) value as verification. The AUC value ranges from 0 to 1, with a higher AUC closer to 1 indicating higher prediction accuracy [29]. Drawing on the habitat suitability classification methods from previous studies, the suitability levels for Chinese sturgeon habitats in this study were categorized into four groups: unsuitable (<0.10), low suitability (0.10–0.30), medium suitability (0.30–0.50), and high suitability (0.50–1.0) [55,56]. To ensure the reliability of the model, statistical methods were employed to filter out these potentially inaccurate data, with a focus on the relationship between recapture intervals and spatial movement distances. This approach guarantees that the results of the MaxEnt model accurately reflect the effective distribution of Chinese sturgeons in nearshore areas, thereby emphasizing the importance of these regions for targeted conservation initiatives.

3. Results

3.1. Marine Habitat Preference of Chinese Sturgeon

The transparency inversion method employed in this study has been validated in prior studies [47]. It is an improved ensemble model based on the semi-analytical algorithm developed by Lee15. To evaluate the accuracy of the improved method, Cao et al. compared its inversion results with in situ measurements and conducted performance comparisons with both the original Lee15 model and a new three-class model [46,47,50]. Validation outcomes indicate that, compared to the Lee15 model, the root mean square error (RMSE) and mean absolute error (MAE) of the improved model decreased from 1.49 to 1.32 and from 0.96 to 0.94, respectively (Figure 2a,c). More importantly, the coefficient of determination (R²) increased from 0.63 to 0.70 (Figure 2a,c). The accuracy of the new three-class model, with an R² of 0.65, RMSE of 1.68, and MAE of 1.11 (Figure 2b), remains inferior to the results of the improved model. These findings further affirm the reliability of the improved model in handling complex water conditions. Overall, the improved model used in this study demonstrated high accuracy and reliability, making it suitable for the extraction and analysis of habitat factors for the Chinese sturgeon [47].

The results revealed the habitat preferences of Chinese sturgeon by analyzing the relationship between their release points and various habitat factors (Figure 3). The sea surface and bottom temperatures at the release points of the Chinese sturgeon ranged from 8.76 °C to 29.40 °C and from 14.06 °C to 31.39 °C, respectively, with the majority concentrated at 8.76–22.36 °C and 14.06–21.81 °C. After release, Chinese sturgeons predominantly migrated to sea surface temperatures of 10–30 °C and bottom temperatures of 10–25 °C, which accounted for 74.64% and 86.76% of the tag pop-up and recapture positions, respectively (Figure 3a,b). The ocean salinity at the Chinese sturgeon release points varied from 0.02‰ to 30.05‰, with the majority concentrated between 5.11‰ and 30.05‰. After release, Chinese sturgeons tended to migrate to areas with salinity of 10–35‰, making up 88.23% of the tag pop-up and recapture positions (Figure 3c). The transparency at release points was primarily within 0.31–2.70 m. After release, Chinese sturgeon preferred to migrate to waters with transparency of no more than 3.00 m, which accounted for 85.00% of the tag pop-up and recapture positions, with the highest catch probability within the 2.00 m transparency range (Figure 3d). The eastward and northward flow velocities at the release points of Chinese sturgeons were concentrated between −0.01–0.05 m/s and −0.07–0.01 m/s, respectively. The EV and NV at the Chinese sturgeon release points ranged from −0.12 m/s to 0.19 m/s and from −0.12 m/s to 0.22 m/s, respectively, with the predominant range concentrated between −0.01 m/s and 0.05 m/s, and between −0.07 m/s and −0.01 m/s. After release, Chinese sturgeons mainly tended to migrate towards the EV of −0.20–0.15 m/s and the NV of −0.20–0.20 m/s, accounting for 95.59% and 73.53% of the tag pop-up and recapture positions, respectively (Figure 3e,f). The depth at the release point of Chinese sturgeon ranges from 3.25 m to 7.75 m. After release, Chinese sturgeons primarily migrate to areas with a water depth of no more than 20 m, which accounts for 70.59% of the marked detachment or retrieval points, with the highest probability in the 0–15 m depth range (58.82% of tag pop-up and recapture positions) (Figure 3g). The NPP and ZOOC at the Chinese sturgeon release points ranged from 1265.07 mg/m² to 4120.92 mg/m² and from 0.75 g/m² to 13.95 g/m², respectively, with the majority within 2385.93–3814.90 mg/m² and 1.96–5.86 g/m². After release, Chinese sturgeons tended to migrate to areas with NPP of 1000–3000 mg/m² and ZOOC of 0.20–6.00 g/m², which account for 76.47% and 60.29% of the tag pop-up and recapture positions, respectively (Figure 3h,i).

3.2. Importance Ranking of Marine Habitat Factors on Chinese Sturgeon

The MaxEnt model’s prediction accuracy evaluation shows AUC values for spring, summer, autumn, and winter at 0.93, 0.92, 0.91, and 0.91, respectively, indicating high prediction accuracy across different seasons (Figure 4). To further validate the accuracy of the MaxEnt model’s predictions, we performed a spatial overlay analysis using the independent validation dataset and the combined suitability maps for all four seasons. Specifically, medium- and high-suitability areas from spring, summer, autumn, and winter were merged to form an overall map representing the medium-to-high-suitability habitat range for the Chinese sturgeon, while the remaining areas were classified as low-suitability or unsuitable habitats (Figure 5). This composite suitability map was then overlaid with 36 occurrence records obtained from the publicly available literature. The results showed that 75% of the occurrence points fell within the high-suitability areas, and the remaining 25% were located in medium-suitability areas. No occurrence points were found in low-suitability or unsuitable areas (Figure 5). These findings provide additional evidence supporting the reliability and ecological validity of the MaxEnt model outputs.

The jackknife method of the MaxEnt model evaluates the importance of habitat factors on the distribution of Chinese sturgeons based on the size of training scores for both the presence and absence of this habitat factor. Using the jackknife method, the model assesses habitat factor importance by analyzing training scores with and without each factor. The key habitat factors for each season, based on their impact on training model scores when used alone, are as follows: spring—salinity, net primary productivity, and depth, all with training scores above 0.80 (Figure 6a); summer and autumn—bottom temperature, net primary productivity, and depth, with training scores above 0.75 and 0.70, respectively (Figure 6b,c); winter—salinity, depth, and transparency, all with scores exceeding 0.70 (Figure 6d).

3.3. Marine Habitat Suitability Distribution of Chinese Sturgeon Across Seasons

The results indicate that the suitable habitat areas for Chinese sturgeon exhibit significant spatiotemporal differences across different seasons (Figure 7). In this study, a suitable habitat refers to the combined area of high-, medium-, and low-suitability zones. In spring, unsuitable areas account for the largest proportion (54.45%), followed by low-suitability areas (30.32%) and medium-suitability areas (11.35%), with high-suitability areas making up the smallest proportion (3.88%). High-suitability areas are primarily concentrated in the Yangtze River estuary and the northern Jiangsu coastal waters, while low-suitability areas are mainly distributed in the coastal waters between the Bohai Sea and the southern Shandong Peninsula to the Yangtze River estuary. In summer, the proportion of unsuitable areas decreases to 41.41%, while the proportions of medium- and low-suitability areas increase to 33.55% and 16.76%, respectively. High-suitability areas still constitute the smallest proportion (8.28%) (Figure 8). Notably, the coastal suitability areas show a clear expansion trend, with high-suitability areas increasing significantly near the Yellow Sea, Bohai Sea, Yangtze River estuary, and southern coastal waters, and low-suitability areas extending into the central waters of the East China Sea. In autumn, the spatial pattern of suitable habitats undergoes a significant change. Low-suitability areas make up the highest proportion (42.48%), followed by unsuitable areas (28.86%) and medium-suitability areas (21.90%). The proportion of high-suitability areas is 6.76%. Medium- and low-suitability areas exhibit large-scale eastward expansion, particularly in the Yellow Sea, but with a reduction in coastal high-suitability areas compared to summer. By winter, the proportion of unsuitable areas rises again to 34.43%. The proportions of low- and medium-suitability areas are similar (29.01% and 28.41%, respectively), while high-suitability areas remain the smallest proportion (8.15%). During this season, the low-suitability zone in the Yellow Sea and East China Sea contracts, the medium-suitability zone in the eastern part of the Yangtze River estuary shows a northward migration trend, and the high-suitability zone area increases compared to autumn. All suitable areas show a consistent increasing trend from spring to autumn, with a slight decrease in winter. Specifically, in spring, the proportion of all suitable areas is around 45%, rising to approximately 58% in summer, and peaking with a significant increase to about 70% in autumn. However, in winter, there is a slight decrease to around 65%. This trend indicates that the overall suitable habitat area for Chinese sturgeon is most expansive in autumn, providing the most favorable environmental conditions for the species during this season.

4. Discussion

4.1. Interpretation of Results and Comparison with Previous Studies

Our study’s findings on Chinese sturgeon habitat preferences align with previous research, confirming the importance of sea surface temperature, bottom temperature, salinity, and water depth in shaping anadromous fish distributions [57,58,59,60] (Table 2). Seasonal variations in these factors’ influence on Chinese sturgeon distribution have been less explored. For instance, salinity and water depth are more critical in spring and winter, while BT and NPP are key in summer and autumn. We consider that the seasonal changes in key habitat factors are primarily driven by variations in precipitation and temperature. During the winter and spring, reduced precipitation over the continent leads to lower freshwater input, resulting in higher salinity levels in nearshore waters during these seasons, while salinity is relatively lower in summer and autumn [61]. Chinese sturgeons require increased energy expenditure to regulate osmotic pressure in high-salinity environments, making salinity a critical limiting factor for their survival in winter and spring [62]. In contrast, during the summer and autumn, bottom temperatures reach their annual peak, with nearshore bottom temperatures frequently exceeding 20 °C and sometimes surpassing 30 °C, exceeding the tolerance range of the Chinese sturgeon. Higher bottom temperatures generally lead to reduced dissolved oxygen levels [63], and given the large size and high oxygen demand of Chinese sturgeons, hypoxic conditions further exacerbate their survival challenges. Additionally, NPP in nearshore waters peaks during the summer and autumn, and excessive NPP may cause a decline in dissolved oxygen concentrations, making bottom temperature and NPP critical factors for Chinese sturgeon survival during these seasons.

The preferred sea surface temperature range of Chinese sturgeons is 10–30 °C, notably higher than the theoretical optimum temperature, yet consistent with their core habitat layer in the middle and lower ocean layers. The bottom temperature in summer is usually 5 °C lower than the surface layer, keeping the actual underwater environment temperature below 25 °C, which aligns with their optimal range [64,65]. In terms of salinity, the preferred range differs from the survival range of 8-month-old fry (0–25‰) raised in hatcheries [66]. The present study reports, under natural marine conditions, that adult Chinese sturgeon (>3 years old) prefer salinities of 10–35‰—a range markedly higher than the 0–25‰ reported for 8-month-old hatchery fry [66]. The green sturgeon (Acipenser medirostris) occurring along the Pacific Coast of North America exhibits a similar adult salinity tolerance up to 34‰ [67]. It reflects genuine ontogenetic shifts in osmoregulatory capacity. Young-of-the-year fish possess incompletely developed branchial ion-transport systems and rely on hypo-osmotic environments to minimize metabolic costs. By contrast, sub-adults and adults exhibit increased densities of chloride cells and enhanced renal concentrating ability, enabling homeostasis at higher salinities [65]. All individuals used in our habitat modeling were either satellite-tagged or mark-recaptured mature or near-mature fish; their salinity requirements therefore exceed those of early life stages [32,68]. Water transparency, which reflects water quality, shows that Chinese sturgeons prefer waters with transparency not exceeding 3 m. Nearshore waters of the Bohai Sea and Yellow Sea generally have lower transparency, not exceeding 5 m [47]. The preference for transparency may also relate to food resource distribution and salinity conditions [66]. The preferred velocity range for Chinese sturgeon in the sea is lower than that required for spawning in the Yangtze River (0.81–1.98 m/s) (Table 2) [32].

The preference of the Chinese sturgeon for flow velocity may be related to whether it is in the spawning season. From Table 2, sexually mature individuals during the spawning period exhibit a preference for higher flow velocities (>0.73 m/s), as higher flow velocities facilitate gonadal development and the attachment of fertilized eggs [69]. However, during the non-spawning period, the flow velocity requirement of the Chinese sturgeon is lower, with only a certain flow rate needed to meet its survival needs [70]. Studies indicate that juvenile Chinese sturgeon tend to select the slow-flow and shoal areas in the Yangtze River estuary as their habitat [13]. The monitoring data from the Donghai Bailongyu Marine Ranch, with a relatively low flow velocity (0.01–0.04 m/s), also supports the environmental conditions suitable for the survival of the Chinese sturgeon [70], demonstrating its strong adaptability to low-flow environments. In addition, Figure 2 shows that anomalous environmental values were observed at the recapture points of the Chinese sturgeon, primarily related to the recapture locations. Most sturgeons are recaptured within three months, primarily in nearshore waters, while a few individuals farther offshore may experience environmental factors outside the normal range. Some anomalies may result from early tag loss, such as detachment near the release site, potentially causing abnormal sea surface temperatures near 0 °C. However, these outliers did not affect the overall preference range. The observed migration of the species to deeper offshore waters provides important insights for a comprehensive understanding of its spatial distribution in nearshore environments.

Chinese sturgeon reproduce exclusively in the Yangtze River; after downstream migration, they spend 5–15 years in marine waters feeding and growing to maturity [8]. Diet studies based on stomach-content and stable-isotope analyses reveal that sub-adult and adult sturgeons are demersal piscivores and benthos feeders, preying predominantly on small benthic fish (e.g., gobies, sand lances), shrimp and polychaetes in shallow coastal habitats [7,12]. The nutrient availability and phytoplankton activity in the bottom layer of the Chinese Mainland continental shelf are high, corresponding to the areas identified by our MaxEnt models as having high suitability [71]. The spatial overlap between predicted high-suitability zones (Yangtze Estuary, northern Jiangsu coast, Zhoushan Archipelago) and documented demersal prey hotspots thus provides indirect, but strong, support that the environmental signature we map corresponds to functionally important feeding grounds.

Overall, these findings highlight the dynamic nature of the marine environment and the adaptability of sturgeon. Crucially, the high-suitability ranges predicted by our models, particularly in terms of water column homogeneity and flow dynamics, were strongly corroborated by the observed migration paths of satellite-tagged individuals released in the Yangtze Estuary, which were tracked moving upstream through the middle and lower reaches of the Yangtze River, Zhoushan and Shengsi Islands, and Chongming Island [49]. This convergence between model predictions and empirical tracking data significantly improves our understanding of their precise ecological requirements in both marine and freshwater phases of their life cycle, thereby informing more targeted conservation strategies.

Table 2. Comparison of marine habitat factors preference of Chinese sturgeon with existing studies.

Habitat Factors	This Study			Other Studies
	Preference Range	Environment Condition	Individual Situation	Preference Range	Environment Condition	Individual Situation	References
Temperature	10–30 °C (SST) 10–25 °C (BT)	China sea	Over 3 years old	17–26 °C	Laboratory test	Eight months old	Li Dapeng et al., 2008 [72]
				20 °C	Laboratory test	Seven months old	Feng Guangpeng et al., 2010 [73]
Depth	0–20 m	China sea	Over 3 years old	22.19 m	the Coastal Waters of China	Over 3 years old	Wang Chengyou et al., 2016 [7]
Salinity	10–35‰	China sea	Over 3 years old	0–25‰	Hatchery	Eight months old	Zhao et al., 2011 [12]
				25‰	Laboratory test	1.5 years old	Qin Shaozong, 2020 [65]
Velocity	0–0.20 m/s	Yangtze River	Over 3 years old	0.81–1.98 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Wei Qiwei et al., 2020 [8]
				0.73–1.75 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Zhang hui et al., 2007 [74]
				1.30–1.50 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Ban xuan et al., 2011 [26]
Transparency	0.40–3.00 m	China sea	Over 3 years old	/	/	/	/
NPP	1000–3000 mg/m2	China sea	Over 3 years old	/	/	/	/
ZOOC	0.20–6.00 g/m2	China sea	Over 3 years old	/	/	/	/

Note: “/” indicates that no relevant literature was retrieved.

Table 2. Comparison of marine habitat factors preference of Chinese sturgeon with existing studies.

Habitat Factors	This Study			Other Studies
	Preference Range	Environment Condition	Individual Situation	Preference Range	Environment Condition	Individual Situation	References
Temperature	10–30 °C (SST) 10–25 °C (BT)	China sea	Over 3 years old	17–26 °C	Laboratory test	Eight months old	Li Dapeng et al., 2008 [72]
				20 °C	Laboratory test	Seven months old	Feng Guangpeng et al., 2010 [73]
Depth	0–20 m	China sea	Over 3 years old	22.19 m	the Coastal Waters of China	Over 3 years old	Wang Chengyou et al., 2016 [7]
Salinity	10–35‰	China sea	Over 3 years old	0–25‰	Hatchery	Eight months old	Zhao et al., 2011 [12]
				25‰	Laboratory test	1.5 years old	Qin Shaozong, 2020 [65]
Velocity	0–0.20 m/s	Yangtze River	Over 3 years old	0.81–1.98 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Wei Qiwei et al., 2020 [8]
				0.73–1.75 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Zhang hui et al., 2007 [74]
				1.30–1.50 m/s	Spawning ground	Mature sturgeon (Over 15 years old)	Ban xuan et al., 2011 [26]
Transparency	0.40–3.00 m	China sea	Over 3 years old	/	/	/	/
NPP	1000–3000 mg/m2	China sea	Over 3 years old	/	/	/	/
ZOOC	0.20–6.00 g/m2	China sea	Over 3 years old	/	/	/	/

Note: “/” indicates that no relevant literature was retrieved.

4.2. Correlation and Covariance Matrix Analysis Between Environmental Variables

To comprehensively assess the multicollinearity among the nine selected environmental predictor variables, we conducted Pearson correlation and covariance analyses (

Table 3. The correlation between environmental factors.

	SST	BT	Salinity	Transparency	Depth	EV	NV	NPP	ZOOC
SST	1.00	−0.12	0.41	0.71	−0.59	0.57	0.53	−0.64	−0.8
BT	−0.12	1.00	−0.03	−0.44	0.63	−0.17	−0.23	0.42	0.01
Salinity	0.41	−0.03	1.00	0.58	−0.27	0.34	0.29	−0.53	−0.44
Transparency	0.71	−0.44	0.58	1.00	−0.58	0.49	0.43	−0.84	−0.68
Depth	−0.59	0.63	−0.27	−0.58	1.00	−0.3	−0.34	0.49	0.39
EV	0.57	−0.17	0.34	0.49	−0.3	1.00	0.69	−0.43	−0.5
NV	0.53	−0.23	0.29	0.43	−0.34	0.69	1.00	−0.39	−0.48
NPP	−0.64	0.42	−0.53	−0.84	0.49	−0.43	−0.39	1.00	0.72
ZOOC	−0.80	0.01	−0.44	−0.68	0.39	−0.5	−0.48	0.72	1.00

and

Table 4. Covariance matrix of environmental variable.

	SST	BT	Salinity	Transparency	Depth	EV	NV	NPP	ZOOC
SST	7.08	−1.91	4.75	15.64	−1797.65	0.16	0.19	−1061.90	−5.34
BT	−1.91	35.14	−0.85	−21.63	4259.49	−0.10	−0.18	1542.37	0.21
Salinity	4.75	−0.85	18.98	20.62	−1365.31	0.16	0.17	−1452.90	−4.83
Transparency	15.64	−21.63	20.62	67.52	−5442.68	0.42	0.48	−4329.09	−13.88
Depth	−1797.65	4259.49	−1365.31	−5442.68	1,315,440.30	−36.13	−53.39	348,679.31	1110.84
EV	0.16	−0.10	0.16	0.42	−36.13	0.01	0.01	−27.94	−0.13
NV	0.19	−0.18	0.17	0.48	−53.39	0.01	0.02	−32.62	−0.16
NPP	−1061.90	1542.37	−1452.90	−4329.09	348,679.31	−27.94	−32.62	388,847.83	1116.22
ZOOC	−5.34	0.21	−4.83	−13.88	1110.84	−0.13	−0.16	1116.22	6.25

). The Pearson correlation analysis and covariance matrix were computed in Python 3.9. The results of the correlation analysis indicated a positive correlation between SST and transparency, as well as between NPP and ZOOC (r = 0.71 and r = 0.72, respectively) (Table 3). In contrast, SST exhibited negative correlations with both ZOOC and transparency, and with NPP (r = −0.80 and r = −0.84, respectively) (Table 3). The correlation between other factors was relatively weak. Additionally, the covariance analysis further revealed the co-variation characteristics among the different factors, such as the inverse relationship between SST and depth, as well as between SST and ZOOC (Table 4).

Although strong correlations exist between certain factors, which could theoretically lead to multicollinearity issues, this study did not exclude these factors during the modeling process after weighing the relationship between collinearity and ecological interpretability. This decision was primarily based on the following considerations: Firstly, in agreement with existing studies, SST and key factors such as depth and transparency exhibit strong correlations with multiple environmental variables. However, the goal of this study was to determine the suitability ranges of different parameters, enabling a multidimensional discussion when interpreting species habitat suitability [75]. In practice, selecting multiple strongly correlated environmental factors in model analysis is a common approach in species habitat modeling and biomass assessment. For example, despite the strong correlation between transparency and SST, both factors were included in the model when studying the effects of environmental variables on fish populations in the southern Yangtze River Estuary [76]. Similarly, although chlorophyll and net primary productivity are closely related, both were included as independent variables in the analysis of environmental influences on squid catch in the Pearl River Estuary [77]. Furthermore, previous studies have included SST, depth, and salinity as independent variables after conducting collinearity analysis, which is very similar to our approach in variable selection.

Secondly, different variables often correspond to distinct ecological processes. For instance, SST and BT are closely related, yet they describe the thermal structure of the ocean’s surface and bottom layers, both of which are crucial for species’ vertical habitat selection. Similarly, SST and transparency are closely related, but SST reflects surface temperature and its impact on the physiological state of the Chinese sturgeon, while transparency provides a direct and comprehensive indicator of water quality, highlighting the species’ perception and adaptation to light availability and turbidity [47]. The productivity status of aquatic ecosystems is often reflected in the abundance of plankton and zooplankton. NPP represents the primary carbon supply, while ZOOC reflects the energy transfer to higher trophic levels, each playing a distinct role in the food supply for the Chinese sturgeon [78].

Moreover, research data on the marine life history of the Chinese sturgeon is relatively scarce, and available parameters have certain spatial resolution limitations. Therefore, it is necessary to retain as much diverse information as possible at this stage to comprehensively capture the multidimensional ecological processes affecting habitat suitability. In areas with strong spatial heterogeneity and significant concentration gradients, the contribution of different factors may vary, and deleting or merging highly correlated variables could obscure the unique ecological significance of certain factors. Therefore, we chose to retain highly correlated factors to provide a more comprehensive and multifaceted interpretive framework. In future studies, our goal is to obtain more biologically relevant data and higher-resolution environmental variables, and to select variables based on the ecological adaptation mechanisms and regional environmental characteristics of the Chinese sturgeon revealed in this study, using a scientifically rigorous approach.

4.3. Study Implications in Chinese Sturgeon Conservation

The insights gained from our study of Chinese sturgeon habitat preferences have implications that extend beyond conservation efforts for this species alone. The integrated approach of remote sensing and ecological modeling provides a robust framework for evaluating the suitability of habitats for other marine species that are under threat [79]. The observed seasonal patterns of habitat suitability could have a significant impact on biodiversity and ecosystem services in the East China Sea [75]. Based on the identified salinity preferences, we propose adopting the 10–35‰ isohalines as dynamic outer and inner boundaries for core nearshore habitats. In the Yangtze Estuary, for example, the 35‰ isohaline is typically located near 123.0 °E during low-discharge periods and can shift westward to 122.5°E during summer floods; accordingly, the eastern margin of any seasonal reserve should be flexibly adjusted between 122.5°E and 123.0°E to encompass optimal adult foraging and overwintering grounds. For the highly suitable northern Jiangsu coast, the 10‰ isohaline can serve as the inner boundary to prevent osmotic stress from excessive freshwater inflow. Future reserve design should integrate these salinity thresholds with depth, temperature, and primary-productivity envelopes to construct multidimensional, seasonally adaptive boundaries that enhance resilience to climate change and anthropogenic pressures.

To translate our habitat suitability maps into actionable conservation, we propose an immediate layered strategy and dynamic containment and control: permanently safeguard all grid cells with a seasonal mean suitability ≥ 0.7 within 20 km of the Yangtze Estuary mouth and the adjacent northern Jiangsu shoals by banning bottom trawling and aggregate dredging and enforcing ≤ 10 kn speed limits for commercial ships within a 9 km corridor [80]; for cells with a suitability of 0.5–0.7 that expand 2–3-fold in autumn—mainly the eastern Yellow Sea plume front and Zhoushan–Shengsi Archipelago—introduce dynamic monthly closures updated with near-real-time CMEMS forecasts and time–area fishing restrictions from August to November; and maintain 9 km wide connectivity corridors along the 10–35‰ isohaline envelope linking these zones, incorporating them into provincial Marine Spatial Plans and requiring cumulative-impact assessments for new infrastructure. Furthermore, seasonal changes can affect the habitat distribution of fish, and seasonal adjustment of conservation strategies can help the conservation efforts of species [75]. The results of this study showed that the seasonal variation in the suitable habitat of Chinese sturgeon was greater in autumn and winter than in spring and summer. Therefore, release activities could be carried out in cooler seasons to improve the survival rate of Chinese sturgeon and supplement Chinese sturgeon resources.

Although this study focuses on the critically endangered Chinese sturgeon, the integrated framework—combining multi-source remote sensing products with high-resolution species distribution modeling—can be exported to a wide range of data-poor, long-distance migrants that share the same shelf-sea environment. The Yangtze Estuary–East China Sea corridor, for example, also serves as critical habitat for the Indo-Pacific humpback dolphin (Sousa chinensis), the scalloped hammerhead (Sphyrna lewini), and the green turtle (Chelonia mydas), all of which are classified as Endangered or Vulnerable by the IUCN and are known to concentrate in shallow, productive waters [81,82,83]. The nine environmental layers we generated (SST, BT, salinity, depth, transparency, NPP, ZOOC, east- and northward current velocities) correspond closely to the core niche axes identified for these taxa in global habitat suitability reviews [84]. By substituting occurrence records from telemetry or citizen-science platforms, managers can rapidly replicate the MaxEnt workflow to delineate seasonal hotspots, identify climate-driven range shifts, and prioritize dynamic no-take zones for multiple species simultaneously. The approach is particularly powerful for taxa with sparse or presence-only datasets—such as the East Asian finless porpoise (Neophocaena asiaeorientalis sunameri) and the oceanic sunfish (Mola alexandrini)—that currently lack comprehensive habitat maps [85]. Thus, the present study not only advances Chinese sturgeon conservation but also provides a transferable, cost-effective toolkit for evidence-based management of the broader threatened megafauna community across China’s continental shelf.

4.4. The Impacts of Global Climate Change and Human Activities on Marine Habitat

Beyond characterizing baseline habitat suitability, our results must be interpreted against the backdrop of accelerating global change. The East China Sea—core feeding and growth habitat for >90% of the Chinese sturgeon life cycle—has warmed at ~0.25 °C decade⁻¹ since the 1980s [86]. CMIP6 projections indicate that sea surface temperature warming in the Asian Marginal Seas will significantly exceed the global average, with some regions warming up to 2.6 times faster. Under the SSP585 scenario, the warming may reach 3.8–6.5 °C per century [16]. This will shift the current optimal thermal window (SST 10–30 °C, BT 10–25 °C) north-eastward, compressing high-suitability areas into a narrower band off the Jiangsu and Shandong coasts and potentially eliminating suitable winter habitats south of 30°N.

Human pressures compound climatic stressors. Bottom-trawling intensity in the northern Jiangsu and Yangtze Estuary fishing zones is high, physically damaging the benthic prey assemblages on which sturgeon rely [87]. Aggregate dredging for construction sand has lowered seabed rugosity and reduced benthic biomass in key habitat corridors [88]. Nutrient loading from agricultural runoff has driven a significant increase in summer phytoplankton blooms since 2010; extreme bloom years correspond to a decline in sturgeon recapture rates, consistent with hypoxia-induced habitat compression [89,90]. Shipping traffic along the Yangtze Estuary–Zhoushan route is projected to triple by 2050 [91,92], elevating the risk of propeller strikes and acoustic disturbance during critical foraging periods. Finally, planned offshore wind farms intersect with the current autumn high-suitability polygon [93]; pile-driving noise can mask benthic prey detection and displace sturgeons from otherwise optimal grounds [94]. Collectively, climate-driven shifts in temperature, salinity, and oxygen, superimposed on escalating anthropogenic impacts, are rapidly eroding the spatiotemporal stability of the habitat mosaic identified in this study [85]. Without immediate, climate-smart management—including dynamic reserve boundaries linked to real-time ocean forecasts, stricter trawl exclusion zones, and cumulative-impact assessments for coastal infrastructure—the remaining marine habitat of the critically endangered Chinese sturgeon could contract within two decades. Our habitat suitability maps therefore provide not only a conservation baseline, but an urgent call to action.

4.5. Future Research Directions

Looking ahead, future research should build upon this study by expanding the temporal scope to uncover long-term trends and the impacts of climate change, while explicitly incorporating seasonal dynamics. Our results indicate significant differences in the importance ranking of marine habitat factors on Chinese sturgeon across seasons. Integrating multiple data sources and extending time series will allow for seasonal modeling. There is still a need for more data on the spatial distribution of Chinese sturgeons to refine habitat models. Some recapture/marker detachment data were found near release points, which may be due to factors such as unstable marker devices or the impact of water flow [11]. With increased nearshore distribution data, it will potentially be possible to conduct monthly-scale analysis and construct a comprehensive, multi-scale habitat prediction modeling system.

Owing to the broad spatial extent and long temporal coverage of the study area, high-resolution datasets that simultaneously provide long-term continuity are extremely scarce. In this study, the dataset obtained from CMEMS had a spatial resolution of 0.083°, whereas the original SST and remote sensing reflectance data used for transparency inversion had a resolution of 1 km. To ensure consistency among model inputs, the high-resolution SST and reflectance data were resampled to 0.083°. Although this resolution is sufficient for extracting environmental variables at Chinese sturgeon occurrence points, it presents limitations for fine-scale habitat modeling in nearshore areas. A resolution of 9 km, for instance, may be suitable for large-scale spatial patterns but often fails to capture the finer nuances of habitat variability, particularly in coastal regions where sharp gradients in environmental variables, such as salinity, temperature, and transparency, are common. This study provides a preliminary assessment of large-scale marine habitat suitability for the Chinese sturgeon. Future studies in coastal regions, particularly in medium- and high-suitability zones, should prioritize higher-resolution datasets to improve model accuracy. To further refine the habitat suitability model, future work should integrate eDNA and telemetry for more accurate presence–absence data, incorporate additional high-resolution environmental variables (oxygen, pH, sediment) to enhance stressor representation, extend the framework to include co-occurring endangered species, and deploy animal-borne cameras with benthic sampling to validate habitat–fitness links.

It is essential to incorporate additional critical environmental variables, such as ocean acidification, pollution, and the deterioration of coastal water quality (eutrophication, red tides, etc.), alongside addressing the significant challenge of accurately monitoring the environmental factors vital for this bottom-dwelling species (e.g., BT, salinity, and sediment type), which currently rely on limited techniques such as site monitoring, tag tracking, and acoustics monitoring [13]. These techniques lack the efficiency and resolution required for large-scale observation, so achieving efficient remote sensing of benthic elements is a crucial future goal. Empirical validation of model predictions through field surveys, such as tag tracking studies, is vital for improving the reliability of habitat suitability maps. Furthermore, integrating genetic data could shed light on population connectivity and diversity across suitable habitats. Dedicated research into the spatiotemporal patterns of habitat patches, particularly regarding the impact of human-induced fragmentation on connectivity and survival, is imperative to evaluate habitat quality, levels of ecosystem disturbance, ecological carrying capacity, and overall suitability. This could be achieved by leveraging spatiotemporal pattern analysis theory to provide precise scientific findings.

5. Conclusions

This study integrated satellite remote sensing data with the MaxEnt ecological niche model to assess the spatiotemporal dynamics of marine habitat suitability for the critically endangered Chinese sturgeon in China sea (2003–2020). By deriving nine key marine habitat variables and analyzing their relationship with Chinese sturgeon distribution data obtained through mark-recapture techniques, we systematically characterized the marine habitat preferences and identified critical factors shaping Chinese sturgeon seasonal distribution. Our key findings are as follows:

(1)

Chinese sturgeon exhibited distinct environmental preferences after release, primarily inhabiting shallow coastal waters (≤20 m depth) with defined ranges, sea surface temperature of 10–30 °C, bottom temperature of 10–25 °C, salinity of 10–35‰, transparency of 0.40–3.00 m, eastward current velocity of −0.20–0.15 m/s, northward current velocity of −0.20–0.20 m/s, net primary productivity of 1000–3000 mg/m², and zooplankton carbon content of 0.20–6.00 g/m².

(2)

The importance ranking of marine habitat factors on Chinese sturgeon exhibited seasonal shifts—salinity, net primary productivity, and depth dominated in spring; bottom temperature, net primary productivity, and depth in summer/autumn; salinity, depth, and transparency in winter.

(3)

Spatially, high-suitability areas peaked in autumn and concentrated in the Yangtze Estuary, northern Jiangsu coast, and Zhoushan Archipelago, with notable seasonal dynamics: summer expanded coastal suitability zones, autumn triggered eastward expansion of medium-/low-suitability areas into the East Sea, and winter saw northward extension of medium-suitability areas.

This research provides a comprehensive, multi-seasonal assessment of marine habitat suitability for Chinese sturgeon using an integrated remote sensing and ecological modeling approach. It significantly advances our understanding beyond previous studies focused primarily on freshwater stages. This approach has significant potential for evaluating habitat suitability and informing conservation strategies for Chinese sturgeon and other marine megafauna with limited data, which face similar threats in complex coastal ecosystems. Future efforts should focus on expanding temporal coverage to assess long-term trends and the impact of climate change, incorporating high-resolution benthic variables (dissolved oxygen, pH, sediment type), assimilating eDNA and telemetry data for presence–absence records, and extending the framework to quantify habitat connectivity under scenarios of human-induced fragmentation, thereby supporting adaptive management for the broader threatened megafauna community.