Assessing the Impact of Climate Change on the Distribution of Portunus trituberculatus in Zhoushan Fishing Ground by Using the Maximum Entropy Method (MaxEnt)

Abstract

Based on previous studies and the ecological characteristics of Portunus trituberculatus, we hypothesized that climate change could substantially reshape its suitable habitat in Zhoushan fishing ground. Under present-day climate conditions (2010–2020), P. trituberculatus exhibits a distinct seasonal distribution pattern in this region. However, its potential spatial response to future climate change, and whether suitable habitat will remain available, remains poorly understood. To address this gap, we combined species occurrence records with environmental variables from the Bio-ORACLE v3.0 database, including benthic temperature, benthic salinity, benthic current velocity, primary productivity, bathymetry, topographic slope, and topographic aspect, to develop a maximum entropy (MaxEnt) model and predict the potential distribution of suitable habitat for P. trituberculatus under present-day conditions and future SSP1-2.6 and SSP2-4.5 scenarios for 2030–2040, 2040–2050, and 2090–2100. Model performance was high across all seasons, with area under the curve values exceeding 0.80. Primary productivity and benthic temperature were the dominant environmental predictors, highlighting the joint influence of trophic conditions and thermal constraints on habitat suitability. Future projections revealed pronounced seasonal reorganization of suitable habitat rather than a uniform range shift. Spring suitable habitat expanded consistently under both scenarios, with the magnitude of expansion increasing toward the end of the century and reaching 46.9% by 2100 under SSP2-4.5, likely because warming relaxed low-temperature limitation during the early seasonal transition. In contrast, suitable habitat in autumn and winter generally contracted. Autumn losses were moderate but persistent, ranging from 5.4% to 16.4%, whereas the strongest declines occurred in winter, particularly under SSP2-4.5, where habitat reductions exceeded 30% after mid-century. These contractions were likely associated with cumulative thermal stress and related environmental changes under continued warming. Summer responses were scenario-dependent, showing weak gains or net declines under SSP1-2.6 but substantial expansion under SSP2-4.5 after mid-century, reaching up to 23.6% by 2050, suggesting that habitat suitability in this season is shaped by interactions among thermal conditions, trophic support, and habitat characteristics. Overall, these findings reveal strong seasonal asymmetry in habitat responses to climate change and provide a scientific basis for seasonally adaptive management of P. trituberculatus resources in Zhoushan fishing ground.

1. Introduction

Climate change has driven persistent and frequently irreversible increases in ocean temperature and acidification, along with sea level rise and declining dissolved oxygen [1]. The magnitude and extent of these effects are anticipated to increase over time. In response to climate change, marine species may adopt various strategies, including adaptation or acclimatization, migration, and shifts in their distribution ranges. For example, in South Australia, the sea urchin Centrostephanus rodgersii spread from the mainland to Tasmania in the late 1970s and subsequently increased in both range and abundance coincident with regional warming [2]. Notably, certain species may be at risk of extinction [3].

Using a seasonally explicit MaxEnt framework under the CMIP6 SSP1-2.6 and SSP2-4.5 scenarios, the present study investigated potential distributional changes in P. trituberculatus in Zhoushan fishing ground at a regional scale. While previous studies have suggested a northward shift in this species under climate change, accompanied by pronounced seasonal variability, they have mainly emphasized broad-scale range shifts and overall habitat loss, with limited attention to fine-scale seasonal redistribution and latitudinal centroid shifts. As research on climate change intensifies, the Coupled Model Intercomparison Project Phase 5 (CMIP5) is demonstrating its limitations in the study of species habitat change. Climate scenario modeling is of central importance to research in the field of climate change in relation to fisheries and aquatic conservation. The Coupled Model Intercomparison Project Phase 6 (CMIP6) dataset integrates a range of updated variables compared with CMIP5. In particular, spatial and temporal variability in the physical and biogeochemical properties of the global ocean is a subject of interest [4]. A comparison of the present study with earlier Representative Concentration Pathway (RCP) scenarios demonstrates that, despite the similarity in terms of radiative forcing pathways, the CMIP6 Shared Socioeconomic Pathways (SSPs) framework employs updated and more detailed socioeconomic assumptions, allowing for a more complete representation of future global change [5].

Portunus trituberculatus belongs to the subphylum Crustacea, order Decapoda, family Portunidae, and genus Portunus.

It is a eurythermal and euryhaline crab species characterized by a relatively short life cycle, rapid growth, high population turnover, and short-distance seasonal migration. Its spatial distribution is strongly influenced by environmental factors, particularly water temperature and salinity [6,7]. Here, “eurythermal” refers to its broad thermal tolerance range, whereas its high thermal sensitivity is reflected in the pronounced responses of its seasonal distribution, migration, and habitat use to temperature variation within that range. In Zhoushan fishing ground, P. trituberculatus is one of the most important commercial crab species and plays a key role in regional fishery production and the marine economy [8]. Owing to its high resource availability, wide distribution, and sensitivity to environmental variability, this species has long been an important target of trawl and pot fisheries in Zhoushan and adjacent waters [9]. Moreover, given its relatively short life span (generally 1–3 years) and strong sensitivity to environmental change, especially temperature, variations in its spatial and temporal distribution and abundance can effectively reflect the impacts of environmental variability and climate change on marine fishery resources. Consequently, P. trituberculatus is widely regarded as an important indicator species for assessing ecological change and the effectiveness of sustainable fishery management in Zhoushan fishing ground. Previous studies have suggested that the distribution of P. trituberculatus may shift under climate change, with pronounced seasonal variability in the magnitude and direction of habitat change. For example, Liu, based on an ensemble species distribution model, reported that P. trituberculatus may exhibit a northward distributional shift under future climate scenarios. Among the four seasons, summer was projected to experience the most substantial contraction in suitable habitat, with losses ranging from 45.23% under the RCP4.5 scenario to 88.26% under the RCP8.5 scenario by the 2100s. Habitat loss was predicted to occur mainly in the East China Sea and the southern Yellow Sea, whereas slight habitat expansion may emerge in the northern Bohai Sea [10]. Such projected redistribution is unlikely to be driven by warming alone. It may instead result from the combined effects of seawater warming, changes in trophic conditions, salinity, hydrodynamic processes, and habitat characteristics such as bathymetry. In particular, seasonal differences in habitat responses may reflect the varying relative influence of benthic primary productivity, benthic temperature, current velocity, salinity, and seafloor topography across seasons. Together, these factors likely shape the spatially and seasonally heterogeneous habitat dynamics of P. trituberculatus under climate change.

The maximum entropy (MaxEnt) model is a machine learning approach derived from statistical mechanics that applies the principle of maximum entropy to presence-only (PO) species occurrence data. MaxEnt estimates response functions that describe how the probability of species occurrence changes with each environmental variable, thereby characterizing the species’ ecological niche and predicting its potential geographic distribution [11]. MaxEnt has been widely adopted in marine species distribution modeling due to its high predictive accuracy and robustness when occurrence data are limited. Kerry applied the MaxEnt model to assess the impacts of climate change on the spatial distribution of the whitefin swellshark (Cephaloscyllium albipinnum), and the results indicated potential future shifts in habitat suitability under different emission scenarios [12]. Research has identified water depth and sea surface temperature as the two primary environmental factors that influence species distribution. Building upon these methodological advances and empirical findings, the MaxEnt framework provides a robust tool for assessing how marine species distribution may respond to environmental variability and the ongoing climate change.

To increase understanding of the potential impact of climate change on P. trituberculatus, we aimed to predict the current and potential future distributions of this species within Zhoushan fishing ground by using species distribution models (SDMs). The seasonal distribution characteristics and presence of P. trituberculatus under the climate scenarios for 2030, 2050, and 2100 were predicted (SSP1-2.6 and SSP2-4.5).

Zhoushan fishing ground was selected as the focal study area because it is one of the most important fishing grounds for P. trituberculatus in the East China Sea and plays an important role in the spawning, feeding, and overwintering processes of this species. As an intensively exploited coastal system that is sensitive to ongoing environmental change, this region provides a biologically and managerially meaningful setting for assessing climate-driven habitat redistribution. We therefore hypothesized that projected changes in benthic temperature and primary productivity under future climate scenarios would asymmetrically alter the seasonal distribution of suitable habitat for P. trituberculatus in Zhoushan fishing ground.

2. Materials and Methods

2.1. Data Sources

2.1.1. Occurrence Data of P. trituberculatus and Spatial Extent

Occurrence records of P. trituberculatus were compiled from multiple authoritative sources, including online databases such as the Global Biodiversity Information Facility (GBIF; https://www.gbif.org (accessed on 5 November 2025), GBIF Occurrence Download https://doi.org/10.15468/dl.f6c5p9 (accessed on 5 November 2025)), fishery resource survey datasets, and relevant peer-reviewed literature included in the Web of Science database.

Fishery survey data for P. trituberculatus were obtained from bottom trawl surveys conducted in the northern East China Sea in August 2006 and January, May, and November 2007. The survey area covered waters between 29°30′ N and 32°00′ N and west of 127°00′ E, encompassing fishing grounds located within the trawling-prohibited zone for powered trawlers. The surveys were conducted using a research vessel with a main engine power of 184 kW and a gross tonnage of 100 t. Sampling gear consisted of a beam shrimp trawl with a beam length of 30 m. Trawling was conducted at each station for approximately 1 h at a towing speed of 2.0 knots.

The study extent was defined to include Zhoushan fishing ground and its adjacent waters in order to encompass the principal habitat, seasonal migration space, and potential redistribution area of P. trituberculatus within a biologically meaningful regional background. This extent also reduces the risk of over-constraining model calibration to a narrowly delimited fishing area while retaining ecological relevance to the focal population. To reduce spatial clustering and sampling bias from heterogeneous data sources (GBIF, fishery surveys, and literature), occurrence records were spatially rarefied using a 0.05° thinning distance to ensure only one record per grid cell, consistent with the environmental layer resolution. This step reduced over-sampling in intensively surveyed areas and improved model independence. Given the well-defined geographic scope of Zhoushan fishing ground and consistent trawl survey design, target-group background sampling was not applied, as uniform background selection across the study area sufficiently represents accessible environmental conditions. These steps mitigate sampling bias and ensure robust model fitting for P. trituberculatus habitat modeling. The final extent was used in the analysis. As shown in Figure 1, 52 occurrence records of P. trituberculatus were included in spring, 66 in summer, 64 in autumn, and 66 in winter.

2.1.2. Sources of Environmental Data

The spatial and temporal distributions of P. trituberculatus are principally influenced by seasonal variation in benthic environmental conditions, particularly benthic temperature, as well as seasonally varying predictors such as benthic salinity, benthic current velocity, and primary productivity [13]. Therefore, the average sea surface temperature was used in spring and autumn to represent the seasonal average surface temperature. The average temperature of the hottest month was used in summer to represent the environmental layer, while the average temperature of the coldest month was used in winter [10]. The remaining environmental variables were treated in the same manner.

All marine environmental data were downloaded from the Bio-ORACLE v3.0 database [14]. On the basis of the life history and biological characteristics of P. trituberculatus [15], we used mean benthic temperature, mean benthic salinity, mean benthic current velocity, primary productivity, bathymetry, topographic slope, and topographic aspect. Because hydrodynamic conditions can influence larval transport, habitat connectivity, and local environmental stability. Primary productivity was included as a proxy for trophic support, given its importance in structuring benthic food availability. Bathymetry was selected because P. trituberculatus exhibits depth-related spawning, feeding, and overwintering habitat preferences. Topographic slope and topographic aspect were incorporated to characterize seafloor heterogeneity, which may influence substrate conditions, local circulation, and habitat suitability. Together, these variables capture the major thermal, trophic, hydrodynamic, and topographic dimensions relevant to the habitat requirements of P. trituberculatus. The Pearson correlation coefficients for summer and autumn were also presented. The spatial resolution of these layers was set at 0.05° (approximately 5.5 km at the equator) to predict the spatial distribution of P. trituberculatus under climate change scenarios.

These datasets can be used for current purposes and for the latest assessment report of the Intergovernmental Panel on Climate Change on three future periods (2020–2030, 2040–2050, and 2090–2100) under SSPs [16]. The current study posits that medium forcing scenarios (SSP2-3.4 and SSP2-4.5) are more reasonably predicted in species distribution models, predicting global warming within the range of 1.5–4 °C. The high-emission scenario (SSP5-8.5) was the most frequently employed in studies, but it was determined to be less feasible [17]. Although SSP3 7.0 is a typical medium forcing scenario, it assumes uncoordinated development and high regional uncertainty, which is less suitable for coastal ecosystem assessment. SSP2 4.5 represents a moderate, middle-of-the-road pathway and is widely recommended for regional impact studies. SSP1 2.6 was selected as the low forcing scenario to reflect ambitious mitigation targets. Therefore, SSP1 2.6 and SSP2 4.5 were adopted for future habitat projections.

To avoid multicollinearity among predictor variables, Pearson correlation analyses were conducted separately for each season. Variables with pairwise Pearson correlation coefficients exceeding |r| = 0.8 were considered highly collinear and were excluded from further analysis [18]. In the present study, the retained variables showed low intercorrelation across all four seasons (Figure 2), indicating limited collinearity among predictors and supporting the robustness and interpretability of the MaxEnt model output. Then, we further calculated the variance inflation factor (VIF) for more robust validation. All retained environmental variables had VIF < 10, indicating no severe multicollinearity. This combined approach ensures the robustness and interpretability of the MaxEnt model.

2.2. Model Construction

The MaxEnt model is a commonly used machine learning-based SDM that has been demonstrated to outperform other models in previous studies [19]. This model is widely used in marine, terrestrial [20,21], and freshwater realms. Its robustness for small sample sizes and capability to model predictions based on PO data contribute to MaxEnt’s popularity [22]. To estimate P. trituberculatus habitat distribution in Zhoushan fishing ground, the MaxEnt model version 3.4.4 was used. The MaxEnt model requires PO data and environment covariates to estimate presence probabilities. It uses environmental covariates at species presence and background points to find the probability distribution of maximum entropy and then constrains the distribution by using environmental covariates at species presence. P. trituberculatus presence points were randomly partitioned into a training set (75%) and a test set (25%), possibly achieving approximately 10,000 background points [23]. To reduce the uncertainty of the MaxEnt model, 10 replicated runs of cross-validation were used to build the model, and the final results were the average of the 10 replicates. MaxEnt was run using the default auto-feature setting, which allows nonlinear responses and may include quadratic features where supported by the data. Default regularization parameters were applied, as they have been widely validated for marine crustacean SDMs and reduce overfitting risks. We acknowledge that MaxEnt outputs are sensitive to background selection, regularization, and bounding box definition; these settings were held consistent across all seasonal models to ensure comparability. Future studies may conduct explicit sensitivity tests across parameter sets to further quantify uncertainty.

We evaluated the predictive capability of MaxEnt by using the size of the area under the receiver operating characteristic curve (AUC) [24], which is the most used method for assessing SDM performance [25]. The results were as follows: 0.5 ≤ AUC < 0.6, the simulation is considered unsuccessful; 0.6 ≤ AUC < 0.7 indicates poor simulation performance; 0.7 ≤ AUC < 0.8 suggests moderate simulation performance; 0.8 ≤ AUC < 0.9 implies good simulation performance; and 0.9 ≤ AUC < 1 indicates excellent simulation performance [12]. This metric quantifies the model’s capability to distinguish between actual species occurrence points and pseudo-absences, with an AUC value closer to 1 indicating higher predictive accuracy [26]. As a modeling method based solely on presence data, specific thresholds can also be derived from the logistic output to assess the model’s performance [24]. Moreover, the response curves of the variables and the jackknife test were calculated. Maximum training sensitivity plus specificity (MTSS) was used as the threshold to classify suitable regions [27]. The response curves demonstrate how variations in environmental factors affect the predicted probability of P. trituberculatus presence. Values that exceed the MTSS threshold (0.4) indicate suitable habitat conditions for P. trituberculatus. Environmental variable values were considered optimal when the response curves reached their maximum. In addition, the relative contribution of each environmental variable was extracted from the MaxEnt output as percent contribution values, which were used to evaluate the importance of different predictors in each seasonal model. Variable importance was further assessed using the jackknife test, and response curves were generated to describe the relationships between environmental variables and habitat suitability.

3. Results

3.1. Evaluation of the Model’s Predictive Performance and Variable Importance

The AUC values for the training and test datasets in each season were above 0.8, indicating that the seasonal MaxEnt model exhibited a high level of accuracy in model predictions (

Table 1. Performance measures of the MaxEnt model.

Season	Training AUC	Test AUC	AUC Standard Deviation	Training TSS	Test TSS
Spring	0.93	0.91	0.03	0.82	0.78
Summer	0.90	0.86	0.03	0.77	0.72
Autumn	0.91	0.87	0.04	0.79	0.74
Winter	0.89	0.83	0.05	0.75	0.69

). This result indicated that the selected environmental variables played significant roles in habitat selection for P. trituberculatus. The habitat suitability simulation results demonstrated high discriminatory capability, suggesting that the model was adequate for studying the distribution of suitable habitat for P. trituberculatus in Zhoushan fishing ground. Furthermore, the TSS values for each season were high, indicating good model performance.

In this study, a jackknife test was used to evaluate the influence of environmental variables on the MaxEnt prediction of suitable habitat for P. trituberculatus (Figure 3). In spring, when used individually, bathymetry made the greatest contribution, followed by primary productivity and mean benthic salinity, whereas terrain slope had the smallest effect. In summer, when used individually, bathymetry contributed the most, followed by mean benthic temperature and primary productivity, whereas terrain aspect had the smallest effect. In autumn, when used individually, bathymetry again made the greatest contribution, followed by primary productivity and mean benthic salinity, whereas terrain aspect had the smallest effect. In winter, when used individually, bathymetry remained the most important variable, followed by mean benthic temperature, mean benthic salinity, and primary productivity, whereas terrain slope had the smallest effect.

In spring, three environmental factors jointly explained 81.2% of the model prediction for P. trituberculatus. Benthic primary productivity contributed the most (38%), followed by mean benthic temperature (22.1%) and bathymetry (21.1%). The remaining four variables collectively contributed 18.8% to the model prediction. In summer, three environmental factors together accounted for 88.1% of the model’s predicted variation in P. trituberculatus. The highest contribution was from benthic primary productivity (43.6%), with bathymetry (32.8%) and mean benthic salinity (11.7%) as the next most influential factors. The other four variables together contributed the remaining 11.9% of the total prediction. In autumn, three environmental factors jointly explained 77% of the model prediction for P. trituberculatus, with benthic primary productivity being the highest contributor (44.4%), followed by bathymetry (19.2%) and mean benthic salinity (13.4%). The other four variables collectively contributed the remaining 23% of the total prediction. In winter, three environmental factors together accounted for 74.7% of the model’s predicted variation in P. trituberculatus, with benthic primary productivity contributing the most (32.3%), followed by mean benthic current velocity (24.5%) and mean benthic temperature (17.9%). The other four variables collectively contributed the remaining 25.3% of the total prediction (

Table 2. Estimates of the relative contributions of the environmental variables to the MaxEnt model.

Environmental Variables	Spring	Summer	Autumn	Winter
Benthic temperature	22.1	2.3	11.4	17.9
Benthic salinity	6.8	11.7	13.4	6.2
Benthic current velocity	1.5	4.4	3	24.5
Primary productivity	38	43.6	44.4	32.3
Bathymetry	21.1	32.8	19.2	7
Topographic slope	2.6	1.8	1.7	3
Topographic aspect	7.9	3.4	6.8	9

).

The response curves illustrate how variations in environmental factors influence the predicted probability of P. trituberculatus occurrence. Values exceeding the MTSS threshold (0.4) indicate suitable habitat conditions for this species. Environmental variable values at which the response curves reach their maximum are considered optimal. Consequently, the result shows the following: benthic temperature between 16.7 °C and 19.0 °C in spring, with an optimal temperature around 17.6 °C; benthic temperature between 23.0 °C and 30.0 °C in summer, with an optimal temperature around 26.9 °C; benthic temperature between 15.8 and 19.0 °C in autumn, with an optimal temperature around 17.7 °C; and benthic temperature between 6.3 °C and 14.1 °C in winter, with an optimal temperature around 7.1 °C (Figure 4).

3.2. Current and Future Distributions

Figure 5 shows the predicted probability of presence and suitable habitat for P. trituberculatus under present (2010–2020) climate conditions.

Projections under CMIP6 climate scenarios revealed pronounced seasonal and scenario-dependent changes in suitable habitat areas within Zhoushan fishing ground (

Table 3. Species range change in suitable habitat for P. trituberculatus.

Species Range Change (%)	Spring	Summer	Autumn	Winter
SSP126-2030	6.98	−3.56	−6.43	−23.03
SSP126-2050	18.60	0.55	−5.43	8.82
SSP126-2100	28.42	−11.77	−11.73	−18.96
SSP245-2030	15.63	−5.29	−0.67	−8.06
SSP245-2050	5.43	23.63	−16.35	−34.50
SSP245-2100	46.90	15.51	−13.47	−29.76

; Figure 6). Among the four seasons, spring habitat consistently expanded under SSP1-2.6 and SSP2-4.5 throughout all future periods, with the magnitude of expansion increasing toward the end of the century. Notably, spring suitable habitat areas increased by up to 46.9% by 2100 under the moderate-emission scenario (SSP2-4.5). By contrast, autumn and winter habitats generally contracted under future climate conditions. Autumn habitat losses were moderate but persistent across scenarios (−5.4% to −16.4%), while winter exhibited the strongest declines, particularly under SSP2-4.5, where habitat reductions exceeded 30% after the middle of the century. Summer responses were more complex and scenario-dependent. Under SSP1-2.6, summer habitat demonstrated weak gains or net declines. Under SSP2-4.5, substantial expansion occurred after the middle of the century, reaching an increase of 23.6% by 2050.

Future predictions under SSP1-2.6 and SSP2-4.5 scenarios showed that the suitable habitat of P. trituberculatus will shift southward (lower latitude); by contrast, the centroid in spring will shift northward (2050s and 2100s) under the SSP2-4.5 scenario. The distribution centroid will shift to about 0.006° south in spring, 0.073° in summer, 0.235° in fall, and 0.06° in winter. Meanwhile, it will shift further southward over time and in other climate scenarios. However, the north shift in the centroid in spring is 0.02° (2050s) and 0.018° (2100s) under the SSP2-4.5 scenario. Notably, the latitudinal centroid shifts more significantly in autumn than in other seasons under the different climate scenarios (

Table 4. Changes in the latitudinal centroid of P. trituberculatus under current and future climate conditions.

	Spring	Summer	Autumn	Winter
SSP126-2030	−0.006	−0.073	−0.235	−0.060
SSP126-2050	−0.009	−0.043	−0.169	−0.038
SSP126-2100	−0.070	−0.141	−0.165	−0.057
SSP245-2030	−0.014	−0.081	−0.047	−0.060
SSP245-2050	0.020	−0.047	−0.119	0.042
SSP245-2100	0.018	−0.043	−0.218	−0.021

). Under the SSP1-2.6 scenario, the centroid in autumn will shift 0.235° southward by the 2030s, 0.169° southward by the middle of the 21st century, and then 0.165° southward by the late 21st century. Under the SSP2-4.5 scenario, the south shift in the centroid in autumn is 0.047° (2030s), 0.119° (2050s), and 0.218° (2100s).

4. Discussion

4.1. Habitat Preferences

P. trituberculatus exhibited different distribution trends in Zhoushan fishing ground based on seasons and habitats. Benthic primary productivity was the most important factor that influenced the distribution of P. trituberculatus. Several biotic factors have been previously shown to influence the distribution of crabs, such as bottom temperature [28], seawater velocity [29], salinity [30], and chlorophyll a [31]. In marine ecosystems, the rate and distribution of primary production play a fundamental role in structuring marine food webs [32], indirectly regulating the habitat suitability of benthic crabs, such as P. trituberculatus. This species inhabits environments with a salinity range of 30–35 and exhibits distinct reproductive and overwintering migratory behavior. During spring and summer, spawning typically occurs in shallow coastal waters (depth: 3–5 m), while overwintering takes place in sandy mud substrates at depths of 10–30 m [15]. This finding is largely consistent with our model predictions. In addition, the habitat preferences of P. trituberculatus are affected by depth and sediment texture. Chou [33] revealed that the crustacean community composition in the southwestern waters of Taiwan was primarily determined using three factors: large-scale differences in water depth, characteristics of the bottom substrate, and topography, which influences water circulation.

Our results showed that the main drivers of habitat suitability for Portunus trituberculatus varied among seasons, with benthic primary productivity consistently contributing the most, bathymetry becoming particularly important in summer and autumn, and benthic temperature showing marked seasonal differences in contribution. This pattern indicates that the habitat distribution of P. trituberculatus is shaped not only by physiological tolerance to benthic environmental conditions but also by seasonal variation in trophic support and habitat structure.

The dominant contribution of primary productivity likely reflects the fundamental role of trophic support in structuring benthic food webs and prey availability. In a highly productive coastal system such as Zhoushan fishing ground, spatial heterogeneity in productivity may translate into corresponding differences in habitat suitability for this commercially important crab. This interpretation is broadly consistent with previous studies showing that productivity-related processes can strongly influence marine consumer distributions.

Bathymetry ranked second in summer and autumn, suggesting that depth-related habitat structure becomes especially important during the warm seasons. This may be associated with seasonal shifts in spawning, feeding, and habitat use, as well as with depth-dependent differences in hydrography, substrate, and thermal buffering. In summer, deeper benthic habitats may provide relatively stable environmental conditions and reduce exposure to unfavorable shallow-water warming.

These findings are broadly consistent with previous studies identifying temperature, salinity, depth, and hydrodynamic conditions as important drivers of crab distribution, but they also suggest that, within Zhoushan fishing ground, trophic support and seasonal habitat structure can be at least as important as thermal conditions. Overall, our results emphasize that the habitat preferences of P. trituberculatus arise from the interaction of trophic, thermal, and topographic controls rather than from a single environmental gradient.

4.2. Future Predicted Distribution of P. trituberculatus

A previous study showed that climate change may substantially reshape the seasonal habitat suitability of P. trituberculatus [34]. Under future climate scenarios, habitat suitability exhibits pronounced seasonal asymmetry rather than a uniform shift. Spring habitat expands persistently under both SSP1-2.6 and SSP2-4.5, with greater increases toward the end of the century, likely because early seasonal warming relaxes low-temperature constraints and allows suitable habitat to emerge earlier when trophic conditions remain favorable. By contrast, autumn and winter habitat contracts consistently, particularly under SSP2-4.5, indicating stronger sensitivity to cumulative thermal stress and hydrodynamic change during cooler seasons. Winter appears to be the most vulnerable season, whereas summer responses are more scenario dependent, with limited change under SSP1-2.6 but marked expansion under SSP2-4.5 after mid-century. Although previous studies have emphasized broad-scale distributional shifts, they have provided limited insight into season-specific habitat reorganization within Zhoushan fishing ground under CMIP6 scenarios. The contribution of the present study therefore lies not only in its finer spatial and seasonal resolution but also in its ability to reveal asymmetric seasonal responses and intra-regional redistribution at a local fishery-ground scale, providing a more practical basis for seasonally adaptive fishery management under climate change.

Empirical and theoretical studies suggest that marine species respond to ocean warming by shifting ranges poleward and/or into deeper depth [35]. Under CMIP5, 80% of species with current distributional centroid in the northern hemisphere, and all species with current distributional centroid in the southern hemisphere, demonstrate potential southward distributional shift in the environmental suitability centroid [36].

Notably, within the spatial extent of Zhoushan fishing ground, our findings show a southward displacement of the latitudinal centroid of P. trituberculatus under CMIP6, with relatively small centroid shifts overall, indicating that climate impacts in this regional context are expressed primarily through intra-regional redistribution rather than large poleward range expansion. In addition to thermal change, projected variation in benthic primary productivity may play a key role in driving the future redistribution of P. trituberculatus. As the dominant predictor across seasons, benthic primary productivity likely reflects the trophic support available to benthic food webs and, indirectly, the prey resources required by this species. Climate-driven changes in productivity may therefore modify habitat suitability by altering energy supply and food availability within Zhoushan fishing ground. From this perspective, the projected habitat shifts observed in our study are unlikely to be explained by warming alone, but rather by the interaction between thermal constraints and trophic conditions under future climate scenarios.

Our results show a southward shift in the distribution centroid of P. trituberculatus under future climate scenarios, which is ecologically unexpected given that most marine species in the Northern Hemisphere shift poleward under ocean warming. This novel pattern can be explained by three key factors: (1) Zhoushan fishing ground and the Yangtze Estuary may retain relatively stable benthic temperatures and high primary productivity under future climate conditions, thereby providing more favorable habitat conditions for P. trituberculatus. (2) Strong warming in the northern part of the study area exceeds the thermal tolerance of P. trituberculatus, reducing habitat suitability. (3) Regional hydrodynamic conditions, salinity stratification, and coastal topography in the southern part of the study area provide more favorable and stable habitats under future scenarios. Together, these drivers lead to a centroid shift toward the south rather than a typical poleward shift, revealing a region-specific adaptive response that differs from large-scale global trends.

4.3. Implication for Fishery Management

The projected habitat changes in P. trituberculatus present significant implications for fishery and resource management in the region. The Yangtze Estuary and Zhoushan fishing ground are primary spawning grounds for P. trituberculatus and serve as destinations and relay stations for overwintering crabs [37]. First, seasonal fishing regulations should be adjusted according to climate-driven habitat shifts: spring habitat expansion supports appropriate fishing pressure, while strict catch limits and temporary fishing bans should be implemented in autumn and winter to reduce the impact of continuous habitat contraction. Second, key habitats such as spawning grounds in the Yangtze Estuary and overwintering grounds in Zhoushan fishing ground should be designated as climate refugia and included in marine protected areas (MPAs) with long-term conservation measures. Third, stock enhancement activities should be optimized by releasing seeds in regions with high and stable habitat suitability under future climate scenarios. Fourth, a real-time monitoring and early warning system for habitat suitability should be established to support adaptive management under ocean warming. These targeted measures can improve the resilience of P. trituberculatus resources and ensure the sustainability of fisheries in Zhoushan fishing ground.

4.4. Limitations of the Study

Although ensemble modeling may improve the robustness of species distribution projections in some cases, it does not always outperform single-algorithm models [38,39]. Previous studies have shown that some single-model approaches can provide comparable predictive performance while requiring less computation and offering greater interpretability [40]. In the present study, MaxEnt was selected because it is well suited to presence-only data, performs robustly with relatively small sample sizes, and has been widely applied in marine species distribution modeling. Nevertheless, the exclusive use of a single model remains a limitation of this study. Future work should therefore compare MaxEnt with other SDMs or apply ensemble modeling frameworks to evaluate the robustness of projected habitat changes.

One limitation of this study is the potential temporal mismatch between species occurrence records and environmental predictors. Key environmental variables affecting P. trituberculatus, including temperature, salinity, and primary productivity, may vary over time in the dynamic East China Sea. Such temporal inconsistency may introduce uncertainty into the species–environment relationships estimated by the MaxEnt model and the resulting habitat suitability prediction. Another limitation of this study is that future MaxEnt projections implicitly assume unlimited dispersal. This assumption may not fully reflect the ecology of P. trituberculatus, which exhibits relatively short-distance migratory behavior. Therefore, projected habitat gains under future climate scenarios, particularly under SSP2-4.5 in spring and summer, may be overestimated to some extent. Future studies should incorporate dispersal constraints to provide a more ecologically realistic assessment of potential distribution shifts under climate change. A further limitation of this study is that spatial uncertainty maps were not provided for the projected habitat suitability patterns. Such maps would be valuable for evaluating the spatial reliability of model outputs, particularly in marginal suitability areas where management decisions may be more sensitive to projection uncertainty. Future studies should therefore incorporate spatial uncertainty mapping to improve the interpretation and practical value of habitat projections under climate change.

5. Conclusions

Using CMIP6 climate projections and seasonally explicit MaxEnt modeling, this study suggests that climate change may reshape the habitat suitability of P. trituberculatus in Zhoushan fishing ground through pronounced seasonal reorganization. Spring habitat was projected to expand under SSP1-2.6 and SSP2-4.5, whereas autumn and winter habitats generally contracted, particularly under the medium-emission scenario. Summer responses were more scenario-dependent. Benthic primary productivity and benthic temperature emerged as the dominant predictors of habitat suitability, indicating that both trophic support and thermal conditions are important in shaping future habitat redistribution. These findings highlight the need for regionally adaptive and seasonally responsive fishery management under ongoing climate change.