Predicting the Current and Future Suitable Habitat Distribution of Swimming Crab (Portunus trituberculatus) using the Species Distribution Model under Climate Change

Species have shown their habital variations in responding to climate change, especially during the spring and summer spawning seasons. The species distribution model (SDM) is considered the most favorable tool to study the potential effects of climate change on species distribution. Therefore, we developed the ensemble SDM to predict the changes in species distribution of Portunus trituberculatus among different seasons in 2050 and 2100 under the climate scenarios RCP4.5 and RCP8.5. The results of SDM indicate that the distribution of this species will move northward and have obviouse seasonal variations. Meanwhile, the suitable habitat for the species will be significantly reduced in summer, with loses rates ranging from 45.23% (RCP4.5) to 88.26% (RCP.8.5) by 2100s. Habitat reduction will mainly occur in the East China Sea and southern part of the Yellow Sea, while there will be a small increase in the northern Bohai Sea. These findings will be important to manage the ecosystem and fishery, provide an information forecast of this species in the future, and maintain species diversity if the seawater temperature rises.

factors, previous studies demonstrated a few variables could accurately predict species distribution [21,23,24,25]. Considering the biological relevance and data availability under current and future climate scenarios, five predictor variables sea surface temperature, surface salinity, current velocity, offshore distance, and depth were used in the SDM. Sea surface temperature, surface salinity, current velocity, offshore distance were extracted from the Bio-Ocean Rasters for Analysis of Climate and Environment (Bio-ORACLE, http://www.bio-oracle.org) [25,26]. Environmental variable depth was extracted from Global Marine Environment Dataset (http://gmed.auckland.ac.nz).
Temperature directly affected the physiological habit of swimming crab and varied greatly among seasons [27]. However, only mean annual sea surface temperature, the average temperature of the warmest month and coldest month are available in Bio-ORACLE. Therefore, the mean annual sea surface temperature represented the mean season surface temperature in spring and autumn, and the average temperature of the warmest month and coldest month were used for the environmental layers in summer and winter, respectively. We adopted the spatial resolution of 5×5 arcmin or higher and the geographic coverage of approximately 9.2×9.2 km 2 to match the five environmental data layers. The all pairwise Pearson correlation coefficient of the environmental data was below |0.7| to remove prohibitive levels of redundancy among layers [16].
We retrieved the future temperature and salinity data layers under the typical concentration path emission scenarios (RCP) of RCP4.5 and RCP8.5 from Bio-ORACLE to predict the spatial distribution of P. trituberculatus in the coastal areas of China under climate change. Future layers were produced by averaging data from distinct AOGCMs provided by the CMIP5 [28]. This database provides two time series of years 2050 (2040-2050) and 2100 (2090-2100) with the same spatial resolution of 5 arcmin for prediction environment variables [29]. The two-time series were representative of mid and long-term future climate conditions. RCP4.5 are intermediate emission scenarios, and RCP8.5 is a pessimistic scenario with higher concentrations. Therefore, two-time series (2050 and 2100) and two RCPs (RCP4.5 and RCP8.5) were considered to investigate potential future distributional shifts of swimming crab.

Ensemble prediction model
Nine models (GLM, generalized linear models; GBM, generalized boosted models; GAM, generalized additive models; CTA, classification tree analysis; ANN, artificial neural networks; SRE, surface range envelope; FDA, flexible discriminant analysis; RF, random forest; MaxEnt, maximum entropy) provided by Biomod2 software package were used to predict the suitable habitat of this species.
Among ccurrence records, 80% of the data were used for modeling, and 20% were used for model validation. Each model was run three times, with a total of 27 times [30]. Kappa coefficient, true skill statistic (TSS), and receiver operating characteristic (ROC) were selected to evaluate the accuracy of the model [31]. We screen single models with TSS > 0.80 and AUC > 0.9 and use the weighted average method to construct a combined forecasting model. The weight of calculation results of single model was determined by its TSS value. The normalized results of a single model were multiplied by the corresponding weights, and the combined model was constructed to calculate the potential suitable habitat Index of the swimming crab in the study area.
The ensemble model predicted the probability (0-100%) of the presence of P. trituberculatus in each grid at 5 arcmin resolution of the study area. To produce a presence/absence map of P. trituberculatus, continuous probability values were converted to binary predictions based on a Pearson correlation analysis was applied to assess the relative importance of each variable in predicting species distribution [17]. Besides, to study the temperature sensitivity of species suitable habitat in summer, we simulated the trend of narrowing of species suitable habitat in summer under 0.2-2.0 °C warming, using 0.2 °C as a temperature increase step. Table 1 The importance of environmental variables in the potentially suitable distribution of swimming crab.

Model assessment and environmental variable factor contribution
The

Current and future potential distributions
The predicted current suitable habitats of swimming crab under current climate conditions are shown in Figure 4. All the occurrence records were covered in the predicted current suitable habitat.
Compared the four seasons of predicted suitable habitats of swimming crab, the Yellow Sea, and the East China Sea were suitable for this species with high habitat suitability in four seasons. The Bohai Sea showed low habitat suitability in the Fall season. Under future climate scenarios, the suitable habitat for swimming crab will decrease in spatial extent, especially in summer ( Figure 5). Under RCP4.5 climate change scenario, a decrease of 4.47% (winter)-35.29%(summer) were predicted in the suitable habitat in 2050s, while in 2100s, the loss of suitable habitat will be predicted from 3.37% (winter)to 45.23% (summer) ( Table 2). The loss of potential distribution will be more aggravated under We also investigated the temperature increase sensitivity of swimming crab in summer. It was found that the rate of habitat loss increased exponentially with increasing temperature. When the temperature rises by 0.2°C to 2°C, the loss rate will rise from 2.8% to 80.5% ( Figure 6)

Effects of climate change on suitable habitats
The spatial variability of P. trituberculatus suitable habitat sustained previous empirical studies showing the effects of climate change on marine nekton distributional ranges and abundance [32,33].
The present model predicts a decrease in P. trituberculatus habitat suitability in the East China Sea, which indicates that climate change will result in a spatial redistribution of this species and a serious history.
The shift of the centroid of the summer suitable habitat to the north may be one of its reproductive strategies to cope with climate change, but there will also be inevitable losses. Simply changing the timing and location of spawning may result in a mismatch between feeding grounds and areas of high phytoplankton density. Moreover, it is difficult to predict whether new spawns can provide necessary sediments and corresponding hydrological characteristics for egg deposition [34]. Therefore, the study on the change of suitable habitat in summer needs to be further deepened.

Conflicts of Interest:
The authors declare no conflict of interest. Table 2 Range variation of swimming crab in different seasons and periods based on ensemble SDM Table 3 The latitudinal centroid of swimming crab under current and future climate conditions