Method for predicting mandarin coral endangered grade under future climate change based on species distribution model

Abstract

The invention discloses a method for predicting an endangered grade of madrepore under future climate change based on a species distribution model. The method comprises the following steps of (1) species distribution information collection, (2) environmental factor selection, (3) model construction and model performance evaluation, and (4) species eradication risk evaluation under climate change. The invention discloses a method for evaluating the endangered grade of madrepore based on a species distribution model for the first time, and the method makes up for the dependence of the traditional IUCN evaluation on the field population quantity and trend data. According to the scheme for predicting the suitable habitat of the species under the climate change of the madrepore, the endangered grade of the madrepore under the climate change can be evaluated. The method provides beneficial enlightenment for the endangered level in evaluating or updating the eradication risk state of marine species and designing protective measures.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of marine biological endangerment level prediction, and in particular to a method for predicting the endangerment level of stony corals under future climate change based on species distribution models. BACKGROUND

[0002] In the era of global biodiversity and climate crisis, assessing the extinction risk of species under climate change is of great significance to guide conservation planning. A large number of studies have shown that the multidimensional impact of climate change has significantly affected biodiversity (Bellard et al., 2012; Pecl et al., 2017). At the same time, climate change has also become a driving factor of contemporary global extinction events, and is predicted to lead to higher extinction rates in the future (Román-Palacios & Wiens, 2020).

[0003] Among numerous marine organisms, stony corals (Anthozoa-Corallia) are particularly vulnerable. On the one hand, they are typical foundation species, supporting the most diverse marine ecosystems globally and providing important ecosystem services; on the other hand, they are highly sensitive to environmental changes (Carpenter et al., 2008). In recent years, driven by atmospheric CO2 and greenhouse gas emissions, sea water warming has triggered large-scale coral bleaching events worldwide, severely weakening coral coverage and ecological function (Hughes et al., 2017). In addition, the synergistic effects of ocean warming and acidification lead to inhibited coral growth and increased mortality (Pandolfi et al., 2011). However, despite the importance and vulnerability of corals, there is currently a lack of global systematic assessment of their extinction risk under climate change.

[0004] Scholars have proposed various methods for assessing species vulnerability under climate change, among which the method based on species distribution models (SDMs) to predict future distribution changes has received widespread attention. SDMs have been successfully applied to risk assessment of mammals (Taylor et al., 2017), birds (Runge et al., 2015), and plants (Peng et al., 2023). However, for marine species, especially stony corals which are highly sensitive to climate change, relevant research is still significantly insufficient (Pacifici et al., 2015).

[0005] In addition, existing assessment methods fail to effectively address the following technical problems and deficiencies:

[0006] Ignoring future threat factors: over-reliance on historical / current data, underestimating long-term extinction risk brought by climate change.

[0007] Data update lag: The risk status of marine species is updated with a lag, and the lack of assessment data often leads to ambiguous classification (e.g., high proportion of "data deficiency" category).

[0008] Spatial dynamic missing: Traditional assessments do not consider the dynamic shift of species distribution areas with climate change, which is not conducive to the formulation of precise protection measures. SUMMARY

[0009] As key builders of global tropical and subtropical marine ecosystems, coral reefs provide important support for marine biodiversity and ecosystem services. However, under the multiple pressures of climate change-driven sea water warming, acidification, and hypoxia, coral reef habitats are rapidly disappearing, and global population resources are facing serious threats. The future risk status of coral reefs still lacks systematic assessment.

[0010] The purpose of the present invention is to provide an innovative method for predicting changes in suitable habitat area of coral reefs under climate change based on species distribution models and determining the endangered grade, in order to realize the quantitative assessment of the extinction risk of coral species. This method can make up for the shortcomings of ignoring future climate threats, lack of systematization and quantitative standards in existing assessments, and provide scientific basis for the protection management and policy making of coral reefs.

[0011] The present invention provides a method for predicting the endangered grade of coral reefs under future climate change based on species distribution models, which comprises the following steps:

[0012] (1) Collection of species distribution information:

[0013] Expert range maps are used as modeling data. After obtaining and cropping the species distribution range, each distribution polygon is rasterized;

[0014] (2) Selection of environmental factors:

[0015] Select marine prediction variables, calculate the pairwise Pearson correlation coefficients r between prediction variables, and select the correlation coefficients with absolute value |r| less than 0.7. The prediction factors with important ecological significance to the selected species are retained for model construction;

[0016] (3) Construction of the model and evaluation of the performance of the model:

[0017] The species distribution model is constructed by MaxEnt algorithm, and the calibration area of the model is defined by recording the buffer area with a radius of 1000 kilometers around each distribution point of the species; 10000 points are randomly extracted as background data in the calibration area; the MaxEnt model parameters are selected by five-fold random cross-validation method using ENMeval package in R, and the best model is selected based on the omission rate and AUC; for the best model, in addition to AUC, the prediction performance is further estimated by true skill statistics TSS and continuous Boyce index, and the model with AUC > 0.7, TSS > 0.4 and continuous Boyce index > 0.4 is selected as suitable for habitat suitability prediction; considering two diffusion assumptions: no diffusion ability and infinite diffusion ability, the continuous habitat suitability prediction is converted into binary data by the threshold of the 10th percentile of the habitat suitability value corresponding to the species occurrence record; the change of suitable habitat of the species CSH is calculated using the following formula:

[0018] CSH = (Areafuture – Areapresent) / Areapresent × 100%

[0019] Wherein, Areapresent represents the area of the current suitable habitat, and Areafuture represents the area of the future suitable habitat, and the positive CSH value indicates that the species will expand its geographical distribution range in the future, and the negative CSH value indicates that the species will experience the contraction of the distribution range under climate change;

[0020] The MaxEnt model parameters include feature classes and regularization multiplier;

[0021] (4) Assessment of extinction risk of species under climate change:

[0022] The extinction risk of species is assessed using the IUCN Red List A3c standard, and the species suitable habitat change CSH predicted by the species distribution model SDM is associated with the threat level of the species under climate change, according to the IUCN A3c standard, the threat level of the species under climate change is simulated: extinction: CSH = -100%; critically endangered: -100% < CSH ≤ -80%; endangered: -80% < CSH ≤ -50%; vulnerable: -50% < CSH ≤ -30%; near threatened: -30% < CSH < 0%; no danger: CSH ≥ 0%.

[0023] Preferably, the obtaining is obtaining the expert range map of Acropora elegans from the IUCN Red List.

[0024] Preferably, the clipping species distribution range is clipped from the water depth distribution range of 20-163 m of the stony coral Acropora elegans in the IUCN Red List.

[0025] Preferably, the gridding of each distribution polygon is gridded to 0.1° resolution, and in order to avoid model overfitting and ensure environmental representativeness, a stratified sampling method is adopted: when the number of grids exceeds 8000, 1000 grids are randomly selected; when the number of grids is 1001-8000, 12.5% are selected; when the number of grids is 501-1000, 25% are selected; when the number of grids is 101-500, 50% are selected; when the number of grids is less than 100, all grids are retained; in order to ensure the reliability of the model, it is necessary to ensure that the number of grids is greater than 10.

[0026] Preferably, the marine prediction variables are selected from the Bio-ORACLE database version 3.0, including geographic prediction variables and marine environmental prediction variables.

[0027] Preferably, the geographic prediction variables are terrain ruggedness and water depth, and the marine environmental prediction variables are annual mean, maximum and minimum values of temperature, salinity, pH, dissolved oxygen and flow rate.

[0028] Preferably, the prediction factors with important ecological significance are maximum water flow velocity, water depth, terrain ruggedness, minimum dissolved oxygen, minimum pH, minimum salinity and maximum temperature.

[0029] Preferably, the selection of environmental factors takes into account two shared socio-economic path scenarios, SSP2-4.5 and SSP5-8.5, in the mid-21st century.

[0030] Preferably, the 900-1100 km is 1000 km, and the 9,000-11,000 points are 10,000 points.

[0031] Preferably, the specific steps are as follows:

[0032] (1) Collection of species distribution information:

[0033] The expert range map of coral Acropora elegans was obtained from the IUCN Red List, and the accuracy of the species name was checked through the World Register of Marine Species. To reduce the inappropriate areas that may be included in the expert range map, the species was cropped according to the depth distribution range of 20-163 m in the IUCN Red List. After obtaining and cropping the species distribution range, each distribution polygon was rasterized to 0.1° resolution. In order to avoid overfitting of the model and ensure environmental representativeness, stratified sampling method was used: when the number of grids is more than 8000, 1000 grids are randomly selected; when the number of grids is 1001-8000, 12.5% are selected; when the number of grids is 501-1000, 25% are selected; when the number of grids is 101-500, 50% are selected; when the number of grids is less than 100, all grids are reserved; in order to ensure the reliability of the model, it is necessary to ensure that the number of grids is greater than 10;

[0034] (2) Selection of environmental factors:

[0035] 17 marine prediction variables were selected from Bio-ORACLE database version 3.0, including 2 geographic prediction variables of terrain ruggedness and water depth, and 15 marine environmental prediction variables of annual mean, maximum and minimum values of temperature, salinity, pH, dissolved oxygen and flow rate. The pairwise Pearson correlation coefficients r between the prediction variables were calculated, and the prediction variables with absolute value |r| less than 0.7 were selected. Finally, 7 prediction factors with important ecological significance to corals were retained for model construction, which were maximum water flow rate, water depth, terrain ruggedness, minimum dissolved oxygen, minimum pH, minimum salinity and maximum temperature. The future data layer of marine prediction variables came from Bio-ORACLE database version 3.0. Two shared socio-economic path scenarios SSP2-4.5 and SSP5-8.5 in the mid-21st century (average of 2040-2050) were considered;

[0036] (3) Construction of the model and evaluation of the performance of the model:

[0037] The MaxEnt algorithm was used to construct the species distribution model. For each species, a buffer zone with a radius of 1000 kilometers was generated around each distribution point of the species to define the model calibration area. 10,000 points were randomly selected as background data within these calibration areas. Using the ENMeval package in R, the Feature Classes and Regularization Multiplier parameters of the MaxEnt model were selected by five-fold random cross-validation method, and the best model was selected based on the omission rate and AUC. For the best model, in addition to AUC, its prediction performance was further estimated by TSS and continuous Boyce index, and the model with AUC > 0.7, TSS > 0.4 and continuous Boyce index > 0.4 was suitable for habitat suitability prediction map; Given the difficulty in determining the dispersal ability of stony corals, two extreme dispersal assumptions were considered when estimating the future suitable habitat of stony corals: no dispersal ability and unlimited dispersal ability. The change in suitable habitat of the species CSH was calculated using the following formula:

[0038] CSH = (Areafuture – Areapresent) / Areapresent × 100%

[0039] Where Areapresent represents the current suitable habitat area, Areafuture represents the future suitable habitat area, and a positive CSH value indicates that the species will expand its geographic distribution range in the future, while a negative CSH value indicates that the species will experience a contraction of its distribution range under climate change.

[0040] (4) Assessment of extinction risk of species under climate change:

[0041] The improved IUCN Red List A3c standard was used to assess the extinction risk of stony corals under future climate change scenarios. The calculation was based on "one generation = 10 years", covering a prediction period of three future generations to meet the specification requirements of the A3c standard. The species distribution model SDMs was used to simulate the change in suitable habitat area CSH of stony corals under future climate scenarios, and the change range was compared with the IUCN threatened grade division standard. The judgment rule of species threat grade is:

[0042] Extinct (EX): CSH = -100%;

[0043] Critically endangered (CR): -100% < CSH ≤ -80%;

[0044] Endangered (EN): -80% < CSH ≤ -50%;

[0045] Vulnerable (VU): -50% < CSH ≤ -30%;

[0046] Near risk (NT): -30% < CSH < 0%;

[0047] No Risk (LC): CSH ≥ 0%.

[0048] This invention discloses for the first time a method for assessing the endangered status of stony corals based on species distribution models. This method overcomes the reliance of traditional IUCN assessments on in-situ population size and trend data. The scheme for predicting suitable habitats for stony corals under climate change can assess their endangered status under climate change conditions. This method provides valuable insights for assessing or updating the extinction risk status of marine species and designing conservation measures. Attached Figure Description

[0049] Figure 1 It is a correlation analysis of 17 predictor variables. Detailed Implementation

[0050] This invention provides the following technical solution:

[0051] (1) Collection of species distribution information:

[0052] Considering the significant biases in marine species distribution records, expert range maps were used as modeling data. Expert range maps of stony coral species were obtained from the IUCN Red List, and the accuracy of species names was verified against the World Marine Species List. To reduce potentially unsuitable areas included in the expert range maps, the distribution maps were cropped based on the species' depth distribution range. After obtaining and cropping the species distribution ranges, the distribution polygons were rasterized to a 0.1° resolution and converted into point records. To avoid model overfitting and ensure environmental representativeness, stratified sampling was used to retain points. Simultaneously, to ensure model reliability, the number of species distribution point records was ensured to be greater than or equal to 10.

[0053] (2) Selection of environmental factors:

[0054] Based on literature evidence, marine predictor variables were selected from the Bio-ORACLE database version 3.0. To prevent model overfitting, pairwise Pearson correlation coefficients (r) between predictor variables were calculated, and correlation coefficients with an absolute value |r| less than 0.7 were selected. Finally, predictor factors with significant ecological importance to the selected species were retained for model construction. The future data layer for marine environmental predictor factors came from the Bio-ORACLE database version 3.0. Two shared socioeconomic path scenarios (SSPs, SSP2-4.5, and SSP5-8.5) were considered for the mid-21st century (average from 2040 to 2050).

[0055] (3) Model construction and model performance evaluation:

[0056] The MaxEnt algorithm was used to construct the species distribution model. The calibration area of the model was defined by generating a 1000-kilometer radius buffer zone around each distribution point of the species. Within these calibration areas, 10,000 points were randomly selected as background data. Using the ENMeval package in R, the MaxEnt model parameters (Feature Classes and Regularization Multiplier) were selected through five-fold random cross-validation method, and the best model was selected based on omission rate and AUC (Area Under the Curve of Receiver Operating Characteristic Curve, ranging from 0 to 1). For the best model, in addition to AUC, its prediction performance was further evaluated by True skill statistics (TSS, ranging from -1 to 1) and Continuous Boyce Index (Boyce, ranging from -1 to 1). According to the recommendation of Engler et al. (2011), models with AUC > 0.7, TSS > 0.4, and Continuous Boyce Index > 0.4 are suitable for habitat suitability prediction. Two diffusion assumptions were considered: no diffusion ability and unlimited diffusion ability. Given that there was only presence data in this study, the continuous habitat suitability prediction was converted to binary data by using the threshold of the 10th percentile of the habitat suitability value of the corresponding species occurrence record (also known as 10% omission rate). The following formula was used to calculate the change in species suitable habitat (CSH) in percentage:

[0057] CSH = (Areafuture - Areapresent) / Areapresent x 100%

[0058] Where Areapresent represents the current suitable habitat area, and Areafuture represents the future suitable habitat area. A positive CSH value indicates that the species will expand its geographic distribution in the future, while a negative CSH value indicates that the species will experience a contraction of its distribution range under climate change.

[0059] (4) Assessment of species extinction risk under climate change:

[0060] The extinction risk of a species was assessed using the IUCN Red List A3c criteria, and the species threat level under climate change was linked to the change of species suitable habitat (CSH) predicted by species distribution models (SDMs). According to the IUCN A3c criteria, the species threat level under climate change was simulated: extinct (CSH = -100%), critically endangered (-100% < CSH ≤ -80%), endangered (-80% < CSH ≤ -50%), vulnerable (-50% < CSH ≤ -30%), near threatened (-30% < CSH < 0%), and least concern (CSH ≥ 0%).

[0061] The scheme of the present application for predicting the suitable habitat of a species under climate change of corals can assess the endangered level of corals under climate change.

[0062] The following examples are further illustrations of the present application and are not intended to limit the present application.

[0063] Example 1: Collection of species distribution information

[0064] (1) Collection of species distribution information:

[0065] Considering the significant bias in marine species distribution records, expert range maps were used as modeling data. The expert range map of the coral Acropora elegans was obtained from the IUCN Red List, and the accuracy of the scientific name of the species was checked through the World Register of Marine Species. To reduce the possible inclusion of unsuitable areas in the expert range map, the species was cropped according to its depth distribution range (20-163 m) in the IUCN Red List. After obtaining and cropping the species distribution range, each distribution polygon was rasterized to a resolution of 0.1°. To avoid overfitting of the model and ensure environmental representativeness, stratified sampling was used: when the number of grids was more than 8000, 1000 grids were randomly selected; when the number of grids was between 1001 and 8000, 12.5% were selected; when the number of grids was between 501 and 1000, 25% were selected; when the number of grids was between 101 and 500, 50% were selected; and when the number of grids was less than 100, all grids were retained. At the same time, to ensure the reliability of the model, the number of grids must be greater than 10. Finally, 292 grids of the coral were retained for subsequent analysis.

[0066] (2) Selection of environmental factors:

[0067] Based on the literature evidence, 17 marine predictors were selected from the Bio-ORACLE database version 3.0, including 2 geographic predictors (topographic roughness and water depth) and 15 marine environmental predictors (annual mean, maximum and minimum of temperature, salinity, pH, dissolved oxygen and current velocity). To prevent overfitting of the model, the pairwise Pearson correlation coefficients r between the predictors were calculated, and the predictors with absolute value |r| less than 0.7 were selected as Figure 1 Finally, 7 predictors with important ecological significance for corals were retained for the model construction, including the maximum current velocity, water depth, topographic roughness, minimum dissolved oxygen, minimum pH, minimum salinity and maximum temperature. The future data layers of marine environmental predictors were from the Bio-ORACLE database version 3.0. Two shared socio-economic pathway scenarios (SSPs) SSP2-4.5 and SSP5-8.5 in the mid-21st century (average of 2040-2050) were considered.

[0068] (3) Model construction and evaluation of model performance:

[0069] The environmental variable values corresponding to the 7 environmental factors selected above were used as independent variables, and the species distribution point information was used as the response variable. The MaxEnt algorithm was used to construct the species distribution model. For each species, the calibration area of the model was defined by generating a buffer zone with a radius of 1000 kilometers around each distribution point of the species. Within these calibration areas, 10,000 points were randomly selected as background data. Using the ENMeval package in R, the MaxEnt model parameters (Feature Classes and Regularization Multiplier, parameter values are L and 205, respectively) were selected by five-fold random cross-validation method, and the best model was selected based on the omission rate and AUC (Area Under the Curve, range from 0 to 1). In addition to AUC, the prediction performance of the best model was further evaluated by True skill statistics (TSS, range from -1 to 1) and Continuous Boyce Index (Boyce, range from -1 to 1), as shown in Table 1.

[0070] Following the recommendations of Engler et al. (2011), models with AUC > 0.7, TSS > 0.4, and continuous Boyce index > 0.4 were considered suitable for habitat suitability prediction maps. Given the difficulty in determining the dispersal ability of SPS, two extreme dispersal assumptions were considered when estimating the future suitable habitat of SPS: no dispersal ability and unlimited dispersal ability. The change in species suitable habitat (CSH) was calculated using the following formula (in percentage):

[0071] CSH = (Areafuture - Areapresent) / Areapresent x 100%

[0072] where Areapresent represents the current suitable habitat area, and Areafuture represents the future suitable habitat area. A positive CSH value indicates that the species will expand its geographic distribution in the future, while a negative CSH value indicates that the species will experience a contraction in its distribution range under climate change.

[0073] Table 1. SDMs performance of SPS Acropora elegans

[0074]

[0075] (4) Assessment of species extinction risk under climate change:

[0076] This technology uses the IUCN Red List A3c standard to assess the extinction risk of SPS under future climate change scenarios. This standard emphasizes the determination of species threat levels based on model-predicted future population or habitat changes, with an assessment time scale covering three generations or ten years (whichever is longer) of the species, and a maximum of one hundred years. SPS belongs to a group of corals with relatively long lifespans, with a generation turnover time generally around 10 years. According to the requirements of the IUCN manual, this technology calculates with "one generation = 10 years" in the assessment, so it should cover a prediction period of three generations (i.e., 30 years) to meet the normative requirements of the A3c standard.

[0077] In the specific implementation process, the area change of suitable habitat of SPS (Climate Suitable Habitat, CSH) under future climate scenarios is simulated using species distribution models (SDMs), and the change amplitude is compared with the IUCN threatened grade division standard.

[0078] In this technology, the determination rule of species threat level is:

[0079] Extinction (EX): CSH = -100%;

[0080] Critically Endangered (CR): -100% < CSH ≤ -80%;

[0081] Endangered (EN): -80% < CSH ≤ -50%;

[0082] Vulnerable (VU): -50% < CSH ≤ -30%;

[0083] Near Threatened (NT): -30% < CSH < 0%;

[0084] Least Concern (LC): CSH ≥ 0%.

[0085] The results showed that future climate change would increase the threat level of the coral species Acropora elegans. For example, in the case of the results of no dispersal ability in 2040-2050 under the SSP2-4.5 scenario, the originally listed as Least Concern was classified as Vulnerable species; in the case of the results of no dispersal ability in 2040-2050 under the SSP2-8.5 scenario, the originally listed as Least Concern was classified as Endangered species; in the case of the results of unlimited dispersal ability in 2040-2050 under the SSP2-4.5 scenario, the originally listed as Least Concern was classified as Vulnerable species; in the case of the results of unlimited dispersal ability in 2040-2050 under the SSP2-8.5 scenario, the originally listed as Least Concern was classified as Endangered species, as shown in Table 2.

[0086] Table 2. IUCN Red List threat level assessment of coral species based on SDMs

[0087]

Claims

1. A method for predicting the endangered status of stony corals under future climate change based on a species distribution model, characterized in that, It includes the following steps: (1) Collection of species distribution information: Using the expert range map as modeling data, after obtaining and cropping the species distribution range, each distribution polygon is rasterized; (2) Selection of environmental factors: Select marine predictor variables, calculate the pairwise Pearson correlation coefficient r between the predictor variables, and select the correlation coefficients with absolute value |r| less than 0.7, and retain the predictor factors that are of important ecological significance to the selected species for model construction; (3) Construction of the model and evaluation of model performance: Construct the species distribution model through the MaxEnt algorithm. For each species, define the model calibration area by recording buffers with a radius of 900 - 1100 kilometers around each distribution point where the species appears; randomly select 9,000 - 11,000 points as background data within these calibration areas; use the ENMeval package in R to select the MaxEnt model parameters through the five-fold random cross-validation method, and select the best model based on the omission rate and AUC; for the best model, in addition to AUC, further estimate its prediction performance through the true skill statistic TSS and the continuous Boyce index, and select the model with AUC > 0.7, TSS > 0.4 and continuous Boyce index > 0.4 as applicable to habitat suitability prediction; consider two diffusion assumptions: no diffusion ability and infinite diffusion ability, and convert the continuous habitat suitability prediction into binary data through the threshold of the 10th percentile of the habitat suitability value corresponding to the species occurrence record; Calculate the change in species suitable habitat CSH using the following formula: CSH = (Areafuture – Areapresent) / Areapresent × 100% Where, Areapresent represents the area of the current suitable habitat, Areafuture represents the area of the future suitable habitat. A positive CSH value indicates that the species will expand its geographical distribution range in the future, while a negative CSH value indicates that the species will experience a contraction in the distribution range under climate change; The MaxEnt model parameters include Feature Classes and RegularizationMultiplier; (4) Assessment of species extinction risk under climate change: Use the IUCN Red List A3c standard to assess the extinction risk of species, link the change in species suitable habitat CSH predicted by the species distribution model SDMs with the species threat level under climate change, and simulate the species threat level under climate change according to the IUCN's A3c standard: Extinct: CSH = -100%; Critically Endangered: -100% < CSH ≤ -80%; Endangered: -80% < CSH ≤ -50%; Vulnerable: -50% < CSH ≤ -30%; Near Threatened: -30% < CSH < 0%; Least Concern: CSH ≥ 0%.

2. The method according to claim 1, characterized in that, The acquisition refers to the expert range map of Acropora elegans obtained from the IUCN Red List.

3. The method according to claim 1, characterized in that, The species distribution range for the cropping was determined based on the depth range of Acropora elegans (20–163 m) as listed in the IUCN Red List.

4. The method according to claim 1, characterized in that, The process involves rasterizing each distributed polygon to a resolution of 0.1°. To avoid model overfitting and ensure environmental representativeness, a stratified sampling method is used: when the number of grid cells exceeds 8000, 1000 grid cells are randomly selected; when the number of grid cells is 1001-8000, 12.5% ​​are selected; when the number of grid cells is 501-1000, 25% are selected; when the number of grid cells is 101-500, 50% are selected; and when the number of grid cells is less than 100, all grid cells are retained. To ensure the reliability of the model, the number of grid cells must be greater than 10.

5. The method according to claim 1, characterized in that, The marine predictor variables were selected from the Bio-ORACLE database version 3.0 and include geographic predictor variables and marine environmental predictor variables.

6. The method according to claim 5, characterized in that, The geographical prediction variables are topographic ruggedness and water depth, and the marine environmental prediction variables are the annual average, maximum, and minimum values ​​of temperature, salinity, pH, dissolved oxygen, and current velocity.

7. The method according to claim 1, characterized in that, The ecologically significant predictors are the maximum water flow velocity, water depth, topographic ruggedness, minimum dissolved oxygen, minimum pH, minimum salinity, and maximum temperature.

8. The method according to claim 1, characterized in that, The selection of environmental factors also takes into account two shared socioeconomic path scenarios, SSP2-4.5 and SSP5-8.5, in the mid-21st century.

9. The method according to claim 1, characterized in that, The 900-1100 km range is 1000 km, and the 9,000-11,000 points are 10,000 points.

10. The method according to claim 1, characterized in that, The specific steps are as follows: (1) Collection of species distribution information: Expert range maps of the stony coral *Acropora elegans* were obtained from the IUCN Red List, and the accuracy of the species name was verified using the World Register of Marine Species. To reduce potentially unsuitable areas in the expert range maps, they were cropped based on the species' depth distribution range of 20–163 m as listed in the IUCN Red List. After obtaining and cropping the species distribution range, each distribution polygon was rasterized to a resolution of 0.1°. To avoid model overfitting and ensure environmental representativeness, a stratified sampling method was used: when the number of raster cells exceeded 8000, 1000 raster cells were randomly selected; when the number of raster cells was 1001–8000, 12.5% ​​were selected; when the number of raster cells was 501–1000, 25% were selected; when the number of raster cells was 101–500, 50% were selected; and when the number of raster cells was less than 100, all raster cells were retained. To ensure the reliability of the model, the number of raster cells had to be greater than 10. (2) Selection of environmental factors: Seventeen marine predictor variables were selected from the Bio-ORACLE database version 3.0, including two geographic predictor variables (topographic ruggedness and water depth) and fifteen marine environmental predictor variables (annual mean, maximum, and minimum values ​​of temperature, salinity, pH, dissolved oxygen, and current velocity). Pairwise Pearson correlation coefficients (r) were calculated among the predictor variables, and predictor variables with an absolute value |r| less than 0.7 were selected. Finally, seven predictor factors with significant ecological importance for stony corals were retained for model construction: maximum current velocity, water depth, topographic ruggedness, minimum dissolved oxygen, minimum pH, minimum salinity, and maximum temperature. The future data layer for the marine predictor variables came from the Bio-ORACLE database version 3.

0. Two shared socioeconomic pathway scenarios, SSP2-4.5 and SSP5-8.5, were considered for the mid-21st century. (3) Model construction and performance evaluation: Species distribution models were constructed using the MaxEnt algorithm. For each species, a model calibration region was defined by generating a buffer zone with a radius of 1000 km around each distribution point. 10,000 points were randomly selected within these calibration regions as background data. Using the ENMeval package in R, the feature classes and regularization multiplier in the MaxEnt model parameters were selected using a five-fold randomized cross-validation method. The optimal model was chosen based on the omission rate and AUC. For the optimal model, in addition to AUC, its predictive performance was further estimated using the True Skill Statistics (TSS) and the Continuous Boyce Index. Models with AUC > 0.7, TSS > 0.4, and Continuous Boyce Index > 0.4 were selected for habitat suitability prediction maps. Given the unknown dispersal capacity of stony corals, two extreme dispersal assumptions were considered when estimating their future suitable habitats: no dispersal capacity and unlimited dispersal capacity. The change in suitable habitat (CSH) for the species was calculated using the following formula: CSH = (Areafuture – Areapresent) / Areapresent × 100% Where Areapresent represents the area of ​​currently suitable habitat, and Areafuture represents the area of ​​future suitable habitat. A positive CSH value indicates that the species will expand its geographical range in the future, while a negative CSH value indicates that the species will experience a shrinkage of its range due to climate change. (4) Species extinction risk assessment under climate change: This technology uses the improved IUCN Red List A3c standard to assess the extinction risk of stony corals under future climate change scenarios. The assessment uses a "one generation = 10 years" calculation, covering the predicted timeframe of three future generations to meet the normative requirements of the A3c standard. Species distribution models (SDMs) are used to simulate the change in suitable habitat area (CSH) of stony corals under future climate scenarios, and the magnitude of this change is correlated with the IUCN threat level classification criteria. The rules for determining the species threat level are as follows: Extinction (EX): CSH = -100%; Critically Hazardous (CR): -100% < CSH ≤ -80%; Endangered (EN): -80% < CSH ≤ -50%; Vulnerable (VU): -50% < CSH ≤ -30%; Near risk (NT): -30% < CSH < 0%; No Risk (LC): CSH ≥ 0%.

Images