Modeling The Impacts Of Marine Subsurface Environmental Variables On Swordfish (Xiphias Gladius) Habitat Suitability In The Western Indian Ocean

The study of species habitat distribution can contribute to the conservation and management of species.In the case of fisheries,predicting the habitat distribution of harvested fish species can maximize the value of harvesting.Species Distribution Model（SDM）,as a common model and core of basic theoretical and applied research in ecology and biogeography,is also an important tool for fisheries ecologists and managers to assess and manage the habitat and distribution of marine organisms,and is widely used in many aspects,especially in the field of species habitat distribution studies.The use of the data is also an important tool for fisheries ecologists and managers to assess and manage marine habitat and distribution.With the development of methodologies and techniques,many new modeling approaches have emerged,providing new ideas and experiences for habitat studies in complex situations（e.g.,limited sample size of species data and only species presence data）.However,in the past,most of the studies on species habitat distribution treated the habitats of species as two-dimensional planes,ignoring the three-dimensional nature of the real environment.Compared with terrestrial environments,the marine environment lacks absolute geographic barriers,and marine species in oceanic habitats use the pelagic water column vertically by moving vertically,especially for large species.Although many scholars have considered this issue for a long time and have made some studies and attempts,more research and experience are needed for the vastness of the ocean and the richness of species.In this study,the swordfish（Xiphias gladius）was selected as the subject of study in an attempt to model its three-dimensional habitat.As a typical highly migratory large pelagic fish,the role and importance of various marine subsurface environmental variables in modeling its habitat suitability has not been well understood in previous studies.Moreover,for similar large marine species,there are differences in habitat environmental preferences at different stages of their life histories,particularly between individual adults and juveniles,which have not been distinguished in previous studies.Therefore,this study screened adult swordfish data from the Chinese Indian Ocean Tuna longline fishery observer data from 2017-2019 as species distribution data,and seawater temperature（T）,sea water salinity（S）,dissolved oxygen（DO）,net primary production（NPP）,northward sea water velocity（NV）,eastward sea water velocity（EV）and ocean mixed layer depth（MLD）are used as environmental factors,and among them,T,S,DO and NPP have subsurface data.Based on the small sample size of observer data but accurate distribution records,the model results generated by comparing subsurface variables with surface-only variables using the maximum entropy model（Max Ent）.The model performance was evaluated by combining AUC_train,AUC_test,Akaike information criterion（AIC）,Bayesian information criteria（BIC）,Sensitivity,and True Skill Statistics（TSS）to investigate whether the addition of subsurface variables could improve the model performance and affect the habitat suitability simulation results.The results of this study showed that（1）with the addition of subsurface environmental variables,all indicators showed that the model’s effectiveness was significantly improved and reached the optimal level at 300 m depth;（2）The results of the importance of environmental factors based on the Jackknife method and the contribution show that among all environmental variables,T,DO,NPP and MLD are the four most important factors,T₀ and T₁₀₀ are the first important factors in the surface and100m depth models,respectively,but as the water layer deepens,dissolved oxygen in the deep layer will replace temperature as the first important environmental factor;（3）T There is a regularity in the spatial distribution of swordfish habitat,and in the model results at the optimal level,it is mainly distributed in two regions in the northwest Indian Ocean,one is the region between 10°N-10°S near the eastern coast of the African continent,and the other is the region where the Arabian Sea meets the Indian Ocean at the same latitude range,and its longitude range is roughly 55°E-70°E;（4）Indian Ocean swordfish show a clear seasonal migration in the time series,showing a relatively stable trend of change between May-September in summer for bait migration and October-April in winter for spawning migration;（5）In the third dimension of the spatial scale,the habitat distribution of swordfish at different depths in the same period has obvious changes,compared with the surface model,with the addition of subsurface variables,the model results in different degrees of contraction of the habitat suitability area at 300m depth,and the location and extent of the high suitability habitat area in some cases have changed significantly,and in some months there are obvious high suitability areas from none to none;（6）there was a significant positive correlation between the suitability results predicted by the model and swordfish abundance,but it was not strong,and suitability could not perfectly explain the changes in abundance.Therefore,this study concluded that the inclusion of marine subsurface environmental variables would have a significant impact on habitat modeling of swordfish in the Western Indian Ocean and would enhance the prediction of distribution models.The results of this study can provide practical experience to better understand the selection of important environmental variables and improvement of model performance in 3D habitat modeling of oceanic fishes,and can serve as an example of the need to consider 3D environments in habitat studies of marine species.However,there are limitations in this study and more practice is needed for the rest of the methods and for realistic habitat distribution studies of other species.