Study On The Habitat Suitability Of Coilia Nasus In The Yangtze River Estuary Based On Different Species Distribution Models

The Yangtze River estuary is the largest estuary in China and even in Asia.It is a highly dynamic salt and freshwater confluence area,where complex interactions such as material exchange and energy flow take place frequently and fully,making the Yangtze River estuary a nutrient-rich and highly productive water,providing a physiological buffer for many aquatic organisms to adapt such as baiting,nursing,spawning and migrating.However,due to environmental pollution,the construction of water conservancy projects and continuous intensive fishing,the ecological environment of the Yangtze River estuary has gradually deteriorated,and fishery resources have been in serious decline,with some aquatic organisms even on the verge of extinction.Coilia nasus,historically one of the most important fishing targets in the Yangtze River estuary,has been declining year by year from the end of the 20 th century to the last decade or so,with the amount of fish caught declining sharply and the fishing season nearly disappearing.In order to protect Coilia nasus and other fish resources,the Ministry of Agriculture and Rural Affairs has imposed a ban on fishing in the Yangtze River basin,starting from 1 January 2021,in the waters within the no-take management zone of the Yangtze River estuary.Therefore,in the context of the "Great Protection of the Yangtze River",it is necessary to adopt a scientific approach to understand the habitat preference and spatial and temporal distribution of Coilia nasus in the Yangtze River estuary,so as to provide a scientific basis for the management and conservation of Coilia nasus resources and the identification of its habitat,and the results of the study will also be of positive significance in enhancing the understanding of the restoration of the aquatic ecosystem in the Yangtze River estuary.Species distribution models are a common and effective tool for investigating the relationship between the spatial distribution of species and environmental variables,and for predicting the potential habitat of species.Due to the diversification of statistical theory and the development of computer technology,a variety of species distribution models have emerged that are based on different theories,assumptions and algorithms and have improved functionality and scope of application to suit different study scales,objectives and species habitats.Species distribution models have been widely used in both terrestrial and marine ecology.The use of species distribution models to investigate the spatial distribution of fish and other aquatic organisms in ecologically sensitive estuaries such as the Yangtze River estuary is relatively rare,and the species distribution models used in these studies are relatively traditional and homogeneous.Different species distribution models exhibit different simulation performance and applicability under different study species and external conditions.Therefore,selecting a suitable species distribution model is an essential process for modelling the spatial and temporal distribution of fish and other aquatic organisms in the Yangtze estuary and assessing habitat suitability.Most studies on Coilia nasus in the Yangtze estuary have focused on the physiological and reproductive characteristics and population structure of Coilia nasus,and there is almost no research on the spatial distribution of Coilia nasus using a variety of species distribution models with different algorithms and comparing the performance of the models.In this study,seven common single-species distribution models were applied,including Artificial Neural Networks(ANN),Classification Tree Analysis(CTA),Flexible Discriminant Analysis(FDA),and Generalized Additive Model(GAM),Generalized Boosting Models or called Boosted Regression Trees(GBM/BRT),Random Forest(RF),Surface Range Envelope(SRE)and 1 Combined Model,The spatial distribution and habitat preferences of winter Coilia nasus in the mouth of the Yangtze River were simulated,the stability and accuracy performance of each model were evaluated and compared from different indicators,the model with the best simulation effect was selected,and the prediction results were verified by using the actual resource density data of knifefish.The results of this study show that:(1)According to two model evaluation criteria,True Skill Statistic(TSS)and Area Under the Curve(AUC),the single model had some predictive ability,except for the SRE model,which failed completely,and the RF and GBM models had high model accuracy in all seasons.The AUC and TSS values of the combined model were higher than those of all single species distribution models.The RF model,GBM model and the combined model,which performed the best in predicting the spatial distribution of Coilia nasus,were compared using the actual resource density of Coilia nasus.The actual resource density of Coilia nasus was in good agreement with the predicted results,and Coilia nasus was mostly found in the areas where the model predicted the possible distribution of Coilia nasus,and the actual distribution of Coilia nasus was more concentrated and the resource density was higher in the areas where the model predicted a higher probability of distribution of Coilia nasus.(2)According to the results of the simulation of the distribution of Coilia nasus in the Yangtze River estuary by single-species distribution model and combined model,Coilia nasus occurs in the area where salt and fresh water meet in the Yangtze River estuary all year round,and in winter,the distribution of Coilia nasus is more scattered,with more distribution near the north and south branches of the estuary.In summer,the distribution of Coilia nasus extends from near the gateway of the north port of the south branch to the mouth along the waterway in the south of Chongming Island and the north of Changxing Island,with a dense distribution,and only sporadic distribution near the north branch and its connection to the outer sea;the distribution area and resource density of Coilia nasus near the outer sea of the north port of the south branch reaches its maximum in autumn.(3)Different environmental variables contributed differently to each model.All the environmental variables involved in the modelling had an impact on the models,while depth and salinity contributed more to the different models in each season,i.e.salinity and depth were the two most important environmental factors affecting the distribution of Coilia nasus in each season.In this study,different species distribution models,including a new machine learning model and a traditional statistical model,were used to simulate and compare the distribution of Coilia nasus in the Yangtze River estuary,providing a multi-model simulation experience and a relatively suitable species distribution model for the study of the spatial distribution and habitat suitability of Coilia nasus in the Yangtze River estuary.The study provides a good example and a more effective modelling experience for the study of fish and other aquatic organisms in the Yangtze River estuary.In addition,it also provides research and theoretical reference for the formulation and implementation of management and conservation strategies for fish resources represented by Coilia nasus in the Yangtze River Estuary,and provides some guidance for the spatial planning and management of the Yangtze River Estuary Watershed Protection Area.