Construction Of Bayesian Network Model (FBM) Based On Fuzzy Membership Degree And Analysis Of Species Habitat Suitability

The distribution of species is affected by their surrounding environmental factors.Species distribution models(SDMs)are models for studying the relationship between environmental factors and habitat suitability.With the development of SDMs,their accuracy and uncertainty are also receiving more and more attention.In this study,a Bayesian network model(FBM)based on fuzzy membership is established.The fuzzy mathematics and Bayesian theory are used to reduce the uncertainty of the model,simulate the relationship between environmental variables,and explore the relationship between environmental variables and the growth of vegetation to predict the habitat suitability of vegetation and reinterpret the potential suitability of vegetation.This paper uses the existing literature and experience to construct three design frameworks for FBM.Taking the medicinal plant A.sinensis as an example.From three aspects,such as climate,soil,and topography to select 29 kinds of environment variables,after screening retained the eight kinds of environment variables,including Biol(Annual mean air temperature),Bio10(Mean temperature of warmest quarter),Biol 2(Annual precipitation),Bio15(Precipitation seasonality),Bio18(Precipitation of the warmest quarter),SOM(Soil organic matter),Totalrad13(Annual total solar radiation)and EL(Elevation above sea level)to participate in modeling,combining 90 sample sites data of A.sinensis,nine sample sites sequences were generated,and A,B and C three model frameworks of A.sinensis FBM were constructed.The predicted sequence of 27 habitat suitability was calculated for 50 predicted sites.The accuracy and stability of the model are analyzed by using the sequence of 27 prediction results,Log-likelihood and AIC indices.The reliability of FBM was verified by comparing the results of FBM and FME(Fuzzy matter element model)at 10 detection sites.In the process of sensitivity analysis of model network nodes,the influence of single environmental variables(SEVs)and integrated environmental variables(IEVs)on the habitat suitability of A.sinensis was explained in detail.Main conclusions of this paper:1.In the A.sinensis FBM network constructed,model C has the best prediction effect,followed by model B and model A.A.sinensis FBM network is divided into three layers,the outer layer is 8 kinds of SEVs nodes,the middle layer is 3 kinds of IEVs nodes,and the core layer is A.sinensis habitat suitability node.Based on model A,model C considers the influence of temperature on precipitation,temperature and precipitation on soil,and highlights the importance of elevation above sea level,annual mean air temperature and annual precipitation on A.sinensis growth.2.Model test indicators show that with the increase of sample sites involved in modeling,the stability and accuracy of the model gradually become stable.The amount of basic sample site data required by the three frameworks of A.sinensis FBM model should be more than 40.When the sample data volume reaches 70 and above,the model prediction results tend to be stable.3.The optimal prediction of the habitat suitability of A.sinensis is the result of sequence 9 of FBM model C.It can be seen that the highly suitability habitat locations of A.sinensis are mainly distributed in Gansu and Shaanxi,which is consistent with the authentic production areas of A.sinensis.4.Using the A.sinensis FBM probability limit setting to analyze the environmental variables,Biol,Bio 12 and SOM are important and sensitive environmental variables affecting the suitability of A.sinensis habitats from the perspective of the influence of single environmental variables on the suitable habitat of A.sinensis.When the annual mean air temperature is between 5.92 ?-9.05 ?,the annual precipitation is between 568.79 mm-791.33 mm,and the soil organic matter ranges from 83.91 ‰-102.21‰,these three variables have a strong promoting effect on the highly suitable growth of A.sinensis.5.Compared with the FME model prediction results,A.sinensis FBM retained the prediction accuracy of the FME model and gave an accurate probability evaluation for each habitat suitability level.In this study,a species distribution model FBM with a new prediction mechanism was constructed.Based on the fuzzy membership function,the model was combined with the maximum entropy principle,fuzzy mathematics comprehensive evaluation and Bayesian network framework to predict the habitat suitability of species.FBM combines species characterization data,environmental data and expert experience to observe the occurrence of habitat suitability at each level from the perspective of probability,which can reduce data and system errors,solve the problem that researchers lack sample data of a large number of species,and simulate relationships and parameters that may not be effectively quantified between environmental variables in a species habitat.The expression of network graphics can help researchers to clarify research ideas while exploring probability transmission,and has broad application prospects.