Coupling Multivariate Statistical Prediction And Factorial Analysis Approaches For Identifying Guangzhou-Foshan Forest Coverage Variation

In recent years,due to the rapid growth of population,accelerated pace of urban construction and the development of wood-processing industry,natural forest coverage decreased sharply,affecting the ecosystem structure and function in China.The contradiction between forest timber resources and urban development is increasingly sticking out,which brings the challenge for protecting urban ecological red line.As big and medium-sized cities(Guangzhou and Foshan),whose forest coverage have greatly changed in resent years,gather large amount of population and industrial bases.For propelling the construction of "forest agglomeration" in Pearl River Delta,identifying Guangzhou-Foshan forest coverage variation rule and main driving factors,managing and protecting forest resource are necessary.However,the relationship between complex forest ecosystem and social-economy-natural processes is unclear;there may be interaction between various influencing factors and a single analysis method is unable to fully reveal the above complex relationship.Therefore,aiming at the above problems,in this study,a set of integrated method system-coupling multivariate statistical prediction and factorial analysis(MSPFA)approaches is developed.In this approach,four multivariate statistical prediction methods,including multiple linear regression(MLR),multiple quantile regression(MQR),support vector machine(SVM)and stepwise cluster analysis(SCA),are coupled with factorial analysis(FA)for dealing with the linear,nonlinear,continuous and discrete complex relations between forest area and factors such as human society and natural process,and for quantifying factors interactions on forest coverage change.The developed MSPFA method was applied to Guangzhou-Foshan to explore the complex relationship and main driving force between the variation of forest coverage and human social activities and natural processes.The results showed that the simulation efficiency of forest coverage variation was different in different multivariate statistical prediction methods.Specifically,SCA and SVM provide better simulation than MLR and MQR.This is because SCA and SVM can fully represent the nonlinear relationship of the system,indicating that the variation of urban forest area is a complex nonlinear process.Population,the consumption of log,the stock of log and precipitation are the main driving factors affecting the variation of forest area,and population contributes the most the variation of forest area.High population density and urbanization will affect the local climate(such as precipitation),then affecting the growth conditions of forest plants and finally changing the urban forest area.The population of Guangzhou and the demand for timbers will greatly affect the variation of forest area between Guangzhou and Foshan,because Guangzhou has a higher population and the demand for timbers resources than Foshan.Therefore,in order to manage the forest area of Guangzhou-Foshan,we should reasonably control the population growth of Guangzhou and reduce log consumption.In this paper,the results analyze the change of forest coverage in Guangzhou-Foshan,summarize the main influencing factors,and provide theoretical basis and effective information for the management of the forest ecosystem in Guangzhou-Foshan.