A Numerical Study Of The Ecosystem Dynamics Of Ecological Variables In Xiamen Bay

Xiamen Bay,located in the southeast coast of China,is an important port of opening-up for Fujian Province.However,due to the rapid development of economy and the continuous advancement of industrialization,the increase of pollutants into the sea,sea reclamation and overexploitation caused a sharp deterioration in the ecosystem service function of Xiamen Bay.The fundamental cause of this is that we lack comprehensive and in-depth understanding about offshore marine ecosystem.An in-depth understanding about the changing laws of ecological variables in the Marine ecosystem of Xiamen Bay has important guiding significance about the protection of ecological environment and can provide theoretical support for improving the ecological environment quality of Xiamen Bay.In this paper,a 3D hydrodynamic model was coupled with the NPZD ecosystem dynamical model to establish a 3D marine ecosystem dynamical model of Xiamen Bay.The model was used to simulate the seasonal variation of ecological variables,to analyze the impact of physical processes and ecological processes on the changes of nutrients and chemical oxygen demand,and to study the balance of nutrients and chemical oxygen demand in Xiamen Bay.Later,a prediction model of ecological variables was established by adding neural network technology.Firstly,a 3D hydrodynamic model of Xiamen Bay is built to simulate the hydrodynamic environment of four seasons.The results show that semi-diurnal tide is the main tidal component in the Xiamen Bay;Salinity is generally low in Jiulong river Estuary,but high outside the bay,which is affected by river runoff and tide.The average temperature is highest in summer,then spring and autumn,and lowest in winter.Next,the NPZD ecosystem dynamical model of Xiamen Bay is constructed,and the calculation results show that the overall distribution trend of each ecological variable is decreasing from inner bay to the outer bay,and from the surface to the bottom.The seasonal changes of ecological variables in Xiamen Bay are obvious.The biomass of phytoplankton is higher in summer and winter,and lower in autumn and spring.The seasonal variation of chemical oxygen demand is consistent with phytoplankton.The biomass of zooplankton is consistent with the seasonal variation of temperature,i.e.,the highest in summer,then in spring and autumn,and the lowest in winter.The concentration of inorganic phosphorus was higher in autumn and summer,lower in spring and lowest in winter.The concentration of inorganic nitrogen is greatly affected by the runoff of the Jiulong river.The concentration of inorganic nitrogen in the estuary area and the surrounding sea area is directly proportional to the runoff(i.e.,higher in spring and summer,lower in autumn and winter),and inversely proportional to the biomass of phytoplankton in other sea areas(i.e.,higher in spring and autumn,lower in summer and winter).The concentration of organic carbon in estuary area is directly proportional to the runoff(i.e.,high in spring and summer,low in autumn and winter),while in other sea area is consistent with the biomass of phytoplankton(i.e.,high in summer and winter,low in spring and autumn).Dissolved organic carbon was higher in spring and summer,and lower in autumn and winter.The daily change of ecological variables is significantly affected by tide.There are two peak values and two valley values in a day,and the change amplitude is positively correlated with tidal amplitude.The influence of the advection process is one order of magnitude larger than that of the diffusion process and two orders of magnitude larger than that of the ecological process.Under the combined action of physical and ecological processes,the amount of inorganic phosphorus,inorganic nitrogen and chemical oxygen demand in Xiamen bay increased by 0.56t,2.64t and 4.7t per day.Finally,the neural network technology is combined with the data output from the ecosystem dynamical model to construct a predictive model for ecological variables.The results show that the prediction model works well in learning the relationship between input variables and output variables.Comparing three type of neural network,BP-LM network has the best performance,and CNN and improved BPNN perform similarly.