Impacts Of Physical Parameter Uncertainties On Marine Ecosystem Simulation

The uncertainty of model parameters has been one of the important research issues in numerical simulation.There are many parameters in marine ecosystem models and the values of these parameters show large uncertainties due to limited observation or different data sources.Using conditional nonlinear optimal perturbation related to parameter(CNOP-P)approach,the impacts of model parameter(especially physical parameters)uncertainties on the simulation of phytoplankton biomass are studied.Firstly,the effects of sea surface temperature(SST)and mixed layer depth(MLD)are explored using the NPZD model.We find that the uncertainty of MLD significantly affects the simulation of phytoplankton biomass at downstream of the Kerguelen plateau(50°S,80E;60°S,80E),whereas SST exhibits limited influence.Specifically,the differences of SST originated from satellite remote sensing or buoy observation are very small(no more than 1°C).Hence,the corresponding SST perturbations do not have much impact on the activities of photosynthetic enzymes and the phytoplankton biomass changes little.In contrast,the differences of MLD are much larger because of inconsistent definitions and imperfect observations.The large uncertainty of MLD greatly affects the light intensity of the ecosystem as well as turbulent mixing.This further results in profound changes of phytoplankton biomass.In most oceans and lakes,the maximum of phytoplankton biomass often occurs in the subsurface rather than the surface of water body,which is called deep chlorophyll maximum(DCM).Phytoplankton in the subsurface contributes most of the primary productivity.Hence,the vertical distribution and formation mechanism of phytoplankton have attracted more attention.By using a one-dimensional NP model,the uncertainty of vertical turbulent diffusion coefficient on DCM simulation was studied.Increasing(decreasing)the vertical turbulent diffusion coefficient leads to a shallower and stronger(deeper and weaker)DCM.A sufficiently small vertical turbulent diffusion coefficient can motivate the oscillation of DCM.Particularly,the uncertainty of vertical turbulence diffusion in 50-100 m leads to relatively larger DCM change.Therefore,in the simulation of DCM,we should pay more attention to the uncertainty of vertical turbulent diffusion in subsurface layer.Furthermore,the effects of uncertainties related to single parameter and parameter combination(including four parameters)on DCM simulation are studied.The purpose is to compare the importance of vertical turbulent diffusion coefficient with other parameters and to explore the nonlinear interaction among them.In the analysis of single parameter,the vertical turbulence diffusion coefficient is the second important when perturbation constraint is small.It becomes the fourth important as perturbation constraint increases.In the analysis of parameter combination,the vertical turbulent diffusion coefficient is included in the optimal parameter combination that causes the largest DCM change.The nonlinear interactions between the vertical turbulent diffusion coefficient and the other three parameters promote the change of DCM.Sensitivity experiments show the numerical simulation skills of DCM can be improved more effectively(up to 80%)by reducing the error of optimal parameter combination.However,the improvement of DCM numerical simulation is not obvious(< 5%)when reducing the error of the remaining six parameters.Hence,the parameters identified through the CNOP-P approach are important for improving the DCM simulation.This paper explored the impacts of the parameter uncertainties on phytoplankton biomass,and emphasized the nonlinear interactions among key physical processes and biochemical processes.The study deepens the understanding of the dynamical relationship between important physical processes and ecological phenomena in the marine ecosystem,and provides theoretical guidance on the identification of key parameters and further optimized observation design of parameters.