Inversion Study And Applications Of Parameters In A Dynamical Marine Ecosystem Model

Nonlinearity and hierarchy are two important characteristics to the marine ecosystem. Nonlinearity causes the uncertain ecological parameters in the marine ecosystem model and the real marine world can be not simulated accurately. Identification of ecological parameters is an important issue but has not been solved until now. Therefore, it is important to find an effective method to identify these parameters, and in this study the adjoint method is employed to invert the spatio-temporal ecological parameters.In our study, a Nitrogen-limited NPZD marine ecosystem model is constructed and the corresponding ecological parameters are inverted by the adjoint method by assimilating the surface chlorophyll a data. Five regions in the whole global field are selected as our main study areas. These regions are away from the land and close to temperate zone, in which ecological parameters are not influenced by the land condition. In the equatorial and bipolar regions, iron-limited condition and seasonal variation of ecological parameters are not obvious, while in our selected regions the above problems can be avoid. After assimilation, both the cost functions and the MAEs of observation data and model result are significantly decreased, which demonstrate that the adjoint method is an effective approach in the NPZD ecosystem model and in the five regions the ecosystem model has strong applications. Model results that ecological parameters are spatio-temporal are more accurate than those are spatial and temporal, indicating that the fact that spatio-temporal of parameters is an important trend in the future. Five spatio-temporal ecological parameters can be divided into two classes:Gm and Dp that have consistent variations can be treated as one group, and Vm, Dz, and e can be seen as the other group because of uniform changes. Parameters in these two groups vary oppositely. These trends are in accidence with the internal regulation and feedback mechanisms, and they also indicate that the NPZD model is accurate in some extent. Parameters in the same attitudinal regions have similar variations while they have the opposite various in the regions symmetric to equatorial, we can say that the ecological parameters distribute according to attitude. This distribution is obvious because temperature, light, and chlorophyll are distributed by attitude.Then the adjoint method is used to assimilate the pollutant transport in the Bohai Sea. The NPZD ecosystem model is modified only including the physical term. By this simplification, the inversion of first condition and source and sink term can be not affected by the ecological parameters to a great extent. Based on the Penglai 19-3 oil spill accident of the Bohai Sea, we assume that the center part of the Bohai Sea has sudden pollution accident. The accident can occur instantaneously and also last a long period. Therefore, in order to acknowledge the influence of pollution accident, it is necessary to identify the above pollution cases. Two classes of experiments are carried out to investigate the Bohai pollution status, and they are the inversion of the first field and the inversion of point source, respectively. Twin experiments demonstrate that the prescribed first field and the point source are inverted successfully by the adjoint method, and the cost function and the MAE of the prescribed distribution and inversion results are decreased significantly. In the estimation of first field, three optimization algorithms are simultaneously used to evaluate the inversion result, and it is found that though the classical optimization algorithms are well convergence and have better applications in other cases, they are not practical and feasible in our model. While some simple algorithms such as the gradient descent method in this study, easy to implement and operate, should be emphasized in the future practical cases. In the point source inversion experiment, independent point scheme has great superiority and can improve the inversion result and model precision.Finally, we aims to assess the potential and applications of the adjoint method to estimate initial value and pollution sources by using routine monitoring data in the Bohai Sea. In the twin experiences, the adjoint method has a strong ability of inverting the prescribed nonpoint source term. In order to compare performances of initial value and source term optimization in the practical experiments, three experiments are carried out, with initial value, source term and joint initial value and source term optimization. Experimental results experimental results demonstrate that simulation precision with only source term optimization is much higher precision than that with only initial condition optimization and extremely close to that with joint initial value and source term rate optimization, which tells us that source term optimization is more essential in the simulation of pollution than initial value and initial value obtained by interpolating the routine monitoring data is somewhat reasonable for our study in this model. Finally, by inverting the nonpoint source term and initial condition, the pollutant distribution are somewhat accurately obtained.