南海北部海洋生态模型的参数分析及遗传算法优化

海洋生态系统动力学模型是研究海洋生态环境的重要手段。随着模型的发展, 生态参数取值不确定性增加, 对模型结果的影响逐渐增大, 因此模型参数优化显得尤为重要。本研究在南海北部应用一维物理-生态耦合模型, 通过对模型生态参数进行敏感性分析, 获取关键生态参数, 利用遗传算法对参数进行优化。结果表明, 模型中的敏感参数主要集中于浮游植物生长和浮游动物生长、摄食和死亡以及碎屑沉降等过程。针对以上参数利用遗传算法优化, 发现仅加入表层卫星数据, 模型表层和垂向模拟误差分别降低27.80%和21.40%; 加入垂向观测数据, 表层和垂向模拟误差分别降低14.90%和32.70%。遗传算法应用于海洋生态模型的关键参数优化研究, 所获取的参数对模型有明显的改善效果, 提高了耦合模型对生态系统的模拟精度, 为参数优化在三维模型中的应用提供了依据。

Parameter analysis and optimization using genetic algorithm in a marine ecosystem model of the northern South China Sea

Marine ecosystem dynamics model is an important means to study marine ecological environment. As the model complexity increases, the number and uncertainty of biological parameters increase, which has a great impact on model results; therefore, optimization of model parameters is particularly important. In this paper, a one-dimensional physical-biological model is applied to the northern South China Sea, and the key biological parameters obtained through sensitivity analysis are optimized by using genetic algorithm. The results show that the sensitive parameters in the model are related to phytoplankton growth, zooplankton growth, feeding and death, and detritus sinking. Based on the genetic algorithm optimization of the above-mentioned parameters, we find that the surface and vertical simulation errors of the model are reduced by 27.80% and 21.40%, respectively, by using only surface satellite data; the surface and vertical simulation errors are reduced by 14.90% and 32.70%, respectively, by adding observed profile data. The success of applying genetic algorithm in the one-dimensional model provides the basis for its further application in three-dimensional marine ecosystem models.