海洋生态系统模型中参数空间变化的伴随法反演研究  

Numerical Study on Spatially Varying Parameters of a Marine Ecosystem Model with the Adjoint Method

作　　者：吕咸青[1] 樊伟[1,2] 

机构地区：[1]中国海洋大学物理海洋实验室,山东青岛266100 [2]中国科学院海洋研究所,山东青岛266071

出　　处：《中国海洋大学学报（自然科学版）》2009年第5期846-854,共9页Periodical of Ocean University of China

基　　金：国家高技术研究发展计划项目(2007AA09Z118);国家重点基础研究发展计划项目(2005CB422308);国家基金委"创新研究群体科学基金"(40821004)资助

摘　　要：以气候耦合模式FOAM(Fast Ocean-Atmosphere Model)的气候态为背景场,基于参数的空间分布,定量地研究通过伴随同化方法反演空间变化的参数的能力,探讨影响反演结果的主要因素。通过孪生数值实验同化浮游植物资料,发现在单独反演1个空间变化的参数时,两种给定的参数空间变化都可以被很好地反演出来,参数的平均相对误差在4%以内,说明通过伴随同化方法可以有效地反演出空间变化的参数;而当同时反演作为控制变量的五个参数时,各参数变化趋势的搭配对反演结果影响很大,只有当这种搭配与模型中影响浮游植物生物量变化的生态机制一致时,5个参数才能较准确地得到反演。同化含有10%随机误差的观测数据对反演结果的影响不大,进一步说明了利用伴随同化方法反演参数空间变化的有效性。Based on the spatial parameterizations, we studied the estimation capability on spatially varying pa- rameters quantitatively by using the adjoint method in a climatological physical environment which provided by a global climate model FOAM（Fast Ocean-Atmosphere Model）, and figured out the factors which influenced the inversion results. According to the twin experiments that the observations of phytoplankton were assimilated, when estimating each parameter separately, the two given types of its spatial variations could be reproduced, and the mean relative error was less than 4 %, which indicated that spatially variable parameters were sufficiently estimated. However, when estimating all the five parameters which were served as control variables simultaneously, the combination of the changing trend of each parameter influenced the estimation results significantly. Only when this combination coincided with the ecological mechanisms which influenced the growth of the phytoplankton in the model, could the five parameters be estimated. Assimilating observations with random errors of 10 % had less influence on the inversion results of the parameters, which further validated the efficiency of estimating spatially varying parameters by the adjoint method.

关 键 词：海洋生态系统模型 伴随同化方法 空间变化的参数 

分 类 号：Q178.53[生物学—水生生物学]