易家成¹, 高艳秋², 张继才¹
1.浙江大学 海洋学院 物理海洋研究所, 浙江 舟山 316021;2.国家海洋局第二海洋研究所 卫星海洋环境动力学国家重点实验室, 浙江 杭州 310012

摘要:
为了得到一种有效的计算海水涡动黏性的途径，本文基于伴随同化方法，研究了Ekman模型中垂向涡动黏性系数（vertical eddy viscosity coefficient，VEVC）的时间变化。本文推导了时间变化VEVC的优化关系式，并利用理想实验对三个影响因素进行了探究，包括优化算法、初始猜测、观测深度，其主要结论是：（1）梯度下降法的优化效果优于共轭梯度法和有限记忆BFGS（limited-memory Broyden-Fletcher-Goldfarb-Shanno）法；（2）初始猜测值与实际值接近时，收敛速度更快，反演误差更小；（3）反演结果对表层和次表层的流速更为敏感。本文从百慕大试验站锚泊系统（Bermuda Testbed Mooring，BTM）的数据中提取了Ekman流速，并反演了VEVC的时间变化，实际实验结果表明：（1）对于实测数据而言，仍是梯度下降法优化效果最好；（2）将VEVC设置为时间变化型的反演策略，反演效果优于常数型和深度变化型；（3）该地区VEVC在0.01 m²/s左右。该研究为Ekman流的数值模拟提供了一种确定VEVC时间变化的有效方法，对于其他动力机制的研究具有一定的参考价值。
关键词:参数反演Ekman模型伴随同化涡动黏性
DOI：10.11759/hykx20180408002
分类号:P731.2
基金项目:国家重点研发计划“全球变化及应对”重点专项（2017YFA0604100）；国家重点研发计划“海洋环境安全保障”重点专项（2017YFC1404000）；国家自然科学基金（41876086）；中央高校基本科研业务费专项资金

Estimation of the time-varying vertical eddy viscosity coeffi-cient in an Ekman layer model through the adjoint data as-similation approach
YI Jia-cheng¹, GAO Yan-qiu², ZHANG Ji-cai¹
1.Institute of Physical Oceanography, Ocean College, Zhejiang University, Zhoushan 316021, China;2.State Key Laboratory of Satellite Ocean Environment Dynamics, Second Institute of Oceanography, Hangzhou 310012, China
Abstract:
A method for estimating the temporal distribution of the vertical eddy viscosity coefficient (VEVC) in an Ekman model is developed based on the adjoint data assimilation approach to obtain an effective approach for computing oceanic eddy viscosity. Three factors that affect the estimation results, optimization algorithm, initial guess, and number of observations, are investigated through ideal experiments. The main conclusions are that the following:(1) the gradient descent (GD) algorithm is better than the quasi-Newton conjugate gradient and limited-memory Broyden-Fletcher-Goldfarb-Shanno algorithms in optimizing the VEVC, (2) the closer the initial guess and given value are to each other, the higher the convergence rate and accuracy, and (3) the inversion results are more sensitive to the observations in the upper layers than those in the lower layers. The Ekman currents are extracted from the measurements obtained from Bermuda Testbed Mooring. Then, the VEVC is inverted in practical experiments. The main conclusions are that (1) the GD algorithm is still the best among the three algorithms in practice experiments; (2) the strategy that assumes that the VEVC is time-varying exhibits a better performance than the two other strategies that assume that the VEVC is constant or depth-varying; and (3) the VEVC of the study area is approximately 0.01 m²/s. The study extends our understanding of the Ekman layer model and provides an effective approach to determine the temporal variations of the VEVC.
Key words:parameter estimationEkman layer modeladjoint assimilationeddy viscosity





PDF全文下载地址:

点我下载PDF