AUTHOR=Nian Rui , Geng Xue , Zhang Zhengguang , Yuan Minghan , Fu Zhen , Xu Hengfu , Yang Hua , Lai Qi , He Hui , Su Chi Wei , Zang Lina , Yuan Qiang , He Bo 

TITLE=Toward Exploring Topographic Effects on Evolution and Propagation of Ocean Mesoscale Eddies Through Life Cycle Across Izu-Ogasawara Ridge in Northwestern Pacific Ocean

JOURNAL=Frontiers in Marine Science

VOLUME=Volume 8 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/marine-science/articles/10.3389/fmars.2021.672272

DOI=10.3389/fmars.2021.672272

ISSN=2296-7745

ABSTRACT=Ocean mesoscale eddies contribute significantly to water transport on a global scale, constituting the ubiquitous, irregular, discrete components. In this paper, we propose to make an attempt to explore whether and how the topographic effects, could exert considerable influences on the evolution and propagation of mesoscale eddies through life cycle, from multiple perspectives, namely the amplitude, the rotation speed, and the radius. We quantitatively investigate the known variability of ocean mesoscale eddies, by taking the Izu-Ogasawara Ridge in Northwest Pacific Ocean as an example. 69 mesoscale eddy trajectories in total, derived by multimission satellite altimetry from 1993 to 2018, of a life cycle more than 6 months, have been systematically examined by mathematical statistics and wavelet coherence analysis. We further employed a deep learning architecture, Long Short Term Memory (LSTM), to capture the predictability of mesoscale eddy trajectories, equipped with attention mechanism at multi-stage, to identify the relevant attributes dominantly correlative with the time-varying evolution process, where the amplitude, rotation speed, radius, the latitude and longitude at the geographical location, the zonal displacement and meridional displacement, the variation of the bathymetric topography have been all taken as the input variables for training. It has been revealed in our experiment that three intrinsic attributes of mesoscale eddies, all demonstrate strong negative correlation tendency with the variation of topographic relief within 25 years over the ridge region. When we intentionally feed with the topography variation as the additional inputs, the attention LSTM behaved well in predicting the zonal displacement of mesoscale eddy trajectories, indicating its effectiveness, in determining the evolution and propagation process to be predicted. The attention LSTM achieved better prediction performances, compared to the competitive extra tree model, with MAE, RMSE 0.15, 0.22 in LSTM and 0.25, 0.38 in extra tree model, when applying multivariate prediction in a longer-term for 14-day duration. The developed scheme could integrate more evidences on how mesoscale eddies response to the topographic effects during their evolution and propagation process, and help provide opportunities to potentially identify the underlying patterns and mechanism that mesoscale eddies engage in ocean dynamics, and the related physical, chemical, biological processes.