UACE2017 Proceedings: A new method of data assimilation based on ensemble Kalman filter with application to coastal acoustic tomography data

• Session:
  Ocean Acoustic Tomography - Various application and data
• Paper:
  A new method of data assimilation based on ensemble Kalman filter with application to coastal acoustic tomography data
• Author(s):
  Minmo Chen, Arata Kaneko, Ju Lin, Chuanzheng Zhang
• Abstract:
  A new method of data assimilation (DA) based on ensemble Kalman filter with much less consuming time is proposed here. In this method, model error covariance is determined from perturbed model state vectors at each DA time without forecasting the ensemble model state vectors. A smooth correlation function is introduced in the Kalman gain to localize the tomography domain. The model state vectors are perturbed by a N-ensemble of smooth pseudorandom fields. Sub-vector in each model grid has zero mean and time invariant variance. Furthermore, the covariance is related to the decorrelation length. Thus, computational time is much reduced in comparison with the conventional method in which model error covariance is calculated through an ensemble of the model growth at every time step. The new method was successfully applied to assimilate the 2013 Hiroshima Bay coastal acoustic tomography (CAT) data into a baroclinic ocean model. A coastal upwelling, generated along the northern shore of Hiroshima Bay by a northerly wind from a typhoon, was well structured in a two-layer system of current and salinity. Data assimilation results were validated with the path-average CAT and CTD data and the estimated errors were smaller than the variation ranges of current and salinity, associated with the upwelling.
•   Download the full paper

Contact details

• Contact person:
  Dr Minmo Chen
• e-mail:
  
• Affiliation:
  Graduate School of Engineering, Hiroshima University
• Country:
  Japan