Machine Learning Methods for Data Assimilation


Data assimilation is a vital step in numerical modeling, particularly in the atmospheric sciences and oceanography. It allows for problems with uneven spatial and temporal data distribution and redundancy to be addressed such that models can ingest information. Conventional methods for assimilation include Kalman filters and variational approaches. They have increased in sophistication to better fit their application requirements and circumvent their implementation issues. Nevertheless, these approaches are incapable of overcoming fully their unrealistic assumptions, particularly linearity, normality, Markovian processes, knowledge of underlying mathematical models and zero error covariances. This paper introduces a family of learning algorithms inspired by support vector machines capable of assisting or replacing the aforementioned traditional methods in assimilating data and making forecasts, without the assumptions of the conventional methods. The application of these algorithms to the processing of the states of a Lorenz 96 model show improvements in speed, efficiency and accuracy in recovering unperturbed state trajectories.

  • Abstract
  • Introduction
  • Kernel-Based Regression Procedure for the Interpolation of State Trajectories
  • Smoothing State Trajectories and Predicting Their Evolution
  • Numerical Applications
  • Conclusions
  • Acknowledgments
  • References

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In