Multiscale Representation of Observation Error Statistics in Data Assimilation

Accounting for realistic observation errors is a known bottleneck in data assimilation, because dealing with error correlations is complex. Following a previous study on this subject, we propose to use multiscale modelling, more precisely wavelet transform, to address this question. This study aims to investigate the problem further by addressing two issues arising in real-life data assimilation: how to deal with partially missing data (e.g., concealed by an obstacle between the sensor and the observed system), and how to solve convergence issues associated with complex observation error covariance matrices? Two adjustments relying on wavelets modelling are proposed to deal with those, and offer significant improvements. The first one consists of adjusting the variance coefficients in the frequency domain to account for masked information. The second one consists of a gradual assimilation of frequencies. Both of these fully rely on the multiscale properties associated with wavelet covariance modelling. Numerical results on twin experiments show that multiscale modelling is a promising tool to account for correlations in observation errors in realistic applications.

[1]  Marc Bocquet,et al.  Data Assimilation: Methods, Algorithms, and Applications , 2016 .

[2]  Frédéric Chevallier,et al.  Impact of correlated observation errors on inverted CO2 surface fluxes from OCO measurements , 2007 .

[3]  S. Dance,et al.  A pragmatic strategy for implementing spatially correlated observation errors in an operational system: An application to Doppler radial winds , 2019, Quarterly Journal of the Royal Meteorological Society.

[4]  Sarah L. Dance,et al.  The conditioning of least‐squares problems in variational data assimilation , 2018, Numer. Linear Algebra Appl..

[5]  Jean-Noël Thépaut,et al.  The Spatial Structure of Observation Errors in Atmospheric Motion Vectors from Geostationary Satellite Data , 2003 .

[6]  Olivier Talagrand,et al.  On extending the limits of variational assimilation in nonlinear chaotic systems , 1996 .

[7]  Alison M. Fowler,et al.  Data compression in the presence of observational error correlations , 2019, Tellus A: Dynamic Meteorology and Oceanography.

[8]  Philippe Courtier,et al.  Unified Notation for Data Assimilation : Operational, Sequential and Variational , 1997 .

[9]  Nancy Nichols,et al.  Data assimilation with correlated observation errors: experiments with a 1-D shallow water model , 2013 .

[10]  N. Papadakis,et al.  Accounting for observation errors in image data assimilation , 2015 .

[11]  P. Courtier,et al.  Correlation modelling on the sphere using a generalized diffusion equation , 2001 .

[12]  Nancy Nichols,et al.  Correlated observation errors in data assimilation , 2008 .

[13]  S. Gratton,et al.  Modelling spatially correlated observation errors in variational data assimilation using a diffusion operator on an unstructured mesh , 2019, Quarterly Journal of the Royal Meteorological Society.

[14]  Craig H. Bishop,et al.  The benefits of correlated observation errors for small scales , 2015 .

[15]  L. Berre,et al.  Filtering of Background Error Variances and Correlations by Local Spatial Averaging: A Review , 2010 .

[16]  Jemima M. Tabeart,et al.  The impact of using reconditioned correlated observation‐error covariance matrices in the Met Office 1D‐Var system , 2019, Quarterly Journal of the Royal Meteorological Society.

[17]  Francisco Tirado,et al.  Parallel wavelet transform for large scale image processing , 2002, Proceedings 16th International Parallel and Distributed Processing Symposium.