Variational data assimilation using targetted random walks
The variational approach to data assimilation is a widely used methodology for both online prediction and for reanalysis (offline hindcasting). In either of these scenarios it can be important to assess uncertainties in the assimilated state. Ideally it would be desirable to have complete informatio...
| Main Authors: | , , , |
|---|---|
| Format: | Journal article |
| Published: |
2010
|
| _version_ | 1797057601511358464 |
|---|---|
| author | Cotter, S Dashti, M Robinson, J Stuart, A |
| author_facet | Cotter, S Dashti, M Robinson, J Stuart, A |
| author_sort | Cotter, S |
| collection | OXFORD |
| description | The variational approach to data assimilation is a widely used methodology for both online prediction and for reanalysis (offline hindcasting). In either of these scenarios it can be important to assess uncertainties in the assimilated state. Ideally it would be desirable to have complete information concerning the Bayesian posterior distribution for unknown state, given data. The purpose of this paper is to show that complete computational probing of this posterior distribution is now within reach in the offline situation. In this paper we will introduce an MCMC method which enables us to directly sample from the Bayesian posterior distribution on the unknown functions of interest, given observations. Since we are aware that these methods are currently too computationally expensive to consider using in an online filtering scenario, we frame this in the context of offline reanalysis. Using a simple random walk-type MCMC method, we are able to characterize the posterior distribution using only evaluations of the forward model of the problem, and of the model and data mismatch. No adjoint model is required for the method we use; however more sophisticated MCMC methods are available which do exploit derivative information. For simplicity of exposition we consider the problem of assimilating data, either Eulerian or Lagrangian, into a low Reynolds number (Stokes flow) scenario in a two dimensional periodic geometry. We will show that in many cases it is possible to recover the initial condition and model error (which we describe as unknown forcing to the model) from data, and that with increasing amounts of informative data, the uncertainty in our estimations reduces. |
| first_indexed | 2024-03-06T19:38:43Z |
| format | Journal article |
| id | oxford-uuid:1fea8385-fe53-456a-bf13-cea7e6ba992f |
| institution | University of Oxford |
| last_indexed | 2024-03-06T19:38:43Z |
| publishDate | 2010 |
| record_format | dspace |
| spelling | oxford-uuid:1fea8385-fe53-456a-bf13-cea7e6ba992f2022-03-26T11:24:41ZVariational data assimilation using targetted random walksJournal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:1fea8385-fe53-456a-bf13-cea7e6ba992fMathematical Institute - ePrints2010Cotter, SDashti, MRobinson, JStuart, AThe variational approach to data assimilation is a widely used methodology for both online prediction and for reanalysis (offline hindcasting). In either of these scenarios it can be important to assess uncertainties in the assimilated state. Ideally it would be desirable to have complete information concerning the Bayesian posterior distribution for unknown state, given data. The purpose of this paper is to show that complete computational probing of this posterior distribution is now within reach in the offline situation. In this paper we will introduce an MCMC method which enables us to directly sample from the Bayesian posterior distribution on the unknown functions of interest, given observations. Since we are aware that these methods are currently too computationally expensive to consider using in an online filtering scenario, we frame this in the context of offline reanalysis. Using a simple random walk-type MCMC method, we are able to characterize the posterior distribution using only evaluations of the forward model of the problem, and of the model and data mismatch. No adjoint model is required for the method we use; however more sophisticated MCMC methods are available which do exploit derivative information. For simplicity of exposition we consider the problem of assimilating data, either Eulerian or Lagrangian, into a low Reynolds number (Stokes flow) scenario in a two dimensional periodic geometry. We will show that in many cases it is possible to recover the initial condition and model error (which we describe as unknown forcing to the model) from data, and that with increasing amounts of informative data, the uncertainty in our estimations reduces. |
| spellingShingle | Cotter, S Dashti, M Robinson, J Stuart, A Variational data assimilation using targetted random walks |
| title | Variational data assimilation using targetted random walks |
| title_full | Variational data assimilation using targetted random walks |
| title_fullStr | Variational data assimilation using targetted random walks |
| title_full_unstemmed | Variational data assimilation using targetted random walks |
| title_short | Variational data assimilation using targetted random walks |
| title_sort | variational data assimilation using targetted random walks |
| work_keys_str_mv | AT cotters variationaldataassimilationusingtargettedrandomwalks AT dashtim variationaldataassimilationusingtargettedrandomwalks AT robinsonj variationaldataassimilationusingtargettedrandomwalks AT stuarta variationaldataassimilationusingtargettedrandomwalks |