Imperial College London
Alternate

Dr Nataliya Le Vine (née Bulygina)

Faculty of Engineering, Department of Civil and Environmental Engineering

Honorary Research Fellow
 
 
 
//

Contact

 

+44 (0)20 7594 6019n.le-vine CV

 
 
//

Assistant

 

Miss Judith Barritt +44 (0)20 7594 5967

 
//

Location

 

405Skempton BuildingSouth Kensington Campus

//

Summary

 

Publications

Citation

BibTex format

@article{Bulygina:2010:10.1007/s00477-010-0387-y,
author = {Bulygina, N and Gupta, H},
doi = {10.1007/s00477-010-0387-y},
journal = {SERRA},
title = {How Bayesian Data Assimilation can be used to estimate mathematical structure of a model},
url = {http://dx.doi.org/10.1007/s00477-010-0387-y},
volume = {477},
year = {2010}
}
Download

RIS format (EndNote, RefMan)

TY  - JOUR
AB  - In previous work, we presented a method forestimation and correction of non-linear mathematical modelstructures, within a Bayesian framework, by merginguncertain knowledge about process physics with uncertainand incomplete observations of dynamical input-stateoutputbehavior. The resulting uncertainty in the modelinput-state-output mapping is expressed as a weightedcombination of an uncertain conceptual model prior and adata-derived probability density function, with weightsdepending on the conditional data density. Our algorithm isbased on the use of iterative data assimilation to update aconceptual model prior using observed system data, andthereby construct a posterior estimate of the model structure(the mathematical form of the equation itself, not just itsparameters) that is consistent with both physically basedprior knowledge and with the information in the data. Animportant aspect of the approach is that it facilitates a cleardifferentiation between the influences of different typesof uncertainties (initial condition, input, and mappingstructure) on the model prediction. Further, if some priorassumptions regarding the structural (mathematical) formsof the model equations exist, the procedure can help revealerrors in those forms and how they should be corrected. Thispaper examines the properties of the approach by investigatingtwo case studies in considerable detail. The resultsshow how, and to what degree, the structure of a dynamicalhydrological model can be estimated without little or no priorknowledge (or under conditions of incorrect prior information)regarding the functional forms of the storage–streamflowand storage–evapotranspiration relationships. Theimportance and implications of careful specification of themodel prior are illustrated and discussed.
AU  - Bulygina,N
AU  - Gupta,H
DO  - 10.1007/s00477-010-0387-y
PY  - 2010///
TI  - How Bayesian Data Assimilation can be used to estimate mathematical structure of a model
T2  - SERRA
UR  - http://dx.doi.org/10.1007/s00477-010-0387-y
VL  - 477
ER  -
Download