Sequential Importance Resampling Filtering in Ecosystem Modelling


Contact
sloza [ at ] awi-bremerhaven.de

Abstract

Sequential Important Resampling filter (SIRF) isapplied for assimilating time-series data into an ecosystem model.Advantage of this Monte-Carlo based data assimilation approachfor combined state and parameter estimation in ecosystem modellinghas been already demonstrated in previous studies (Losa et al., 2003).Some aspects of the SIRF implementation for the highly non-linear system,however, still remain to be worked out. The filter is known to suffer fromdegeneration of the ensemble if either the system noise does notprovide sufficient spreading of states which are resampled severaltimes or the ensemble badly approximates the true prior distribution(the distance between the best member and the true state is too big).This problem is even more pronounced in the case of simultaneousstate-parameter estimation where regeneratingthe number of samples in the parameter space is needed. In this study, we arefocusing on the model noise optimization. Investigating the system noisewould, probably, allow us to explain the notable seasonality obtainedfor some of the optimized parameters in our previous study (Losa et al., 2003).



Item Type
Conference (Poster)
Authors
Divisions
Programs
Peer revision
Not peer-reviewed
Publication Status
Published
Event Details
European Geosciences Union General Assembly, ViennaApril 2005..
Eprint ID
12744
Cite as
Loza, S. , Kivman, G. , Schröter, J. and Wenzel, M. (2005): Sequential Importance Resampling Filtering in Ecosystem Modelling , European Geosciences Union General Assembly, ViennaApril 2005. .


Download
[img]
Preview
PDF (Fulltext)
Los2005e.pdf

Download (186kB) | Preview
Cite this document as:

Share

Research Platforms
N/A

Campaigns


Actions
Edit Item Edit Item