|
A key focus of our research is developing ways to bring together physical
understanding and observations, to get the best from both. We call this hybrid
modelling, and we are actively pursuing a range of ideas.
The
figure opposite describes the concept of a hybrid model. The key features
are the feedback between the statistical and physical model components,
and an active link back to the physical parameters representing learning
from observations.
Our research is focused
on the use of Bayesian Hierarchical
Models to develop hybrid models, particularly for seasonal forecasting
of climate. More widely we are using hierarchical modelling as a way to
integrate multiple sources of information (sub-models, data sources …).
A key benefit of this
approach is that uncertainties in parameters and forecasts are captured
directly as probability distributions. This is highly valuable for robust
decision-making, especially as the probability distributions integrate
both physical knowledge and observations.
Useful Links:
http://www.ioci.org.au
http://www.stat.ohio-state.edu/~sses/collab_enso.php
Contact:
Dr
Eddy Campbell
|
This figure depicts a
hybrid forecasting scheme. There are historically two approaches: using
physical models (left hand side) or empirical approaches using statistical
models (right hand side). In hybrid modelling we provide for feedback
between the statistical and physical models, and use the data to learn
about the physical parameters. The resulting forecasts incorporate
information from both observations and physics.
|
|
Some references:
ADDIN ENBbu ́ Berliner, L. M. (2003). Physical-statistical
modeling in geophysics. Journal of
Geophysical Research- Atmospheres, 108
(D24).
Berliner,
L. M., Milliff, R. F. and Wikle, C. K. (2003). Bayesian hierarchical
modeling of air-sea interaction. Journal
of Geophysical Research, 108,
1-1:1-18.
Berliner,
L. M., Wikle, C. K. and Cressie, N. (2000). Long-lead prediction of
Pacific SST via Bayesian dynamic modeling. Journal
of Climate, 13, 3953-3968.
Campbell,
E. P. (2004). An
introduction to physical-statistical modelling using Bayesian methods.
CSIRO Mathematical and Information Sciences, Perth, Western Australia,
Technical Report 2004/49, 18 pages.
Royle,
J. A., Berliner, L. M., Wikle, C. K. and Milliff, R. F. (1999) In Case
Studies in Bayesian Statistics IV(Ed, Gatsonis, C.) Springer-Verlag,
New York, pp. 367-382.
Wikle,
C. K. (2003). Hierarchical models in environmental science. International
Statistical Review, 71,
181-199.
Wikle,
C. K., Berliner, L. M. and Milliff, R. F. (2003). Hierarchical Bayesian
approach to boundary value problems with stochastic boundary conditions. Monthly
Weather Review, 131,
1051-1062.
|
