Forecasting%20and%20Uncertainties

1
Forecasting and Uncertainties
GLOBEC Program DiLorenzo Bond Ballerini Brodeur C
ollie Hastings Kimmel Ribic Strub Wiebe
2
What we learned from GLOBEC
Improvement of physical/biological dynamical
model (e.g. ROMS, FVCOM, NPZD, IBM) Trained a
generation of multi-disciplinary (e.g. from
observationalist to modelers, from biologist to
physical scientist) Appreciation of the
importance of forecast
3
What processes can we model? using Dynamical and
Statistical models
1) Processes that we understand and model that
can lead to forecast. 2) How to propagate
uncertainties in current and future states of the
physical/biological system, both observed and
modeled. 3) Still limitations due to lack of
observations to assemble statistics.
4
Dynamical and Ecosystem Regional Models
upwelling winds boundary layer dynamics (e.g.
fog) heat/fresh water fluxes
COAMPS, RSM
Satellite products (winds, SST, SSH, CHL-a)
upwelling transport dynamicschanges in property
distribution vertical distribution and mix
layer stratification
ROMS 3D circulation model
FVCOM
tidal and estuarine environment surface currents
and transport baroclinic eddy circulation
CHL-a distribution Nutrient distributionsZooplank
ton Parameters uncertainty
NPZD
5
What is the role of the dynamical models in
forecasting?
Large-scale variability ENSO, NAO, SAM, PDO,
NPGO, etc.
dynamical model 20
regression model 80
forecast the delayed ecosystem response
nowcast of unobservable states
6
What is the role of the dynamical models in
forecasting?
A possible approach
Dynamical model can be used to compile
statistics and constrain the processes that we
understand
Statistical characterizations of things we cannot
model
Bayesian/Hybrid Modeling Frameworks
7
Need for specific examples of forecasting
Outcomes we learn from trying relative merit
of different approaches synthesis activity in
that it forces us to define what we really
understand and model
8
Recommendation for present projects Obligation
to assess uncertainties in models Sources of
error and measures of skill Sources of
uncertainties and relative importance Summary of
modeling applications
9
Future recommendations Pilot forecasting
experiments with interdisciplinary team. Real
time basin-wide physical-biological model in
assimilative mode to give a first order estimate
of the states. Continue the development of
low-dimensional or simple models. GLOBEC
involved in IPCC assessments
10
END