new functionalities in openda: an overvie functionalities in openda: an overview martin verlaan stef...
TRANSCRIPT
New functionalities in OpenDA:an overview
Martin VerlaanStef Hummel, Julius Sumihar, Albrecht Weerts,
Ghada El Serafy, Herman Gerritsen, ...Deltares & TU Delft
www.openda.org
Data-assimilation example
Use observations for a betterstart for forecasting
Stormsurge error estimate
data-assimilation
• What is OpenDA?• An example of calibration for Delft3D-flow• Kalman filtering• Other developments
Outline
OpenDA
What is OpenDA?• A generic toolbox for data-assimilation
• set of interfaces that define interactions between components
• library of data-assimilation algorithms• Open source
Why OpenDA?• More efficient than development
for each application• Avoid spaghetti code• Shared knowledge between applications• Development of algorithms with eg universities• Easier to test, which should result in fewer bugs
www.openda.org
-Download-Documentation-Association
Additionalwrapper moduleneeded foreach model
OpenDA application progam
- Linux, windows and mac versions- One application for all algorithms, models, observations- With and without GUI
Object-oriënted approach
• Clear interfaces support clean code• Building blocks implemented once • Better testing, fewer bugs• Easier to try different combinations, e.g. new algrithms
Data assimilation scope
• Calibration of parameters• Real time forecasting updates
• Kalman filtering
• Variational algorithms• Uncertainty analysis• Study potential value of new observations• Estimating sources
A model wrapper is needed for each model
1. Blackbox (uses files and model-executable)2. Native toolbox (eg link to existing fortran code)3. Simple java models for testing and research4. Direct implementation of java interface
(maximum flexibility)
- Specification of uncertainty = control variables
Model wrappers
OpenDA black-box wrapper
• Uses input and output files of the model
• No source code of the model is needed
• Provides support for specification of uncertainty
Delft3d calibration
Friction/roughness, depth
Observations
Tides at open boundaries
Control parameters
Tides before and after calibration
Observed Initial After calibration
K1
Kalman filter for Delft3D-flow
Observations
Under development: first applications ongoing
Improve initial state
Average accuracy of storm surge forecasts
lead-time
Impact of data-assimilation
Ongoing developments
• Parallel computing• Observation sensitivity and design of monitoring networks• Coupling to FEWS• Various model wrappers: SWAN, Delft-3D flow and waq, Sobek...• Introduction of Delta-shell• New public release of OpenDA
Parallel Computing
• Model timestepping often most time consuming part of algorithm (90-99%)
• Often natural parallelism in the model timestepping• Parallel time stepping for all native OpenDA models (OO)
01-11-11 16
Model Calibration Kalman filtering
Delft3D-flow
Delft3D-waq
Sobek
SWAN
WAQUA
HBV
WANDA
Modflow
PCRaster
MSette
/
EDFC, HSPF, MCRM, Lotos-euros, Chimere, ...
Two-sample approach
FEWS-OpenDA connector
Questions?