1. Progress on work package 1
Changing Coastlines: data assimilation for morphodynamic prediction and predictability
Progress on work package 1
University of Reading: Tania Scott, Sarah Dance, Mike Baines, Amos Lawless, David Mason, Nancy Nichols, Polly Smith and Peter Sweby
POL: Roger Proctor

2. Will it work?
Data assimilation for a coastal area morphodynamic model: Morecambe Bay Tania Scott & David Mason Objective: to improve a medium-term morphodynamic model of Morecambe Bay by assimilating waterline observations of bathymetry into the model run

3. The method We developed a morphodynamic model to
predict changes in Morecambe Bay bathymetry caused by tide induced sediment transport
This model was very simplistic compared to the best engineering models
because we wanted to investigate the benefits of using data assimilation in the model
rather than develop a state-of-the-art model in itself
We ran the model for 3 years
with and without data assimilation

4. The data - waterlines An image of the land-sea boundary
Water elevation
( bathymetry)
An image of the land-sea boundary
(satellite SAR image © ESA)
Water elevation predicted along the boundary by a hydrodynamic model with surge component

5. The data assimilation For the data assimilation we used
13 waterlines acquired at intervals over the 3-year period
Each waterline was assimilated at its validity time
using a method called optimal interpolation
More weight was given to the observations than to the model background
Information from each waterline was spread to a distance of approximately three model grid cells

6. Examples of assimilated waterlines
Difference between initial and observed bathymetries
cool colours – model bathymetry needs to be lower
green – model bathymetry is good
warm colours – model bathymetry needs to be higher

7. Results for our 3-year model run
Observed change in bathymetry
Modelled change in bathymetry without data assimilation
Modelled change in bathymetry with data assimilation
Qualitative assessment: data assimilation successfully introduced erosion along the Ulverston channel
Quantitative assessment: data assimilation improved both the current bathymetry and the predictive performance of the model

8. Conclusions Which led to the Changing Coastlines project!
We showed that a morphodynamic model of Morecambe Bay could be improved by using data assimilation
Results suggested that data assimilation could benefit morphodynamic models by
improving the current model bathymetry
helping to better predict future bathymetry
Thus we should continue to develop this technique
Which led to the Changing Coastlines project!

9. Work package 1 Develop morphodynamic models for each study site
Develop data assimilation schemes for each model
Acquire and process data for
generation of initial bathymetries
data assimilation
model validation
Evaluate model performance
predictability and uncertainty
with and without data assimilation
Develop an aggregate morphological model

10. The models The Morecambe Bay model The Dee estuary model
will be based on our old Morecambe Bay model
The Dee estuary model
will be similar to the Morecambe Bay model
they are both enclosed tide dominated environments
The east Lincolnshire coast model
will be different from the others
it is an open coastline environment
different physical processes may dominate
e.g. waves are likely to be more important

11. The Morecambe Bay model
We have been considering improvements to our old Morecambe Bay model, in consultation with
the Proudman Oceanographic Laboratory
HR Wallingford
The consensus is that the old model works OK
so don’t break it by making it too complicated
However we should
increase the resolution (what impact does this have?)
include a critical current for sediment transport
possibly include waves (by possibly using SWAN)

12. Computing platforms
We have ported the old Morecambe Bay model to newer platforms:
My Linux workstation
the model now runs much faster than before (approximately 12 hours per model year)
which is good for development and testing
The University of Reading Condor pool
(with assistance from the Reading e-science Centre)
on which we can run large numbers of models simultaneously
which will be good for generating data assimilation statistics

13. The data We need lots of observations of bathymetry for
generation of initial model bathymetries
data assimilation
model validation
Possible data sources are
airborne LiDAR
ship-based swath bathymetry
waterlines
ground surveys
shore-based X-band radar

14. The Dee estuary Each LiDAR survey covers most of the estuary
Date
Data type
Source
April 2003
LiDAR
Environment Agency
July 2003
Swath bathymetry of main channel
February 2004
May 2004
October 2006
Apr 2003 – Aug 2006
30 SAR images
European Space Agency ?
X-band radar of estuary mouth
Proudman Oceanographic Laboratory
Each LiDAR survey covers most of the estuary
We will run the Dee model from April 2003 to October 2006
the April 2003 LiDAR data will be used for initial model bathymetry
the October 2006 LiDAR data will be used for model validation
all other data will be used for both model validation and data assimilation

15. LiDAR images of the Dee estuary
April 2003
October 2006

16. Observed change in bathymetry, April 2003 to October 2006
Dee estuary
Observed change in bathymetry, April 2003 to October 2006

17. SAR image - georeferencing
Get a root mean square location error of about 25m (2 SAR image pixels)
Digital Ordnance Survey map from EDINA Digimap
Look along edge of estuary for
road/rail/ditch junctions
field corners

18. SAR image - waterline extraction
Automatic waterline extraction software previously developed by David Mason
The extracted waterline needs correcting for
false waterlines
- missing waterlines
The corrected waterline then needs to be assigned height values, by running a hydrodynamic model with surge component

19. Morecambe Bay Each LiDAR survey covers only a small amount of the Bay
Date
Data type
Source
December 2003
LiDAR
Environment Agency
November 2004
March 2005
November 2005
February - May 2006
2003 -
n SAR images
European Space Agency
Each LiDAR survey covers only a small amount of the Bay
mainly along the shore
but some over the channels
We will start the Morecambe Bay model run sometime in 2003
the initial model bathymetry will be generated from waterlines

20. LiDAR image of the Ulverston channel, Morecambe Bay, November 2004

21. Next steps We need to get the models up and running for
the Dee estuary
Morecambe Bay
To do this we need to
acquire SAR images for Morecambe Bay
generate initial bathymetries (which includes lots of SAR processing for Morecambe Bay)
determine tidal forcing parameters at the open boundaries
implement a critical current for sediment transport
possibly include waves
Then we can start work on the data assimilation

22. Morecambe Bay at sunset
Scott T.R. and Mason D.C., 2007.
Data assimilation for a coastal area morphodynamic model: Morecambe Bay. Coastal Engineering, 54(2),