DATA ASSIMILATION FOR A COASTAL AREA MORPHODYNAMIC MODEL

Coastal area morphodynamic models are an important tool for coastal management. Operated over timescales of months to decades, they can provide information on how the morphology of a coastline should evolve in response to natural or man-made forcing. These models often perform rather poorly in detail, partly because the physical processes that drive morphological change occur on much shorter timescales than the changes themselves. State-of-the-art models are becoming increasingly sophisticated in their attempts to model sediment transport accurately. However, an alternative or complementary solution may be to integrate a model with observations of the morphology, using data assimilation techniques. Data assimilation is routinely used to keep atmospheric and oceanographic models on track, but has rarely been used in coastal area morphodynamic modelling. This is despite the fact that a variety of methods for sampling a subset of the model bathymetry in space and time exist. These include ground survey along beach transects, ship-based echo sounding, satellite and airborne measurement of waterlines (instantaneous land-sea boundaries), airborne LiDAR survey of the inter-tidal zone, and bathymetric measurement using shore-based radars. These observations would generally not provide full coverage of the model domain such as provided by a ship or bathymetric LiDAR survey, but would be less expensive.

The talk will describe the use of data assimilation in morphodynamic modelling using Morecambe Bay in north-west England as a study site. A simple model of the Bay has been enhanced with a data assimilation scheme to predict large-scale changes in bathymetry observed in the Bay over a 3-year period. The 2DH decoupled morphodynamic model developed for the work will be described, together with the optimal interpolation scheme used to assimilate waterline observations into the model run. Each waterline was acquired from a SAR satellite image and essentially provides a quasi-contour of the bathymetry at some level within the inter-tidal zone. The use of data assimilation is shown to successfully compensate for a particular failing of the model, and to help keep the model bathymetry on track. It also improves the ability of the model to predict future bathymetry.