Application of data assimilation in portable operational forecasting systems—the DATools assimilation environment

The first part of the paper describes a portable and flexible data assimilation environment (DATools) for easy application of data assimilation and calibration techniques to models that are used in smaller-scale engineering applications, for example to guide temporary offshore construction works, marine surveying and salvage operations. These applications are characterized by the need for detailed forecasts of often strongly non-linear marine behaviour in shallow-shelf seas under constraints of limited field measurements and absence of large-scale computing facilities. The applications are often prepared, and their results are interpreted by end-user engineers in field offices, not well acquainted with the data assimilation theory. The DATools data assimilation environment has been designed to facilitate such applications. It presently features an ensemble Kalman filter and two particle filters. These can be coupled to any process model in a standardized way through the so-called Published Interface that is used in an increasing number of flood-forecasting applications across Europe. The design of the system and its main modules are discussed, looking at the system from a non-specialist user perspective and focusing on modularity, transparency, user guidance, intuitive and flexible uncertainty prescription. The second part of the paper describes a typical example application of data assimilation in an engineering environment for which the modelling environment has been developed: the daily forecast of current and salinity profiles to guide construction works in Osaka Bay. An ensemble Kalman filter-based steady-state Kalman filter is developed for assimilation of salinity and horizontal currents into an existing three-dimensional flow model for the highly non-linear stratified shallow bay. Calibration results are summarised for both hindcast mode and forecast mode. After assessment of the results, a second, improved model setup is described, including now assimilation of temperature and water level as well as currents and salinity. The calibration results are shown to improve further. Both applications show that significant forecast improvements can be possible with data assimilation in typical engineering applications, notwithstanding operational constraints such as limited measurement data and restricted computational infrastructure. It is argued that the described portable data assimilation environment is well suited for temporary engineering applications as it provides typical end-user functionalities for data assimilation and will so allow the user to focus on the characteristics of the physical processes and interpretation of the results instead of on details of the mathematical techniques. This should make practical data assimilation accessible for a much wider (end) user community in oceanography and offshore.