Regional and Inverse Modelling
Processes in atmospheric chemistry are impacted by a large range of spatial scales, from intercontinental transport to very small scale surface patterns. Along these scales the quality of air is influences by numerous factors: nearby exhaust from combustion driven vehicles, industrial emissions from near and far, even from overseas, chemical conversions in gas phase and on aerosols, transport, turbulent redistribution, and removal of trace gases and particles from the atmosphere by plants, precipitation, or sedimentation.
Forecasting of air quality thus involves the ability to simulate many chemical and physical processes and combine them with their interactions.
The working group "Regional and Inverse Modelling” addresses mainly two objectives:
- provision of numerical analyses of case studies and field missions, where in cooperation with experimentally working groups and their measurements, the scope and the limits of the current understanding of tropospheric chemistry processes is analysed,
- development and implementation of operational air quality forecasts and analyses in the frame of European Global Monitoring for Environment and Security (GMES) projects.
The essential and novel approach for successful accomplishments of the above two objectives rest on Inverse Modelling and Data Assimilation. Here, methods from optimisation theory and statistics are applied to combine model results with observations, to optimally infer atmospheric chemistry states and parameter estimates. The identification of emission strengths with the aid of observations figures prominently in this context.
The underlying model and inversion system is the EURopean Air pollution Dispersion – Inverse Model (EURAD-IM). It has been developed at the Rhenish Institute for Environmental Research at the University of Cologne (RIU), with whom a close collaboration is practised. The tangent-linear and adjoint form of EURAD-IM are key ingredients of the development of advanced data assimilation and inversion algorithms, like the 4-dimensional variational approach. . In this model up to 120 variables per grid point are calculated , even if only a small fraction of which, like Nitrogen Oxides, Ozone, and particulate matter are published daily and used by environmental protection agencies or the public.
The complexity of the models and the inversion methods enforces the application of high performance computing facilities, as they are hosted at the research centre Jülich.