Personal tools
You are here: Home Science & Projects Deisa Extreme Computing Initiative Projects 2006 - 2007 Innovative Chemical Assimilation of Remote Sensing Observations of the Stratosphere

Innovative Chemical Assimilation of Remote Sensing Observations of the Stratosphere

Project ICAROS
Scientific Discipline Atmospheric chemistry, stratospheric ozone research, climate change
Principal Investigator(s) Frank Baier
  • DLR , Germany


To better characterize errors in chemical reanalyses of stratospheric ozone derived from remote sensing data a combination of four-dimensional variational (4Dvar) data assimilation with model ensemble calculations is proposed. In 4Dvar the initial conditions for a model forecast are adjusted to better comply with observations for the whole assimilation time window (e.g., 24 hours). Taking into account the known instrument and model uncertainties the most likely distribution of trace gases is derived. The resulting analyses are a prerequisite to study chemistry and dynamics of the stratosphere. In close cooperation DLR and national partners have developed a novel 4Dvar chemical data assimilation system within the BMBF project SACADA (Synoptic analysis of Chemical Constituents by Advanced Data Assimilation). Meteorological and chemical differential equations are solved in the physical domain, pioneering a novel icosahedral spherical grid.

By distributed computing SACADA offers the possibility to derive consistent long-term reanalyses of reactive and conservative chemical constituents in the stratosphere using current and historic satellite borne data. Time series of global trace gas fields are crucial in determining the anthropogenic impact on atmospheric chemistry and the climate system, and for improving atmospheric models.

4Dvar numerical operations are still computationally demanding. This project aims at optimizing the scaling characteristics of SACADA to make best use of the exceptional computing resources of DECI. This would allow doing ensemble calculations to determine the behaviour and evolution of errors in the system. In this way, it would be possible to improve modelling of a-priori errors and their correlations, which significantly influence the overall quality of 4Dvar results. SACADA is written in pure Fortran 90 and uses MPI for parallelization. Hence, we do not expect major difficulties in porting the system.

Document Actions