Computationally efficient atmospheric chemical kinetic modeling by means of High Dimensional Model Representation (HDMR)
S.W. Wang, P.G. Georgopoulos (EOHSI, UMDNJ - R.W. Johnson Medical School
and Rutgers University);
G. Li, H. Rabitz (Department of Chemistry, Princeton University)
Chemical kinetics calculations can consume as much as 90% of the total CPU time in simulations that employ chemistry-transport modules with comprehensive photochemical air quality models. To relieve this computational burden, this work introduces an efficient High Dimensional Model Representation (HDMR) method to perform the chemical kinetics calculations. An efficient HDMR for these types of calculations is based on expressing the concentrations of the chemical species as a hierarchical correlated function expansion which captures the chemical kinetic input-output relationships. Therefore, an HDMR expansion is used as an input-output chemical kinetics map, with the input being initial chemical concentrations and perhaps other variables (e.g., solar intensity for photochemical reactions), and the output similarly being chemical concentrations at a later time. Thus, by repeating this process for successive times, an HDMR effectively can act as an integrator, with perhaps a very large time step size. The computational efficiency of an HDMR is due to its operations only involving very rapid and stable algebraic manipulations, and the large time steps could in turn allow for the inclusion of enhanced chemical mechanisms. The test application of the HDMR method on atmospheric chemistry presented here focuses on a photochemical box model study of alkane photochemistry. It is shown that the FEOM calculations of chemical species concentrations can maintain accuracy comparable to conventional chemistry solvers (e.g., the Gear-type implicit solvers), while being orders of magnitude faster.