A comparison of numerical techniques for solution of atmospheric kinetic equations

Numerical modeling of atmospheric chemistry is a computationally intensive problem. The equations describing the interaction among various modeled chemical species are coupled, nonlinear ordinary differential equations. Spatial dependencies in comprehensive three-dimensional air quality models require the solution of this system at thousands of spatial points. Even with increasing computer power, there is a need for efficient and accurate numerical solvers with expanded capabilities, since the next generation of air quality simulation models needs to address the increasingly complex chemistry issues emerging in new model applications. Variants of the commonly used quasi steady-state approximation and the hybrid methods currently used in several modeling systems are examined against a reference mechanism describing chemical interactions related to tropospheric oxidant and acid formation. Additional modifications to the methods are incorporated to yield more robust integration techniques. The chemistry solution methodology used in the regional acid deposition model is also incorporated in this comparison as a base methodology for representing the reference chemical mechanism. The methods are tested against the Gear integration scheme for a variety of test cases including traditional box-model calculations and detailed three-dimensional simulations, and their relative accuracies and efficiencies are investigated. Performance and implementation issues related to chemical integration schemes are examined in the context of the demands and needs of the chemistry component of future comprehensive atmospheric chemistry/transport simulation models.