On the performance of numerical solvers for a chemistry submodel in three-dimensional air quality models 1. Box model simulations

The performance of numerical techniques in solving differential equations o f the gas phase chemistry submodel (i.e., the chemical solver) is one of th e most important factors in determining the overall computational cost for a three-dimensional (3-D) Air Quality Model (AQM). The estimated performanc e of a chemical solver in an AQM is often obtained by using simple box mode l analysis. In the present work some essential characteristics of the compu tational environment of any AQM, the operator splitting technique, have bee n identified and shown that different evaluation procedures will result in different conclusions for the relative performances of chemical solvers. A new box model evaluation procedure incorporating the impact of operator spl itting has been designed to better mimic the true performances of various c hemical solvers. Among the chemical solvers tested, the Hertel solver has t he best overall performance and is the most robust in dealing with diverse computational environments.