An asynchronous time-stepping (ATS) integrator for atmospheric applications: Aerosol dynamics

Abstract

Mixed time integration methods, also known as multiple time-step methods (MTS), have been proposed and implemented successfully to reduce the stiffness of problems in molecular dynamics and solid mechanics, where different time scales are used to integrate different motions or meshes. We apply the similar concept and develop an asynchronous time-stepping (ATS) integrator for atmospheric application. This paper first presents ATS as a generalized ordinary differential equation solver and describes its key components including: (1) global and local time management; (2) dynamic time-scale ordering; and (3) integration and iteration methods. Then we apply ATS to solve stiff problems in aerosol dynamics, where we further introduce several techniques such as grouping particulate species by particle sections to speed up computations. The comparisons between ATS and standard ODE solvers including LSODE, LSODES and VODE in aerosol dynamics simulation are presented, which show that ATS is robust and able to achieve a significant digit average (SDA) value of 2.0 to 3.0 while its speed is around 10 to 100 times faster than LSODE with equivalent accuracy. ATS has been implemented in CMAQ-UCD, a sectional, dynamic aerosol model built on USEPA's CMAQ platform.