On the influence of chemical initial and boundary conditions on annual regional air quality model simulations for North America

The influence of chemical initial conditions and chemical lateral boundary conditions (CLBCs) on long-term regional air quality model simulations was investigated using outputs from an annual simulation of the year 2002 on a North American domain. This simulation was carried out using the AURAMS regional air quality model. It was subdivided into three multi-month segments with two overlap periods (May 15–30 and September 1–30) to allow the segments to be run in parallel. For this approach to work, model predictions had to match very closely by the end of the two-week and four-week overlap periods. The time required for the values of daily domain-average surface PM_2.5 concentration to match for the two simulation segments associated with each of the two overlap periods was four and six days, respectively. For individual locations within the model domain, however, the required spin-up period was as much as nine days, considerably longer than the 2–4-day spin-up period usually assumed in the literature. For ozone, on the other hand, the daily domain-average surface ozone concentration values did not converge for either overlap period. A zero-gradient CLBC had been used for all run segments and species. When a time-invariant CLBC for ozone was used instead, the daily domain-average surface ozone concentration values behaved more realistically and did converge after fewer than three days. A similar improvement was also obtained for individual locations, but with spin-up periods of up to nine days. Model chemical spin-up time thus seems to be dependent on the species considered, the time required for the influence of the inflow boundaries to reach all locations within the domain, and the impact of local emissions sources. These results suggest the use of a spin-up period of longer than one week for a large (continental) domain and long-term simulation of PM_2.5 and O₃ rather than the 2–4 days commonly assumed in the literature.