PNNL-13501 - Pacific Northwest National Laboratory

developing code architecture to minimize numerical
diffusion in the transport algorithms
• ensure mass conservation among the governing
equations for reactive trace gas species
• maintain the efficiency of numerical solvers for
systems of coupled sets of nonlinear differential
equations
• maintain scalability.
Results and Accomplishments
A central component of any air chemistry model is the
mathematical description of the chemical processes
themselves. The results from these algorithms provide
input for subsequent mathematical descriptions of
transport and removal. Many chemical mechanisms have
been put forward in the open literature. The diversity of
mechanisms represent efforts to balance the thousands of
known reactions against finite computer resources. In
principle, each mechanism is tailored to the scientific
problem of concern to the investigator. The chemical
mechanisms vary widely in the detail with which they
treat the dominant long-lived chemical species (such as
ozone or various nitrogen oxides), and key short-lived
species (such as HO2 or OH) that affect these longer-lived
trace gases.
Recent work on the development of PEGASUS has
included the insertion of a new mechanism designed to
balance details against finite computational resources.
This newly developed photochemical mechanism, called
the Carbon-Bond Mechanism – Zaveri or CBM-Z (Zaveri
and Peters 1999) extends the framework of an earlier
lumped-structure mechanism to function properly at
larger spatial and longer timescales, of interest for a
variety of problems of concern. As a preliminary quality
control check, the performance of CBM-Z has been
compared with an older mechanism based on Lurmann et
al. (1986). As illustrated in Figure 1, CBM-Z produces
lower peak ozone concentrations than the old mechanism.
We attribute these differences to the refined treatment of
isoprene, one of the most reactive biogenic hydrocarbon
species in the U.S., in CBM-Z.
Keeping in mind the need for computational efficiency to
offset the added detail in CBM-Z, we have also added a
novel numerical solver that is regime-dependent. With
this new algorithm, the CBM-Z mechanism is optimized
at each grid point and at every transport time step. The
optimization works by solving only the necessary
208 FY 2000 Laboratory Directed Research and Development Annual Report
LLA Mechanism (old)
CBM-Z Mechanism (new)
Time of Day (hour)
Figure 1. Comparison of O3 production by the method of
Lurmann et al. (1986) (top) and CBM-Z chemical
mechanisms in PEGASUS (bottom). Differences are
primarily due to the refined treatment of hydrocarbons,
especially isoprene, in CBM-Z.
chemical reactions. For example, dimethylsulfide
chemistry is turned on only when the dimethylsufide
concentrations are appreciable. Similarly, isoprene
chemistry is turned on only when the dimenthylsulfide are
appreciable. This approach will reduce the overall
computer time by about 30% in a large-scale simulation.
The “A” in PEGASUS stands for “aerosol.” One of the
most challenging tasks for a large model like PEGASUS
is the prediction of the equilibrium phase of the aerosol
(solid, liquid, or mixed) and the associated water content
as a function of ambient relative humidity. While
methods based on the direct minimization of Gibbs free
energy of the aerosol system are available (Ansari and
Pandis 1999), they are computationally extremely
expensive and therefore not amenable for use in large
three-dimensional Eulerian air-quality models. We have
developed an alternative solution algorithm, which
appears to be much faster than the existing methods.
Instead of searching for the minimum Gibbs free energy
of the system to calculate the equilibrium phase state, we
recast the problem in terms of set of hypothetical,
dynamic ordinary differential equations, and allow its
solution to reach equilibrium.