One measure of our knowledge comes from the convergence in predictions made
by a variety of models. Difficulties in the analysis and evaluation of such
comparisons can result from models employing different emissions, meteorological
fields etc. A set of standardised input was provided for the Comparison of Large
Scale Sulphate Aerosol Models (COSAM) workshop which took place in 1998 and
1999 (see Barrie et al., 2001). Ten models participated in this comparison.
As noted above in Section 126.96.36.199, the simulation
of the processes determining the sulphate concentration differed considerably
between models. The fraction of sulphur removed by precipitation ranged from
50 to 80% of the total source of sulphur. The fraction of total chemical production
of sulphate from SO2 that took place in clear air (in contrast to
in-cloud) ranged from 10 to 50%. This latter variability points to important
uncertainties in current model capability to predict the indirect forcing by
anthropogenic sulphate, because the mechanism of sulphate production determines
the number of CCN produced (Section 5.3.3).
The ability of the models to predict the vertical distributions of aerosols was examined by comparing model predictions of the vertical distribution of SO42-, SO2 and related parameters such as ozone, hydrogen peroxide and cloud liquid-water content to mean profiles taken during aircraft campaigns at North Bay, a remote forested location (in southern Nova Scotia) 500 km north of the city of Toronto (Lohmann et al., 2001). For SO42-, the models were within a factor of 2 of the observed mean profiles (which were averages of 64 and 46 profiles for North Bay and Novia Scotia, respectively) but there was a tendency for the models to be higher than observations. The sulphur dioxide concentration was generally within a factor of two of the observations. Those models that overpredicted SO42- also underpredicted SO2, demonstrating that the model treatment for the chemical transformation of SO2 to SO42- is a source of uncertainty in the prediction of the vertical distribution of SO42-.
A comparison of modelled and observed ground level sulphate at 25 remote sites showed that on average, most models predict surface level seasonal mean SO42- aerosol mixing ratios to within 20%, but that surface SO2 was overestimated by 100% or more. A high resolution limited area model performed best by matching both parameters within 20%. This was consistent with the large variation in the ability of models to transport and disperse sulphur in the vertical.
Both regional source budget analyses (Roelofs et al., 2001) as well as long-range
transport model tests (Barrie et al., 2001) suggest that the dominant cause
of model-to-model differences is the representation of cloud processes (e.g.
aqueous-phase sulphate production rates, wet deposition efficiency and vertical
transport efficiencies) and horizontal advection to remote regions.
In summary, the COSAM model comparison showed that both convective transport and oxidation of sulphur dioxide to sulphate are processes that are not well simulated in large-scale models. In addition, the treatment of dry and wet deposition of aerosols and aerosol precursors continues to lead to important variations among the models. These variations lead to uncertainties in atmospheric sulphate concentrations of up to a factor of 2. The uncertainties in SO2 are larger than those for SO42-.
Other reports in this collection