The horizontal advection is one of the most important physical processes in an air pollution model. While it is clear how to describe mathematically this process, the computer treatment of the arising first-order partial differential equation (PDE) causes great difficulties. It is assumed that the spatial derivatives in this equation are discretized either by finite differences or by finite elements. This results in a very large system of ordinary differential equations (ODEs). The numerical treatment of this system of ODEs is based on the application of a set of predictor-corrector (PC) schemes with different absolute stability properties. The PC schemes can be varied during the time-integration. Schemes, which are computationally cheaper, are selected when the stability requirements are not stringent. If the stability requirements are stringent, then schemes that are more time-consuming, but also have better stability properties, are chosen. Some norms of the wind velocity vectors are calculated and used in the check of the stability requirements. Reductions of the time-step size are avoided (or, at least, reduced considerably) when the PC schemes are appropriately varied. This leads to an increase of the efficiency of the computations in the treatment of large-scale air pollution models. The procedure is rather general and can also be used in the computer treatment of other large-scale problems arising in different fields of science and engineering.

"Models are often the only way of interpreting measurements to in vestigate long-range transport, and this is the reason for the emphasis on them in many research programs". B. E. A. Fisher: "A review of the processes and models of long-range transport of air pollutants", Atmospheric Environment, 17(1983), p. 1865. Mathematical models are (potentially, at least) powerful means in the efforts to study transboundary transport of air pollutants, source-receptor relationships and efficient ways of reducing the air pollution to acceptable levels. A mathematical model is a complicated matter, the development of which is based on the use of (i) various mechanisms describing mathematically the physical and chemical properties of the studied phenomena, (ii) different mathematical tools (first and foremost, partial differenti al equations), (iii) various numerical methods, (iv) computers (especially, high-speed computers), (v) statistical approaches, (vi) fast and efficient visualization and animation techniques, (vii) fast methods for manipulation with huge sets of data (input data, intermediate data and output data).

This book constitutes the thoroughly refereed post-proceedings of the Third International Conference on Large-Scale Scientific Computing, LSSC 2001, held in Sozopol, Bulgaria, in June 2001. The 7 invited full papers and 45 selected revised papers were carefully reviewed for inclusion in the book. The papers are organized in topical sections on robust preconditioning algorithms, Monte-Carlo methods, advanced programming environments for scientific computing, large-scale computations in air pollution modeling, large-scale computations in mechanical engineering, and numerical methods for incompressible flow.

Abstract: "Due to the large number of chemical species and the three space dimensions, off-the-shelf stiff ODE integrators are not feasible for the numerical time integration of stiff systems of advection-diffusion-reaction equations [formula] from the field of air pollution modelling. This has led to the use of special time integration techniques. This paper is devoted to a survey of such techniques, encompassing stiff chemistry solvers, positive advection schemes, time or operator splitting, implicit-explicit methods and approximate matrix factorization solutions. Of great importance in practice is high performance computing due to the huge problem scales, in particular for global models. We will therefore also report on experiences with vector/parallel shared memory and massively parallel distributed memory architectures and clusters of workstations. The survey is not entirely unique to air pollution models and biased towards work done at CWI over approximately the last 5 years."

Proceedings of the NATO Advanced Research Workshop, Sofia, Bulgaria, 6-10 July 1998

Abstract: "We investigate numerical algorithms for use in air pollution models. The emphasis lies on time integration aspects in connection with advection, vertical turbulent diffusion and stiff chemical transformations. The time integration scheme considered is a 2nd-order implicit-explicit BDF scheme which handles advection explicitly and vertical turbulent diffusion and chemistry implicitly and coupled. The investigation is divided into three parts. In the first part we propose a Gauss-Seidel technique for the implicit solution of the chemistry and vertical turbulent diffusion. For a hypothetical 1D model, based on a 66- species EMEP photochemical ozone chemistry scheme, this technique is shown to be significantly more efficient than the usual approach of using modified Newton with a linear band solver. For the stiff chemistry the Gauss-Seidel iteration is effectively explicit. For the diffusion the implicitness is retained, which gives rise to tridiagonal linear systems. In the second part we discuss stability and consistency properties of the implicit-explicit BDF scheme, assuming the 3rd-order upwind biased finite- difference discretization of the advection operator. In the third part we apply the implicit-explicit scheme to a hypothetical 3D model based on the same photochemical ozone chemistry. Here we employ vectorization over the horizontal grid. Grid-vectorization is of large practical interest as it leads to significantly higher efficiency on a vector computer. Dependent of the process at hand, parallelization is obtained either over the horizontal or over the vertical grid."

The air pollution, and especially the reduction of the air pollution to some acceptable levels, is an important environmental problem, which will become even more important in the next 10-20 years. This problem can successfully be studied only when high-resolution comprehensive models are developed and used on a routinely basis. However, such models are very time-consuming, also when modern high-speed computers are available. Indeed, if an air pollution model is to be applied on a large space domain by using fine grids, then its discretization will always lead to huge computational problems. Assume, for example, that the space domain is discretized by using a (480×480) grid and that the number of chemical species studied by the model is 35. Then several systems of ordinary differential equations containing 8064000 equations have to be treated at every time-step (the number of time-steps being typically several thousand). If a three-dimensional version of the same air pollution model is to be used, then the above figure must be multiplied by the number of layers. It is extremely difficult to treat such large computational problems; even when the fastest computers that are available at present are used.There is an additional great difficulty which is very often underestimated (or even neglected) when large application packages are moved from sequential computers to modern parallel machines. The high-speed computers have normally a very complicated memory architecture and, therefore, the task of producing an efficient code for the particular high-speed computer that is available is both extremely hard and very laborious.The use of standard parallelization tools in the solution of the problems sketched above is discussed in this paper. Results obtained on different types of parallel computers are given. It is demonstrated that the new efficient parallel algorithms allow us to solve more problems and bigger problems.