variables is measured. A subset of the observations and the relevant theories is then used to predict values for a “closure variable,” which is also measured independently. The result is a test of both measurements and theory and an opportunity to evaluate the quality of our understanding. With the instrumentation now available, closure experiments can be performed on aerosol number concentration (using a variety of sizing instruments), mass (based on measurements of relevant inorganic and organic species), radiative properties (using chemical composition, relative humidity, and Mie theory), and the integrated column effect of aerosols on short- and long-wave radiation. Closure experiments on aerosol mass can help answer questions about chemical composition, since missing species will make closure impossible. Theories about the impacts of aerosols on radiative climate forcing can also be tested by local and column closure experiments. Most of the aerosol experiments planned for the next decade depend heavily on this strategy because it offers a rigorous test of both measurements and the process models on which more comprehensive models depend.

The other new observational strategy is to observe the evolution of aerosols and their precursor gases in a Lagrangian reference frame. The idea of Lagrangian experiments is not new, and variations on this theme have been used occasionally. Recently, however, there has been considerable work on tagging air masses with balloons and chemical tracers, so that aircraft carrying large suites of instruments can revisit the air mass over a period of days to observe changes with time. Although these experiments cannot eliminate the effects of dispersion and vertical mixing on concentrations, with ample dynamical measurements, they make it possible to sort out the chemical and physical processes that cause changes in aerosols. These processes include gas-to-particle conversion, chemical transformations, wet and dry deposition, entrapment of air from other strata, and mixing through the sides of the “air mass” (dispersion). These experiments tend to be complex and expensive (at least one ship and one or two aircraft are required), but they offer the potential to test the aerosol models that now exist and that will be developed from future laboratory work and other process studies.

The overall strategic goal for the next two decades should be a predictive model to calculate atmospheric temperature and chemical species concentration fields and from that information to derive new aerosol particle formation rates and predict the chemical content and size distribution of the aerosol fields. Because current atmospheric models generally impose rather than predict aerosol distributions, significantly more sophistication will be needed in future models to represent precursor gas and gas/particle kinetics, nucleation and agglomeration kinetics, and vapor/particle interactions. One way to stimulate the needed improvements in aerosol modeling is to encourage the modeling community to participate directly in the planning, execution, and data analysis for the strategic field measurement programs described above.

Furthermore, predictive aerosol models will require currently unavailable quantitative mechanistic and kinetic input data describing a large number of