An Adaptive Grid and Associated Advanced GIS Techniques for Air Quality Models

Maudood Khan, M. Talat Odman, Hassan A. Karimi, Michael F. Goodchild

Hassan A. Karimi
Dept of Information Science and Telecommunications
University of Pittsburgh
Pittsburgh, PA 15260
Tel: (412) 624-4449
Fax: (412) 624-2788

E-mail: hkarimi@sis.pitt.edu


Air quality models (AQMs) simulate the transport, transformation, and fate of pollutants emitted into the atmosphere from various sources. They are increasingly used to design emission control strategies. The accuracy of AQMs depends on the resolution of relevant atmospheric processes. Inaccurate results may lead to huge investments in ineffective strategies and still leave the public health at risk. Current AQMs have fixed grids with predetermined resolution. However, there is a need for increased resolution in places of high chemical activity. The location of such places may vary throughout the simulation as the meteorology and emission patterns change. To reduce the uncertainty in the AQM results, we have developed an adaptive grid model and a set of advanced geographic information systems (GIS) techniques and tools that together can provide the optimal resolution automatically. One main characteristic of the adaptive grid technique is that it preserves the topology of the grid, that is each cell always has the same neighbors.

One of the most important issues in developing the adaptive grid AQM is the processing of emissions. Current GIS, which are used to retrieve emissions data (source location and emission rates) and map them onto fixed grids, cannot be used. Much more efficient GIS techniques are necessary for real-time emission processing over adaptive grids that evolve during the simulation. Every time the adaptor changes the grid point locations, the following three-stage processing is performed: (1) point sources are allocated into the appropriate grid cells, (2) line sources are intersected with grid cells to determine their contribution to each cell, and (3) area sources are mapped onto the grid cells. It is desirable to complete this processing in a fraction of the time required by the other AQM calculations. For this, we have designed very efficient search and intersection algorithms that take advantage of the unchanging topology of the grid.

Another important issue is the identification of criteria to drive the adaptor. A linear combination of solution features such as the gradients and curvatures may be used to generate a weight function. The adaptor will then move the grid points, clustering them around the regions where the weight function has high values. This will increase the resolution where it is needed more. Since the number of grid points is constant, this can only be achieved by decreasing the resolution in other places where it is needed less but the final result is a more favorable positioning of grid points such that the errors are reduced. Deciding which criteria to use is not a trivial matter in AQMs. There are too many species with very different spatial distributions and different contributions to pollution. Including the gradients of all the species in the weight function may lead to a more uniformly spaced grid than it is desired. We are developing criteria that take into account the chemical reactivity of the atmosphere and give more weighting to those species that are more important in the formation of pollution.

The performance of the adaptor and its efficiency to cluster grid nodes where needed, were evaluated using surface elevation data for the island of Hawaii. Initially, surface elevation values were specified on a 4x4 km resolution grid. Then two grids were created with 25% of the number of original data points: a fixed grid with 8x8 km resolution and an adaptive grid with variable resolution. The adaptive grid retained the data values much better than the fixed grid. The adaptor clustered the grid points near sloping terrain including the coastline and mountainous regions. The result was a reduction of the overall absolute and normalized errors as well as the local maximum error.

The development of other pieces of the adaptive grid AQM, such as the meteorological data processor, are underway. Full-blown regional-scale air quality simulations based on the developed adaptive grid and advanced GIS techniques will be performed in the near future.