Reports and Papers
- Fedra, K. (1995)
- Decision support for natural resources management:
Models, GIS and expert systems.
AI Applications, 9/3 (1995) pp 3-19.
An Air Quality Management System
Another typical example of a natural resource in dire need of better
management is clean air.
The problem owner here is usually a regulatory agency that
controls various sources of emissions, or a major industry or thermal
power plant that has to comply with regulatory standards.
While several novel approaches, including direct market mechanisms and
economic incentives have been discussed, the majority of systems still
rely on simple source by source emission control versus ambient air
An example system designed for a typical regulatory agency
integrates a set of simulation and optimization models for
air quality management.
They are built for the assessment of emission sources and the design
and evaluation of pollution control strategies.
The system brings together data bases (emission inventories,
meteorological data, and model scenarios), a geographical information
system, simulation models, an optimization model,
and an expert system for the estimation of point and area source
The geographical information system provides tools to
display, access, and manipulate spatially distributed data
that are either used for the models directly (eg., a digital elevation
model, or land cover used to estimate surface roughness and surface
temperature differentials), data used to derive emission estimates (eg.,
urban areas and major traffic arteries used to estimate area source
emission), data for impact assessment such as different land use of
different vulnerability to various pollutants, and finally geographical
background data for the spatial orientation of the user, the location of
sources, and as a spatial reference for model results.
The emission inventories are available either through the
map display by picking sources for display and editing of their
characteristics, or from a parallel listing of named sources.
Basic source characteristics such as location, emission for various
pollutants, stack parameters, and cost functions for alternative pollution
abatement technologies that are applicable for a given source are stored
in the inventories. An embedded expert system can be used to derive
emission estimates from basic technological data such as fuel consumption
and characteristics, or production technologies and volumes.
The meteorological data base allows the display of weather data
and the selection of either a particular set of observations
as the basis for a short-term model run, or the definition of
a longer period, usually an entire year, for long-term simulation.
In the latter case, a pre-processor generates the frequency distributions
required by the long-term model from the time series of observations
selected by the user.
The simulation models of the system include an implementation of EPA's
Industrial Source Complex model ISC, a Gaussian model that can be
run both for short episodes and long-term frequency data (USEPA 1979).
Alternatively, a three layer finite element model, used conjunctively with
a spatially distributed wind field generator can be used for dynamic
short-term runs over a few days.
These models describe pollutants such as SO2, NOx, or dust.
For summer smog and ground level ozone, a photochemical box
model simulating daily episodes based on EPA's PBM code is used
(USEPA 1984). It is driven by the same weather
scenarios and shares the emission inventories for point and area sources
of NOx. Emissions volatile organic compounds are again estimated with
an embedded expert system, using emission coefficients and a set of rules.
The output of any long-term model can be used as an emission and
impact scenario for the optimization module. Using a source-receptor
and a spatially distributed, non-linear environmental impact function
that can assign different weights to different land use or population
this component finds cost-effective strategies for pollution abatement.
Each controlled source has a number of alternative control technologies
available including the option in some cases to close a plant.
Each option is associated with costs, and for a given overall budget
the model finds the most effective (in terms of the environmental impact
or just bulk emissions) investment strategy.
Varying the budget, or the time horizon and discount rate for the cost
estimates results in a large number of scenarios, that can be further
analyzed by a discrete multi-criteria optimization tool (Zhao et al.,
An ``optimal'' emission scenario can then be used again at the level of
the simulation model, and tested with a broad range of individual
short-term weather scenarios (rather than the frequency data used for
the long-term model) to test the abatement strategy under specific,
including worst case assumptions.
Model results are displayed as color coded overlays over the
background maps in 2D and pseudo 3D displays over a digital
as a set of symbols representing
emission reductions at the source locations in the optimization model,
or as a set of time series displays and diagrams for the ozone model.