Dynamic air pollution modeling and GIS

The pollutants are gases, aerosols or solid substances, which represent a permanent component of the Earth atmosphere. Their concentration changes with place and time. Pollutants come into atmosphere from different sources. The diffusion of pollutants in the air is influenced by many factors. These factors causing pollutant concentration changes in any point of space, can be divided into three groups:

• source factors
• atmosphere factors
• environmental factors

All mentioned influences work at the same time in a complex composition, which is possible to describe both physically and mathematically.

In the current practice, two principally different types of models are applied:

1. statistical models;
2. dynamic models.

These model types differ in their dissemination and also by the level of difficulty of their mathematical basis, their computational difficulty and the method of entry of their input parameters.

Models of the first type use analytical solution based on the equation of diffusion simplified by the application of practical observation. Models of the second type are based on direct application of basic hydrodynamic and thermodynamic equations, which are solved numerically.

The statistical models describe real flow in a simplified way. They are based on simplified preconditions and a set of boundary conditions:

1. The pollution source is a point with constant permanent escape of pollutants.
2. There are no spatial limiting factors.
3. The modeled case is static.
4. The diffusion of pollutants has a statistical distribution in the direction of the y and z-axis.

Mathematically, the statistical models are based on the solution of the diffusion equation, where prevailing transport in the wind direction is assumed. In the vertical and horizontal direction perpendicular to the direction of the flow the turbulent diffusion is statistically described by normal distribution.

The Ministry of the Environment of the Czech Republic recommends the SYMOS 97 methodology for air pollution modeling. The standard Gaussian dispersion model, based on the Sutton equation, forms the basis of this methodology, which is also available in software form. The main drive behind the effort to interconnect this model with GIS was to make possible its comparison with the results of dynamic modeling.

The mathematical model is formed by a system of algebraic and partial differential equations, which is solved by the method of final volumes. The definition of border conditions on the modeling area borders enables us to introduce the physical basis of the process into the model. That includes the real characteristics of the spatial elements limiting the flow, meteorological characteristics - particularly the distribution of temperature, distribution of horizontal flow speed and so on. There are several Computer Fluid Dynamics (CFD) S/W packages available for numerical solving of the equations. These packages usually offer many options for the input of border conditions. Those can be defined as constants, functions, and eventually derivations of given quantities.

The task of practical solution was to create and to verify prototype interconnection of GIS and mathematic models for solving of accidental escape of danger pollutant.

For practical solving was chosen an example of real technology in Biocel Paskov a.s. cellulose factory. The reason for choosing of this factory is, that around is flat terrain and this way is possible to use a statistical Gaussian model for comparison.

The potential source of pollutant is the broken pipe in ammonium storage. The maximum short-term (30-min.) concentration of ammonium (NH3) safety for irreversible health changes or death (IDLH) is 500 ppm . The sure death concentration is 5000 ppm.

ArcInfo version 7.2.1 was used for solving of interconnection and data pre processing. The ArcView 3D Analyst was used for output presentation.

To represent a group of statistical models, the SYMOS 97 by Idea Envi, s.r.o., was chosen. It implements the official method of the current methodology for pollutants dispersion modeling, issued by the Ministry of the Environment of the Czech Republic.

The Panache software by Transoft International was selected to represent dynamic models. It provides a specialized version of the pollutants diffusion dynamic model for fluid circulation with variable types of pollution sources.

For model realization, the following equipment of the GIS lab at the Institute of the Environment Protection in the Industry of our Faculty has been used:

1. Pentium-based personal computers, running Microsoft Windows NT 4.0, were used to run the Panache software, at this stage available to us only in PC version.
2. The GIS software was running on graphical workstations of following types:

• Silicon Graphics OCTANE 2x MIPS R 12000, 300 MHz, 256 MB RAM, OS IRIX 6.4,
• Silicon Graphics 02, procesor MIPS R 5000, 200 MHz, 64 MB RAM, OS IRIX 6.3

1. Output devices:

• Printer HP 850C
• Plotter HP 450C

To define the spatial border conditions, the Fluidyn Panache s/w uses geometrical characteristics and their properties influencing the flow. These spatial data can be provided by the GIS. For these purposes the Panache model specifies the following spatial entities:

1. DOMAIN = modeling area

• An ashlar parallel with the x-axis, defined by its center co-ordinates, the height of its base above the ground and its height.

2. CURVE = contour line

• A line formed by a list of points, defined by points' co-ordinates their height above the sea.

3. BUILDING = building

• An ashlar parallel with the x-axis, defined by its center co-ordinates, the height of its base above the ground and its height.

4. FIELD = vegetation

• A polygon defined by the co-ordinates of its border points and the attributes describing the average height of vegetation, density of foliage, roughness coefficient.

5. FOREST = forest

• A polygon defined by the co-ordinates of its border points and the attributes characterizing the average height of the trees, the average height of the undergrowth, the foliage density both at the trunk and at the tree-top level, roughness coefficient.

6. WATER BODY = water surface

• A polygon defined by the co-ordinates of its border points.

7. URBAN AREA = intravilan

• A polygon defined by the co-ordinates of its border points and attributes characterizing the average height of buildings and the direction of the heat flow.

8. MONITOR POINT = receptor

• A point defined by its x,y,z co-ordinates.

9. ARBITRARY BUILDING = random building

• A polygon defined by the co-ordinates of its border points, the height of its base above the ground and its height.

10. METEO STN = meteostation

• A point defined by its x,y,z co-ordinates and attributes containing up-to-date meteorological data.

These entities are populated by the GIS data. It was necessary to choose such source of digital spatial data, witch would satisfy the model requirements. Data precision and the level of generalization have to meet the required modeling detail and the size of the modeling area. At the same time, the data need to be generally available and standard so that the GIS - Panache interconnection can be applicable throughout the Czech Republic.

The combination of the following two digital spatial data sources meets these requirements:

1. The cadastral map.

2. The basic map of Czech Republic, scale 1:10 000.

After consulting the Fluidyn Panache provider (Transoft International, Paris) a text file data format was agreed to as the most suitable for the data interchange.

The objective was to automate the data transfer from GIS to Panache to enable the continuous dynamic model computation on the GIS background, without the GIS operator intervention and without having the operating knowledge of the Panache software. Hence, a program module system was created within Arc/Info. Modules from this system make use of Arc/Info analytical capabilities for spatial data manipulation to enable correct transport of all Panache data types (features) while respecting the restrictions of this s/w. The restrictions for Panache features (entities):

1. Non-existence of topological relations. Each polygon must be defined separately with an integral border. The whole area enclosed by the border line constitutes a feature - "islands" are not allowed in this area.
2. No element can be formed by more then 200 breaking points.
3. No more than 128 elements can be used to define one entity.
4. The modeling area must lie in the positive co-ordinates.

Considering the way the BUILDING and DOMAIN features are defined, it is advisable to include another condition:

5. The x-axis should be parallel with as many buildings represented by the BUILDING feature as possible.

Furthermore, it was necessary to automatically test whether to put a given building polygon into the category "BUILDING" or "ARBITRARY BUILDING". A module was created, which tested GIS data and selects the appropriate category in following steps:

1. Check if the categorization is contained directly in the attribute database.

2. Check if the building is a quadrangle.

3. Check if it is a parallelogram (with determined tolerance).

4. Check if it is a rectangle or a square.

5. Check if one of its sides is parallel with the x-axis (with determined tolerance).

During the modeling in Panache the outputs are produced in individual time steps as a binary and textual output files containing the complete information about the modeled case. AML modules were used to extract selected data from these files:

Flow speed vectors were selected for clear presentation. Panache produces speed vector components according the x,y,z axis for each receptor. The graphic presentation of these outputs in GIS are spatially localized vectors in selected scale.

Receptors are in Panache determined by the crossing of spatial lines of calculated net, which is defined by steps in the direction of the individual axis. Absolute coordinates of these points are part of the output files. The presentation of receptors in GIS are points with given coordinates.

Out of the scalar quantities the concentration of pollutants is the most one. The dynamic model output contains concentrations according to the defined system, which depends on arrangement of the receptors in the net.

It was decided, that standard text output files will be used for data transfer (files for solved example count ten thousands lines). There are standard character strings in the files, which are invariable in all outputs. These was applied as a labels for quantity reading.

The result of previous steps is a series of different types of spatial data:

Furthermore the data files containing coordinates of receptors and information about the terrain, which serves as input for statistic model SYMOS 97 was created. This way is secured, that both models statistic and dynamic are working with the same input data and the outputs are evaluated on the same receptors.

For the mentioned data operations between both systems of modeling the series of AML ArcInfo modules was created.

At this stage, it is not technically feasible to use expert systems based on this solution for real world crises situations. Partial improvements of the response times can be achieved by data transfer optimization and by the introduction of tailor-made prototype programs. Further speed gain can be achieved by the use of faster hardware, possibly shortening the times by several times. However, it will still be necessary to expect an overall system response time within tens of minutes to hours.

It can be expected that the upcoming development of technical equipment (particularly graphic and computation power of computers) and mathematical algorithms for spatial dynamic models solution (further new codes for parallel computing) will bring dramatic improvements in this area.

Petr Jancik, Ph.D., head of Laboratory of GIS

Dpt. of Environmental Protection in Industry, VSB – Technical University of Ostrava, 17.listopadu 15, 70833, Ostrava, Czech Republic