|Conducting Environmental Impact Assessment in Developing Countries (UNU, 1999, 375 pages)|
|5. EIA tools|
|5.1 Impact prediction|
The essence of EIA is predicting future environmental conditions, either with the proposed development or without. A comparison of the two predicted situations is also often made with the present. Prediction is the process of determining the nature and extent of the environmental changes that will result from a proposed activity. From an examination of the methods available for predicting different effects, the following main "types'' of methods could be identified: physical models, experimental methods, and mathematical models.
Physical models, illustrative or working scale models constructed to represent the environment, may include visual representations of the environment by picture, photograph, film, or three-dimensional model and working models of the environment using, for example, wind tunnels or wave chambers.
Experimental methods, practical field or laboratory work, include field experiments, in which tests are carried out at the proposed site, and laboratory experiments, in which tests are carried out in the laboratory in conditions simulating the environment.
Mathematical models, in which the relationship between the cause and effect is represented by one or more mathematical relationships, may be either empirical (black box) models, where the relationships between inputs and outputs are established from statistical analysis of observations in the environment, or "internally descriptive'', where the mathematical relationships within the model are based on an explicit representation of the mechanisms of the processes occurring within the environment. These latter models range from simple formulations, which can be applied manually, to complex dynamic models, which require computer application.
Evidence from case studies suggests that the methods most often used in the Environmental Protection Agency (EPA) are the simpler methods, for example:
· steady-state, single source, Gaussian plume dispersion models for air quality;
· simple run-off and leachate models based on catchment area and rainfall;
· simple dilution and steady-state dispersion models for water quality;
· simple inventory approaches for both direct and higher-order effects on receptors (e.g., man, plants, habitats, etc.).
These methods can usually be applied manually or graphically, or using simple computer programs. The predictions obtained using these methods are usually very approximate, although the quality of results will depend on the particular problem and circumstances for which the method is used.
Environmental Resources Limited, UK, commissioned a study to examine the predictive methods used in EIA. The study covered 140 EIAs and environmental planning studies. There were 910 predictions of 36 different types of impact processes or effects. Mathematical, physical, or experimental modelling methods were used in 25% of these examples; in a further 15%, simple methods such as inventories of the factor affected (e.g., numbers of people or properties, area of habitat, etc.) were used to describe effects. In the remaining cases, other approaches not involving the use of formal methods were adopted.
Some of the reasons for the lack of application of more complex prediction models in EIA are related to their time and resource requirements for data input, calibration, and application, and the diminishing returns achieved in terms of the quality of result relative to their resource requirements, compared with simpler forms of predictive methods.
Types of methods less frequently used in EIA include:
· working physical models of atmospheric, aquatic, and acoustic effects (wind tunnels, hydraulic models, etc.);
· field and laboratory experiments (tracer experiments, bioassays, etc.);
· site-specific mathematical models and dynamic mathematical models.
These types of methods take into account more of the specific characteristics of the particular activity and environment, and make fewer generalizing assumptions. They may also take into account complex sources. As a result they have greater input and resource requirements for calibration and application (unless an existing model is available for use in the study area). As a general rule, they provide more detailed information about effects, but not necessarily more accurate information.