Impact prediction, evaluation and mitigation - Introduction to Environmental Impact Assessment - page 135

5.2.2 Methods and models for prediction (how to predict)
There are many possible methods to predict impacts; a study undertaken by
Environmental Resources Ltd for the Dutch government in the early 1980s identified 150
different prediction methods used in just 140 EIA studies from The Netherlands and
North America (VROM 1984). None provides a magic solution to the prediction
problem.
All predictions are based on conceptual models of how the universe
functions; they range in complexity from those that are totally intuitive to
those based on explicit assumptions concerning the nature of
environmental processes…the environment is never as well behaved as
assumed in models, and the assessor is to be discouraged from accepting
off-the-shelf formulae (Munn 1979).
Predictive methods can be classified in many ways; they are not mutually exclusive. In
terms of scope, all methods are partial in their coverage of impacts, but some seek to be
more holistic than others. Partial methods may be classified according to type of project
(e.g. retail impact assessment) and type of impact (e.g. wider economic impacts). Some
may be extrapolative, others may be more normative . For extrapolative methods,
predictions are made that are consistent with past and present data. Extrapo lative
methods include, for example, trend analysis (extrapolating present trends, modified to
take account of changes caused by the project), scenarios (common-sense forecasts of
future state based on a variety of assumptions), analogies (transferring experience from
elsewhere to the study in hand) and intuitive forecasting (e.g. the use of the Delphi
technique to achieve group consensus on the impacts of a project) (Green et al. 1989).
Normative approaches work backwards from desired outcomes to assess whether a
project, in its environmental context, is adequate to achieve them. For example, a desired
socio-economic outcome from the construction stage of a major project may be 50 per
cent local employment. The achievement of this outcome may necessitate modifications
to the project and/or to associated employment policies (e.g. on training). Various
scenarios may be tested to determine the one most likely to achieve the desired outcomes.
Methods can also be classified according to their form, as the following types of
model illustrate.
Mathematical and computer-based models
Mathematical models seek to represent the behaviour of aspects of the environment
through the use of mathematical functions. They are usually based upon scientific laws,
statistical analysis or some combination of the two, and are often computer based. The
underpinning functions can range from simple direct input-output relationships to more
complex dynamic mathematical models with a wide array of interrelationships.
Mathematical models can be spatially aggregated (e.g. a model to predict the survival rate
of a cohort population, or an economic multiplier for a particular area), or more
locationally based, predicting net changes in detailed locations throughout a study area.
Of the latter, retail impact models, which predict the distribution of retail expenditure
using gravity model principles, provide a simple example; the comprehensive land-use
locational models of Harris, Lowry, Cripps et al., provide more holistic examples