Factorial or indicator modelling
Since many of the processes and factors that influence a particular type of environmental
problem are well known, it is possible to rank individual factors by the strength of their
association with the problem, providing a series of indicators. For example, climatic
indices may be based on the frequency of high-intensity precipitation, and on the
degree of aridity or rainfall seasonality. Soil indicators may reflect the tendency to
crusting and sealing. Similar rank indicators may be developed for parent materials,
topographic gradients and other factors.
Individual indicators may be mapped separately, but combining the factors into a
single scale –by adding or multiplying suitably weighted indicators for each individual
factor– is more problematic. There are difficulties both about the individual weightings
and about the assumed linearity and statistical independence of the separate factors. The
method may be effective for identifying the extremes of high and low susceptibility, but
is less satisfactory in identifying the gradations between the extremes.
Despite these theoretical limitations, factor or indicator mapping has the considerable
advantage that it can be widely applied using data that are available at a regional or even
continental scale. There is a continuous spectrum between indicators based on simple
ranking and those based on equations with a more explicit physical or empirical basis.
Process modelling
A process model consists of an equation or a set of equations designed to represent the
process and its behaviour under study. A large variety of models have been developed
for describing processes. Depending on the level of process understanding and data
availability, process models may be characterized in different ways. Mechanistic
models, as opposed to empirical models, describe the process on the basis of physical
understanding. Deterministic models are capable of producing quantitative predictions,
whereas stochastic models also incorporate associated probability distributions and
random elements, so that not only a value can be predicted but also its uncertainty. The
technical complexity of stochastic models justifies in some cases the combination of
Monte Carlo simulation and deterministic models to deal with uncertainties. Dynamic
models, in contrast to static models, incorporate time as an explicit component in
describing processes. A review of many models dealing with changes in soil properties
can be found in Young (1994).
Process models have the potential to respond explicitly and rationally to changes in
input variables, e.g. climate, and so have promise for developing scenarios of change,
and what-if analyses of policy or economic options. Set against this advantage, process
models generally cannot assess environmental processes up to the present time, and can
only incorporate the impact of past events where this is recorded, as in soil databases,
yield statistics or hydrological records. Also, models generally are simplifications of
the set of processes operating, so they may not be appropriate under particular local
circumstances.
Chia sẻ với bạn bè của bạn: |