Environmental Engineering Reference
In-Depth Information
variables. It gives us much more clarity when we start
to assemble the causal links. This is highlighted in two
points by Dorner (1996):
+
Problem
definition
Apply
+
R
Conclude
+
We need to know the variables on which those that we
want to influence depend. We need to understand, in
other words, how the causal relationships among the
variables in a system work together in the system.
•
+
Questions
+
+
Knowledge
Develop
mental model
R
+
We need to know how the individual components
of a system fit into a hierarchy of broad and narrow
concepts. This information can help us fill in by analogy
those parts of a structure unfamiliar to us.
•
+
+
Test
+
Established
knowledge
B
−
After performing basic sorting, previously unrecog-
nized interactions between variables may require alter-
ation or revision of the hypothesis. Some light may have
been shed on the larger complexes outside the defined sys-
tem boundaries, in which elements of the defined system
are embedded. This process is iterative and may require
several revisions before the model is fixed according to
desired standards.
Data and
information
+
+
Data from
experiments
Data from
literature
Figure 17.7
The learning loop is an iterative process were the
steps from building a mental model and producing results can
always be improved.
it. Then a specific demand for data will arise. Not all
data are needed - only what is relevant to our problem.
The rest must be sorted. From the test we will learn
and redevelop our understanding, our 'model'. Model
development requires iteration in such a cycle several
times and will continue as long as there is sufficient data
from experiments or literature to support the testing
of the model. In that way the communication of the
model is effective, both for the user and the developer.
Furthermore, it enables us to communicate the success
and problems encountered.
17.7 The process
Generalization is often the key to understanding complex
systems. Modelling is, without exception, based on some
sort of recipe. Whatever approach is used, all methods
focus on answering a specific question. We propose that
the basic process should start by defining and confining
system boundaries according to the specific question, or
questions requiring an answer. The whole process should
be considered iterative (Figure 17.7) where experience
from previous evaluation is used to redefine the problem
(Haraldsson, 2004, 2005; Haraldsson
et al
., 2007). The
learning experience gained and the conclusions from the
testing will help us re-evaluate our problem and our
mental model.
Modelling never starts with data, contrary to what
many believe. Just collecting 'even more' data without
a clear model, does not create greater understanding
or clarity - it merely adds to general confusion. Not
everybody would agree, but the track record is just too
consistent to accept any objections. All research starts
with a definition of a problem and through that an
understanding of that problem. This is 'the model', the
mental image of how the problem is to be understood.
Thus, no models are 'wrong'. If something does not work,
the cause is to be found in faulty understanding. After
a model has been created, we can test and challenge
17.8 Biogeochemical models
We define biogeochemical models as models that describe
the connection from biology through chemistry, to geol-
ogy and back. Geology (the solid phase) is only visible to
biology through its expression in chemistry. The only way
that geology (solids) affects chemistry is through chemical
weathering and ion exchange. Ion exchange is always in
the form of reversible reactions, whereas chemical weath-
ering reactions are almost always irreversible and slow.
Decomposition of organic matter, however, we regard as
part of the biotic cycle. For all biogeochemical models
where the connection between biological and geological
aspects is significant, the representation of the weathering
processes will be a vulnerable point for overall model
performance (Figure 17.8).
Search WWH ::
Custom Search