Work
Package 2 (WP2): Development of case study Bayesian networks
Start
date or starting event: Month 6
Objectives
and input to workpackage
- The
objective of this workpackage is to develop Bayesian networks for
each of the four study sites. The four catchments we have selected
are in different parts of Europe and subject to contrasting water
resource pressures. The catchments are, the Wye (UK); Jucar (Spain);
Northeast Zealand (Denmark) and Abruzzo (Italy). The specific issues
to be addressed by each network will emerge from the stakeholder meetings
described in WP1.
Methodology
/ Description of work
- Design
and construction of the Bns for each site will start at month 6, which
is shortly before the final input from WP1 is due in month 9. But
before this phase begins Aalborg will host an internal seminar to
familiarise the other partners with the software to be used for the
exercise. Aalborg will provide expert support to all participants
throughout the course of the project. Development of the networks
will start when the deliverables from WP1 become available. It is
anticipated that the case studies will be complete by month 24.
- A Bayesian
network will be constructed for each site. A network will be composed
of a series of nodes representing variables such as 'river flow' (environmental),
'water demand' (economic) and 'population density' (social). These
variables are linked together so that a change in one results in a
chain reaction of impacts on all the linked variables. Variables which
affect the state of another (e.g. rainfall on river flow) are known
as 'parent' variables. The probability that a variable is in any particular
state, given the state of its 'parent' variables, is expressed as
a conditional probability table (cpt). These tables may be derived
from data sets, the output of complex models, or expert opinion. The
ability of Bayesian networks to integrate information lies with these
cpts, which are constructed for each variable.
- At the
four study sites the variables within the networks will be those identified
by the stakeholder groups in WP1. Data from which to construct the
cpts will be provided by the data collection exercise carried out
as WP3.
- Once
complete Bns can be used to interrogate the response of systems to
a variety of management or planning scenarios. Changing the status
of one variable will cause a knock on effect and change the status
of all those variables linked to it. Uncertainty in the system is
explicitly expressed since the likelihood of a variable being in any
particular state is given as a probability distribution.
|
