The increasing need for groundwater as a source for fresh water and
the continuous deterioration in many places around the world of that
precious source as a result of anthropogenic sources of pollution
highlights the need for efficient groundwater resources management. To
be efficient, groundwater resources management requires efficient access
to reliable information that can be acquired through monitoring. Due to
the limited resources to implement a monitoring program, a groundwater
quality monitoring network design should identify what is an optimal
network from the point of view of cost, the value of information
collected, and the amount of uncertainty that will exist about the
quality of groundwater. When considering the potential social impact of
monitoring, the design of a network should involve all stakeholders
including people who are consuming the groundwater.
This research introduces a methodology for groundwater quality
monitoring network design that utilizes state-of-the-art learning
machines that have been developed from the general area of statistical
learning theory. The methodology takes into account uncertainties in
aquifer properties, pollution transport processes, and climate. To check
the feasibility of the network design, the research introduces a
methodology to estimate the value of information (VOI) provided by the
network using a decision tree model. Finally, the research presents the
results of a survey administered in the study area to determine whether
the implementation of the monitoring network design could be supported.
Applying these methodologies on the Eocene Aquifer, Palestine
indicates that statistical learning machines can be most effectively
used to design a groundwater quality monitoring network in real-life
aquifers. On the other hand, VOI analysis indicates that for the value
of monitoring to exceed the cost of monitoring, more work is needed to
improve the accuracy of the network and to increase people’s awareness
of the pollution problem and the available alternatives.