Application of bayesian networks to assess water poverty

The conventional approaches to water assessment are inappropriate for describing the increasing complexity of water issues. Instead, an integrated and holistic framework is required to capture the wide range of aspects which are influencing sustainable development of water resources. It is with this in mind that the Water Poverty Index (WPI) was created, as an interdisciplinary policy tool to assess water stress that links physical estimates of water availability with the socio-economic drivers of poverty. In parallel, in light of the investments envisaged for the next decade to reach the sector targets set by the Millennium Development Goals (MDGs), appropriate Decision Support Systems (DSS) are required to inform about the expected impacts to be achieved throughout these interventions. This would provide water managers with adequate information to define strategies that are efficient, effective, and sustainable. The paper explores the use of object oriented Bayesian networks (ooBn) as a valid approach for supporting decision making in water resource planning and management. On the basis of the WPI, a simple ooBn model has been designed and applied to reflect the main issues that determine access to safe water and improved sanitation. A pilot case study is presented for the Turkana district, in Kenya, where the Government has launched a national program to meet sector targets set out in the MDGs. Main impacts of this initiative are evaluated and compared with respect to the present condition. The study concludes that this new approach is able to accommodate local conditions and represent an accurate reflection of the complexities of water issues. Such a tool helps decision-makers to assess the effects of sector-related development policies on the variables of the index, as well as to analyse different future scenarios