Evacuations occur when human safety is compromised by disasters, such as floods. Shelters play a crucial role in providing protection for individuals who have been displaced or have lost their housing, emphasizing the requirement for secure accessibility. This paper introduces an systematic optimization tool, utilizing mathematical programming, to assist decision-makers in designing an effective shelter network. It is effective preparedness decision as it minimizes the total risk of the network considering its flooding condition during the response phase while maximizing coverage for those most exposed and vulnerable to hazard. We propose a risk-based approach, wherein the inherent risks of the shelter network (i.e., population, shelter, and evacuation risks) have been thoroughly assessed and measured to consider the impacts of floods based on empirical research outputs. To formulate a well-parameterized and valid problem, extensive data collection and processing, incorporating the use of geographic information system (GIS) for data management, have been conducted. In collaboration with the World Bank and Government of Haiti, this project contributes to a development initiative focused on strengthening disaster response capacity and infrastructure for Haiti, experiencing recurrent devastating floods and in need of enhancing its existing shelter network. Detailed computational results highlight the value of our risk-based methodology compared to more common approaches, emphasizing contributions to addressing real humanitarian problems.