Belirsiz bilgi verici varlığı altında afet sonrası dayanıklılık değerlendirmesi için seçici rota planlama

After a disaster, rapid assessment reports provide invaluable insights into immediate needs during the first response, while long-term needs assessment studies are essential for understanding the evolving vulnerabilities of populations during the recovery and reconstruction phases. In this study, we focus on a selective routing problem under uncertainty, where field teams visit earthquake-affected sites to conduct long-term needs assessment studies with key informants with uncertain presence, about the resilience of agricultural and educational activities to disasters. Specifically, we select a sample of sites for visits and prepare daily routes for the field teams, considering the possibility of absent key informants. There are two key challenges in this problem: (1) ensuring sample representativeness in site selection, and (2) accounting for the uncertain presence of key informants in preparing efficient routing plans. To address representativeness concerns, we use a stratified sampling approach leveraging a clustering algorithm to define groups of similar sites and ensure a proportional number of visits to each. The uncertainty in the key informant's presence is tackled by a novel main-backup assignment strategy, where each selected main site is paired with a backup site from the same cluster, enabling field teams to maintain representativeness. We develop a 2-phase matheuristic approach to solve this problem. It combines two enumeration strategies with a set partitioning model to generate feasible routes and select efficient ones, ensuring representativeness. The developed solution approach is used on a large dataset of 11 provinces in Türkiye to support a real-life assessment project following the February 2023 earthquakes. The findings show that different enumeration strategies can be more advantageous in terms of solution times and expected routing costs when the probability of the key informant's presence varies.