MSc Student Technion - Israel Institute of Technology
Floods are the most common and among the deadliest natural disasters in the world. As global warming continues to exacerbate floodplains are expected to grow. Even with the growing efforts to mitigate floods through infrastructure investments, flooding cannot be wholly prevented. Thus, arises the need for search and rescue (S&R) efficient operations in case of a flood. Typically, flood affects a large geographical area, in which S&R resources are usually limited. This study aims at pinpointing locations in which flood victims, who have been carried by the flow, are probable to be found by combining both physical models and machine learning approaches into one novel, conceptual framework, dubbed machine education. The term education here is used as the learning task is not left for the machine to do by itself, but it is guided, or educated, by a set of physical rules, i.e., a physical model. Thus, the machine does not only decern patterns from the data, but also from a model that describes the physical governing forces. The machine-educating approach is presented and is used here to analyze flooding datasets. The outcome of the study advances and improves S&R civil and military efforts in case of a flooding emergency.
