Evaluating Lost Person Behavior Models
Elena Sava, Charles Richard Twardy, Robert J. Koester, Mukul Sonwalkar · Transactions in GIS · 2015
Abstract USwilderness search and rescue consumes thousands of person‐hours and millions of dollars annually. Timeliness is critical: the probability of success decreases substantially after 24 hours. Although over 90% of searches are quickly resolved by standard “reflex” tasks, the remainder require and reward intensive planning. Planning begins with a probability map showing where the lost person is likely to be found. TheMapScore project described here provides a way to evaluate probability maps using actual historical searches. In this work we generated probability maps theEuclidean distance tables in (Koester ), and usingDoke's ( ) watershed model.Watershed boundaries follow high terrain and may better reflect actual barriers to travel. We also created a third model using the joint distribution usingEuclidean and watershed features. On a metric where random maps score 0 and perfect maps score 1, the Euclidean distance model scored 0.78 (95%CI: 0.74–0.82, on 376 cases). The simple watershed model by itself was clearly inferior at 0.61, but the Combined model was slightly better at 0.81 (95%CI: 0.77–0.84).