Robert Koester PhD


Robert Koester PhD

Robert J. Koester first joined the Appalachian Search & Rescue Conference in 1981 and since then has participated in hundreds of searches. He holds a PhD in search theory from the University of Portsmouth and is CEO of dbS Productions. His contributions to search and rescue include seminal research on lost person behavior (with emphasis on dementia), the International Search and Rescue Incident Database (ISRID), a the tactical decision aid software – FIND. A Search Mission Coordinator, instructor for the Virginia Department of Emergency Management, and past president of the Virginia Search and Rescue Council (15 years), Robert has also worked for the United States Coast Guard, National Aeronautics and Space Administration, National Park Service, Department of Homeland Security, and Federal Emergency Management Agency. Author of numerous books and articles on search and rescue. He has presented in Aruba, Australia, Iceland, Ireland, New Zealand, Poland, the United Kingdom, and throughout the United States and Canada.

Enhancements to statistical Probability of Area Models based upon updated ISRID data collection

12 Oct 3:00 PM - 3:45 PM

The International Search and Rescue Incident Database (ISRID) previously contained 50,000 incidents which was the basis for information found in Lost Person Behavior. A new round of data collection increased the size of ISRID to 145,000 incidents. This latest effort resulted in new data standards and a new data collection tool called Search and Rescue data Collection & Analysis Tool (SARCAT) to improve data quality. A new spatial model called the point model was created which provides the probability that the subject is found within 100 meters of the IPP or the destination. The value of combining different spatial models such as the ring model and distance from IPP as scored by MapScore has been determined to statistically significant. Several new subject categories are defined including multiple scenario based profiles (medical, trauma, investigative, avalanche, etc.). Additional new subject categories will be presented. A more detail analysis of Autism profile is also provided. Analysis of new emerging trends to predict spatial model statistics when data is missing for temperate or dry domain specific expression has been determined. Additional polar domain data also offers new insights. The new ISRID data has also been used to determine enhanced and more detail survivability curves out to the 95%. The integration of multiple models together along with expression of the Probability of Success Rate (PSR) in the FIND tactical decision aid will be discussed




New sweep widths values, correction factors, models, UAVs values, and detection model validation

13 Oct 11:00 AM - 11:45 AM

Search theory allows for correction factors to account for conditions (such as night) that affect the sweep width value. Search theory also predicts that the coverage is proportional to the probability of detection by either the inverse cube curve or exponential function (random search). The objective is to determine the correction factors from night searching, use of IR detection devices and validate the coverage curves.
Effective Sweep Width experiments were conducted in the same location with the same medium visibility adult-sized targets during both the day and the night in a temperate forest. Additional experiments examined the impact of one, two, and three person teams. In addition, high and low visibility clue-sized objects were placed directly on the trail. Low visibility human subjects were used for infrared detection.
We found an effective sweep width of 64 meters during the daytime and 22 meters at night for a correction factor of 0.34 for the adult-sized targets. Both high (100% vs 94%) and low (83% vs 43%) visibility clues were more detectable during the day versus night (P<0.001). Searchers with dim flashlights (<200 lux at one meter) resulted in an additional correction factor of 0.5. The use of IR at night increased the sweep width to 68 meters. Two searchers increased the sweep width value by 1.3, however, three searchers didn’t see a statistical increase. The probability of detection versus coverage plots of both day and night experiments fell between the inverse cube and exponential curves.
Additional experiments have been conducted to examine the sweep width factors generated from both electro-optical and Infrared sensors on a UAV platform. Sweep width values and important correction factors will be examined which will impact search patterns.This single experiment for only one visibility class of search target showed that visual searching is significantly degraded by searching at night. The daytime coverage suggests the inverse cube model while the night coverage suggests an interim result. The use of effective sweep width, correction factors, and validated coverage curves can lead to more accurate assessments of the probability of detection.