Doctoral student Paul Doherty, under the supervision of Professor Qinghua Guo in the School of Engineering at the University of California Merced will develop a system designed to estimate the location of injured and/or lost people in Yosemite National Park based on historical data of where people have been injured and/or found in past rescue events. This project has three research components. The first component involves the process of georeferencing historic incidents from text-based information. This is a common challenge for users of geographic information (ecologists, historians, curators) who rely on spatial data to make decisions. A probability-field georeferencing technique will be used to map historic search and rescue incidents in Yosemite National Park. The second component of this project evaluates the geographic one-class data issue faced by geographers with presence-only test data. More specifically, a model whose parameters are defined by expert knowledge (helicopter managers) will be compared to a model that uses presence-only test points (known landing areas) to develop helicopter landing area suitability layers in Yosemite National Park. This will project will provide a novel testing scenario for dealing with presence-only or geographic one-class data using machine-learning algorithms. The third component of this project will address the problem of locating moving objects in spatiotemporal space. This entails creating a travel-cost layer that uses slope, vegetation density, and path network presence to estimate the time required to cross a known distance within Yosemite National Park. Once this has been completed and tested, an object-oriented model will be constructed to use this travel-cost layer to generate isochrones (time-distance rings). This will allow a user to enter a geographic point and time elapsed to determine the maximum distance a human can travel from a known location.
The logical problem to be addressed by this research is an entirely spatial one, "how can developments in geography and spatial sciences help rescuers find and rescue visitors in Yosemite National Park more effectively?" In essence, this research will develop tools for emergency service providers to better do their job in three ways: enable institutional knowledge from incident history in a spatially-enabled digital library format, generate a visualization tool for emergency communication dispatchers and helicopter pilots to locate landing areas, and give search managers a tool for defining the outer limits of their search area when persons go missing. The classic dilemma of search and rescue is a puzzle that professional rescuers, their rescuees, and loved-ones take very seriously. Geographic Information Systems can provide a platform for spatial analyses to help solve problems so that others may live. This research will introduce science-based techniques to provide society with safer and more effective rescue techniques.
The process of searching for and rescuing people in distress provides an appealing spatial problem for geographers to support and allows for the testing theoretical developments in the real-world. Essentially, the fundamental goals of search and rescue are to: locate persons in need, care for them, and extract them from dangerous situations. More recently, SAR researchers and professionals have also cited a need for proactive incident prevention as a critical responsibility, known as preventative search and rescue (PSAR). Each component of SAR is inherently spatial and should be evaluated with respect to emerging technology and theoretical advances in spatial sciences. If Geographic Information Science (GIScience) is the theory behind the development, use, and application of geographic information systems (GISystems), then SAR is an ideal topic for GIScientists to study. Below are abstracts from our NSF-funded research and just a few examples of how GIS can be applied to SAR and how we can learn much about GIScience in the process. Future research will focus on search theory and more advanced forecasting analyses. Georeferencing from text The Search and Rescue (SAR) of individuals who become lost, injured, or stranded in wilderness presents a unique and worthwhile spatiotemporal challenge to investigate. Once incidents are georeferenced they can be spatially queried and analyzed. However, one major challenge for evaluating SAR in a spatial context is the lack of explicitly spatial data (addresses or coordinates) for historic incidents; they must be georeferenced from textual descriptions. This study implemented two established approaches for georeferencing incidents, the ‘Point-Radius’ and ‘Shape’ methods. Incorporating uncertainty measurements into a spatial database allows for more appropriate analyses of spatial dependence and the spatial distribution of incidents. From 2005 – 2010, 1271 of 1356 Yosemite Search and Rescue YOSAR incidents (93.7%) could be georeferenced using the Point-Radius Method, with mean uncertainty radius = 560 ± 51 meters. However, when the Shape Method was applied to six case studies by considering the reference object shape, the uncertainty areas were reduced considerably but required a significant increase in time investment. For practical purposes, to generate a SAR database from archived text, the Point-Radius Method is recommended. This is the first spatially?explicit study of SAR incidents and yields valuable insights into the role of georeferenced data in emergency preparedness. Helicopter Landing Suitability mapping using machine-learning Landing a rescue helicopter in a wilderness environment, such as Yosemite National Park, requires suitable areas that are flat, devoid of tree canopy, and not within close proximity to other hazards. The objective of this study was to identify helicopter landing that areas are most likely to exist based on available geographic data using two GIScience methods. The first approach produced an expert model that was derived from pre-defined feature constraints based on existing knowledge of helicopter landing area requirements (weighted overlay algorithm). The second model is derived using a machine learning technique (maximum entropy algorithm, Maxent) that derives feature constraints from existing presence-only points, i.e. geographic one-class data. Both models yielded similar output and successfully classified test coordinates, however Maxent was more efficient and required no user-defined weighting that is typically subject to human bias or disagreement. The pros and cons of each approach are discussed and the comparison reveals important considerations for a variety of future land suitability studies, including ecological niche modeling. The conclusion is that the two approaches complement each other. Overall, we produced an effective GISystem product to support the identification of suitable landing areas in emergent rescue situations. To our knowledge, this is the first GIScience study focused on estimating the location of landing zones for a search and rescue application. Forecasting events & Geographic One-Class Data In order to address a spatio-temporal challenge such as incident prevention, we need information about the time and place where previous incidents have occurred. Using geographic coordinates of incidents occurring in the past in coincidence with spatial layers corresponding to environmental variables, we can produce probability maps in geographic and temporal space. Here we evaluate spatial statistic and machine learning approaches to answer an important space-time question: where and when are search and rescue (SAR) incidents most likely to occur within Yosemite National Park in the future? We have produced a probability map based on the presence and background learning algorithm (PBL) that successfully forecasts the location of future SAR incidents based on environmental variables (distance to anthropogenic & natural features, vegetation, elevation, and slope). This will allow decision-makers to allocate resources where and when incidents are most likely to occur. In the process we not only answered questions related to a real-world problem, we used novel space-time analyses that help us learn more about what factors are associated with SAR occurrence. The GIScience findings from this applied research have implications for best-practices in future space-time research in the fields of epidemiology and ecological niche modelling.
