Applications of location analysis (Q501104)

As a companion to their theory oriented book [Foundations of location analysis. Berlin: Springer (2011; Zbl 1351.90007)] the same editors collect here 16 contributions on typical and important applications of this active subfield of operations research, emphasizing the particularities of each sector and the modeling and solving tools effectively used in the decision-making process. Each chapter focuses on one particular application area, surveys the literature, usually offering a classification and/or details of existing model types, often proposes an additional encompassing model together with an explicit case and mostly avoids the analysis of mathematical properties and/or detailed solution techniques. We briefly indicate the main topic of each contribution below. Bank branches ask for locations maximizing net profit, which is almost impossible to predict. The case uses instead multicriteria estimated relative influence of transaction volumes, obtained through segmented study of customer location and behavior, and locational costs due to proximity between open branches, opening new and closing old ones. The resulting large scale MIP model is solved heuristically with subsequent sensitivity analysis. With the claim `location in practice is 80\ is described as a four step qualitative managerial approach, comprising 1.~situational analysis of the region of interest, 2.~initial selection of candidate sites, 3.~readiness assessment of each site on multiple factors and 4.~a comparative summary stating strengths and weaknesses of each site. The wide spatial dimension in industrial forestry leads to a hierarchy of interrelated mostly large scale MIP location models covering the strategic long term decisions of where/when to harvest and to process the wood, tactical layout of roads for transports and the shorter term operational organisation within each harvest area. Time efficiency within a hospital calls for an internal layout aimed at reducing travel-distances of patients, personnel and material during the multiple flow processes that constitute health care. But the surveyed and already wealthy literature on optimized layout planning seems not yet to have convinced the main actors in the sector. Anti-covering location models allow to determine the maximum capacity of an area for a territorial species, e.g.~owls. These seek the maximum number among potential sites that lie sufficiently far from each other. The dramatic effects of forest fires may be reduced by preventive fuel reduction treatment of some areas. A decision support system was designed for the selection and scheduling of these areas, optimizing several user-specified objectives such as protected volume, dispersion or concentration of treatments in time and space, possibly with prioritized areas. Fusion of data improves crash detection on roads by automated sensors. Location of sensors should therefore optimize single and double (or higher) coverage. Three steps of increased approximation quality are proposed: node-based exact optimization, a node and path-based heuristic and a simulation based optimization step. Design of a preventive health care network should maximize the level of voluntary participation primarily by its accessibility, the main location theoretic objective, but practice shows that workload needs to be taken into account, as well by a minimum activity level for cost-effectiveness, as through capacity to avoid discouraging waiting queue length. Retail activity revenue modeling needs not only readily available residence location data but also usually unknown employment sites and choice behavior data. It is shown these latter may, however, be estimated by methods inspired by econometrics. Rural school system design is basically a location-allocation question, but has to account for many complicating factors, such as parental choice, size constraints, shift allocation, transportation, staffing, districting rules, continuous serving zones, student profile balancing, yielding complex large scale MIP models. Fire stations should in principle be sited so as to make their full response region reachable within a critical time, a goal usually unreachable at acceptable cost. Models of maximum covering type are therefore used in practice. These depend on socio-economic spatial data, that change with urban development, so should be regularly reevaluated. Real-time estimation of road travel times may be done by way of automated vehicle recognition sensors. Models for optimally locating such sensors consider origin-destination pairs and the routes between them to be covered by at least a pair of sensors, the more the better, and the distance between them may be used as quality measure of the time-estimation. Districts of all kinds have often to be designed so as to be balanced in terms of volume (measured geographically, by population or workload), contiguous (`of one piece') and compact (not elongated). Traditional \(p\)-median like models are strongly complicated by incorporation of these partly conflicting objectives. Siting and sizing prisons is mostly handled politically, often mainly considering local opposition. A model mixing several other objectives is discussed, followed by minimization of regret based on an extensive scenario analysis. The many covering type models for emergency services may be adapted to maritime search and rescue, provided distance is calculated by land-avoidance, and incident severity and rescue vessel heterogeneities are incorporated. Finally, the military field offers many very disparate application opportunities for location modelling. These vary e.g.~from material and weapon positioning, through protection, search and rescue aircraft siting, to closure of military bases.