Stochastic simulation and probabilistic location allocation modeling for a case study of missing source points. George Charisoulis Barbara P. Buttenfield Yi Qiang

Keywords: Stochastic simulation, location allocation, missing aggregated data, bootstrap simulation
Abstract Type: Paper Abstract

Abstract

Missing or aggregated source data has long hampered reliable determination of location allocation modeling results (Goodchild 1979, Murray and Gottsegan 1997). Lacking precise locations of source points, shortest path distances are either inferred or else identified at extremely coarse spatial resolution. This paper reports a pilot study utilizing stochastic simulation to identify probable locations of source points. The case study describes a location modelling problem to determine a suitable location for an anerobic digestor in a semi-rural county. Anaerobic digestion is tightly coupled with farming due to reliance on organic waste management. The US Department of Agriculture (USDA) distributes farm locations in a size-stratified format aggregated by county. Individual farm locations are available from local tax assessors or nationally from private firms in parcel databases that are proprietary and often expensive. The solution to be presented generates candidate polygons by GIS modeling of several constraining data layers such as land use and ownership, distance to water and road networks, terrain and other variables. Farms are allocated stochastically to candidate polygons in a Monte Carlo simulation matching the USDA size stratification, building a frequency distribution of farm acreage, biowaste volume (based on farm type and size) and potential energy yield. The allocation process is validated by comparing probabilistic results with actual parcel data provided through a university partnership with a national firm. Location modeling proceeds by aggregating shortest path travel distances from candidate polygons to identify possible locations for the anaerobic digestor that meet one of two solutions (efficiency and equity).