PARENT SESSION
Organized Oral Session 9: Multiscale advanced raster map analysis for ecosystem health monitoring, assessment, and management in the 21st Century
Organizer(s): GP Patil and W Myers
Monday, August 8, 1:30 PM - 5:00 PM, Meeting Room 511a, Level 5, Palais des congrès de Montréal

Detection and delineation of hotspots in the Florida Everglades.

Rathbun, Stephen^*,1, ¹ Penn State University, University Park, Pennsylvania, USA

ABSTRACT- The delineation of hotspots is essential for optimal allocation of environmental remediation resources. The spatial scan statistic has been developed for delineating circular hotspots. The upper level set scan statistic can delineate arbitrarily-shaped hotspots, but is only appropriate for data collected on a lattice such as formed by political (e.g., zip codes, counties, states, etc.), or natural units (e.g., watersheds, river networks, etc.). Moreover, both scan statistics assume that the data are independently distributed, an assumption that may be untenable for georeferenced data. A Bayesian framework is introduced for hotspot delineation that can be applied to geostatistical data collected in continuous space. The data are modeled in three stages: 1) a prior model for the spatial distribution of hotspots, 2) a data model for the spatial distribution of the observed data, conditional on the realized hotspot distribution, and 3) a prior model. A spatial probit model is used to describe the spatial distribution of hotspots, while a geostatistical model is considered for the conditional distribution of the observed data given the hotspot delineation. A Monte Carlo Markov Chain algorithm is developed for hotspot delineation. Under the Bayesian framework, all sources of variation are taken into account when assessing the uncertainty of spatial predictions. The proposed approach is illustrated using data collected from the South Florida Ecosystem Assessment Project.

Key words: Bayesian hotspot delineation, spatial scan statistics, spatial statistics