Spatial Optimization with Respect to Extreme Weather and Human Health: A Zoning with Multidimensional Objective for Environmental Studies

• Activity Number: 452
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Alexander Liss*+ and Elena Naumova
• Companies: Tufts University and Tufts University
• Keywords: zoning ; climate ; extreme weather ; environmental statistics ; spatial ; time series

Abstract

The spatial zoning, an NP-hard problem, is often discussed in relation to school or political redistricting with respect to population density. We discuss a heuristic spatial optimization algorithm to partition a large spatial extent (ex. Continental US) into p independent sub-regions suitable for environmental research. We initially relax the constraint of contiguity within each region, and use spatial Principal Component Analysis (sPCA) to reduce dimensionality. We built an objective function to balance conflicting requirements of environmental health studies, such as reducing climate heterogeneity with the desire to increase population within each region. We optimized number of regions by maximizing variance ratio criterion. We applied this methodology to partition Continental US into 8 regions based on population density and remote sensing vegetation index data and compared results with conventional climatic zoning. We demonstrated a superiority of suggested method to existing zoning for environmental health studies of extreme weather events based on vulnerability assessment of elderly population to extreme weather. We plan to extend this method to autoregressive time series.