Geographic sampling of urban soils for contaminant mapping: how many samples and from where

Properly sampling soils and mapping soil contamination in urban environments requires that impacts of spatial autocorrelation be taken into account. As spatial autocorrelation increases in an urban landscape, the amount of duplicate information contained in georeferenced data also increases, whether an entire population or some type of random sample drawn from that population is being analyzed, resulting in conventional power and sample size calculation formulae yielding incorrect sample size numbers vis-à-vis model-based inference. Griffith (in Annals, Association of American Geographers, 95, 740-760, 2005) exploits spatial statistical model specifications to formulate equations for estimating the necessary sample size needed to obtain some predetermined level of precision for an analysis of georeferenced data when implementing a tessellation stratified random sampling design, labeling this approach model-informed, since a model of latent spatial autocorrelation is required. This paper addresses issues of efficiency associated with these model-based results. It summarizes findings from a data collection exercise (soil samples collected from across Syracuse, NY), as well as from a set of resampling and from a set of simulation experiments following experimental design principles spelled out by Overton and Stehman (in Communications in Statistics: Theory and Methods, 22, 2641-2660). Guidelines are suggested concerning appropriate sample size (i.e., how many) and sampling network (i.e., where).