Measuring differences between overlapping but unpaired spatial surveys using a geographic information system

To compare the results from two overlapping but unpaired synoptic surveys, a method using a geographic information system is proposed. The new method uses a GIS to create (i) minimum convex polygons (MCP) enclosing each set of survey data and (ii) Voronoi tesselations assigning area weights to each datum. The pairs of MCP and Voronoi maps are overlaid to produce a mosaic of polygons each with one datum assigned from each survey. The differences between the pairs weighted by the polygon areas provides the basis for statistical testing. Area-weighted means and variances of paired differences are computed and a z-statistic measures the significance of differences for the whole intersection area. A cross-product autocorrelation statistic provides an assessment of the spatial distribution of differences. Alternate, conventional methods are compared with the new method: analysis of variance (ANOVA), analysis of covariance (ANCOVA), and contour overlays. The four methods were applied to the comparison of two macrobenthic surveys conducted in Lake Erie. Measures for five different taxa were examined. The ANOVA and ANCOVA methods found many significant differences between surveys. Both methods were judged inappropriate as sampling data are not expected to be drawn from normally-distributed populations in spatial surveys. Differences between surveys were detected but were difficult to assess using contour overlays. The new method only found a significant difference in one taxon, an expected difference as members of that taxon were absent from most sites in one survey.