Stochastic and Partly-determined Statistical Perturbation of Points, Polygons and Surfaces to Assess Certainty of Environmental Modelling

摘要

With the advent of geographical information systems (GIS) many practicioners are attempting to quantify spatial associations between natural features. In some cases the base data have been generated at different scales and the tools provided to allow manipulation of spatially-referenced data allow overlay of almost any coverages. Moreover, the accuracy of the spatial data in any coverage is variable and often not quantified. We present here an approach that relies on perturbing the underlying spatial data and comparing measurements of spatial association before and after perturbation. The stability of the measurement of the spatial association is shown by the change in the value and, in some cases, its proximity to a result represented by randomness. In cases where confidence limits can be generated by iteration, we term the bounding of the original function by the limits the 'uncertainty envelope'. Examples of perturbation of points and polygons are presented. We also briefly explore the use of perturbation of continuous data as a means of studying land management options.