Abstract
Many organisations collect spatial data at a range of scales for a variety of purposes. Scientists apply new environmental knowledge either by improving existing simulation models or developing new ones. Technological advances, and the increasing availability of both models and the data required to run them, are potentially putting better information in the hands of environmental policy analysts, decision makers and land managers to solve environmental problems. A review of the literature, however, shows that each of this trio of Data–Model--Problem is scale dependent, and that this scale dependence can be very significant, potentially leading to the generation of information that is misleading or simply invalid. The research objective is to contribute to the area of spatial modelling by investigating how a best practice of applying environmental data and models, which considers the effects of scale, might be achieved. It is proposed that a systematic framework (dubbed the 'Scale Matcher') can be developed to identify and match the scale requirements of a problem with the scale limitations of spatial data and models. A bottom-up approach is taken by breaking scale into its constituent parts (extent, accuracy and precision), each of which is then analysed in some depth to understand the various ways in which scale can impact on data and simulation models. Each component of scale is quantified by identifying suitable metrics from the literature.
Two new indices that take into account attribute imprecision are introduced. Spatial confusion measures the degree of overlap of the attribute values between mapunits. Colour purity measures the proportion of a map that is incorrectly shaded due to within-mapunit imprecision. Both measures can be calculated for continuous and categorical, raster and vector data, providing there is a parametric or empirical estimate of the distribution of sub-mapunit variability.
The Scale Matcher comprises a set of 11 comparisons, derived from a systematic match between each combination of the scale components of data, model, and problem. Each match is named and described. In some cases the matches are simple comparisons of the relevant metrics. Others require the combination of data variability and model sensitivity to be investigated. By randomly simulating data and model imprecision and inaccuracy, sensitivity analysis can then be used to identify scales over which scale effects are minor or stable. The Scale Matcher procedure is specified in a flowchart format that leads the user through the relevant sequence of matches. Listing the results as a set of scale assumptions helps to draw attention to them making users more aware of the limitations of spatial modelling.
The Scale Matcher is evaluated by assessing scale compatibility in a case study of nitrate leaching vulnerability. Some hypothetical examples are used to verify other routes through the procedure. The case study requires the development of some new representations of soil imprecision and inaccuracy, and loosely coupling a sensitivity analysis tool with a geographic information system. It was concluded that the scale-matching framework successfully broke down the scale issue into a series of comparisons, which, if performed should give the user more confidence in the scale validity of model output for a given problem.