Aquifer vulnerability mapping and GIS: A proposal to monitor uncertainty associated with spatial data processing

Abstract

An aquifer assessment was undertaken by the Geological Survey of Canada to estimate the sustainability and aquifer vulnerability in the St. Lawrence Lowlands of south western Quebec. The DRASTIC model and GIS was used to calculate and produce vulnerability maps. A detailed monitoring of data processing was performed to control the accuracy of the vulnerability maps. Overall estimates involved identifying errors and uncertainty associated with spatial and descriptive data used to run the model. The data analysed was related to wells, drillings, thematic maps, and also multiple processing data including errors and uncertainty attributed to calculations of the hydraulic conductivity, data interpolations, intersections of spatial data layers, etc. A categorization system using the Unified Modeling Language (UML) was proposed to categorize spatial data with respect to the degree and sources of possible uncertainties. This article presents the categorization system used, an example of an application for an study area and a discussion around its usefulness in controlling data processing (GIS and model integration). This work shows that uncertainty associated with spatial data processing and integrating data to a numerical system can be very significant, the main ambiguity occurring when cleaning data, interpolating, classifying and overlaying. Uncertainty characterization on the data processes was a valuable source of information. Monitoring the uncertainty associated with spatial data processing is almost more important to assemble than the model itself. However uncertainty monitoring may be complex and subjective and in fact it is rarely done on a regular basis mainly because it requires much more efforts compare to simply running the model.