A. Mkrtchian

Lviv National Ivan Franko University, Lviv, Ukraine


All maps are based on some primary data, which often have been derived from different sources characterized with varied accuracy, reliability, and precision. If this is the case, an obtaining of a map layer requires some kind of integration of primary data layers. Thus, the quality of semantic information conveyed by a map will depend both on characteristics of primary data layers and the method of their integration. GIS provides means to formalize the latter procedure, making it more rigorous and objective.

The methodology for data integration will differ for quantitative and qualitative (nominal) data. The general measure of accuracy for quantitative data is error variance measured for vector feature of raster pixel. Thus, the method of quantitative data different layers integration should account for this by assigning weights to data layers that are inversely proportional to their estimated error variance. We proposed such technique (realized in GIS environment) to integrate two layers of data on prospective plant species distribution, one of which (derived by a regression model) is characterized by a uniform error variance, while other (derived by kriging) has a spatially distributed variance. The resulting layer has smaller error variance than any of the primary layers. This approach can also be utilized for the mapping of quantitative soil characteristics, meteorological elements, ground water levels, etc.

The qualitative data layers can best be integrated using Bayes approach, when different data layers contain data on possibilities of a nominal categories set encountering for vector features or raster pixels. When primary data layers contain different sets of features, data integration will involve the spatial overlay of these layers; thereby the resulting possibilities will be calculated for the intersections of the features of primary layers. Thus, a map of the soil categories distribution can be created using two indicator layers: the layer of land surface morphology elements and the layer of vegetation types derived from satellite image classification.

The map layers obtained by such approaches are supplied by the measures of their uncertainty (often in form of the separate spatial layers), which greatly facilitates their utilization for decision making, safeguarding from making wrong decisions based on unreliable data, and also optimizes the process of data acquisition, allowing to identify areas with a deficiency of data.