Errors are ubiquitous in current GIS. The word, error, can be used in a variety of senses. In 1995, Unwin defined it as the difference between reality and representation of the reality. Defined in this way, error is related to accuracy. Accuracy is the degree with the values or descriptions ofthe real-world features that they represent.
When field data are incorporated into a GIS, a common mistake is to assume that the error can be simply equated to the measurement error at the sampled points and quoted as a simple global statement, which does not address the spatial variation. Monckton in 1994 criticized that the spatially uniform error assumption was untenable. A complete specification of the error should include not only the spatial field of its mean but also its variance and spatial dependence. The problems of error and uncertainty in field models have been proved to be much hard-addressed in principle using well-developed theory.
The process of integrating remote sensing data into a GIS usually includes five steps, viz. data acquisition, data processing, data analysis, error assessment, and final product presentation. Error may accumulate throughout the process in an additive or multiplicative fashion. Data acquisition error may be from geometric aspects, sensor system, platforms, ground control, and/or scene considerations. Data processing error may be caused by geometric rectification such as resampling and data conversion such as from raster to vector format and from vector to raster format. Data analysis error may be from classification systems, data generalization, and quantitative analysis of relationships between data variables and the subsequent inferences that may be developed. Error of error assessment might be mainly produced by expression of locational accuracy, discrete multivariate, and reporting standards. Final product presentation error includes attribute error and spatial error that may be introduced through the use of base maps with different scales, different national horizontal datum in the source materials, and different minimum mapping units which are then resampled to a final minimum mapping unit.
The errors can be distinguished into inherent ones and operational ones. The inherent error is the error present in source documents. The operational error is accumulated through data capture and manipulation functions of a GIS. The inherent errors include errors from sampling and attribute errors in data source. Operational errors include positional errors and identification errors. Positional errors stem from inaccuracies in the horizontal placement of boundaries and identification errors occur when there is a mislabeling of areas on thematic maps. Spatial models as simulations of the real world often simplify the complexity of the real world and are therefore obviously open to errors. The inherent errors can be propagated through the simulation process and become manifest in the final products. Although there are many types and sources of error and uncertainty in geographical data and their processing, the problem is not simply technical and it arises from an evident inability of GIS. Integration of data from different sources and indifferent formats, at different original scales, plus inherent errors, can yield a product of questionable accuracy. Manipulation of thematic overlays within GIS to derive model variables are susceptible to inherent and operational errors, from which results may have such error margins as to be useless for specific applications. Any decision based on such products would thus be flawed.
The explosion of interest in scale has created many methods for scaling. For instance, interpolation brings multiple phenomena measured at different resolutions into a common coordinate grid with a single size. Multiple-variable scaling method simultaneously examines each variable at different scales. Spatially explicit models are simply maps of actual or simulated phenomena to demonstrate scale-sensitive issues. Fractal geometry is used to treat the dependence of various phenomena on scales. Resampling techniques are used to frame samples within a hierarchical framework to assess how scale and sequence of assembly affect ecosystem characteristics. Geostatistical techniques employ knowledge of the spatial covariance to produce a spatial model. Neural models are developed to test scale effects resulting from changes in grain size and spatial structure. Hierarchy theory is employed to address issues of spatial scale, which implies that an ecosystem is composed of interacting components and is itself a component of a larger system. However, they are not generalized in GIS as module.
Was this article helpful?