|
|
Error Issues | |
|
|
Understanding precision and accuracy of the data is important for any project as it determines the types of questions that can be asked and analyses that can be preformed. For our project, although precision is less of an issue than accuracy; both are discussed separately below. Precision A high degree of digitizing precision was not critical given the scale, the somewhat inaccurate data source from the highly generalized, small-scale 1:500000 map, and our level of inquiry. The intended use of this dataset is, as already discussed, to determine the general spatial patterns and relationships between elevation and features which may possibly contribute to a much finer grained, larger-scale hypothesis and research study being done for the area. Data accuracy was an issue at all levels of this project. As the Shetsen (1987) base map was already full of data inaccuracies, with geology attributes derived from little ground truth, very old geology maps, and air photo interpretations, the digitized dataset would be inheriting the original errors. Data inaccuracy affects the reliability of our own interpretations. Further, it is important that any future persons using the created project data are aware of the errors behind it so as to avoid inappropriate data usage. Therefore, the next section below (5.3) is a detailed discussion of the various error sources in our project. Despite the wide range of error sources, we do not feel that, given the scale and analysis objective, the accumulated error negates the intent for which this project is intended; that is, to identify large patterns and relationships through display. According to ArcInfo Help sources, acceptable RMS errors should range between 0.000 – ±0.0050 whereas the readings that we obtaining ranged from –0.056 - 0.063. About five or six attempts were consequently made in reregistering the master tick coverage and subsequent base map coverages, trying to lower the RMS errors. Although low RMS errors were achieved with some ticks (ranging from 0.000-0.005 in digitizer units) a few were between –0.040 –±0.044, values that were subsequently accepted for the new coverages (see Appendix 3, Fig. 2). About a dozen polygons were unintentionally removed during the final dissolve operation, where adjacent polygons were improperly labeled with the same soiltype id. As well, one polygon remains unlabeled and is given a soiltype id value of 0, and six slivers are undeleted. These errors were not corrected due to pressing time constraints and are in the coverages used for our project analysis.
|
