Читать книгу GIS Research Methods - Steven J. Steinberg - Страница 98

Because a GIS involves technology that can use data acquired without ever leaving your office, it is theoretically possible to complete an entire analysis without ever leaving your desk. However, it would be foolish to believe such an analysis would be without flaws (especially considering the variable quality, scale, projections, data formats, etc., we discuss throughout this book). Even if all of your data can be acquired without fieldwork, it is a good idea to ground truth at key points in the process. Whenever you use GIS, you want to make sure that you ground truth, not only to ensure that the data you are using are appropriate to your question but also to validate that the results obtained in the GIS match what you find in the field. (Note that ground truthing of results, by necessity, comes after the analysis described in step 9, “analyze the data,” but we discuss it here because the issues are similar for both input data and results.)

In most cases, ground truthing means actually traveling to the place where the study is located to get a visual on the data or results. In situations where physically visiting the site may be difficult, alternative sources might be used to cross-validate your data against a second, reliable source. For example, if you are studying urban development patterns, you might be able to use current aerial photography or satellite images to confirm that a particular location has or has not been developed. If you can go to the field, you would want to take along a map or maps that represent data you are going to use in your GIS analysis and make sure your maps are both valid and accurate. Of course, you cannot check everything; but simply spot-checking or sampling from the map can go a long way toward ensuring that you start with good data.

In our pollution example discussed earlier in the chapter, ground truthing could help to answer questions about the form of the pollution and where it travels. For example, observing a smoke plume from a factory may show that a prevailing wind carries the smoke in a particular direction most of the time. Or a drive through each census block could tell you if the residents are evenly distributed within the polygon or if they are clustered in specific portions.

Finally, ground truthing is essential at the conclusion of an analysis when you are evaluating the results. Most GIS-based studies will pinpoint certain locations as having met a set of criteria. Again, spot-checking in the field can go a long way toward telling you if the results look correct and if they make sense. If your study is large, you may want to spot-check a number of sites to assess the overall accuracy of the analysis (e.g., if nine of ten results are correct when checked on the ground, it would be reasonable to estimate accuracy at 90 percent). The topic of accuracy assessment in spatial data analysis could fill an entire book, so we only mention it here and encourage those who are interested to pursue this topic independently.