Error & Uncertainty
3/11/21, edited 5/25/21
Spatial research will always have uncertainty, but it is not always recognized during analyses. This class and my senior thesis have been the first time I have thought deeply about uncertainty. While it seems like uncertainty is most talked about in relation to measurement and analysis, Longley et al. (2008) describe uncertainty as “filters” between the real world, conception, measurement and representation, and analysis. Conceptions of the real world are steeped in cultural norms and ways of seeing the world. Even science, which is often seen as fact, is often deeply influenced by cultural norms. Then, quantifying one’s conception through measurement and representation can introduce uncertainty because some things can be really difficult to quantify. There can also be uncertainty introduced during the analysis of those quantified values. Analytical uncertainty often shows up in methodologies that perhaps calculate values differently from what you thought, or if you use packages that are not open source, it can be impossible to know how a value was transformed or calculated.
In my senior thesis, I primarily discuss gaps in data and uncertainty in normalization/estimation techniques (measurement and representation), uncertainty in analytical processes (analysis), and uncertainty in how the study I reproduced reported their methods. Uncertainty in reporting is another aspect that Longley does not discuss, which can make it difficult to reproduce a study or even fully understand it conceptually without actually reproducing it.
It is difficult to quantify uncertainty. Since uncertainty can increase at so many different steps of the data collection and analysis projects, and there are often so many numbers already to keep track of in GIS analyses, it sounds pretty difficult to have a running calculation of uncertainty as well. While this would be ideal, I think there may also be ways of estimating uncertainty after an analysis is completed—even if it is just identifying where uncertainty comes from and whether it is big or small. It is important that geographers report potential sources of uncertainty and estimate uncertainties the best they can while maintaining clear, good quality research. I think this is also where we see importance of clear and detailed documentation, so that even if geographers themselves are not calculating uncertainty, other geographers can still get a sense of what the uncertainty may be through looking at the methods and assumptions that went into the work. Creating a norm and ethic of open source work in geography would help with the transparency necessary to calculate, or at the very least better understand, uncertainty.
Sources
- Longley, P. A., M. F. Goodchild, D. J. Maguire, and D. W. Rhind. 2008. Geographical information systems and science 2nd ed. Chichester: Wiley. pp. 127-153