Provenance, Error, and Uncertainty
Longley et. al’s chapter on uncertainty shows the many filters through which reality passes when researchers attempt to distill questions, observations, and findings into a published study. This compounding distortion is represented well in Figure 6.1––as a study passes through each stage, it passes through subjective decisions that ultimately take it further from reality. This illustrates well the importance of reproduction studies. It is impossible to completely eliminate subjectivity in geographic studies; carrying out a study multiple times from a diversity of approaches is the next-best way to generate representative research and ensure a degree of accuracy.
As a student studying Geography and Global Health, I have been made aware of many areas of uncertainty and strategies to minimize it. I very recently joined a research team working with data about households in parts of Africa using data compiled by the Wildlife Conservation Society. Working with information from an outside source limits the scope of our research from the outset, and forces the team to work around its flaws. The step between a larger organization and the smaller team using its data creates areas of uncertainty as to which data is most accurate and narrows the approaches we can take, similar to the graphic shown in Longley et al’s chapter.
There were many places in Malcomb et al’s study that could contribute to levels of uncertainty––all have to do with decisions made by the researchers. One example, a topic that is always a difficult one in geographic research, was deciding how to visualize boundaries. Malcomb et al used a raster grid, but it is unclear how they decided grid cell size. Another example is the normalization of their interview data; in order to compile the data in a common geographic frame they weighted each of eighteen different vulnerability indicators, however data ranged from “type of cooking fuel” to “own a cell phone” to “exposition to drought events. The transformation of these indicators into a weighted hierarchy is not fully described in the paper and is another example of a method involving many opportunities for uncertainty mentioned in Longley et al’s paper, such as vagueness and ambiguity.