https://core.ac.uk/download/pdf/96667096.pdf
In the research paper written by Guttorm Alendal, “Using Bayes Theorem to Quantify and Reduce Uncertainties when Monitoring Varying Marine Environments for Indications of a Leak”, the paper discusses a real-life application of the Bayes theorem. This paper was presented at the 13th International Conference on Greenhouse Gas Control Technologies. The paper is mainly concerned with finding a way to correctly predict whether a CO2 leak is happening on the seafloor while minimizing the number of false positives. Marine operations are very expensive, so false positives are considered a huge sink in the budget. And at the same time, given the severe damage that a CO2 leak can cause to the surrounding marine habitat, they cannot be left alone.
The paper presents the different aspects of a leak that needs to be taken into account to make accurate predictions. The paper discusses how leaks can appear in multiple forms, either dissolved or in bubbles of various configurations. For different environments, there also needs to be different thresholds for concern, which is prior in the Bayes theorem for this study. Sensors already exist to detect whether a leak is present, but Bayes will allow the calculations to produce a probability on how likely there is a leak, as opposed to something else raising the CO2 levels in the area. The Bayes Theorem used here is
Which can be read as given measurement or reading of x, how likely is there not a leak in the area. Looking at historic examples, the calculation here also takes into account the time of the year as the overall concentration of CO2 according to the season. The presented method has a flaw, where the researcher points out that certain months have more accurate predictions than others. This links back to what we’ve discussed in class about the Bayes theorem, and how the underlying force of Bayes theorem is using prior knowledge to inform future actions. The formula is the same as the one we’ve been using in class, where the bottom of the formula is the chance of x happening, and the top of the equation is the chase of x and -L happening together. Using the prior information, in this case, historic data on leaks, the researchers were able to create a different model for predicting the probability of future leaks.