The SGSA algorithm, implemented in R, automates the detection of anomaly thresholds using an iterative statistical approach. It first projects log-transformed gas concentrations onto probability diagrams (Q-Q plots) against the theoretical quantiles of a standard normal distribution. A segmented regression is then applied to the sorted distribution to mathematically identify the inflection point (slope break), objectively defining the cutoff value that separates the background population from anomalous values.
Lefeuvre et al. (2026) studied this question.