Benthic sulfide concentration is a common proxy for dissolved oxygen. It is often used to classify benthic health as Oxic, Hypoxic, or Anoxic for regulatory compliance of marine finfish aquaculture. This site classification can dictate if mitigative action is required, such as changes in farm management practices or production levels.
Extensive variability is common with environmental physicochemical metrics such as sulfides. Accurate measurements necessitate replicate sample collection to determine measures of central tendency (i.e., mean or median). Scientific research typically aims to confirm or refute a hypothesis through inferential statistics based on an acceptable level of confidence. However, most regulatory environmental monitoring does not achieve the statistical power necessary for valid inferential comparisons. This is often because, determining if there is an impact is not in question and measures are used to scale an effect. It could be that a small number of samples are collected in an exploratory nature with results used to trigger a larger sampling regimen, if needed. It may also be that the ideal number of replicates is cost prohibitive or impractical to achieve. Sulfide measurements for benthic classification arguably meet all these criteria, being notoriously difficult to measure, highly variable, and expensive. Consequently, descriptive statistical measures such as means are common sulfide reporting metrics applied for regulatory compliance of marine finfish aquaculture in many jurisdictions. However, sulfide data distributions are also often skewed. This begs the question: would sulfide medians be a more accurate measure for regulatory compliance given data distribution potential for skewness? If so, would medians favour more stringent or more lenient regulatory compliance?
In Nova Scotia, Canada, benthic sulfide data has been collected at marine finfish aquaculture sites annually since 2002. In this jurisdiction, means are applied to replicates to determine sulfide values at individual sample locations, and the mean of sample locations are used to classify farm site condition. A total of 193 historical sulfide site assessments were reanalyzed, applying medians to sample values and sample locations. An R dashboard was created to enable ease of plot generation for individual years and sites, and to track changes in classification, upon the application of medians. If medians were applied to only sample values alone, 3% of sites transitioned to a less impacted classification. If medians were applied to both sample values and sample locations, almost 10% of site classifications changed, with most transitioning to a less impacted classification. Implications on environmental monitoring programs of finfish aquaculture are discussed in the context of traditional sulfide monitoring and new more accurate methods being proposed.