Considerations of bias and precision associated with ISE based measurements of sediment ‘free’ sulfide

Abstract

This document utilizes literature and new data to examine the bias (systematic error), precision (non-systematic or random error; indicated by standard deviations and coefficients of variation (CV)) and total uncertainty or accuracy (includes bias and precision) associated with aspects of Ion Selective Electrode (ISE) measurements as applied to the determination of “free” sulfide in marine surficial sediment associated with aquaculture operations. The examination concludes that a careful and experienced analyst should be able to routinely achieve ISE sulfide measurements that have a total uncertainty of 10% or less when standard solutions, electrodes and samples are not stored, standard solutions are titrated and millivolt readings are made within 2 minutes of electrode insertion into a sample and electrodes are regularly checked and calibrated (at least once every 3 h). The examination also concludes that storage of standard solutions, storage of calibrated electrodes, storage of sediment samples and failing to account for sediment water content can all increase this uncertainty quite significantly, by increasing bias and reducing precision (increasing CVs). Bias is typically less than 10% and often about 5% when storage times are near zero, the bias is routinely greater than 10% when storage times exceed 4 h and can often exceed 20%. The actual magnitude and direction of bias varies between samples and storage times, and is not presently known or predictable, and as a result measurements on stored samples cannot be reliably adjusted for bias after the fact. The precision of ISE sulfide measurements is largely independent of sample storage time with values of the CV ranging from <5% to >30%. The precision associated with well conducted analyses is much greater (lower CVs) than the CVs obtained from multiple measurements from spatially separated in-situ sediment samples. Although highly variable, the CVs associated with spatially separated in situ measurements generally increases with the magnitude of the separation distance between samples. Variation on the scale of a fish farm can be of order 100% whereas the variation associated with a well conducted analysis that is not influenced by storage factors is of order 10%. The high CVs of the spatially separated measurements of in situ sediments is assumed to be generated by inherent natural variability in sediment characteristics on the scales of meters (within station variability) to hundreds of meters (within farm variability). The implication to environmental management is that well defined standard operating procedures need to be adopted that ideally do not include storage of standards, electrodes or samples and that environmental monitoring sample designs and summary statistics for sediment pore water sulfide need to robustly take into consideration the inherent natural variability of the environment so monitoring results are reproducible and hence credible.