Suspended matter concentrations in coastal waters: Methodological improvements to quantify individual measurement uncertainty

Abstract

Measurements of total suspended matter (TSM) concentration and the discrimination of the particulate inorganic (PIM) and organic matter fraction by the loss on ignition methods are susceptible to significant and contradictory bias errors by: (a) retention of sea salt in the filter (despite washing with deionized water), and (b) filter material loss during washing and combustion procedures. Several methodological procedures are described to avoid or correct errors associated with these biases but no analysis of the final uncertainty for the overall mass concentration determination has yet been performed. Typically, the exact values of these errors are unknown and can only be estimated. Measurements were performed in coastal and estuarine waters of the German Bight that allowed the individual error for each sample to be determined with respect to a systematic mass offset. This was achieved by using different volumes of the sample and analyzing the mass over volume relationship by linear regression. The results showed that the variation in the mass offset is much larger than expected (mean mass offset: 0.85 ± 0.84 mg, range: −2.4 – 7.5 mg) and that it often leads to rather large relative errors even when TSM concentrations were high. Similarly large variations were found for the mass offset for PIM measurements. Correction with a mean offset determined with procedural control filters reduced the maximum error to <60%. The determination errors for the TSM concentration was <40% when three different volume were used, and for the majority of the samples the error was <10%. When six different volumes were used and outliers removed, the error was always <25%, very often errors of only a few percent were obtained. The approach proposed here can determine the individual determination error for each sample, is independent of bias errors, can be used for TSM and PIM determination, and allows individual quality control for samples from coastal and estuarine waters. It should be possible to use the approach in oceanic or fresh water environments as well. The possibility of individual quality control will allow mass-specific optical properties to be determined with better resolved uncertainties and, hence, lower statistical variability, greatly improving our capability to model inherent optical properties of natural particles and its natural variability, e.g. dependence on particle size and the complex refractive index.