Evaluation of seawater composition in a vast area from the Monte Carlo simulation of georeferenced information in a Bayesian framework

The detection of composition or pollution trends of vast environmental water areas, from a river, lake or sea, requires the determination of the mean concentration of the studied component in the studied area at defined depth in, at least, two occasions. Mean concentration estimates of a large area are robust to system heterogeneity and, if expressed with uncertainty, allow assessing if observed trends are meaningful or can be attributed to the measurement process. Mean concentration values and respective uncertainty are more accurately determined if various samples are collected from the studied area and if samples coordinates are considered. The spatial representation of concentration variation and the subsequent randomization of this model, given coordinates and samples analysis uncertainty, allows an improved characterization of studied area and the optimization of the sampling process. Recently, this evaluation methodology was described and implemented in a user-friendly MS-Excel file. This tool was upgraded to allow determinations close to zero concentration and "bottom-up" uncertainty evaluations of collected samples analysis. Since concentrations cannot be negative, this prior knowledge is merged with the original measurements in a Bayesian uncertainty evaluation that improves studied area description and sampling modelling. The Bayesian assessment avoids the underestimation of concentrations distribution by assuming that negative concentrations are impossible. This tool was successfully applied to the determination of reactive phosphate concentration in a vast ocean area of the Portuguese coast. The new version of the developed tool is made available as Supplementary Material. (C) 2020 Elsevier Ltd. All rights reserved.