Quantifying variability within water samples: The need for adequate subsampling

Accurate and precise determination of the concentration of nutrients and other substances in waterbodies is an essential requirement for supporting effective management and legislation. Owing primarily to logistic and financial constraints, however, national and regional agencies responsible for monitoring surface waters tend to quantify chemical indicators of water quality using a single sample from each waterbody, thus largely ignoring spatial variability. We show here that total sample variability, which comprises both analytical variability and within-sample heterogeneity, of a number of important chemical indicators of water quality (chlorophyll a, total phosphorus, total nitrogen, soluble molybdate-reactive phosphorus and dissolved inorganic nitrogen) varies significantly both over time and among determinands, and can be extremely high. Within-sample heterogeneity, whose mean contribution to total sample variability ranged between 62% and 100%, was significantly higher in samples taken from rivers compared with those from lakes, and was shown to be reduced by filtration. Our results show clearly that neither a single sample, nor even two sub-samples from that sample is adequate for the reliable, and statistically robust, detection of changes in the quality of surface waters. We recommend strongly that, in situations where it is practicable to take only a single sample from a waterbody, a minimum of three sub-samples are analysed from that sample for robust quantification of both the concentrations of determinands and total sample variability.